diff --git a/.dockerignore b/.dockerignore new file mode 100644 index 0000000..39d8893 --- /dev/null +++ b/.dockerignore @@ -0,0 +1,6 @@ +* +!LICENSE +!Setup.hs +!ShellCheck.cabal +!shellcheck.hs +!src diff --git a/.github/ISSUE_TEMPLATE.md b/.github/ISSUE_TEMPLATE.md index 4fc3f9f..493b465 100644 --- a/.github/ISSUE_TEMPLATE.md +++ b/.github/ISSUE_TEMPLATE.md @@ -1,11 +1,11 @@ #### For bugs -- Rule Id (if any, e.g. SC1000): -- My shellcheck version (`shellcheck --version` or "online"): -- [ ] I read the issue's wiki page, e.g. https://github.com/koalaman/shellcheck/wiki/SC2086 -- [ ] I tried on shellcheck.net and verified that this is still a problem on the latest commit +- Rule Id (if any, e.g. SC1000): +- My shellcheck version (`shellcheck --version` or "online"): +- [ ] The rule's wiki page does not already cover this (e.g. https://shellcheck.net/wiki/SC2086) +- [ ] I tried on https://www.shellcheck.net/ and verified that this is still a problem on the latest commit #### For new checks and feature suggestions -- [ ] shellcheck.net (i.e. the latest commit) currently gives no useful warnings about this +- [ ] https://www.shellcheck.net/ (i.e. the latest commit) currently gives no useful warnings about this - [ ] I searched through https://github.com/koalaman/shellcheck/issues and didn't find anything related diff --git a/.github/dependabot.yml b/.github/dependabot.yml new file mode 100644 index 0000000..81bae9a --- /dev/null +++ b/.github/dependabot.yml @@ -0,0 +1,7 @@ +version: 2 + +updates: + - package-ecosystem: "github-actions" + directory: "/" + schedule: + interval: "daily" diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml new file mode 100644 index 0000000..83269c9 --- /dev/null +++ b/.github/workflows/build.yml @@ -0,0 +1,159 @@ +name: Build ShellCheck + +# Run this workflow every time a new commit pushed to your repository +on: push + +jobs: + package_source: + name: Package Source Code + runs-on: ubuntu-latest + steps: + - name: Install Dependencies + run: | + sudo apt-get update + sudo apt-mark manual ghc # Don't bother installing ghc just to tar up source + sudo apt-get install cabal-install + + - name: Checkout repository + uses: actions/checkout@v4 + with: + fetch-depth: 0 + + - name: Deduce tags + run: | + mkdir source + echo "latest" > source/tags + if tag=$(git describe --exact-match --tags) + then + echo "stable" >> source/tags + echo "$tag" >> source/tags + fi + cat source/tags + + - name: Package Source + run: | + grep "stable" source/tags || ./setgitversion + cabal sdist + mv dist-newstyle/sdist/*.tar.gz source/source.tar.gz + + - name: Upload artifact + uses: actions/upload-artifact@v4 + with: + name: source + path: source/ + + run_tests: + name: Run tests + needs: package_source + runs-on: ubuntu-latest + steps: + - name: Download artifacts + uses: actions/download-artifact@v4 + + - name: Install dependencies + run: | + sudo apt-get update && sudo apt-get install ghc cabal-install + cabal update + + - name: Unpack source + run: | + cd source + tar xvf source.tar.gz --strip-components=1 + + - name: Build and run tests + run: | + cd source + cabal test + + build_source: + name: Build + needs: package_source + strategy: + matrix: + build: [linux.x86_64, linux.aarch64, linux.armv6hf, linux.riscv64, darwin.x86_64, darwin.aarch64, windows.x86_64] + runs-on: ubuntu-latest + steps: + - name: Checkout repository + uses: actions/checkout@v4 + + - name: Download artifacts + uses: actions/download-artifact@v4 + + - name: Build source + run: | + mkdir -p bin + mkdir -p bin/${{matrix.build}} + ( cd bin && ../build/run_builder ../source/source.tar.gz ../build/${{matrix.build}} ) + + - name: Upload artifact + uses: actions/upload-artifact@v4 + with: + name: ${{matrix.build}}.bin + path: bin/ + + package_binary: + name: Package Binaries + needs: build_source + runs-on: ubuntu-latest + steps: + - name: Checkout repository + uses: actions/checkout@v4 + + - name: Download artifacts + uses: actions/download-artifact@v4 + + - name: Work around GitHub permissions bug + run: chmod +x *.bin/*/shellcheck* + + - name: Package binaries + run: | + export TAGS="$(cat source/tags)" + mkdir -p deploy + cp -r *.bin/* deploy + cd deploy + ../.prepare_deploy + rm -rf */ README* LICENSE* + + - name: Upload artifact + uses: actions/upload-artifact@v4 + with: + name: deploy + path: deploy/ + + deploy: + name: Deploy binaries + needs: package_binary + runs-on: ubuntu-latest + environment: Deploy + steps: + - name: Install Dependencies + run: | + sudo apt-get update + sudo apt-get install hub + + - name: Checkout repository + uses: actions/checkout@v4 + + - name: Download artifacts + uses: actions/download-artifact@v4 + + - name: Upload to GitHub + env: + GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} + run: | + export TAGS="$(cat source/tags)" + ./.github_deploy + + - name: Waiting for GitHub to replicate uploaded releases + run: | + sleep 300 + + - name: Upload to Docker Hub + env: + DOCKER_USERNAME: ${{ secrets.DOCKER_USERNAME }} + DOCKER_PASSWORD: ${{ secrets.DOCKER_PASSWORD }} + DOCKER_EMAIL: ${{ secrets.DOCKER_EMAIL }} + DOCKER_BASE: ${{ secrets.DOCKER_USERNAME }}/shellcheck + run: | + export TAGS="$(cat source/tags)" + ( source ./.multi_arch_docker && set -eux && multi_arch_docker::main ) diff --git a/.github_deploy b/.github_deploy new file mode 100755 index 0000000..82c8ec5 --- /dev/null +++ b/.github_deploy @@ -0,0 +1,28 @@ +#!/bin/bash +set -x +shopt -s extglob + +export EDITOR="touch" + +# Sanity check +gh --version || exit 1 +hub release show latest || exit 1 + +for tag in $TAGS +do + if ! hub release show "$tag" + then + echo "Creating new release $tag" + git show --no-patch --format='format:%B' > description + hub release create -F description "$tag" + fi + + files=() + for file in deploy/* + do + [[ $file == *.@(xz|gz|zip) ]] || continue + [[ $file == *"$tag"* ]] || continue + files+=("$file") + done + gh release upload "$tag" "${files[@]}" --clobber || exit 1 +done diff --git a/.gitignore b/.gitignore index 6d5f1ae..cf373a8 100644 --- a/.gitignore +++ b/.gitignore @@ -20,3 +20,4 @@ cabal.config /parts/ /prime/ *.snap +/dist-newstyle/ diff --git a/.multi_arch_docker b/.multi_arch_docker new file mode 100755 index 0000000..81048a2 --- /dev/null +++ b/.multi_arch_docker @@ -0,0 +1,89 @@ +#!/bin/bash +# This script builds and deploys multi-architecture docker images from the +# binaries previously built and deployed to GitHub. + +function multi_arch_docker::install_docker_buildx() { + # Install QEMU multi-architecture support for docker buildx. + docker run --rm --privileged multiarch/qemu-user-static --reset -p yes + + # Instantiate docker buildx builder with multi-architecture support. + docker buildx create --name mybuilder + docker buildx use mybuilder + # Start up buildx and verify that all is OK. + docker buildx inspect --bootstrap +} + +# Log in to Docker Hub for deployment. +function multi_arch_docker::login_to_docker_hub() { + echo "$DOCKER_PASSWORD" | docker login -u="$DOCKER_USERNAME" --password-stdin +} + +# Run buildx build and push. Passed in arguments augment the command line. +function multi_arch_docker::buildx() { + mkdir -p /tmp/empty + docker buildx build \ + --platform "${DOCKER_PLATFORMS// /,}" \ + --push \ + --progress plain \ + -f Dockerfile.multi-arch \ + "$@" \ + /tmp/empty + rmdir /tmp/empty +} + +# Build and push plain and alpine docker images for all tags. +function multi_arch_docker::build_and_push_all() { + for tag in $TAGS; do + multi_arch_docker::buildx -t "$DOCKER_BASE:$tag" --build-arg "tag=$tag" + multi_arch_docker::buildx -t "$DOCKER_BASE-alpine:$tag" \ + --build-arg "tag=$tag" --target alpine + done +} + +# Test all pushed docker images. +function multi_arch_docker::test_all() { + printf '%s\n' "#!/bin/sh" "echo 'hello world'" > myscript + + for platform in $DOCKER_PLATFORMS; do + for tag in $TAGS; do + for ext in '-alpine' ''; do + image="${DOCKER_BASE}${ext}:${tag}" + msg="Testing docker image $image on platform $platform" + line="${msg//?/=}" + printf '\n%s\n%s\n%s\n' "${line}" "${msg}" "${line}" + docker pull -q --platform "$platform" "$image" + if [ -n "$ext" ]; then + echo -n "Image architecture: " + docker run --rm --entrypoint /bin/sh "$image" -c 'uname -m' + version=$(docker run --rm "$image" shellcheck --version \ + | grep 'version:') + else + version=$(docker run --rm "$image" --version | grep 'version:') + fi + version=${version/#version: /v} + echo "shellcheck version: $version" + if [[ ! ("$tag" =~ ^(latest|stable)$) && "$tag" != "$version" ]]; then + echo "Version mismatch: shellcheck $version tagged as $tag" + exit 1 + fi + if [ -n "$ext" ]; then + docker run --rm -v "$PWD:/mnt" -w /mnt "$image" shellcheck myscript + else + docker run --rm -v "$PWD:/mnt" "$image" myscript + fi + done + done + done +} + +function multi_arch_docker::main() { + export DOCKER_PLATFORMS='linux/amd64' + DOCKER_PLATFORMS+=' linux/arm64' + DOCKER_PLATFORMS+=' linux/arm/v6' + DOCKER_PLATFORMS+=' linux/riscv64' + + multi_arch_docker::install_docker_buildx + multi_arch_docker::login_to_docker_hub + multi_arch_docker::build_and_push_all + multi_arch_docker::test_all +} diff --git a/.prepare_deploy b/.prepare_deploy index dcf0346..9f39912 100755 --- a/.prepare_deploy +++ b/.prepare_deploy @@ -1,8 +1,9 @@ #!/bin/bash -# This script packages up Travis compiled binaries +# This script packages up compiled binaries set -ex -shopt -s nullglob -cd deploy +shopt -s nullglob extglob + +ls -l cp ../LICENSE LICENSE.txt sed -e $'s/$/\r/' > README.txt << END @@ -22,29 +23,32 @@ This binary was compiled on $(date -u). $(git log -n 3) END -for file in ./*.exe +for dir in */ do - zip "${file%.*}.zip" README.txt LICENSE.txt "$file" + cp LICENSE.txt README.txt "$dir" done -for file in *.linux-x86_64 -do - base="${file%.*}" - cp "$file" "shellcheck" - tar -cJf "$base.linux.x86_64.tar.xz" --transform="s:^:$base/:" README.txt LICENSE.txt shellcheck - rm "shellcheck" -done +echo "Tags are $TAGS" -for file in *.linux-armv6hf +for tag in $TAGS do - base="${file%.*}" - cp "$file" "shellcheck" - tar -cJf "$base.linux.armv6hf.tar.xz" --transform="s:^:$base/:" README.txt LICENSE.txt shellcheck - rm "shellcheck" + + for dir in windows.*/ + do + ( cd "$dir" && zip "../shellcheck-$tag.zip" * ) + done + + for dir in {linux,darwin}.*/ + do + base="${dir%/}" + ( cd "$dir" && tar -cJf "../shellcheck-$tag.$base.tar.xz" --transform="s:^:shellcheck-$tag/:" * ) + done done for file in ./* do + [[ -f "$file" ]] || continue sha512sum "$file" > "$file.sha512sum" done +ls -l diff --git a/.travis.yml b/.travis.yml deleted file mode 100644 index 8a86307..0000000 --- a/.travis.yml +++ /dev/null @@ -1,80 +0,0 @@ -sudo: required - -language: sh - -services: - - docker - -before_install: - - DOCKER_BASE="$DOCKER_USERNAME/shellcheck" - - DOCKER_BUILDS="" - - TAGS="" - - test "$TRAVIS_BRANCH" = master && TAGS="$TAGS latest" || true - - test -n "$TRAVIS_TAG" && TAGS="$TAGS stable $TRAVIS_TAG" || true - - echo "Tags are $TAGS" - -script: - - mkdir deploy - # Remove all tests to reduce binary size - - ./striptests - # Linux Docker image - - name="$DOCKER_BASE" - - DOCKER_BUILDS="$DOCKER_BUILDS $name" - - docker build -t "$name:current" . - - docker run "$name:current" --version - - printf '%s\n' "#!/bin/sh" "echo 'hello world'" > myscript - - docker run -v "$PWD:/mnt" "$name:current" myscript - # Copy static executable from docker image - - id=$(docker create "$name:current") - - docker cp "$id:/bin/shellcheck" "shellcheck" - - docker rm "$id" - - ls -l shellcheck - - ./shellcheck myscript - - for tag in $TAGS; do cp "shellcheck" "deploy/shellcheck-$tag.linux-x86_64"; done - # Linux Alpine based Docker image - - name="$DOCKER_BASE-alpine" - - DOCKER_BUILDS="$DOCKER_BUILDS $name" - - sed -e '/DELETE-MARKER/,$d' Dockerfile > Dockerfile.alpine - - docker build -f Dockerfile.alpine -t "$name:current" . - - docker run "$name:current" sh -c 'shellcheck --version' - # Linux armv6hf static executable - - docker run -v "$PWD:/mnt" koalaman/armv6hf-builder -c 'compile-shellcheck' - - for tag in $TAGS; do cp "shellcheck" "deploy/shellcheck-$tag.linux-armv6hf"; done - - rm -f shellcheck || true - # Windows .exe - - docker run --user="$UID" -v "$PWD:/appdata" koalaman/winghc cuib - - for tag in $TAGS; do cp "dist/build/ShellCheck/shellcheck.exe" "deploy/shellcheck-$tag.exe"; done - - rm -rf dist shellcheck || true - # Misc packaging - - ./.prepare_deploy - -after_success: - - docker login -u="$DOCKER_USERNAME" -p="$DOCKER_PASSWORD" - - for repo in $DOCKER_BUILDS; - do - for tag in $TAGS; - do - echo "Deploying $repo:current as $repo:$tag..."; - docker tag "$repo:current" "$repo:$tag" || exit 1; - docker push "$repo:$tag" || exit 1; - done; - done; - -after_failure: - - id - - pwd - - df -h - - find . -name '*.log' -type f -exec grep "" /dev/null {} + - - find . -ls - -deploy: - provider: gcs - skip_cleanup: true - access_key_id: GOOG7MDN7WEH6IIGBDCA - secret_access_key: - secure: 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 - bucket: shellcheck - local-dir: deploy - on: - repo: koalaman/shellcheck - all_branches: true diff --git a/CHANGELOG.md b/CHANGELOG.md index 60f890c..ccc8a79 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,17 +1,213 @@ -## ??? +## Git +### Added +- SC2327/SC2328: Warn about capturing the output of redirected commands. +- SC2329: Warn when (non-escaping) functions are never invoked. +- SC2330: Warn about unsupported glob matches with [[ .. ]] in BusyBox. +- SC2331: Suggest using standard -e instead of unary -a in tests. +- SC2332: Warn about `[ ! -o opt ]` being unconditionally true in Bash. +- SC3062: Warn about bashism `[ -o opt ]`. +- Precompiled binaries for Linux riscv64 (linux.riscv64) +### Changed +- SC2002 about Useless Use Of Cat is now disabled by default. It can be + re-enabled with `--enable=useless-use-of-cat` or equivalent directive. +- SC2015 about `A && B || C` no longer triggers when B is a test command. +- SC3012: Do not warn about `\<` and `\>` in test/[] as specified in POSIX.1-2024 +### Fixed +- SC2218 about function use-before-define is now more accurate. +- SC2317 about unreachable commands is now less spammy for nested ones. +- SC2292, optional suggestion for [[ ]], now triggers for Busybox. + +### Removed +- SC3013: removed since the operators `-ot/-nt/-ef` are specified in POSIX.1-2024 + +## v0.10.0 - 2024-03-07 +### Added +- Precompiled binaries for macOS ARM64 (darwin.aarch64) +- Added support for busybox sh +- Added flag --rcfile to specify an rc file by name. +- Added `extended-analysis=true` directive to enable/disable dataflow analysis + (with a corresponding --extended-analysis flag). +- SC2324: Warn when x+=1 appends instead of increments +- SC2325: Warn about multiple `!`s in dash/sh. +- SC2326: Warn about `foo | ! bar` in bash/dash/sh. +- SC3012: Warn about lexicographic-compare bashism in test like in [ ] +- SC3013: Warn bashism `test _ -op/-nt/-ef _` like in [ ] +- SC3014: Warn bashism `test _ == _` like in [ ] +- SC3015: Warn bashism `test _ =~ _` like in [ ] +- SC3016: Warn bashism `test -v _` like in [ ] +- SC3017: Warn bashism `test -a _` like in [ ] + +### Fixed +- source statements with here docs now work correctly +- "(Array.!): undefined array element" error should no longer occur + + +## v0.9.0 - 2022-12-12 +### Added +- SC2316: Warn about 'local readonly foo' and similar (thanks, patrickxia!) +- SC2317: Warn about unreachable commands +- SC2318: Warn about backreferences in 'declare x=1 y=$x' +- SC2319/SC2320: Warn when $? refers to echo/printf/[ ]/[[ ]]/test +- SC2321: Suggest removing $((..)) in array[$((idx))]=val +- SC2322: Suggest collapsing double parentheses in arithmetic contexts +- SC2323: Suggest removing wrapping parentheses in a[(x+1)]=val + +### Fixed +- SC2086: Now uses DFA to make more accurate predictions about values +- SC2086: No longer warns about values declared as integer with declare -i + +### Changed +- ShellCheck now has a Data Flow Analysis engine to make smarter decisions + based on control flow rather than just syntax. Existing checks will + gradually start using it, which may cause them to trigger differently + (but more accurately). +- Values in directives/shellcheckrc can now be quoted with '' or "" + + +## v0.8.0 - 2021-11-06 +### Added +- `disable=all` now conveniently disables all warnings +- `external-sources=true` directive can be added to .shellcheckrc to make + shellcheck behave as if `-x` was specified. +- Optional `check-extra-masked-returns` for pointing out commands with + suppressed exit codes (SC2312). +- Optional `require-double-brackets` for recommending \[\[ ]] (SC2292). +- SC2286-SC2288: Warn when command name ends in a symbol like `/.)'"` +- SC2289: Warn when command name contains tabs or linefeeds +- SC2291: Warn about repeated unquoted spaces between words in echo +- SC2292: Suggest [[ over [ in Bash/Ksh scripts (optional) +- SC2293/SC2294: Warn when calling `eval` with arrays +- SC2295: Warn about "${x#$y}" treating $y as a pattern when not quoted +- SC2296-SC2301: Improved warnings for bad parameter expansions +- SC2302/SC2303: Warn about loops over array values when using them as keys +- SC2304-SC2306: Warn about unquoted globs in expr arguments +- SC2307: Warn about insufficient number of arguments to expr +- SC2308: Suggest other approaches for non-standard expr extensions +- SC2313: Warn about `read` with unquoted, array indexed variable + +### Fixed +- SC2102 about repetitions in ranges no longer triggers on [[ -v arr[xx] ]] +- SC2155 now recognizes `typeset` and local read-only `declare` statements +- SC2181 now tries to avoid triggering for error handling functions +- SC2290: Warn about misused = in declare & co, which were not caught by SC2270+ +- The flag --color=auto no longer outputs color when TERM is "dumb" or unset + +### Changed +- SC2048: Warning about $\* now also applies to ${array[\*]} +- SC2181 now only triggers on single condition tests like `[ $? = 0 ]`. +- Quote warnings are now emitted for declaration utilities in sh +- Leading `_` can now be used to suppress warnings about unused variables +- TTY output now includes warning level in text as well as color + +### Removed +- SC1004: Literal backslash+linefeed in '' was found to be usually correct + + +## v0.7.2 - 2021-04-19 +### Added +- `disable` directives can now be a range, e.g. `disable=SC3000-SC4000` +- SC1143: Warn about line continuations in comments +- SC2259/SC2260: Warn when redirections override pipes +- SC2261: Warn about multiple competing redirections +- SC2262/SC2263: Warn about aliases declared and used in the same parsing unit +- SC2264: Warn about wrapper functions that blatantly recurse +- SC2265/SC2266: Warn when using & or | with test statements +- SC2267: Warn when using xargs -i instead of -I +- SC2268: Warn about unnecessary x-comparisons like `[ x$var = xval ]` + +### Fixed +- SC1072/SC1073 now respond to disable annotations, though ignoring parse errors + is still purely cosmetic and does not allow ShellCheck to continue. +- Improved error reporting for trailing tokens after ]/]] and compound commands +- `#!/usr/bin/env -S shell` is now handled correctly +- Here docs with \r are now parsed correctly and give better warnings + +### Changed +- Assignments are now parsed to spec, without leniency for leading $ or spaces +- POSIX/dash unsupported feature warnings now have individual SC3xxx codes +- SC1090: A leading `$x/` or `$(x)/` is now treated as `./` when locating files +- SC2154: Variables appearing in -z/-n tests are no longer considered unassigned +- SC2270-SC2285: Improved warnings about misused `=`, e.g. `${var}=42` + + +## v0.7.1 - 2020-04-04 +### Fixed +- `-f diff` no longer claims that it found more issues when it didn't +- Known empty variables now correctly trigger SC2086 +- ShellCheck should now be compatible with Cabal 3 +- SC2154 and all command-specific checks now trigger for builtins + called with `builtin` + +### Added +- SC1136: Warn about unexpected characters after ]/]] +- SC2254: Suggest quoting expansions in case statements +- SC2255: Suggest using `$((..))` in `[ 2*3 -eq 6 ]` +- SC2256: Warn about translated strings that are known variables +- SC2257: Warn about arithmetic mutation in redirections +- SC2258: Warn about trailing commas in for loop elements + +### Changed +- SC2230: 'command -v' suggestion is now off by default (-i deprecate-which) +- SC1081: Keywords are now correctly parsed case sensitively, with a warning + + +## v0.7.0 - 2019-07-28 +### Added +- Precompiled binaries for macOS and Linux aarch64 +- Preliminary support for fix suggestions +- New `-f diff` unified diff format for auto-fixes +- Files containing Bats tests can now be checked +- Directory wide directives can now be placed in a `.shellcheckrc` +- Optional checks: Use `--list-optional` to show a list of tests, + Enable with `-o` flags or `enable=name` directives +- Source paths: Use `-P dir1:dir2` or a `source-path=dir1` directive + to specify search paths for sourced files. +- json1 format like --format=json but treats tabs as single characters +- Recognize FLAGS variables created by the shflags library. +- Site-specific changes can now be made in Custom.hs for ease of patching +- SC2154: Also warn about unassigned uppercase variables (optional) +- SC2252: Warn about `[ $a != x ] || [ $a != y ]`, similar to SC2055 +- SC2251: Inform about ineffectual ! in front of commands +- SC2250: Warn about variable references without braces (optional) +- SC2249: Warn about `case` with missing default case (optional) +- SC2248: Warn about unquoted variables without special chars (optional) +- SC2247: Warn about $"(cmd)" and $"{var}" +- SC2246: Warn if a shebang's interpreter ends with / +- SC2245: Warn that Ksh ignores all but the first glob result in `[` +- SC2243/SC2244: Suggest using explicit -n for `[ $foo ]` (optional) +- SC1135: Suggest not ending double quotes just to make $ literal + +### Changed +- If a directive or shebang is not specified, a `.bash/.bats/.dash/.ksh` + extension will be used to infer the shell type when present. +- Disabling SC2120 on a function now disables SC2119 on call sites + +### Fixed +- SC2183 no longer warns about missing printf args for `%()T` + +## v0.6.0 - 2018-12-02 ### Added - Command line option --severity/-S for filtering by minimum severity - Command line option --wiki-link-count/-W for showing wiki links +- SC2152/SC2151: Warn about bad `exit` values like `1234` and `"foo"` - SC2236/SC2237: Suggest -n/-z instead of ! -z/-n - SC2238: Warn when redirecting to a known command name, e.g. ls > rm - SC2239: Warn if the shebang is not an absolute path, e.g. #!bin/sh +- SC2240: Warn when passing additional arguments to dot (.) in sh/dash - SC1133: Better diagnostics when starting a line with |/||/&& + ### Changed - Most warnings now have useful end positions - SC1117 about unknown double-quoted escape sequences has been retired + ### Fixed -- SC2021 no longer triggers for equivalence classes like '[=e=]' +- SC2021 no longer triggers for equivalence classes like `[=e=]` - SC2221/SC2222 no longer mistriggers on fall-through case branches +- SC2081 about glob matches in `[ .. ]` now also triggers for `!=` +- SC2086 no longer warns about spaces in `$#` +- SC2164 no longer suggests subshells for `cd ..; cmd; cd ..` +- `read -a` is now correctly considered an array assignment +- SC2039 no longer warns about LINENO now that it's POSIX ## v0.5.0 - 2018-05-31 ### Added @@ -74,7 +270,7 @@ - SC2204/SC2205: Warn about `( -z foo )` and `( foo -eq bar )` - SC2200/SC2201: Warn about brace expansion in [/[[ - SC2198/SC2199: Warn about arrays in [/[[ -- SC2196/SC2197: Warn about deprected egrep/fgrep +- SC2196/SC2197: Warn about deprecated egrep/fgrep - SC2195: Warn about unmatchable case branches - SC2194: Warn about constant 'case' statements - SC2193: Warn about `[[ file.png == *.mp3 ]]` and other unmatchables @@ -91,7 +287,7 @@ ### Fixed - `-c` no longer suggested when using `grep -o | wc` - Comments and whitespace are now allowed before filewide directives -- Here doc delimters with esoteric quoting like `foo""` are now handled +- Here doc delimiters with esoteric quoting like `foo""` are now handled - SC2095 about `ssh` in while read loops is now suppressed when using `-n` - `%(%Y%M%D)T` now recognized as a single formatter in `printf` checks - `grep -F` now suppresses regex related suggestions @@ -104,7 +300,7 @@ - SC2185: Suggest explicitly adding path for `find` - SC2184: Warn about unsetting globs (e.g. `unset foo[1]`) - SC2183: Warn about `printf` with more formatters than variables -- SC2182: Warn about ignored arguments with `printf` +- SC2182: Warn about ignored arguments with `printf` - SC2181: Suggest using command directly instead of `if [ $? -eq 0 ]` - SC1106: Warn when using `test` operators in `(( 1 -eq 2 ))` @@ -275,7 +471,7 @@ ### Added - SC2121: Warn about trying to `set` variables, e.g. `set var = value` - SC2120/SC2119: Warn when a function uses `$1..` if none are ever passed -- SC2117: Warn when using `su` in interactive mode, e.g. `su foo; whoami` +- SC2117: Warn when using `su` in interactive mode, e.g. `su foo; whoami` - SC2116: Detect useless use of echo, e.g. `for i in $(echo $var)` - SC2115/SC2114: Detect some catastrophic `rm -r "$empty/"` mistakes - SC1081: Warn when capitalizing keywords like `While` @@ -326,7 +522,7 @@ ### Removed - Suggestions about using parameter expansion over basename -- The `jsoncheck` binary. Use `shellcheck -f json` instead. +- The `jsoncheck` binary. Use `shellcheck -f json` instead. ## v0.2.0 - 2013-10-27 diff --git a/Dockerfile b/Dockerfile deleted file mode 100644 index 5ecfd08..0000000 --- a/Dockerfile +++ /dev/null @@ -1,36 +0,0 @@ -# Build-only image -FROM ubuntu:17.10 AS build -USER root -WORKDIR /opt/shellCheck - -# Install OS deps -RUN apt-get update && apt-get install -y ghc cabal-install - -# Install Haskell deps -# (This is a separate copy/run so that source changes don't require rebuilding) -COPY ShellCheck.cabal ./ -RUN cabal update && cabal install --dependencies-only --ghc-options="-optlo-Os -split-sections" - -# Copy source and build it -COPY LICENSE Setup.hs shellcheck.hs ./ -COPY src src -RUN cabal build Paths_ShellCheck && \ - ghc -optl-static -optl-pthread -isrc -idist/build/autogen --make shellcheck -split-sections -optc-Wl,--gc-sections -optlo-Os && \ - strip --strip-all shellcheck - -RUN mkdir -p /out/bin && \ - cp shellcheck /out/bin/ - -# Resulting Alpine image -FROM alpine:latest -LABEL maintainer="Vidar Holen " -COPY --from=build /out / - -# DELETE-MARKER (Remove everything below to keep the alpine image) - -# Resulting ShellCheck image -FROM scratch -LABEL maintainer="Vidar Holen " -WORKDIR /mnt -COPY --from=build /out / -ENTRYPOINT ["/bin/shellcheck"] diff --git a/Dockerfile.multi-arch b/Dockerfile.multi-arch new file mode 100644 index 0000000..217aa74 --- /dev/null +++ b/Dockerfile.multi-arch @@ -0,0 +1,26 @@ +# Alpine image +FROM alpine:latest AS alpine +LABEL maintainer="Vidar Holen " +ARG tag + +# Put the right binary for each architecture into place for the +# multi-architecture docker image. +RUN set -x; \ + arch="$(uname -m)"; \ + echo "arch is $arch"; \ + if [ "${arch}" = 'armv7l' ]; then \ + arch='armv6hf'; \ + fi; \ + url_base='https://github.com/koalaman/shellcheck/releases/download/'; \ + tar_file="${tag}/shellcheck-${tag}.linux.${arch}.tar.xz"; \ + wget "${url_base}${tar_file}" -O - | tar xJf -; \ + mv "shellcheck-${tag}/shellcheck" /bin/; \ + rm -rf "shellcheck-${tag}"; \ + ls -laF /bin/shellcheck + +# ShellCheck image +FROM scratch +LABEL maintainer="Vidar Holen " +WORKDIR /mnt +COPY --from=alpine /bin/shellcheck /bin/ +ENTRYPOINT ["/bin/shellcheck"] diff --git a/LICENSE b/LICENSE index 0507f1f..f288702 100644 --- a/LICENSE +++ b/LICENSE @@ -1,13 +1,3 @@ -Employer mandated disclaimer: - - I am providing code in the repository to you under an open source license. - Because this is my personal repository, the license you receive to my code is - from me and other individual contributors, and not my employer (Facebook). - - - Vidar "koala_man" Holen - ----- - GNU GENERAL PUBLIC LICENSE Version 3, 29 June 2007 @@ -681,4 +671,4 @@ into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Lesser General Public License instead of this License. But first, please read -. +. diff --git a/README.md b/README.md index 3b2ee5e..9b776cf 100644 --- a/README.md +++ b/README.md @@ -1,4 +1,5 @@ -[![Build Status](https://travis-ci.org/koalaman/shellcheck.svg?branch=master)](https://travis-ci.org/koalaman/shellcheck) +[![Build Status](https://github.com/koalaman/shellcheck/actions/workflows/build.yml/badge.svg)](https://github.com/koalaman/shellcheck/actions/workflows/build.yml) + # ShellCheck - A shell script static analysis tool @@ -8,45 +9,45 @@ ShellCheck is a GPLv3 tool that gives warnings and suggestions for bash/sh shell The goals of ShellCheck are -- To point out and clarify typical beginner's syntax issues that cause a shell +* To point out and clarify typical beginner's syntax issues that cause a shell to give cryptic error messages. -- To point out and clarify typical intermediate level semantic problems that +* To point out and clarify typical intermediate level semantic problems that cause a shell to behave strangely and counter-intuitively. -- To point out subtle caveats, corner cases and pitfalls that may cause an +* To point out subtle caveats, corner cases and pitfalls that may cause an advanced user's otherwise working script to fail under future circumstances. See [the gallery of bad code](README.md#user-content-gallery-of-bad-code) for examples of what ShellCheck can help you identify! ## Table of Contents -- [How to use](#how-to-use) - - [On the web](#on-the-web) - - [From your terminal](#from-your-terminal) - - [In your editor](#in-your-editor) - - [In your build or test suites](#in-your-build-or-test-suites) -- [Installing](#installing) -- [Travis CI](#travis-ci) -- [Compiling from source](#compiling-from-source) - - [Installing Cabal](#installing-cabal) - - [Compiling ShellCheck](#compiling-shellcheck) - - [Running tests](#running-tests) -- [Gallery of bad code](#gallery-of-bad-code) - - [Quoting](#quoting) - - [Conditionals](#conditionals) - - [Frequently misused commands](#frequently-misused-commands) - - [Common beginner's mistakes](#common-beginners-mistakes) - - [Style](#style) - - [Data and typing errors](#data-and-typing-errors) - - [Robustness](#robustness) - - [Portability](#portability) - - [Miscellaneous](#miscellaneous) -- [Testimonials](#testimonials) -- [Ignoring issues](#ignoring-issues) -- [Reporting bugs](#reporting-bugs) -- [Contributing](#contributing) -- [Copyright](#copyright) +* [How to use](#how-to-use) + * [On the web](#on-the-web) + * [From your terminal](#from-your-terminal) + * [In your editor](#in-your-editor) + * [In your build or test suites](#in-your-build-or-test-suites) +* [Installing](#installing) +* [Compiling from source](#compiling-from-source) + * [Installing Cabal](#installing-cabal) + * [Compiling ShellCheck](#compiling-shellcheck) + * [Running tests](#running-tests) +* [Gallery of bad code](#gallery-of-bad-code) + * [Quoting](#quoting) + * [Conditionals](#conditionals) + * [Frequently misused commands](#frequently-misused-commands) + * [Common beginner's mistakes](#common-beginners-mistakes) + * [Style](#style) + * [Data and typing errors](#data-and-typing-errors) + * [Robustness](#robustness) + * [Portability](#portability) + * [Miscellaneous](#miscellaneous) +* [Testimonials](#testimonials) +* [Ignoring issues](#ignoring-issues) +* [Reporting bugs](#reporting-bugs) +* [Contributing](#contributing) +* [Copyright](#copyright) +* [Other Resources](#other-resources) ## How to use @@ -54,7 +55,7 @@ There are a number of ways to use ShellCheck! ### On the web -Paste a shell script on https://www.shellcheck.net for instant feedback. +Paste a shell script on for instant feedback. [ShellCheck.net](https://www.shellcheck.net) is always synchronized to the latest git commit, and is the easiest way to give ShellCheck a go. Tell your friends! @@ -76,7 +77,7 @@ You can see ShellCheck suggestions directly in a variety of editors. * Sublime, through [SublimeLinter](https://github.com/SublimeLinter/SublimeLinter-shellcheck). -* Atom, through [Linter](https://github.com/AtomLinter/linter-shellcheck). +* Pulsar Edit (former Atom), through [linter-shellcheck-pulsar](https://github.com/pulsar-cooperative/linter-shellcheck-pulsar). * VSCode, through [vscode-shellcheck](https://github.com/timonwong/vscode-shellcheck). @@ -85,8 +86,46 @@ You can see ShellCheck suggestions directly in a variety of editors. ### In your build or test suites While ShellCheck is mostly intended for interactive use, it can easily be added to builds or test suites. +It makes canonical use of exit codes, so you can just add a `shellcheck` command as part of the process. -ShellCheck makes canonical use of exit codes, and can output simple JSON, CheckStyle compatible XML, GCC compatible warnings as well as human readable text (with or without ANSI colors). See the [Integration](https://github.com/koalaman/shellcheck/wiki/Integration) wiki page for more documentation. +For example, in a Makefile: + +```Makefile +check-scripts: + # Fail if any of these files have warnings + shellcheck myscripts/*.sh +``` + +or in a Travis CI `.travis.yml` file: + +```yaml +script: + # Fail if any of these files have warnings + - shellcheck myscripts/*.sh +``` + +Services and platforms that have ShellCheck pre-installed and ready to use: + +* [Travis CI](https://travis-ci.org/) +* [Codacy](https://www.codacy.com/) +* [Code Climate](https://codeclimate.com/) +* [Code Factor](https://www.codefactor.io/) +* [Codety](https://www.codety.io/) via the [Codety Scanner](https://github.com/codetyio/codety-scanner) +* [CircleCI](https://circleci.com) via the [ShellCheck Orb](https://circleci.com/orbs/registry/orb/circleci/shellcheck) +* [Github](https://github.com/features/actions) (only Linux) +* [Trunk Check](https://trunk.io/products/check) (universal linter; [allows you to explicitly version your shellcheck install](https://github.com/trunk-io/plugins/blob/bcbb361dcdbe4619af51ea7db474d7fb87540d20/.trunk/trunk.yaml#L32)) via the [shellcheck plugin](https://github.com/trunk-io/plugins/blob/main/linters/shellcheck/plugin.yaml) +* [CodeRabbit](https://coderabbit.ai/) + +Most other services, including [GitLab](https://about.gitlab.com/), let you install +ShellCheck yourself, either through the system's package manager (see [Installing](#installing)), +or by downloading and unpacking a [binary release](#installing-a-pre-compiled-binary). + +It's a good idea to manually install a specific ShellCheck version regardless. This avoids +any surprise build breaks when a new version with new warnings is published. + +For customized filtering or reporting, ShellCheck can output simple JSON, CheckStyle compatible XML, +GCC compatible warnings as well as human readable text (with or without ANSI colors). See the +[Integration](https://github.com/koalaman/shellcheck/wiki/Integration) wiki page for more documentation. ## Installing @@ -104,13 +143,13 @@ On systems with Stack (installs to `~/.local/bin`): On Debian based distros: - apt-get install shellcheck + sudo apt install shellcheck On Arch Linux based distros: pacman -S shellcheck -or get the dependency free [shellcheck-static](https://aur.archlinux.org/packages/shellcheck-static/) from the AUR. +or get the dependency free [shellcheck-bin](https://aur.archlinux.org/packages/shellcheck-bin/) from the AUR. On Gentoo based distros: @@ -118,8 +157,8 @@ On Gentoo based distros: On EPEL based distros: - yum -y install epel-release - yum install ShellCheck + sudo yum -y install epel-release + sudo yum install ShellCheck On Fedora based distros: @@ -129,23 +168,49 @@ On FreeBSD: pkg install hs-ShellCheck -On OS X with homebrew: +On macOS (OS X) with Homebrew: brew install shellcheck +Or with MacPorts: + + sudo port install shellcheck + +On OpenBSD: + + pkg_add shellcheck + On openSUSE zypper in ShellCheck -Or use OneClickInstall - https://software.opensuse.org/package/ShellCheck +Or use OneClickInstall - On Solus: eopkg install shellcheck - -On Windows (via [scoop](http://scoop.sh)): - scoop install shellcheck +On Windows (via [chocolatey](https://chocolatey.org/packages/shellcheck)): + +```cmd +C:\> choco install shellcheck +``` + +Or Windows (via [winget](https://github.com/microsoft/winget-pkgs)): + +```cmd +C:\> winget install --id koalaman.shellcheck +``` + +Or Windows (via [scoop](http://scoop.sh)): + +```cmd +C:\> scoop install shellcheck +``` + +From [conda-forge](https://anaconda.org/conda-forge/shellcheck): + + conda install -c conda-forge shellcheck From Snap Store: @@ -154,37 +219,77 @@ From Snap Store: From Docker Hub: ```sh -docker pull koalaman/shellcheck:stable # Or :v0.4.7 for that version, or :latest for daily builds -docker run -v "$PWD:/mnt" koalaman/shellcheck myscript +docker run --rm -v "$PWD:/mnt" koalaman/shellcheck:stable myscript +# Or :v0.4.7 for that version, or :latest for daily builds ``` or use `koalaman/shellcheck-alpine` if you want a larger Alpine Linux based image to extend. It works exactly like a regular Alpine image, but has shellcheck preinstalled. +Using the [nix package manager](https://nixos.org/nix): +```sh +nix-env -iA nixpkgs.shellcheck +``` + +Using the [Flox package manager](https://flox.dev/) +```sh +flox install shellcheck +``` + Alternatively, you can download pre-compiled binaries for the latest release here: -* [Linux, x86_64](https://storage.googleapis.com/shellcheck/shellcheck-stable.linux.x86_64.tar.xz) (statically linked) -* [Linux, armv6hf](https://storage.googleapis.com/shellcheck/shellcheck-stable.linux.armv6hf.tar.xz), i.e. Raspberry Pi (statically linked) -* [Windows, x86](https://storage.googleapis.com/shellcheck/shellcheck-stable.zip) +* [Linux, x86_64](https://github.com/koalaman/shellcheck/releases/download/stable/shellcheck-stable.linux.x86_64.tar.xz) (statically linked) +* [Linux, armv6hf](https://github.com/koalaman/shellcheck/releases/download/stable/shellcheck-stable.linux.armv6hf.tar.xz), i.e. Raspberry Pi (statically linked) +* [Linux, aarch64](https://github.com/koalaman/shellcheck/releases/download/stable/shellcheck-stable.linux.aarch64.tar.xz) aka ARM64 (statically linked) +* [macOS, aarch64](https://github.com/koalaman/shellcheck/releases/download/stable/shellcheck-stable.darwin.aarch64.tar.xz) +* [macOS, x86_64](https://github.com/koalaman/shellcheck/releases/download/stable/shellcheck-stable.darwin.x86_64.tar.xz) +* [Windows, x86](https://github.com/koalaman/shellcheck/releases/download/stable/shellcheck-stable.zip) -or see the [storage bucket listing](https://shellcheck.storage.googleapis.com/index.html) for checksums, older versions and the latest daily builds. +or see the [GitHub Releases](https://github.com/koalaman/shellcheck/releases) for other releases +(including the [latest](https://github.com/koalaman/shellcheck/releases/tag/latest) meta-release for daily git builds). -## Travis CI +There are currently no official binaries for Apple Silicon, but third party builds are available via +[ShellCheck for Visual Studio Code](https://github.com/vscode-shellcheck/shellcheck-binaries/releases). + +Distro packages already come with a `man` page. If you are building from source, it can be installed with: + +```console +pandoc -s -f markdown-smart -t man shellcheck.1.md -o shellcheck.1 +sudo mv shellcheck.1 /usr/share/man/man1 +``` + +### pre-commit + +To run ShellCheck via [pre-commit](https://pre-commit.com/), add the hook to your `.pre-commit-config.yaml`: + +``` +repos: +- repo: https://github.com/koalaman/shellcheck-precommit + rev: v0.7.2 + hooks: + - id: shellcheck +# args: ["--severity=warning"] # Optionally only show errors and warnings +``` + +### Travis CI Travis CI has now integrated ShellCheck by default, so you don't need to manually install it. -If you still want to do so in order to upgrade at your leisure or ensure the latest release, follow the steps to install the shellcheck binary, bellow. +If you still want to do so in order to upgrade at your leisure or ensure you're +using the latest release, follow the steps below to install a binary version. -## Installing the shellcheck binary +### Installing a pre-compiled binary -*Pre-requisite*: the program 'xz' needs to be installed on the system. -To install it on debian/ubuntu/linux mint, run `apt install xz-utils`. -To install it on Redhat/Fedora/CentOS, run `yum -y install xz`. +The pre-compiled binaries come in `tar.xz` files. To decompress them, make sure +`xz` is installed. +On Debian/Ubuntu/Mint, you can `apt install xz-utils`. +On Redhat/Fedora/CentOS, `yum -y install xz`. + +A simple installer may do something like: ```bash -export scversion="stable" # or "v0.4.7", or "latest" -wget "https://storage.googleapis.com/shellcheck/shellcheck-${scversion}.linux.x86_64.tar.xz" -tar --xz -xvf shellcheck-"${scversion}".linux.x86_64.tar.xz -cp shellcheck-"${scversion}"/shellcheck /usr/bin/ +scversion="stable" # or "v0.4.7", or "latest" +wget -qO- "https://github.com/koalaman/shellcheck/releases/download/${scversion?}/shellcheck-${scversion?}.linux.x86_64.tar.xz" | tar -xJv +cp "shellcheck-${scversion}/shellcheck" /usr/bin/ shellcheck --version ``` @@ -196,13 +301,11 @@ This section describes how to build ShellCheck from a source directory. ShellChe ShellCheck is built and packaged using Cabal. Install the package `cabal-install` from your system's package manager (with e.g. `apt-get`, `brew`, `emerge`, `yum`, or `zypper`). -On MacOS (OS X), you can do a fast install of Cabal using brew, which takes a couple of minutes instead of more than 30 minutes if you try to compile it from source. +On macOS (OS X), you can do a fast install of Cabal using brew, which takes a couple of minutes instead of more than 30 minutes if you try to compile it from source. - brew install cask - brew cask install haskell-platform - cabal install cabal-install + $ brew install cabal-install -On MacPorts, the package is instead called `hs-cabal-install`, while native Windows users should install the latest version of the Haskell platform from https://www.haskell.org/platform/ +On MacPorts, the package is instead called `hs-cabal-install`, while native Windows users should install the latest version of the Haskell platform from Verify that `cabal` is installed and update its dependency list with @@ -214,10 +317,6 @@ Verify that `cabal` is installed and update its dependency list with $ cabal install -Or if you intend to run the tests: - - $ cabal install --enable-tests - This will compile ShellCheck and install it to your `~/.cabal/bin` directory. Add this directory to your `PATH` (for bash, add this to your `~/.bashrc`): @@ -238,12 +337,15 @@ may use a legacy codepage. In `cmd.exe`, `powershell.exe` and Powershell ISE, make sure to use a TrueType font, not a Raster font, and set the active codepage to UTF-8 (65001) with `chcp`: - > chcp 65001 - Active code page: 65001 +```cmd +chcp 65001 +``` In Powershell ISE, you may need to additionally update the output encoding: - > [Console]::OutputEncoding = [System.Text.Encoding]::UTF8 +```powershell +[Console]::OutputEncoding = [System.Text.Encoding]::UTF8 +``` ### Running tests @@ -270,6 +372,7 @@ echo 'Don't forget to restart!' # Singlequote closed by apostrophe echo 'Don\'t try this at home' # Attempting to escape ' in '' echo 'Path is $PATH' # Variables in single quotes trap "echo Took ${SECONDS}s" 0 # Prematurely expanded trap +unset var[i] # Array index treated as glob ``` ### Conditionals @@ -288,6 +391,7 @@ ShellCheck can recognize many types of incorrect test statements. [ grep -q foo file ] # Command without $(..) [[ "$$file" == *.jpg ]] # Comparisons that can't succeed (( 1 -lt 2 )) # Using test operators in ((..)) +[ x ] & [ y ] | [ z ] # Accidental backgrounding and piping ``` ### Frequently misused commands @@ -359,6 +463,8 @@ echo "Hello $name" # Unassigned lowercase variables cmd | read bar; echo $bar # Assignments in subshells cat foo | cp bar # Piping to commands that don't read printf '%s: %s\n' foo # Mismatches in printf argument count +eval "${array[@]}" # Lost word boundaries in array eval +for i in "${x[@]}"; do ${x[$i]} # Using array value as key ``` ### Robustness @@ -383,6 +489,7 @@ ShellCheck will warn when using features not supported by the shebang. For examp echo {1..$n} # Works in ksh, but not bash/dash/sh echo {1..10} # Works in ksh and bash, but not dash/sh echo -n 42 # Works in ksh, bash and dash, undefined in sh +expr match str regex # Unportable alias for `expr str : regex` trap 'exit 42' sigint # Unportable signal spec cmd &> file # Unportable redirection operator read foo < /dev/tcp/host/22 # Unportable intercepted files @@ -403,10 +510,15 @@ rm “file” # Unicode quotes echo "Hello world" # Carriage return / DOS line endings echo hello \ # Trailing spaces after \ var=42 echo $var # Expansion of inlined environment -#!/bin/bash -x -e # Common shebang errors +!# bin/bash -x -e # Common shebang errors echo $((n/180*100)) # Unnecessary loss of precision ls *[:digit:].txt # Bad character class globs sed 's/foo/bar/' file > file # Redirecting to input +var2=$var2 # Variable assigned to itself +[ x$var = xval ] # Antiquated x-comparisons +ls() { ls -l "$@"; } # Infinitely recursive wrapper +alias ls='ls -l'; ls foo # Alias used before it takes effect +for x; do for x; do # Nested loop uses same variable while getopts "a" f; do case $f in "b") # Unhandled getopts flags ``` @@ -421,13 +533,13 @@ Alexander Tarasikov, Issues can be ignored via environmental variable, command line, individually or globally within a file: -https://github.com/koalaman/shellcheck/wiki/Ignore + ## Reporting bugs Please use the GitHub issue tracker for any bugs or feature suggestions: -https://github.com/koalaman/shellcheck/issues + ## Contributing @@ -442,11 +554,11 @@ The contributor retains the copyright. ShellCheck is licensed under the GNU General Public License, v3. A copy of this license is included in the file [LICENSE](LICENSE). -Copyright 2012-2018, Vidar 'koala_man' Holen and contributors. +Copyright 2012-2019, [Vidar 'koala_man' Holen](https://github.com/koalaman/) and contributors. Happy ShellChecking! +## Other Resources -## Other Resources * The wiki has [long form descriptions](https://github.com/koalaman/shellcheck/wiki/Checks) for each warning, e.g. [SC2221](https://github.com/koalaman/shellcheck/wiki/SC2221). * ShellCheck does not attempt to enforce any kind of formatting or indenting style, so also check out [shfmt](https://github.com/mvdan/sh)! diff --git a/Setup.hs b/Setup.hs deleted file mode 100644 index e6064f9..0000000 --- a/Setup.hs +++ /dev/null @@ -1,36 +0,0 @@ -import Distribution.PackageDescription ( - HookedBuildInfo, - emptyHookedBuildInfo ) -import Distribution.Simple ( - Args, - UserHooks ( preSDist ), - defaultMainWithHooks, - simpleUserHooks ) -import Distribution.Simple.Setup ( SDistFlags ) - -import System.Process ( system ) - - -main = defaultMainWithHooks myHooks - where - myHooks = simpleUserHooks { preSDist = myPreSDist } - --- | This hook will be executed before e.g. @cabal sdist@. It runs --- pandoc to create the man page from shellcheck.1.md. If the pandoc --- command is not found, this will fail with an error message: --- --- /bin/sh: pandoc: command not found --- --- Since the man page is listed in the Extra-Source-Files section of --- our cabal file, a failure here should result in a failure to --- create the distribution tarball (that's a good thing). --- -myPreSDist :: Args -> SDistFlags -> IO HookedBuildInfo -myPreSDist _ _ = do - putStrLn "Building the man page (shellcheck.1) with pandoc..." - putStrLn pandoc_cmd - result <- system pandoc_cmd - putStrLn $ "pandoc exited with " ++ show result - return emptyHookedBuildInfo - where - pandoc_cmd = "pandoc -s -t man shellcheck.1.md -o shellcheck.1" diff --git a/build/README.md b/build/README.md new file mode 100644 index 0000000..31e8607 --- /dev/null +++ b/build/README.md @@ -0,0 +1,17 @@ +This directory contains Dockerfiles for all builds. + +A build image will: + +* Run on Linux x86\_64 with vanilla Docker (no exceptions) +* Not contain any software that would restrict easy modification or copying +* Take a `cabal sdist` style tar.gz of the ShellCheck directory on stdin +* Output a tar.gz of artifacts on stdout, in a directory named for the arch + +This makes it simple to build any release without exotic hardware or software. + +An image can be built and tagged using `build_builder`, +and run on a source tarball using `run_builder`. + +Tip: Are you developing an image that relies on QEmu usermode emulation? + It's easy to accidentally depend on binfmt\_misc on the host OS. + Do a `echo 0 | sudo tee /proc/sys/fs/binfmt_misc/status` before testing. diff --git a/build/build_builder b/build/build_builder new file mode 100755 index 0000000..b34b996 --- /dev/null +++ b/build/build_builder @@ -0,0 +1,12 @@ +#!/bin/sh +if [ $# -eq 0 ] +then + echo >&2 "No build image directories specified" + echo >&2 "Example: $0 build/*/" + exit 1 +fi + +for dir +do + ( cd "$dir" && docker build -t "$(cat tag)" . ) || exit 1 +done diff --git a/build/darwin.aarch64/Dockerfile b/build/darwin.aarch64/Dockerfile new file mode 100644 index 0000000..7839728 --- /dev/null +++ b/build/darwin.aarch64/Dockerfile @@ -0,0 +1,40 @@ +FROM ghcr.io/shepherdjerred/macos-cross-compiler:latest + +ENV TARGET aarch64-apple-darwin22 +ENV TARGETNAME darwin.aarch64 + +# Build dependencies +USER root +ENV DEBIAN_FRONTEND noninteractive +ENV LC_ALL C.utf8 + +# Install basic deps +RUN apt-get update && apt-get install -y automake autoconf build-essential curl xz-utils qemu-user-static + +# Install a more suitable host compiler +WORKDIR /host-ghc +RUN curl -L "https://downloads.haskell.org/~cabal/cabal-install-3.9.0.0/cabal-install-3.9-x86_64-linux-alpine.tar.xz" | tar xJv -C /usr/local/bin +RUN curl -L 'https://downloads.haskell.org/~ghc/8.10.7/ghc-8.10.7-x86_64-deb10-linux.tar.xz' | tar xJ --strip-components=1 +RUN ./configure && make install + +# Build GHC. We have to use an old version because cross-compilation across OS has since broken. +WORKDIR /ghc +RUN curl -L "https://downloads.haskell.org/~ghc/8.10.7/ghc-8.10.7-src.tar.xz" | tar xJ --strip-components=1 +RUN apt-get install -y llvm-12 +RUN ./boot && ./configure --host x86_64-linux-gnu --build x86_64-linux-gnu --target "$TARGET" +RUN cp mk/flavours/quick-cross.mk mk/build.mk && make -j "$(nproc)" +RUN make install + +# Due to an apparent cabal bug, we specify our options directly to cabal +# It won't reuse caches if ghc-options are specified in ~/.cabal/config +ENV CABALOPTS "--ghc-options;-optc-Os -optc-fPIC;--with-ghc=$TARGET-ghc;--with-hc-pkg=$TARGET-ghc-pkg;--constraint=hashable==1.3.5.0" + +# Prebuild the dependencies +RUN cabal update +RUN IFS=';' && cabal install --dependencies-only $CABALOPTS ShellCheck + +# Copy the build script +COPY build /usr/bin + +WORKDIR /scratch +ENTRYPOINT ["/usr/bin/build"] diff --git a/build/darwin.aarch64/build b/build/darwin.aarch64/build new file mode 100755 index 0000000..ff522ff --- /dev/null +++ b/build/darwin.aarch64/build @@ -0,0 +1,16 @@ +#!/bin/sh +set -xe +{ + tar xzv --strip-components=1 + chmod +x striptests && ./striptests + mkdir "$TARGETNAME" + ( IFS=';'; cabal build $CABALOPTS ) + find . -name shellcheck -type f -exec mv {} "$TARGETNAME/" \; + ls -l "$TARGETNAME" + # Stripping invalidates the code signature and the build image does + # not appear to have anything similar to the 'codesign' tool. + # "$TARGET-strip" "$TARGETNAME/shellcheck" + ls -l "$TARGETNAME" + file "$TARGETNAME/shellcheck" | grep "Mach-O 64-bit arm64 executable" +} >&2 +tar czv "$TARGETNAME" diff --git a/build/darwin.aarch64/tag b/build/darwin.aarch64/tag new file mode 100644 index 0000000..ae93ef3 --- /dev/null +++ b/build/darwin.aarch64/tag @@ -0,0 +1 @@ +koalaman/scbuilder-darwin-aarch64 diff --git a/build/darwin.x86_64/Dockerfile b/build/darwin.x86_64/Dockerfile new file mode 100644 index 0000000..a53245f --- /dev/null +++ b/build/darwin.x86_64/Dockerfile @@ -0,0 +1,33 @@ +FROM liushuyu/osxcross@sha256:fa32af4677e2860a1c5950bc8c360f309e2a87e2ddfed27b642fddf7a6093b76 + +ENV TARGET x86_64-apple-darwin18 +ENV TARGETNAME darwin.x86_64 + +# Build dependencies +USER root +ENV DEBIAN_FRONTEND noninteractive +RUN sed -e 's/focal/kinetic/g' -i /etc/apt/sources.list +RUN apt-get update +RUN apt-get dist-upgrade -y +RUN apt-get install -y ghc automake autoconf llvm curl alex happy + +# Build GHC +WORKDIR /ghc +RUN curl -L "https://downloads.haskell.org/~ghc/9.2.5/ghc-9.2.5-src.tar.xz" | tar xJ --strip-components=1 +RUN ./configure --host x86_64-linux-gnu --build x86_64-linux-gnu --target "$TARGET" +RUN cp mk/flavours/quick-cross.mk mk/build.mk && make -j "$(nproc)" +RUN make install +RUN curl -L "https://downloads.haskell.org/~cabal/cabal-install-3.9.0.0/cabal-install-3.9-x86_64-linux-alpine.tar.xz" | tar xJv -C /usr/local/bin + +# Due to an apparent cabal bug, we specify our options directly to cabal +# It won't reuse caches if ghc-options are specified in ~/.cabal/config +ENV CABALOPTS "--with-ghc=$TARGET-ghc;--with-hc-pkg=$TARGET-ghc-pkg" + +# Prebuild the dependencies +RUN cabal update && IFS=';' && cabal install --dependencies-only $CABALOPTS ShellCheck + +# Copy the build script +COPY build /usr/bin + +WORKDIR /scratch +ENTRYPOINT ["/usr/bin/build"] diff --git a/build/darwin.x86_64/build b/build/darwin.x86_64/build new file mode 100755 index 0000000..058cece --- /dev/null +++ b/build/darwin.x86_64/build @@ -0,0 +1,13 @@ +#!/bin/sh +set -xe +{ + tar xzv --strip-components=1 + chmod +x striptests && ./striptests + mkdir "$TARGETNAME" + ( IFS=';'; cabal build $CABALOPTS ) + find . -name shellcheck -type f -exec mv {} "$TARGETNAME/" \; + ls -l "$TARGETNAME" + "$TARGET-strip" -Sx "$TARGETNAME/shellcheck" + ls -l "$TARGETNAME" +} >&2 +tar czv "$TARGETNAME" diff --git a/build/darwin.x86_64/tag b/build/darwin.x86_64/tag new file mode 100644 index 0000000..237a65c --- /dev/null +++ b/build/darwin.x86_64/tag @@ -0,0 +1 @@ +koalaman/scbuilder-darwin-x86_64 diff --git a/build/linux.aarch64/Dockerfile b/build/linux.aarch64/Dockerfile new file mode 100644 index 0000000..1ffe1bd --- /dev/null +++ b/build/linux.aarch64/Dockerfile @@ -0,0 +1,40 @@ +FROM ubuntu:20.04 + +ENV TARGET aarch64-linux-gnu +ENV TARGETNAME linux.aarch64 + +# Build dependencies +USER root +ENV DEBIAN_FRONTEND noninteractive + +# These deps are from 20.04, because GHC's compiler/llvm support moves slowly +RUN apt-get update && apt-get install -y llvm gcc-$TARGET + +# The rest are from 22.10 +RUN sed -e 's/focal/kinetic/g' -i /etc/apt/sources.list +# Kinetic does not receive updates anymore, switch to last available +RUN sed -e 's/archive.ubuntu.com/old-releases.ubuntu.com/g' -i /etc/apt/sources.list +RUN sed -e 's/security.ubuntu.com/old-releases.ubuntu.com/g' -i /etc/apt/sources.list + +RUN apt-get update && apt-get install -y ghc alex happy automake autoconf build-essential curl qemu-user-static + +# Build GHC +WORKDIR /ghc +RUN curl -L "https://downloads.haskell.org/~ghc/9.2.8/ghc-9.2.8-src.tar.xz" | tar xJ --strip-components=1 +RUN ./boot && ./configure --host x86_64-linux-gnu --build x86_64-linux-gnu --target "$TARGET" +RUN cp mk/flavours/quick-cross.mk mk/build.mk && make -j "$(nproc)" +RUN make install +RUN curl -L "https://downloads.haskell.org/~cabal/cabal-install-3.9.0.0/cabal-install-3.9-x86_64-linux-alpine.tar.xz" | tar xJv -C /usr/local/bin + +# Due to an apparent cabal bug, we specify our options directly to cabal +# It won't reuse caches if ghc-options are specified in ~/.cabal/config +ENV CABALOPTS "--ghc-options;-split-sections -optc-Os -optc-Wl,--gc-sections -optc-fPIC;--with-ghc=$TARGET-ghc;--with-hc-pkg=$TARGET-ghc-pkg;-c;hashable -arch-native" + +# Prebuild the dependencies +RUN cabal update && IFS=';' && cabal install --dependencies-only $CABALOPTS ShellCheck + +# Copy the build script +COPY build /usr/bin + +WORKDIR /scratch +ENTRYPOINT ["/usr/bin/build"] diff --git a/build/linux.aarch64/build b/build/linux.aarch64/build new file mode 100755 index 0000000..3ce61ce --- /dev/null +++ b/build/linux.aarch64/build @@ -0,0 +1,14 @@ +#!/bin/sh +set -xe +{ + tar xzv --strip-components=1 + chmod +x striptests && ./striptests + mkdir "$TARGETNAME" + ( IFS=';'; cabal build $CABALOPTS --enable-executable-static ) + find . -name shellcheck -type f -exec mv {} "$TARGETNAME/" \; + ls -l "$TARGETNAME" + "$TARGET-strip" -s "$TARGETNAME/shellcheck" + ls -l "$TARGETNAME" + qemu-aarch64-static "$TARGETNAME/shellcheck" --version +} >&2 +tar czv "$TARGETNAME" diff --git a/build/linux.aarch64/tag b/build/linux.aarch64/tag new file mode 100644 index 0000000..6788e14 --- /dev/null +++ b/build/linux.aarch64/tag @@ -0,0 +1 @@ +koalaman/scbuilder-linux-aarch64 diff --git a/build/linux.armv6hf/Dockerfile b/build/linux.armv6hf/Dockerfile new file mode 100644 index 0000000..b4d4197 --- /dev/null +++ b/build/linux.armv6hf/Dockerfile @@ -0,0 +1,42 @@ +# This Docker file uses a custom QEmu fork with patches to follow execve +# to build all of ShellCheck emulated. + +FROM ubuntu:24.04 + +ENV TARGETNAME linux.armv6hf + +# Build QEmu with execve follow support +USER root +ENV DEBIAN_FRONTEND noninteractive +RUN apt-get update +RUN apt-get install -y --no-install-recommends build-essential git ninja-build python3 pkg-config libglib2.0-dev libpixman-1-dev python3-setuptools ca-certificates debootstrap +WORKDIR /qemu +RUN git clone --depth 1 https://github.com/koalaman/qemu . +RUN ./configure --static --disable-werror && cd build && ninja qemu-arm +ENV QEMU_EXECVE 1 + +# Convenience utility +COPY scutil /bin/scutil +COPY scutil /chroot/bin/scutil +RUN chmod +x /bin/scutil /chroot/bin/scutil + +# Set up an armv6 userspace +WORKDIR / +RUN debootstrap --arch armhf --variant=minbase --foreign bookworm /chroot http://mirrordirector.raspbian.org/raspbian +RUN cp /qemu/build/qemu-arm /chroot/bin/qemu +RUN scutil emu /debootstrap/debootstrap --second-stage + +# Install deps in the chroot +RUN scutil emu apt-get update +RUN scutil emu apt-get install -y --no-install-recommends ghc cabal-install +RUN scutil emu cabal update + +# Finally we can build the current dependencies. This takes hours. +ENV CABALOPTS "--ghc-options;-split-sections -optc-Os -optc-Wl,--gc-sections;--gcc-options;-Os -Wl,--gc-sections -ffunction-sections -fdata-sections" +# Generated with `cabal freeze --constraint 'hashable -arch-native'` +COPY cabal.project.freeze /chroot/etc +RUN IFS=";" && scutil install_from_freeze /chroot/etc/cabal.project.freeze emu cabal install $CABALOPTS + +# Copy the build script +COPY build /chroot/bin +ENTRYPOINT ["/bin/scutil", "emu", "/bin/build"] diff --git a/build/linux.armv6hf/build b/build/linux.armv6hf/build new file mode 100755 index 0000000..1d496ae --- /dev/null +++ b/build/linux.armv6hf/build @@ -0,0 +1,17 @@ +#!/bin/sh +set -xe +mkdir /scratch && cd /scratch +{ + tar xzv --strip-components=1 + cp /etc/cabal.project.freeze . + chmod +x striptests && ./striptests + mkdir "$TARGETNAME" + # This script does not cabal update because compiling anything new is slow + ( IFS=';'; cabal build $CABALOPTS --enable-executable-static ) + find . -name shellcheck -type f -exec mv {} "$TARGETNAME/" \; + ls -l "$TARGETNAME" + strip -s "$TARGETNAME/shellcheck" + ls -l "$TARGETNAME" + "$TARGETNAME/shellcheck" --version +} >&2 +tar czv "$TARGETNAME" diff --git a/build/linux.armv6hf/cabal.project.freeze b/build/linux.armv6hf/cabal.project.freeze new file mode 100644 index 0000000..183bcc6 --- /dev/null +++ b/build/linux.armv6hf/cabal.project.freeze @@ -0,0 +1,93 @@ +active-repositories: hackage.haskell.org:merge +constraints: any.Diff ==0.5, + any.OneTuple ==0.4.2, + any.QuickCheck ==2.14.3, + QuickCheck -old-random +templatehaskell, + any.StateVar ==1.2.2, + any.aeson ==2.2.3.0, + aeson +ordered-keymap, + any.array ==0.5.4.0, + any.assoc ==1.1.1, + assoc -tagged, + any.base ==4.15.1.0, + any.base-orphans ==0.9.2, + any.bifunctors ==5.6.2, + bifunctors +tagged, + any.binary ==0.8.8.0, + any.bytestring ==0.10.12.1, + any.character-ps ==0.1, + any.comonad ==5.0.8, + comonad +containers +distributive +indexed-traversable, + any.containers ==0.6.4.1, + any.contravariant ==1.5.5, + contravariant +semigroups +statevar +tagged, + any.data-array-byte ==0.1.0.1, + any.data-fix ==0.3.3, + any.deepseq ==1.4.5.0, + any.directory ==1.3.6.2, + any.distributive ==0.6.2.1, + distributive +semigroups +tagged, + any.dlist ==1.0, + dlist -werror, + any.exceptions ==0.10.4, + any.fgl ==5.8.2.0, + fgl +containers042, + any.filepath ==1.4.2.1, + any.foldable1-classes-compat ==0.1, + foldable1-classes-compat +tagged, + any.generically ==0.1.1, + any.ghc-bignum ==1.1, + any.ghc-boot-th ==9.0.2, + any.ghc-prim ==0.7.0, + any.hashable ==1.4.6.0, + hashable -arch-native +integer-gmp -random-initial-seed, + any.indexed-traversable ==0.1.4, + any.indexed-traversable-instances ==0.1.2, + any.integer-conversion ==0.1.1, + any.integer-logarithms ==1.0.3.1, + integer-logarithms -check-bounds +integer-gmp, + any.mtl ==2.2.2, + any.network-uri ==2.6.4.2, + any.parsec ==3.1.14.0, + any.pretty ==1.1.3.6, + any.primitive ==0.9.0.0, + any.process ==1.6.13.2, + any.random ==1.2.1.2, + any.regex-base ==0.94.0.2, + any.regex-tdfa ==1.3.2.2, + regex-tdfa +doctest -force-o2, + any.rts ==1.0.2, + any.scientific ==0.3.8.0, + scientific -integer-simple, + any.semialign ==1.3.1, + semialign +semigroupoids, + any.semigroupoids ==6.0.1, + semigroupoids +comonad +containers +contravariant +distributive +tagged +unordered-containers, + any.splitmix ==0.1.0.5, + splitmix -optimised-mixer, + any.stm ==2.5.0.0, + any.strict ==0.5, + any.tagged ==0.8.8, + tagged +deepseq +transformers, + any.template-haskell ==2.17.0.0, + any.text ==1.2.5.0, + any.text-iso8601 ==0.1.1, + any.text-short ==0.1.6, + text-short -asserts, + any.th-abstraction ==0.7.0.0, + any.th-compat ==0.1.5, + any.these ==1.2.1, + any.time ==1.9.3, + any.time-compat ==1.9.7, + any.transformers ==0.5.6.2, + any.transformers-compat ==0.7.2, + transformers-compat -five +five-three -four +generic-deriving +mtl -three -two, + any.unix ==2.7.2.2, + any.unordered-containers ==0.2.20, + unordered-containers -debug, + any.uuid-types ==1.0.6, + any.vector ==0.13.1.0, + vector +boundschecks -internalchecks -unsafechecks -wall, + any.vector-stream ==0.1.0.1, + any.witherable ==0.5 +index-state: hackage.haskell.org 2024-06-18T02:21:19Z diff --git a/build/linux.armv6hf/scutil b/build/linux.armv6hf/scutil new file mode 100644 index 0000000..a85d810 --- /dev/null +++ b/build/linux.armv6hf/scutil @@ -0,0 +1,48 @@ +#!/bin/dash +# Various ShellCheck build utility functions + +# Generally set a ulimit to avoid QEmu using too much memory +ulimit -v "$((10*1024*1024))" +# If we happen to invoke or run under QEmu, make sure to follow execve. +# This requires a patched QEmu. +export QEMU_EXECVE=1 + +# Retry a command until it succeeds +# Usage: scutil retry 3 mycmd +retry() { + n="$1" + ret=1 + shift + while [ "$n" -gt 0 ] + do + "$@" + ret=$? + [ "$ret" = 0 ] && break + n=$((n-1)) + done + return "$ret" +} + +# Install all dependencies from a freeze file +# Usage: scutil install_from_freeze /path/cabal.project.freeze cabal install +install_from_freeze() { + linefeed=$(printf '\nx') + linefeed=${linefeed%x} + flags=$( + sed 's/constraints:/&\n /' "$1" | + grep -vw -e rts -e base | + sed -n -e 's/^ *\([^,]*\).*/\1/p' | + sed -e 's/any\.\([^ ]*\) ==\(.*\)/\1-\2/; te; s/.*/--constraint\n&/; :e') + shift + # shellcheck disable=SC2086 + ( IFS=$linefeed; set -x; "$@" $flags ) +} + +# Run a command under emulation. +# This assumes the correct emulator is named 'qemu' and the chroot is /chroot +# Usage: scutil emu echo "Hello World" +emu() { + chroot /chroot /bin/qemu /usr/bin/env "$@" +} + +"$@" diff --git a/build/linux.armv6hf/tag b/build/linux.armv6hf/tag new file mode 100644 index 0000000..9172c5c --- /dev/null +++ b/build/linux.armv6hf/tag @@ -0,0 +1 @@ +koalaman/scbuilder-linux-armv6hf diff --git a/build/linux.riscv64/Dockerfile b/build/linux.riscv64/Dockerfile new file mode 100644 index 0000000..d138ff7 --- /dev/null +++ b/build/linux.riscv64/Dockerfile @@ -0,0 +1,46 @@ +FROM ubuntu:24.04 + +ENV TARGETNAME linux.riscv64 +ENV TARGET riscv64-linux-gnu + +USER root +ENV DEBIAN_FRONTEND noninteractive + +# Init base +RUN apt-get update -y + +# Install qemu +RUN apt-get install -y --no-install-recommends build-essential ninja-build python3 pkg-config libglib2.0-dev libpixman-1-dev curl ca-certificates python3-virtualenv git python3-setuptools debootstrap +WORKDIR /qemu +RUN git clone --depth 1 https://github.com/koalaman/qemu . +RUN ./configure --target-list=riscv64-linux-user --static --disable-system --disable-pie --disable-werror +RUN cd build && ninja qemu-riscv64 +ENV QEMU_EXECVE 1 + +# Convenience utility +COPY scutil /bin/scutil +# We have to copy to /usr/bin because debootstrap will try to symlink /bin and fail if it exists +COPY scutil /chroot/usr/bin/scutil +RUN chmod +x /bin/scutil /chroot/usr/bin/scutil + +# Set up a riscv64 userspace +WORKDIR / +RUN debootstrap --arch=riscv64 --variant=minbase --components=main,universe --foreign noble /chroot http://ports.ubuntu.com/ubuntu-ports +RUN cp /qemu/build/qemu-riscv64 /chroot/bin/qemu +RUN scutil emu /debootstrap/debootstrap --second-stage + +# Install deps in the chroot +RUN scutil emu apt-get update +RUN scutil emu apt-get install -y --no-install-recommends ghc cabal-install +RUN scutil emu cabal update + +# Generated with: cabal freeze -c 'hashable -arch-native'. We put it in /etc so cabal won't find it. +COPY cabal.project.freeze /chroot/etc + +# Build all dependencies from the freeze file. The emulator segfaults at random, +# so retry a few times. +RUN scutil install_from_freeze /chroot/etc/cabal.project.freeze retry 5 emu cabal install --keep-going + +# Copy the build script +COPY build /chroot/bin/build +ENTRYPOINT ["/bin/scutil", "emu", "/bin/build"] diff --git a/build/linux.riscv64/build b/build/linux.riscv64/build new file mode 100755 index 0000000..ed9dc27 --- /dev/null +++ b/build/linux.riscv64/build @@ -0,0 +1,21 @@ +#!/bin/sh +set -xe +IFS=';' +{ + mkdir -p /tmp/scratch + cd /tmp/scratch + tar xzv --strip-components=1 + chmod +x striptests && ./striptests + # Use a freeze file to ensure we use the same dependencies we cached during + # the docker image build. We don't want to spend time compiling anything new. + cp /etc/cabal.project.freeze . + mkdir "$TARGETNAME" + # Retry in case of random segfault + scutil retry 3 cabal build --enable-executable-static + find . -name shellcheck -type f -exec mv {} "$TARGETNAME/" \; + ls -l "$TARGETNAME" + "$TARGET-strip" -s "$TARGETNAME/shellcheck" + ls -l "$TARGETNAME" + "$TARGETNAME/shellcheck" --version +} >&2 +tar czv "$TARGETNAME" diff --git a/build/linux.riscv64/cabal.project.freeze b/build/linux.riscv64/cabal.project.freeze new file mode 100644 index 0000000..cbb42e1 --- /dev/null +++ b/build/linux.riscv64/cabal.project.freeze @@ -0,0 +1,93 @@ +active-repositories: hackage.haskell.org:merge +constraints: any.Diff ==0.5, + any.OneTuple ==0.4.2, + any.QuickCheck ==2.14.3, + QuickCheck -old-random +templatehaskell, + any.StateVar ==1.2.2, + any.aeson ==2.2.3.0, + aeson +ordered-keymap, + any.array ==0.5.4.0, + any.assoc ==1.1.1, + assoc -tagged, + any.base ==4.17.2.0, + any.base-orphans ==0.9.2, + any.bifunctors ==5.6.2, + bifunctors +tagged, + any.binary ==0.8.9.1, + any.bytestring ==0.11.5.2, + any.character-ps ==0.1, + any.comonad ==5.0.8, + comonad +containers +distributive +indexed-traversable, + any.containers ==0.6.7, + any.contravariant ==1.5.5, + contravariant +semigroups +statevar +tagged, + any.data-fix ==0.3.3, + any.deepseq ==1.4.8.0, + any.directory ==1.3.7.1, + any.distributive ==0.6.2.1, + distributive +semigroups +tagged, + any.dlist ==1.0, + dlist -werror, + any.exceptions ==0.10.5, + any.fgl ==5.8.2.0, + fgl +containers042, + any.filepath ==1.4.2.2, + any.foldable1-classes-compat ==0.1, + foldable1-classes-compat +tagged, + any.generically ==0.1.1, + any.ghc-bignum ==1.3, + any.ghc-boot-th ==9.4.7, + any.ghc-prim ==0.9.1, + any.hashable ==1.4.6.0, + hashable -arch-native +integer-gmp -random-initial-seed, + any.indexed-traversable ==0.1.4, + any.indexed-traversable-instances ==0.1.2, + any.integer-conversion ==0.1.1, + any.integer-logarithms ==1.0.3.1, + integer-logarithms -check-bounds +integer-gmp, + any.mtl ==2.2.2, + any.network-uri ==2.6.4.2, + any.os-string ==2.0.3, + any.parsec ==3.1.16.1, + any.pretty ==1.1.3.6, + any.primitive ==0.9.0.0, + any.process ==1.6.17.0, + any.random ==1.2.1.2, + any.regex-base ==0.94.0.2, + any.regex-tdfa ==1.3.2.2, + regex-tdfa +doctest -force-o2, + any.rts ==1.0.2, + any.scientific ==0.3.8.0, + scientific -integer-simple, + any.semialign ==1.3.1, + semialign +semigroupoids, + any.semigroupoids ==6.0.1, + semigroupoids +comonad +containers +contravariant +distributive +tagged +unordered-containers, + any.splitmix ==0.1.0.5, + splitmix -optimised-mixer, + any.stm ==2.5.1.0, + any.strict ==0.5, + any.tagged ==0.8.8, + tagged +deepseq +transformers, + any.template-haskell ==2.19.0.0, + any.text ==2.0.2, + any.text-iso8601 ==0.1.1, + any.text-short ==0.1.6, + text-short -asserts, + any.th-abstraction ==0.7.0.0, + any.th-compat ==0.1.5, + any.these ==1.2.1, + any.time ==1.12.2, + any.time-compat ==1.9.7, + any.transformers ==0.5.6.2, + any.transformers-compat ==0.7.2, + transformers-compat -five +five-three -four +generic-deriving +mtl -three -two, + any.unix ==2.7.3, + any.unordered-containers ==0.2.20, + unordered-containers -debug, + any.uuid-types ==1.0.6, + any.vector ==0.13.1.0, + vector +boundschecks -internalchecks -unsafechecks -wall, + any.vector-stream ==0.1.0.1, + any.witherable ==0.5 +index-state: hackage.haskell.org 2024-06-17T00:48:51Z diff --git a/build/linux.riscv64/tag b/build/linux.riscv64/tag new file mode 100644 index 0000000..901eaaa --- /dev/null +++ b/build/linux.riscv64/tag @@ -0,0 +1 @@ +koalaman/scbuilder-linux-riscv64 diff --git a/build/linux.x86_64/Dockerfile b/build/linux.x86_64/Dockerfile new file mode 100644 index 0000000..edafb36 --- /dev/null +++ b/build/linux.x86_64/Dockerfile @@ -0,0 +1,24 @@ +FROM alpine:3.16 +# alpine:3.16 (GHC 9.0.1): 5.8 megabytes +# alpine:3.17 (GHC 9.0.2): 15.0 megabytes +# alpine:3.18 (GHC 9.4.4): 29.0 megabytes +# alpine:3.19 (GHC 9.4.7): 29.0 megabytes + +ENV TARGETNAME linux.x86_64 + +# Install GHC and cabal +USER root +RUN apk add ghc cabal g++ libffi-dev curl bash + +# Use ld.bfd instead of ld.gold due to +# x86_64-linux-gnu/libpthread.a(pthread_cond_init.o)(.note.stapsdt+0x14): error: +# relocation refers to local symbol "" [2], which is defined in a discarded section +ENV CABALOPTS "--ghc-options;-optl-Wl,-fuse-ld=bfd -split-sections -optc-Os -optc-Wl,--gc-sections" + +# Other archs pre-build dependencies here, but this one doesn't to detect ecosystem movement + +# Copy the build script +COPY build /usr/bin + +WORKDIR /scratch +ENTRYPOINT ["/usr/bin/build"] diff --git a/build/linux.x86_64/build b/build/linux.x86_64/build new file mode 100755 index 0000000..099f127 --- /dev/null +++ b/build/linux.x86_64/build @@ -0,0 +1,15 @@ +#!/bin/sh +set -xe +{ + tar xzv --strip-components=1 + chmod +x striptests && ./striptests + mkdir "$TARGETNAME" + cabal update + ( IFS=';'; cabal build $CABALOPTS --enable-executable-static ) + find . -name shellcheck -type f -exec mv {} "$TARGETNAME/" \; + ls -l "$TARGETNAME" + strip -s "$TARGETNAME/shellcheck" + ls -l "$TARGETNAME" + "$TARGETNAME/shellcheck" --version +} >&2 +tar czv "$TARGETNAME" diff --git a/build/linux.x86_64/tag b/build/linux.x86_64/tag new file mode 100644 index 0000000..f0224de --- /dev/null +++ b/build/linux.x86_64/tag @@ -0,0 +1 @@ +koalaman/scbuilder-linux-x86_64 diff --git a/build/run_builder b/build/run_builder new file mode 100755 index 0000000..d6de27b --- /dev/null +++ b/build/run_builder @@ -0,0 +1,30 @@ +#!/bin/bash +if [ $# -lt 2 ] +then + echo >&2 "This script builds a source archive (as produced by cabal sdist)" + echo >&2 "Usage: $0 sourcefile.tar.gz builddir..." + exit 1 +fi + +file=$(realpath "$1") +shift + +if [ ! -e "$file" ] +then + echo >&2 "$file does not exist" + exit 1 +fi + +set -ex -o pipefail + +for dir +do + tagfile="$dir/tag" + if [ ! -e "$tagfile" ] + then + echo >&2 "$tagfile does not exist" + exit 2 + fi + + docker run -i "$(< "$tagfile")" < "$file" | tar xz +done diff --git a/build/windows.x86_64/Dockerfile b/build/windows.x86_64/Dockerfile new file mode 100644 index 0000000..2ae78ac --- /dev/null +++ b/build/windows.x86_64/Dockerfile @@ -0,0 +1,27 @@ +FROM ubuntu:20.04 + +ENV TARGETNAME windows.x86_64 + +# We don't need wine32, even though it complains +USER root +ENV DEBIAN_FRONTEND noninteractive +RUN apt-get update && apt-get install -y curl busybox wine winbind + +# Fetch Windows version, will be available under z:\haskell +WORKDIR /haskell +RUN curl -L "https://downloads.haskell.org/~ghc/8.10.4/ghc-8.10.4-x86_64-unknown-mingw32.tar.xz" | tar xJ --strip-components=1 +WORKDIR /haskell/bin +RUN curl -L "https://downloads.haskell.org/~cabal/cabal-install-3.2.0.0/cabal-install-3.2.0.0-x86_64-unknown-mingw32.zip" | busybox unzip - +RUN curl -L "https://curl.se/windows/dl-8.7.1_7/curl-8.7.1_7-win64-mingw.zip" | busybox unzip - && mv curl-*-win64-mingw/bin/* . +ENV WINEPATH /haskell/bin + +# It's unknown whether Cabal on Windows suffers from the same issue +# that necessitated this but I don't care enough to find out +ENV CABALOPTS "--ghc-options;-split-sections -optc-Os -optc-Wl,--gc-sections" + +# Precompile some deps to speed up later builds +RUN wine /haskell/bin/cabal.exe update && IFS=';' && wine /haskell/bin/cabal.exe install --lib --dependencies-only $CABALOPTS ShellCheck + +COPY build /usr/bin +WORKDIR /scratch +ENTRYPOINT ["/usr/bin/build"] diff --git a/build/windows.x86_64/build b/build/windows.x86_64/build new file mode 100755 index 0000000..22e5b42 --- /dev/null +++ b/build/windows.x86_64/build @@ -0,0 +1,18 @@ +#!/bin/sh +cabal() { + wine /haskell/bin/cabal.exe "$@" +} + +set -xe +{ + tar xzv --strip-components=1 + chmod +x striptests && ./striptests + mkdir "$TARGETNAME" + ( IFS=';'; cabal build $CABALOPTS ) + find dist*/ -name shellcheck.exe -type f -ls -exec mv {} "$TARGETNAME/" \; + ls -l "$TARGETNAME" + wine "/haskell/mingw/bin/strip.exe" -s "$TARGETNAME/shellcheck.exe" + ls -l "$TARGETNAME" + wine "$TARGETNAME/shellcheck.exe" --version +} >&2 +tar czv "$TARGETNAME" diff --git a/build/windows.x86_64/tag b/build/windows.x86_64/tag new file mode 100644 index 0000000..a85921b --- /dev/null +++ b/build/windows.x86_64/tag @@ -0,0 +1 @@ +koalaman/scbuilder-windows-x86_64 diff --git a/doc/shellcheck_logo.svg b/doc/shellcheck_logo.svg new file mode 100644 index 0000000..836aa63 --- /dev/null +++ b/doc/shellcheck_logo.svg @@ -0,0 +1,294 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/manpage b/manpage new file mode 100755 index 0000000..0898092 --- /dev/null +++ b/manpage @@ -0,0 +1,4 @@ +#!/bin/sh +echo >&2 "Generating man page using pandoc" +pandoc -s -f markdown-smart -t man shellcheck.1.md -o shellcheck.1 || exit +echo >&2 "Done. You can read it with: man ./shellcheck.1" diff --git a/nextnumber b/nextnumber index 6b91a92..3bdf10a 100755 --- a/nextnumber +++ b/nextnumber @@ -6,8 +6,8 @@ then exit 1 fi -for i in 1 2 +for i in 1 2 3 do - last=$(grep -hv "^prop" ./**/*.hs | grep -Ewo "$i[0-9]{3}" | sort -n | tail -n 1) + last=$(grep -hv "^prop" ./**/*.hs | grep -Ewo "${i}[0-9]{3}" | sort -n | tail -n 1) echo "Next ${i}xxx: $((last+1))" done diff --git a/quickrun b/quickrun index 172ae88..e0e0547 100755 --- a/quickrun +++ b/quickrun @@ -2,4 +2,12 @@ # quickrun runs ShellCheck in an interpreted mode. # This allows testing changes without recompiling. -runghc -isrc -idist/build/autogen shellcheck.hs "$@" +path=$(find . -type f -path './dist*/Paths_ShellCheck.hs' | sort | head -n 1) +if [ -z "$path" ] +then + echo >&2 "Unable to find Paths_ShellCheck.hs. Please 'cabal build' once." + exit 1 +fi +path="${path%/*}" + +exec runghc -isrc -i"$path" shellcheck.hs "$@" diff --git a/quicktest b/quicktest index 42a7ba6..6a1cf61 100755 --- a/quicktest +++ b/quicktest @@ -3,16 +3,18 @@ # This allows running tests without compiling, which can be faster. # 'cabal test' remains the source of truth. +path=$(find . -type f -path './dist*/Paths_ShellCheck.hs' | sort | head -n 1) +if [ -z "$path" ] +then + echo >&2 "Unable to find Paths_ShellCheck.hs. Please 'cabal build' once." + exit 1 +fi +path="${path%/*}" + + ( - var=$(echo 'liftM and $ sequence [ - ShellCheck.Analytics.runTests - ,ShellCheck.Parser.runTests - ,ShellCheck.Checker.runTests - ,ShellCheck.Checks.Commands.runTests - ,ShellCheck.Checks.ShellSupport.runTests - ,ShellCheck.AnalyzerLib.runTests - ]' | tr -d '\n' | cabal repl 2>&1 | tee /dev/stderr) -if [[ $var == *$'\nTrue'* ]] +var=$(echo 'main' | ghci -isrc -i"$path" test/shellcheck.hs 2>&1 | tee /dev/stderr) +if [[ $var == *ExitSuccess* ]] then exit 0 else diff --git a/setgitversion b/setgitversion new file mode 100755 index 0000000..3afad61 --- /dev/null +++ b/setgitversion @@ -0,0 +1,11 @@ +#!/bin/sh -xe +# This script hardcodes the `git describe` version as ShellCheck's version number. +# This is done to allow shellcheck --version to differ from the cabal version when +# building git snapshots. + +file="src/ShellCheck/Data.hs" +test -e "$file" +tmp=$(mktemp) +version=$(git describe) +sed -e "s/=.*VERSIONSTRING.*/= \"$version\" -- VERSIONSTRING, DO NOT SUBMIT/" "$file" > "$tmp" +mv "$tmp" "$file" diff --git a/shellcheck.1.md b/shellcheck.1.md index 6600613..c768bfe 100644 --- a/shellcheck.1.md +++ b/shellcheck.1.md @@ -29,14 +29,13 @@ will warn that decimals are not supported. + For scripts starting with `#!/bin/ksh` (or using `-s ksh`), ShellCheck will not warn at all, as `ksh` supports decimals in arithmetic contexts. - # OPTIONS **-a**,\ **--check-sourced** : Emit warnings in sourced files. Normally, `shellcheck` will only warn about issues in the specified files. With this option, any issues in - sourced files files will also be reported. + sourced files will also be reported. **-C**[*WHEN*],\ **--color**[=*WHEN*] @@ -44,28 +43,71 @@ not warn at all, as `ksh` supports decimals in arithmetic contexts. is *auto*. **--color** without an argument is equivalent to **--color=always**. +**-i**\ *CODE1*[,*CODE2*...],\ **--include=***CODE1*[,*CODE2*...] + +: Explicitly include only the specified codes in the report. Subsequent **-i** + options are cumulative, but all the codes can be specified at once, + comma-separated as a single argument. Include options override any provided + exclude options. + **-e**\ *CODE1*[,*CODE2*...],\ **--exclude=***CODE1*[,*CODE2*...] : Explicitly exclude the specified codes from the report. Subsequent **-e** options are cumulative, but all the codes can be specified at once, comma-separated as a single argument. +**--extended-analysis=true/false** + +: Enable/disable Dataflow Analysis to identify more issues (default true). If + ShellCheck uses too much CPU/RAM when checking scripts with several + thousand lines of code, extended analysis can be disabled with this flag + or a directive. This flag overrides directives and rc files. + **-f** *FORMAT*, **--format=***FORMAT* : Specify the output format of shellcheck, which prints its results in the standard output. Subsequent **-f** options are ignored, see **FORMATS** below for more information. -**-S**\ *SEVERITY*,\ **--severity=***severity* +**--list-optional** -: Specify minimum severity of errors to consider. Valid values are *error*, - *warning*, *info* and *style*. The default is *style*. +: Output a list of known optional checks. These can be enabled with **-o** + flags or **enable** directives. + +**--norc** + +: Don't try to look for .shellcheckrc configuration files. + +**--rcfile** *RCFILE* + +: Prefer the specified configuration file over searching for one + in the default locations. + +**-o**\ *NAME1*[,*NAME2*...],\ **--enable=***NAME1*[,*NAME2*...] + +: Enable optional checks. The special name *all* enables all of them. + Subsequent **-o** options accumulate. This is equivalent to specifying + **enable** directives. + +**-P**\ *SOURCEPATH*,\ **--source-path=***SOURCEPATH* + +: Specify paths to search for sourced files, separated by `:` on Unix and + `;` on Windows. This is equivalent to specifying `search-path` + directives. **-s**\ *shell*,\ **--shell=***shell* -: Specify Bourne shell dialect. Valid values are *sh*, *bash*, *dash* and *ksh*. - The default is to use the file's shebang, or *bash* if the target shell - can't be determined. +: Specify Bourne shell dialect. Valid values are *sh*, *bash*, *dash*, *ksh*, + and *busybox*. + The default is to deduce the shell from the file's `shell` directive, + shebang, or `.bash/.bats/.dash/.ksh` extension, in that order. *sh* refers to + POSIX `sh` (not the system's), and will warn of portability issues. + +**-S**\ *SEVERITY*,\ **--severity=***severity* + +: Specify minimum severity of errors to consider. Valid values in order of + severity are *error*, *warning*, *info* and *style*. + The default is *style*. **-V**,\ **--version** @@ -78,11 +120,18 @@ not warn at all, as `ksh` supports decimals in arithmetic contexts. **-x**,\ **--external-sources** -: Follow 'source' statements even when the file is not specified as input. +: Follow `source` statements even when the file is not specified as input. By default, `shellcheck` will only follow files specified on the command line (plus `/dev/null`). This option allows following any file the script may `source`. + This option may also be enabled using `external-sources=true` in + `.shellcheckrc`. This flag takes precedence. + +**FILES...** + +: One or more script files to check, or "-" for standard input. + # FORMATS @@ -119,27 +168,59 @@ not warn at all, as `ksh` supports decimals in arithmetic contexts. ... -**json** +**diff** + +: Auto-fixes in unified diff format. Can be piped to `git apply` or `patch -p1` + to automatically apply fixes. + + --- a/test.sh + +++ b/test.sh + @@ -2,6 +2,6 @@ + ## Example of a broken script. + for f in $(ls *.m3u) + do + - grep -qi hq.*mp3 $f \ + + grep -qi hq.*mp3 "$f" \ + && echo -e 'Playlist $f contains a HQ file in mp3 format' + done + + +**json1** : Json is a popular serialization format that is more suitable for web applications. ShellCheck's json is compact and contains only the bare - minimum. + minimum. Tabs are counted as 1 character. + + { + comments: [ + { + "file": "filename", + "line": lineNumber, + "column": columnNumber, + "level": "severitylevel", + "code": errorCode, + "message": "warning message" + }, + ... + ] + } + +**json** + +: This is a legacy version of the **json1** format. It's a raw array of + comments, and all offsets have a tab stop of 8. + +**quiet** + +: Suppress all normal output. Exit with zero if no issues are found, + otherwise exit with one. Stops processing after the first issue. - [ - { - "file": "filename", - "line": lineNumber, - "column": columnNumber, - "level": "severitylevel", - "code": errorCode, - "message": "warning message" - }, - ... - ] # DIRECTIVES -ShellCheck directives can be specified as comments in the shell script -before a command or block: + +ShellCheck directives can be specified as comments in the shell script. +If they appear before the first command, they are considered file-wide. +Otherwise, they apply to the immediately following command or block: # shellcheck key=value key=value command-or-structure @@ -167,19 +248,88 @@ Valid keys are: **disable** : Disables a comma separated list of error codes for the following command. The command can be a simple command like `echo foo`, or a compound command - like a function definition, subshell block or loop. + like a function definition, subshell block or loop. A range can be + be specified with a dash, e.g. `disable=SC3000-SC4000` to exclude 3xxx. + All warnings can be disabled with `disable=all`. + +**enable** +: Enable an optional check by name, as listed with **--list-optional**. + Only file-wide `enable` directives are considered. + +**extended-analysis** +: Set to true/false to enable/disable dataflow analysis. Specifying + `# shellcheck extended-analysis=false` in particularly large (2000+ line) + auto-generated scripts will reduce ShellCheck's resource usage at the + expense of certain checks. Extended analysis is enabled by default. + +**external-sources** +: Set to `true` in `.shellcheckrc` to always allow ShellCheck to open + arbitrary files from 'source' statements (the way most tools do). + + This option defaults to `false` only due to ShellCheck's origin as a + remote service for checking untrusted scripts. It can safely be enabled + for normal development. **source** : Overrides the filename included by a `source`/`.` statement. This can be used to tell shellcheck where to look for a file whose name is determined at runtime, or to skip a source by telling it to use `/dev/null`. +**source-path** +: Add a directory to the search path for `source`/`.` statements (by default, + only ShellCheck's working directory is included). Absolute paths will also + be rooted in these paths. The special path `SCRIPTDIR` can be used to + specify the currently checked script's directory, as in + `source-path=SCRIPTDIR` or `source-path=SCRIPTDIR/../libs`. Multiple + paths accumulate, and `-P` takes precedence over them. + **shell** : Overrides the shell detected from the shebang. This is useful for files meant to be included (and thus lacking a shebang), or possibly - as a more targeted alternative to 'disable=2039'. + as a more targeted alternative to 'disable=SC2039'. + +# RC FILES + +Unless `--norc` is used, ShellCheck will look for a file `.shellcheckrc` or +`shellcheckrc` in the script's directory and each parent directory. If found, +it will read `key=value` pairs from it and treat them as file-wide directives. + +Here is an example `.shellcheckrc`: + + # Look for 'source'd files relative to the checked script, + # and also look for absolute paths in /mnt/chroot + source-path=SCRIPTDIR + source-path=/mnt/chroot + + # Since 0.9.0, values can be quoted with '' or "" to allow spaces + source-path="My Documents/scripts" + + # Allow opening any 'source'd file, even if not specified as input + external-sources=true + + # Turn on warnings for unquoted variables with safe values + enable=quote-safe-variables + + # Turn on warnings for unassigned uppercase variables + enable=check-unassigned-uppercase + + # Allow [ ! -z foo ] instead of suggesting -n + disable=SC2236 + +If no `.shellcheckrc` is found in any of the parent directories, ShellCheck +will look in `~/.shellcheckrc` followed by the `$XDG_CONFIG_HOME` +(usually `~/.config/shellcheckrc`) on Unix, or `%APPDATA%/shellcheckrc` on +Windows. Only the first file found will be used. + +Note for Snap users: the Snap sandbox disallows access to hidden files. +Use `shellcheckrc` without the dot instead. + +Note for Docker users: ShellCheck will only be able to look for files that +are mounted in the container, so `~/.shellcheckrc` will not be read. + # ENVIRONMENT VARIABLES + The environment variable `SHELLCHECK_OPTS` can be set with default flags: export SHELLCHECK_OPTS='--shell=bash --exclude=SC2016' @@ -189,7 +339,7 @@ invocation. # RETURN VALUES -ShellCheck uses the follow exit codes: +ShellCheck uses the following exit codes: + 0: All files successfully scanned with no issues. + 1: All files successfully scanned with some issues. @@ -198,6 +348,7 @@ ShellCheck uses the follow exit codes: + 4: ShellCheck was invoked with bad options (e.g. unknown formatter). # LOCALE + This version of ShellCheck is only available in English. All files are leniently decoded as UTF-8, with a fallback of ISO-8859-1 for invalid sequences. `LC_CTYPE` is respected for output, and defaults to UTF-8 for @@ -206,20 +357,50 @@ locales where encoding is unspecified (such as the `C` locale). Windows users seeing `commitBuffer: invalid argument (invalid character)` should set their terminal to use UTF-8 with `chcp 65001`. -# AUTHOR -ShellCheck is written and maintained by Vidar Holen. +# KNOWN INCOMPATIBILITIES + +(If nothing in this section makes sense, you are unlikely to be affected by it) + +To avoid confusing and misguided suggestions, ShellCheck requires function +bodies to be either `{ brace groups; }` or `( subshells )`, and function names +containing `[]*=!` are only recognized after a `function` keyword. + +The following unconventional function definitions are identical in Bash, +but ShellCheck only recognizes the latter. + + [x!=y] () [[ $1 ]] + function [x!=y] () { [[ $1 ]]; } + +Shells without the `function` keyword do not allow these characters in function +names to begin with. Function names containing `{}` are not supported at all. + +Further, if ShellCheck sees `[x!=y]` it will assume this is an invalid +comparison. To invoke the above function, quote the command as in `'[x!=y]'`, +or to retain the same globbing behavior, use `command [x!=y]`. + +ShellCheck imposes additional restrictions on the `[` command to help diagnose +common invalid uses. While `[ $x= 1 ]` is defined in POSIX, ShellCheck will +assume it was intended as the much more likely comparison `[ "$x" = 1 ]` and +fail accordingly. For unconventional or dynamic uses of the `[` command, use +`test` or `\[` instead. # REPORTING BUGS + Bugs and issues can be reported on GitHub: https://github.com/koalaman/shellcheck/issues +# AUTHORS + +ShellCheck is developed and maintained by Vidar Holen, with assistance from a +long list of wonderful contributors. + # COPYRIGHT -Copyright 2012-2015, Vidar Holen. + +Copyright 2012-2024, Vidar Holen and contributors. Licensed under the GNU General Public License version 3 or later, see https://gnu.org/licenses/gpl.html - # SEE ALSO -sh(1) bash(1) +sh(1) bash(1) dash(1) ksh(1) diff --git a/shellcheck.hs b/shellcheck.hs index 6b9047c..def3654 100644 --- a/shellcheck.hs +++ b/shellcheck.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2019 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -17,6 +17,7 @@ You should have received a copy of the GNU General Public License along with this program. If not, see . -} +import qualified ShellCheck.Analyzer import ShellCheck.Checker import ShellCheck.Data import ShellCheck.Interface @@ -24,12 +25,17 @@ import ShellCheck.Regex import qualified ShellCheck.Formatter.CheckStyle import ShellCheck.Formatter.Format +import qualified ShellCheck.Formatter.Diff import qualified ShellCheck.Formatter.GCC import qualified ShellCheck.Formatter.JSON +import qualified ShellCheck.Formatter.JSON1 import qualified ShellCheck.Formatter.TTY +import qualified ShellCheck.Formatter.Quiet import Control.Exception import Control.Monad +import Control.Monad.IO.Class +import Control.Monad.Trans.Class import Control.Monad.Except import Data.Bits import Data.Char @@ -46,6 +52,7 @@ import System.Console.GetOpt import System.Directory import System.Environment import System.Exit +import System.FilePath import System.IO data Flag = Flag String String @@ -67,17 +74,21 @@ instance Monoid Status where data Options = Options { checkSpec :: CheckSpec, externalSources :: Bool, + sourcePaths :: [FilePath], formatterOptions :: FormatterOptions, - minSeverity :: Severity + minSeverity :: Severity, + rcfile :: Maybe FilePath } defaultOptions = Options { checkSpec = emptyCheckSpec, externalSources = False, + sourcePaths = [], formatterOptions = newFormatterOptions { foColorOption = ColorAuto }, - minSeverity = StyleC + minSeverity = StyleC, + rcfile = Nothing } usageHeader = "Usage: shellcheck [OPTIONS...] FILES..." @@ -87,14 +98,31 @@ options = [ Option "C" ["color"] (OptArg (maybe (Flag "color" "always") (Flag "color")) "WHEN") "Use color (auto, always, never)", + Option "i" ["include"] + (ReqArg (Flag "include") "CODE1,CODE2..") "Consider only given types of warnings", Option "e" ["exclude"] (ReqArg (Flag "exclude") "CODE1,CODE2..") "Exclude types of warnings", + Option "" ["extended-analysis"] + (ReqArg (Flag "extended-analysis") "bool") "Perform dataflow analysis (default true)", Option "f" ["format"] (ReqArg (Flag "format") "FORMAT") $ "Output format (" ++ formatList ++ ")", + Option "" ["list-optional"] + (NoArg $ Flag "list-optional" "true") "List checks disabled by default", + Option "" ["norc"] + (NoArg $ Flag "norc" "true") "Don't look for .shellcheckrc files", + Option "" ["rcfile"] + (ReqArg (Flag "rcfile") "RCFILE") + "Prefer the specified configuration file over searching for one", + Option "o" ["enable"] + (ReqArg (Flag "enable") "check1,check2..") + "List of optional checks to enable (or 'all')", + Option "P" ["source-path"] + (ReqArg (Flag "source-path") "SOURCEPATHS") + "Specify path when looking for sourced files (\"SCRIPTDIR\" for script's dir)", Option "s" ["shell"] (ReqArg (Flag "shell") "SHELLNAME") - "Specify dialect (sh, bash, dash, ksh)", + "Specify dialect (sh, bash, dash, ksh, busybox)", Option "S" ["severity"] (ReqArg (Flag "severity") "SEVERITY") "Minimum severity of errors to consider (error, warning, info, style)", @@ -102,10 +130,13 @@ options = [ (NoArg $ Flag "version" "true") "Print version information", Option "W" ["wiki-link-count"] (ReqArg (Flag "wiki-link-count") "NUM") - "The number of wiki links to show, when applicable.", + "The number of wiki links to show, when applicable", Option "x" ["external-sources"] - (NoArg $ Flag "externals" "true") "Allow 'source' outside of FILES" + (NoArg $ Flag "externals" "true") "Allow 'source' outside of FILES", + Option "" ["help"] + (NoArg $ Flag "help" "true") "Show this usage summary and exit" ] +getUsageInfo = usageInfo usageHeader options printErr = lift . hPutStrLn stderr @@ -114,15 +145,18 @@ parseArguments argv = case getOpt Permute options argv of (opts, files, []) -> return (opts, files) (_, _, errors) -> do - printErr $ concat errors ++ "\n" ++ usageInfo usageHeader options + printErr $ concat errors ++ "\n" ++ getUsageInfo throwError SyntaxFailure formats :: FormatterOptions -> Map.Map String (IO Formatter) formats options = Map.fromList [ ("checkstyle", ShellCheck.Formatter.CheckStyle.format), + ("diff", ShellCheck.Formatter.Diff.format options), ("gcc", ShellCheck.Formatter.GCC.format), ("json", ShellCheck.Formatter.JSON.format), - ("tty", ShellCheck.Formatter.TTY.format options) + ("json1", ShellCheck.Formatter.JSON1.format), + ("tty", ShellCheck.Formatter.TTY.format options), + ("quiet", ShellCheck.Formatter.Quiet.format options) ] formatList = intercalate ", " names @@ -200,7 +234,7 @@ runFormatter sys format options files = do f :: Status -> FilePath -> IO Status f status file = do newStatus <- process file `catch` handler file - return $ status `mappend` newStatus + return $! status `mappend` newStatus handler :: FilePath -> IOException -> IO Status handler file e = reportFailure file (show e) reportFailure file str = do @@ -209,7 +243,7 @@ runFormatter sys format options files = do process :: FilePath -> IO Status process filename = do - input <- siReadFile sys filename + input <- siReadFile sys Nothing filename either (reportFailure filename) check input where check contents = do @@ -225,9 +259,9 @@ runFormatter sys format options files = do else SomeProblems parseEnum name value list = - case filter ((== value) . fst) list of - [(name, value)] -> return value - [] -> do + case lookup value list of + Just value -> return value + Nothing -> do printErr $ "Unknown value for --" ++ name ++ ". " ++ "Valid options are: " ++ (intercalate ", " $ map fst list) throwError SupportFailure @@ -267,10 +301,30 @@ parseOption flag options = } } + Flag "include" str -> do + new <- mapM parseNum $ filter (not . null) $ split ',' str + let old = csIncludedWarnings . checkSpec $ options + return options { + checkSpec = (checkSpec options) { + csIncludedWarnings = + if null new + then old + else Just new `mappend` old + } + } + Flag "version" _ -> do liftIO printVersion throwError NoProblems + Flag "list-optional" _ -> do + liftIO printOptional + throwError NoProblems + + Flag "help" _ -> do + liftIO $ putStrLn getUsageInfo + throwError NoProblems + Flag "externals" _ -> return options { externalSources = True @@ -284,6 +338,12 @@ parseOption flag options = } } + Flag "source-path" str -> do + let paths = splitSearchPath str + return options { + sourcePaths = (sourcePaths options) ++ paths + } + Flag "sourced" _ -> return options { checkSpec = (checkSpec options) { @@ -307,7 +367,39 @@ parseOption flag options = } } - _ -> return options + Flag "norc" _ -> + return options { + checkSpec = (checkSpec options) { + csIgnoreRC = True + } + } + + Flag "rcfile" str -> do + return options { + rcfile = Just str + } + + Flag "enable" value -> + let cs = checkSpec options in return options { + checkSpec = cs { + csOptionalChecks = (csOptionalChecks cs) ++ split ',' value + } + } + + Flag "extended-analysis" str -> do + value <- parseBool str + return options { + checkSpec = (checkSpec options) { + csExtendedAnalysis = Just value + } + } + + -- This flag is handled specially in 'process' + Flag "format" _ -> return options + + Flag str _ -> do + printErr $ "Internal error for --" ++ str ++ ". Please file a bug :(" + return options where die s = do printErr s @@ -319,23 +411,36 @@ parseOption flag options = throwError SyntaxFailure return (Prelude.read num :: Integer) + parseBool str = do + case str of + "true" -> return True + "false" -> return False + _ -> do + printErr $ "Invalid boolean, expected true/false: " ++ str + throwError SyntaxFailure + +ioInterface :: Options -> [FilePath] -> IO (SystemInterface IO) ioInterface options files = do inputs <- mapM normalize files cache <- newIORef emptyCache - return SystemInterface { - siReadFile = get cache inputs + configCache <- newIORef ("", Nothing) + return (newSystemInterface :: SystemInterface IO) { + siReadFile = get cache inputs, + siFindSource = findSourceFile inputs (sourcePaths options), + siGetConfig = getConfig configCache } where emptyCache :: Map.Map FilePath String emptyCache = Map.empty - get cache inputs file = do + + get cache inputs rcSuggestsExternal file = do map <- readIORef cache case Map.lookup file map of Just x -> return $ Right x - Nothing -> fetch cache inputs file + Nothing -> fetch cache inputs rcSuggestsExternal file - fetch cache inputs file = do - ok <- allowable inputs file + fetch cache inputs rcSuggestsExternal file = do + ok <- allowable rcSuggestsExternal inputs file if ok then (do (contents, shouldCache) <- inputFile file @@ -343,14 +448,16 @@ ioInterface options files = do modifyIORef cache $ Map.insert file contents return $ Right contents ) `catch` handler - else return $ Left (file ++ " was not specified as input (see shellcheck -x).") - + else + if rcSuggestsExternal == Just False + then return $ Left (file ++ " was not specified as input, and external files were disabled via directive.") + else return $ Left (file ++ " was not specified as input (see shellcheck -x).") where handler :: IOException -> IO (Either ErrorMessage String) handler ex = return . Left $ show ex - allowable inputs x = - if externalSources options + allowable rcSuggestsExternal inputs x = + if fromMaybe (externalSources options) rcSuggestsExternal then return True else do path <- normalize x @@ -362,6 +469,103 @@ ioInterface options files = do fallback :: FilePath -> IOException -> IO FilePath fallback path _ = return path + + -- Returns the name and contents of .shellcheckrc for the given file + getConfig cache filename = + case rcfile options of + Just file -> do + -- We have a specified rcfile. Ignore normal rcfile resolution. + (path, result) <- readIORef cache + if path == "/" + then return result + else do + result <- readConfig file + when (isNothing result) $ + hPutStrLn stderr $ "Warning: unable to read --rcfile " ++ file + writeIORef cache ("/", result) + return result + + Nothing -> do + path <- normalize filename + let dir = takeDirectory path + (previousPath, result) <- readIORef cache + if dir == previousPath + then return result + else do + paths <- getConfigPaths dir + result <- findConfig paths + writeIORef cache (dir, result) + return result + + findConfig paths = + case paths of + (file:rest) -> do + contents <- readConfig file + if isJust contents + then return contents + else findConfig rest + [] -> return Nothing + + -- Get a list of candidate filenames. This includes .shellcheckrc + -- in all parent directories, plus the user's home dir and xdg dir. + -- The dot is optional for Windows and Snap users. + getConfigPaths dir = do + let next = takeDirectory dir + rest <- if next /= dir + then getConfigPaths next + else defaultPaths `catch` + ((const $ return []) :: IOException -> IO [FilePath]) + return $ (dir ".shellcheckrc") : (dir "shellcheckrc") : rest + + defaultPaths = do + home <- getAppUserDataDirectory "shellcheckrc" + xdg <- getXdgDirectory XdgConfig "shellcheckrc" + return [home, xdg] + + readConfig file = do + exists <- doesFileExist file + if exists + then do + (contents, _) <- inputFile file `catch` handler file + return $ Just (file, contents) + else + return Nothing + where + handler :: FilePath -> IOException -> IO (String, Bool) + handler file err = do + hPutStrLn stderr $ file ++ ": " ++ show err + return ("", True) + + andM a b arg = do + first <- a arg + if not first then return False else b arg + + findM p = foldr go (pure Nothing) + where + go x acc = do + b <- p x + if b then pure (Just x) else acc + + findSourceFile inputs sourcePathFlag currentScript rcSuggestsExternal sourcePathAnnotation original = + if isAbsolute original + then + let (_, relative) = splitDrive original + in find relative original + else + find original original + where + find filename deflt = do + sources <- findM ((allowable rcSuggestsExternal inputs) `andM` doesFileExist) $ + (adjustPath filename):(map (( filename) . adjustPath) $ sourcePathFlag ++ sourcePathAnnotation) + case sources of + Nothing -> return deflt + Just first -> return first + scriptdir = dropFileName currentScript + adjustPath str = + case (splitDirectories str) of + ("SCRIPTDIR":rest) -> joinPath (scriptdir:rest) + _ -> str + inputFile file = do (handle, shouldCache) <- if file == "-" @@ -418,3 +622,14 @@ printVersion = do putStrLn $ "version: " ++ shellcheckVersion putStrLn "license: GNU General Public License, version 3" putStrLn "website: https://www.shellcheck.net" + +printOptional = do + mapM f list + where + list = sortOn cdName ShellCheck.Analyzer.optionalChecks + f item = do + putStrLn $ "name: " ++ cdName item + putStrLn $ "desc: " ++ cdDescription item + putStrLn $ "example: " ++ cdPositive item + putStrLn $ "fix: " ++ cdNegative item + putStrLn "" diff --git a/snap/snapcraft.yaml b/snap/snapcraft.yaml index b7f0e96..f294c4e 100644 --- a/snap/snapcraft.yaml +++ b/snap/snapcraft.yaml @@ -16,31 +16,41 @@ description: | advanced user's otherwise working script to fail under future circumstances. - By default ShellCheck can only check non-hidden files under /home, to make + By default ShellCheck can only check non-hidden files under /home, to make ShellCheck be able to check files under /media and /run/media you must connect it to the `removable-media` interface manually: # snap connect shellcheck:removable-media - + version: git -grade: devel +base: core20 +grade: stable confinement: strict apps: shellcheck: command: usr/bin/shellcheck plugs: [home, removable-media] + environment: + LANG: C.UTF-8 parts: shellcheck: plugin: dump - source: ./ + source: . build-packages: - cabal-install - build: | + stage-packages: + - libatomic1 + override-build: | + # Give ourselves enough memory to build + dd if=/dev/zero of=/tmp/swap bs=1M count=2000 + mkswap /tmp/swap + swapon /tmp/swap + cabal sandbox init cabal update cabal install -j - install: | + install -d $SNAPCRAFT_PART_INSTALL/usr/bin install .cabal-sandbox/bin/shellcheck $SNAPCRAFT_PART_INSTALL/usr/bin diff --git a/src/ShellCheck/ASTLib.hs b/src/ShellCheck/ASTLib.hs index ef72b09..1e1b9cd 100644 --- a/src/ShellCheck/ASTLib.hs +++ b/src/ShellCheck/ASTLib.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2021 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -17,16 +17,27 @@ You should have received a copy of the GNU General Public License along with this program. If not, see . -} +{-# LANGUAGE TemplateHaskell #-} module ShellCheck.ASTLib where import ShellCheck.AST +import ShellCheck.Prelude +import ShellCheck.Regex import Control.Monad.Writer import Control.Monad import Data.Char import Data.Functor +import Data.Functor.Identity import Data.List import Data.Maybe +import qualified Data.List.NonEmpty as NE +import qualified Data.Map as Map +import Numeric (showHex) + +import Test.QuickCheck + +arguments (T_SimpleCommand _ _ (cmd:args)) = args -- Is this a type of loop? isLoop t = case t of @@ -46,13 +57,32 @@ willSplit x = T_BraceExpansion {} -> True T_Glob {} -> True T_Extglob {} -> True + T_DoubleQuoted _ l -> any willBecomeMultipleArgs l T_NormalWord _ l -> any willSplit l _ -> False -isGlob T_Extglob {} = True -isGlob T_Glob {} = True -isGlob (T_NormalWord _ l) = any isGlob l -isGlob _ = False +isGlob t = case t of + T_Extglob {} -> True + T_Glob {} -> True + T_NormalWord _ l -> any isGlob l || hasSplitRange l + _ -> False + where + -- foo[x${var}y] gets parsed as foo,[,x,$var,y], + -- so check if there's such an interval + hasSplitRange l = + let afterBracket = dropWhile (not . isHalfOpenRange) l + in any isClosingRange afterBracket + + isHalfOpenRange t = + case t of + T_Literal _ "[" -> True + _ -> False + + isClosingRange t = + case t of + T_Literal _ str -> ']' `elem` str + _ -> False + -- Is this shell word a constant? isConstant token = @@ -81,7 +111,7 @@ oversimplify token = (T_NormalWord _ l) -> [concat (concatMap oversimplify l)] (T_DoubleQuoted _ l) -> [concat (concatMap oversimplify l)] (T_SingleQuoted _ s) -> [s] - (T_DollarBraced _ _) -> ["${VAR}"] + (T_DollarBraced _ _ _) -> ["${VAR}"] (T_DollarArithmetic _ _) -> ["${VAR}"] (T_DollarExpansion _ _) -> ["${VAR}"] (T_Backticked _ _) -> ["${VAR}"] @@ -110,7 +140,7 @@ getFlagsUntil stopCondition (T_SimpleCommand _ _ (_:args)) = flag (x, '-':'-':arg) = [ (x, takeWhile (/= '=') arg) ] flag (x, '-':args) = map (\v -> (x, [v])) args flag (x, _) = [ (x, "") ] -getFlagsUntil _ _ = error "Internal shellcheck error, please report! (getFlags on non-command)" +getFlagsUntil _ _ = error $ pleaseReport "getFlags on non-command" -- Get all flags in a GNU way, up until -- getAllFlags :: Token -> [(Token, String)] @@ -128,30 +158,146 @@ isFlag token = _ -> False -- Is this token a flag where the - is unquoted? -isUnquotedFlag token = fromMaybe False $ do - str <- getLeadingUnquotedString token - return $ "-" `isPrefixOf` str +isUnquotedFlag token = + case getLeadingUnquotedString token of + Just ('-':_) -> True + _ -> False + +-- getGnuOpts "erd:u:" will parse a list of arguments tokens like `read` +-- -re -d : -u 3 bar +-- into +-- Just [("r", (-re, -re)), ("e", (-re, -re)), ("d", (-d,:)), ("u", (-u,3)), ("", (bar,bar))] +-- +-- Each string flag maps to a tuple of (flag, argument), where argument=flag if it +-- doesn't take a specific one. +-- +-- Any unrecognized flag will result in Nothing. The exception is if arbitraryLongOpts +-- is set, in which case --anything will map to "anything". +getGnuOpts :: String -> [Token] -> Maybe [(String, (Token, Token))] +getGnuOpts str args = getOpts (True, False) str [] args + +-- As above, except the first non-arg string will treat the rest as arguments +getBsdOpts :: String -> [Token] -> Maybe [(String, (Token, Token))] +getBsdOpts str args = getOpts (False, False) str [] args + +-- Tests for this are in Commands.hs where it's more frequently used +getOpts :: + -- Behavioral config: gnu style, allow arbitrary long options + (Bool, Bool) + -- A getopts style string + -> String + -- List of long options and whether they take arguments + -> [(String, Bool)] + -- List of arguments (excluding command) + -> [Token] + -- List of flags to tuple of (optionToken, valueToken) + -> Maybe [(String, (Token, Token))] + +getOpts (gnu, arbitraryLongOpts) string longopts args = process args + where + flagList (c:':':rest) = ([c], True) : flagList rest + flagList (c:rest) = ([c], False) : flagList rest + flagList [] = longopts + flagMap = Map.fromList $ ("", False) : flagList string + + process [] = return [] + process (token:rest) = do + case getLiteralStringDef "\0" token of + "--" -> return $ listToArgs rest + '-':'-':word -> do + let (name, arg) = span (/= '=') word + needsArg <- + if arbitraryLongOpts + then return $ Map.findWithDefault False name flagMap + else Map.lookup name flagMap + + if needsArg && null arg + then + case rest of + (arg:rest2) -> do + more <- process rest2 + return $ (name, (token, arg)) : more + _ -> fail "Missing arg" + else do + more <- process rest + -- Consider splitting up token to get arg + return $ (name, (token, token)) : more + '-':opts -> shortToOpts opts token rest + arg -> + if gnu + then do + more <- process rest + return $ ("", (token, token)):more + else return $ listToArgs (token:rest) + + shortToOpts opts token args = + case opts of + c:rest -> do + needsArg <- Map.lookup [c] flagMap + case () of + _ | needsArg && null rest -> do + (next:restArgs) <- return args + more <- process restArgs + return $ ([c], (token, next)):more + _ | needsArg -> do + more <- process args + return $ ([c], (token, token)):more + _ -> do + more <- shortToOpts rest token args + return $ ([c], (token, token)):more + [] -> process args + + listToArgs = map (\x -> ("", (x, x))) + + +-- Generic getOpts that doesn't rely on a format string, but may also be inaccurate. +-- This provides a best guess interpretation instead of failing when new options are added. +-- +-- "--" is treated as end of arguments +-- "--anything[=foo]" is treated as a long option without argument +-- "-any" is treated as -a -n -y, with the next arg as an option to -y unless it starts with - +-- anything else is an argument +getGenericOpts :: [Token] -> [(String, (Token, Token))] +getGenericOpts = process + where + process (token:rest) = + case getLiteralStringDef "\0" token of + "--" -> map (\c -> ("", (c,c))) rest + '-':'-':word -> (takeWhile (`notElem` "\0=") word, (token, token)) : process rest + '-':optString -> + let opts = takeWhile (/= '\0') optString + in + case rest of + next:_ | "-" `isPrefixOf` getLiteralStringDef "\0" next -> + map (\c -> ([c], (token, token))) opts ++ process rest + next:remainder -> + case reverse opts of + last:initial -> + map (\c -> ([c], (token, token))) (reverse initial) + ++ [([last], (token, next))] + ++ process remainder + [] -> process remainder + [] -> map (\c -> ([c], (token, token))) opts + _ -> ("", (token, token)) : process rest + process [] = [] --- Given a T_DollarBraced, return a simplified version of the string contents. -bracedString (T_DollarBraced _ l) = concat $ oversimplify l -bracedString _ = error "Internal shellcheck error, please report! (bracedString on non-variable)" -- Is this an expansion of multiple items of an array? -isArrayExpansion t@(T_DollarBraced _ _) = - let string = bracedString t in +isArrayExpansion (T_DollarBraced _ _ l) = + let string = concat $ oversimplify l in "@" `isPrefixOf` string || not ("#" `isPrefixOf` string) && "[@]" `isInfixOf` string isArrayExpansion _ = False -- Is it possible that this arg becomes multiple args? -mayBecomeMultipleArgs t = willBecomeMultipleArgs t || f t +mayBecomeMultipleArgs t = willBecomeMultipleArgs t || f False t where - f t@(T_DollarBraced _ _) = - let string = bracedString t in - "!" `isPrefixOf` string - f (T_DoubleQuoted _ parts) = any f parts - f (T_NormalWord _ parts) = any f parts - f _ = False + f quoted (T_DollarBraced _ _ l) = + let string = concat $ oversimplify l in + not quoted || "!" `isPrefixOf` string + f quoted (T_DoubleQuoted _ parts) = any (f True) parts + f quoted (T_NormalWord _ parts) = any (f quoted) parts + f _ _ = False -- Is it certain that this word will becomes multiple words? willBecomeMultipleArgs t = willConcatInAssignment t || f t @@ -159,7 +305,6 @@ willBecomeMultipleArgs t = willConcatInAssignment t || f t f T_Extglob {} = True f T_Glob {} = True f T_BraceExpansion {} = True - f (T_DoubleQuoted _ parts) = any f parts f (T_NormalWord _ parts) = any f parts f _ = False @@ -175,9 +320,13 @@ willConcatInAssignment token = getLiteralString :: Token -> Maybe String getLiteralString = getLiteralStringExt (const Nothing) +-- Definitely get a literal string, with a given default for all non-literals +getLiteralStringDef :: String -> Token -> String +getLiteralStringDef x = runIdentity . getLiteralStringExt (const $ return x) + -- Definitely get a literal string, skipping over all non-literals onlyLiteralString :: Token -> String -onlyLiteralString = fromJust . getLiteralStringExt (const $ return "") +onlyLiteralString = getLiteralStringDef "" -- Maybe get a literal string, but only if it's an unquoted argument. getUnquotedLiteral (T_NormalWord _ list) = @@ -187,6 +336,12 @@ getUnquotedLiteral (T_NormalWord _ list) = str _ = Nothing getUnquotedLiteral _ = Nothing +isQuotes t = + case t of + T_DoubleQuoted {} -> True + T_SingleQuoted {} -> True + _ -> False + -- Get the last unquoted T_Literal in a word like "${var}foo"THIS -- or nothing if the word does not end in an unquoted literal. getTrailingUnquotedLiteral :: Token -> Maybe Token @@ -205,8 +360,11 @@ getTrailingUnquotedLiteral t = getLeadingUnquotedString :: Token -> Maybe String getLeadingUnquotedString t = case t of - T_NormalWord _ ((T_Literal _ s) : _) -> return s + T_NormalWord _ ((T_Literal _ s) : rest) -> return $ s ++ from rest _ -> Nothing + where + from ((T_Literal _ s):rest) = s ++ from rest + from _ = "" -- Maybe get the literal string of this token and any globs in it. getGlobOrLiteralString = getLiteralStringExt f @@ -214,9 +372,24 @@ getGlobOrLiteralString = getLiteralStringExt f f (T_Glob _ str) = return str f _ = Nothing + +prop_getLiteralString1 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\x01") == Just "\1" +prop_getLiteralString2 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\xyz") == Just "\\xyz" +prop_getLiteralString3 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\x1") == Just "\x1" +prop_getLiteralString4 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\x1y") == Just "\x1y" +prop_getLiteralString5 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\xy") == Just "\\xy" +prop_getLiteralString6 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\x") == Just "\\x" +prop_getLiteralString7 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\1x") == Just "\1x" +prop_getLiteralString8 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\12x") == Just "\o12x" +prop_getLiteralString9 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\123x") == Just "\o123x" +prop_getLiteralString10 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\1234") == Just "\o123\&4" +prop_getLiteralString11 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\1") == Just "\1" +prop_getLiteralString12 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\12") == Just "\o12" +prop_getLiteralString13 = getLiteralString (T_DollarSingleQuoted (Id 0) "\\123") == Just "\o123" + -- Maybe get the literal value of a token, using a custom function -- to map unrecognized Tokens into strings. -getLiteralStringExt :: (Token -> Maybe String) -> Token -> Maybe String +getLiteralStringExt :: Monad m => (Token -> m String) -> Token -> m String getLiteralStringExt more = g where allInList = fmap concat . mapM g @@ -246,14 +419,15 @@ getLiteralStringExt more = g '\\' -> '\\' : rest 'x' -> case cs of - (x:y:more) -> - if isHexDigit x && isHexDigit y - then chr (16*(digitToInt x) + (digitToInt y)) : rest - else '\\':c:rest + (x:y:more) | isHexDigit x && isHexDigit y -> + chr (16*(digitToInt x) + (digitToInt y)) : decodeEscapes more + (x:more) | isHexDigit x -> + chr (digitToInt x) : decodeEscapes more + more -> '\\' : 'x' : decodeEscapes more _ | isOctDigit c -> - let digits = take 3 $ takeWhile isOctDigit (c:cs) - num = parseOct digits - in (if num < 256 then chr num else '?') : rest + let (digits, more) = spanMax isOctDigit 3 (c:cs) + num = (parseOct digits) `mod` 256 + in (chr num) : decodeEscapes more _ -> '\\' : c : rest where rest = decodeEscapes cs @@ -261,12 +435,54 @@ getLiteralStringExt more = g where f n "" = n f n (c:rest) = f (n * 8 + digitToInt c) rest + spanMax f n list = + let (first, second) = span f list + (prefix, suffix) = splitAt n first + in + (prefix, suffix ++ second) decodeEscapes (c:cs) = c : decodeEscapes cs decodeEscapes [] = [] -- Is this token a string literal? isLiteral t = isJust $ getLiteralString t +-- Is this token a string literal number? +isLiteralNumber t = fromMaybe False $ do + s <- getLiteralString t + guard $ all isDigit s + return True + +-- Escape user data for messages. +-- Messages generally avoid repeating user data, but sometimes it's helpful. +e4m = escapeForMessage +escapeForMessage :: String -> String +escapeForMessage str = concatMap f str + where + f '\\' = "\\\\" + f '\n' = "\\n" + f '\r' = "\\r" + f '\t' = "\\t" + f '\x1B' = "\\e" + f c = + if shouldEscape c + then + if ord c < 256 + then "\\x" ++ (pad0 2 $ toHex c) + else "\\U" ++ (pad0 4 $ toHex c) + else [c] + + shouldEscape c = + (not $ isPrint c) + || (not (isAscii c) && not (isLetter c)) + + pad0 :: Int -> String -> String + pad0 n s = + let l = length s in + if l < n + then (replicate (n-l) '0') ++ s + else s + toHex :: Char -> String + toHex c = map toUpper $ showHex (ord c) "" -- Turn a NormalWord like foo="bar $baz" into a series of constituent elements like [foo=,bar ,$baz] getWordParts (T_NormalWord _ l) = concatMap getWordParts l @@ -295,25 +511,48 @@ getCommand t = -- Maybe get the command name string of a token representing a command getCommandName :: Token -> Maybe String -getCommandName = fst . getCommandNameAndToken +getCommandName = fst . getCommandNameAndToken False + +-- Maybe get the name+arguments of a command. +getCommandArgv t = do + (T_SimpleCommand _ _ args@(_:_)) <- getCommand t + return args -- Get the command name token from a command, i.e. -- the token representing 'ls' in 'ls -la 2> foo'. -- If it can't be determined, return the original token. -getCommandTokenOrThis = snd . getCommandNameAndToken +getCommandTokenOrThis = snd . getCommandNameAndToken False -getCommandNameAndToken :: Token -> (Maybe String, Token) -getCommandNameAndToken t = fromMaybe (Nothing, t) $ do - (T_SimpleCommand _ _ (w:rest)) <- getCommand t +-- Given a command, get the string and token that represents the command name. +-- If direct, return the actual command (e.g. exec in 'exec ls') +-- If not, return the logical command (e.g. 'ls' in 'exec ls') + +getCommandNameAndToken :: Bool -> Token -> (Maybe String, Token) +getCommandNameAndToken direct t = fromMaybe (Nothing, t) $ do + cmd@(T_SimpleCommand _ _ (w:rest)) <- getCommand t s <- getLiteralString w - if "busybox" `isSuffixOf` s || "builtin" == s - then - case rest of - (applet:_) -> return (getLiteralString applet, applet) - _ -> return (Just s, w) - else - return (Just s, w) - + return $ fromMaybe (Just s, w) $ do + guard $ not direct + actual <- getEffectiveCommandToken s cmd rest + return (getLiteralString actual, actual) + where + getEffectiveCommandToken str cmd args = + let + firstArg = do + arg <- listToMaybe args + guard . not $ isFlag arg + return arg + in + case str of + "busybox" -> firstArg + "builtin" -> firstArg + "command" -> firstArg + "run" -> firstArg -- Used by bats + "exec" -> do + opts <- getBsdOpts "cla:" args + (_, (t, _)) <- find (null . fst) opts + return t + _ -> fail "" -- If a command substitution is a single command, get its name. -- $(date +%s) = Just "date" @@ -330,8 +569,8 @@ getCommandNameFromExpansion t = -- Get the basename of a token representing a command getCommandBasename = fmap basename . getCommandName - where - basename = reverse . takeWhile (/= '/') . reverse + +basename = reverse . takeWhile (/= '/') . reverse isAssignment t = case t of @@ -351,22 +590,34 @@ isOnlyRedirection t = isFunction t = case t of T_Function {} -> True; _ -> False +-- Bats tests are functions for the purpose of 'local' and such +isFunctionLike t = + case t of + T_Function {} -> True + T_BatsTest {} -> True + _ -> False + + isBraceExpansion t = case t of T_BraceExpansion {} -> True; _ -> False -- Get the lists of commands from tokens that contain them, such as --- the body of while loops or branches of if statements. +-- the conditions and bodies of while loops or branches of if statements. getCommandSequences :: Token -> [[Token]] getCommandSequences t = case t of T_Script _ _ cmds -> [cmds] T_BraceGroup _ cmds -> [cmds] T_Subshell _ cmds -> [cmds] - T_WhileExpression _ _ cmds -> [cmds] - T_UntilExpression _ _ cmds -> [cmds] + T_WhileExpression _ cond cmds -> [cond, cmds] + T_UntilExpression _ cond cmds -> [cond, cmds] T_ForIn _ _ _ cmds -> [cmds] T_ForArithmetic _ _ _ _ cmds -> [cmds] - T_IfExpression _ thens elses -> map snd thens ++ [elses] + T_IfExpression _ thens elses -> (concatMap (\(a,b) -> [a,b]) thens) ++ [elses] T_Annotation _ _ t -> getCommandSequences t + + T_DollarExpansion _ cmds -> [cmds] + T_DollarBraceCommandExpansion _ cmds -> [cmds] + T_Backticked _ cmds -> [cmds] _ -> [] -- Get a list of names of associative arrays @@ -374,13 +625,13 @@ getAssociativeArrays t = nub . execWriter $ doAnalysis f t where f :: Token -> Writer [String] () - f t@T_SimpleCommand {} = fromMaybe (return ()) $ do + f t@T_SimpleCommand {} = sequence_ $ do name <- getCommandName t let assocNames = ["declare","local","typeset"] - guard $ elem name assocNames + guard $ name `elem` assocNames let flags = getAllFlags t - guard $ elem "A" $ map snd flags - let args = map fst . filter ((==) "" . snd) $ flags + guard $ "A" `elem` map snd flags + let args = [arg | (arg, "") <- flags] let names = mapMaybe (getLiteralStringExt nameAssignments) args return $ tell names f _ = return () @@ -398,38 +649,36 @@ data PseudoGlob = PGAny | PGMany | PGChar Char -- Turn a word into a PG pattern, replacing all unknown/runtime values with -- PGMany. -wordToPseudoGlob :: Token -> Maybe [PseudoGlob] -wordToPseudoGlob word = - simplifyPseudoGlob . concat <$> mapM f (getWordParts word) - where - f x = case x of - T_Literal _ s -> return $ map PGChar s - T_SingleQuoted _ s -> return $ map PGChar s - - T_DollarBraced {} -> return [PGMany] - T_DollarExpansion {} -> return [PGMany] - T_Backticked {} -> return [PGMany] - - T_Glob _ "?" -> return [PGAny] - T_Glob _ ('[':_) -> return [PGAny] - T_Glob {} -> return [PGMany] - - T_Extglob {} -> return [PGMany] - - _ -> return [PGMany] +wordToPseudoGlob :: Token -> [PseudoGlob] +wordToPseudoGlob = fromMaybe [PGMany] . wordToPseudoGlob' False -- Turn a word into a PG pattern, but only if we can preserve -- exact semantics. wordToExactPseudoGlob :: Token -> Maybe [PseudoGlob] -wordToExactPseudoGlob word = - simplifyPseudoGlob . concat <$> mapM f (getWordParts word) +wordToExactPseudoGlob = wordToPseudoGlob' True + +wordToPseudoGlob' :: Bool -> Token -> Maybe [PseudoGlob] +wordToPseudoGlob' exact word = + simplifyPseudoGlob <$> toGlob word where + toGlob :: Token -> Maybe [PseudoGlob] + toGlob word = + case word of + T_NormalWord _ (T_Literal _ ('~':str):rest) -> do + guard $ not exact + let this = (PGMany : (map PGChar $ dropWhile (/= '/') str)) + tail <- concat <$> (mapM f $ concatMap getWordParts rest) + return $ this ++ tail + _ -> concat <$> (mapM f $ getWordParts word) + f x = case x of - T_Literal _ s -> return $ map PGChar s + T_Literal _ s -> return $ map PGChar s T_SingleQuoted _ s -> return $ map PGChar s - T_Glob _ "?" -> return [PGAny] - T_Glob _ "*" -> return [PGMany] - _ -> fail "Unknown token type" + T_Glob _ "?" -> return [PGAny] + T_Glob _ "*" -> return [PGMany] + T_Glob _ ('[':_) | not exact -> return [PGAny] + _ -> if exact then fail "" else return [PGMany] + -- Reorder a PseudoGlob for more efficient matching, e.g. -- f?*?**g -> f??*g @@ -479,14 +728,199 @@ pseudoGlobIsSuperSetof = matchable matchable (PGMany : rest) [] = matchable rest [] matchable _ _ = False -wordsCanBeEqual x y = fromMaybe True $ - liftM2 pseudoGlobsCanOverlap (wordToPseudoGlob x) (wordToPseudoGlob y) +wordsCanBeEqual x y = pseudoGlobsCanOverlap (wordToPseudoGlob x) (wordToPseudoGlob y) -- Is this an expansion that can be quoted, -- e.g. $(foo) `foo` $foo (but not {foo,})? isQuoteableExpansion t = case t of + T_DollarBraced {} -> True + _ -> isCommandSubstitution t + +isCommandSubstitution t = case t of T_DollarExpansion {} -> True T_DollarBraceCommandExpansion {} -> True T_Backticked {} -> True - T_DollarBraced {} -> True _ -> False + +-- Is this an expansion that results in a simple string? +isStringExpansion t = isCommandSubstitution t || case t of + T_DollarArithmetic {} -> True + T_DollarBraced {} -> not (isArrayExpansion t) + _ -> False + +-- Is this a T_Annotation that ignores a specific code? +isAnnotationIgnoringCode code t = + case t of + T_Annotation _ anns _ -> any hasNum anns + _ -> False + where + hasNum (DisableComment from to) = code >= from && code < to + hasNum _ = False + +prop_executableFromShebang1 = executableFromShebang "/bin/sh" == "sh" +prop_executableFromShebang2 = executableFromShebang "/bin/bash" == "bash" +prop_executableFromShebang3 = executableFromShebang "/usr/bin/env ksh" == "ksh" +prop_executableFromShebang4 = executableFromShebang "/usr/bin/env -S foo=bar bash -x" == "bash" +prop_executableFromShebang5 = executableFromShebang "/usr/bin/env --split-string=bash -x" == "bash" +prop_executableFromShebang6 = executableFromShebang "/usr/bin/env --split-string=foo=bar bash -x" == "bash" +prop_executableFromShebang7 = executableFromShebang "/usr/bin/env --split-string bash -x" == "bash" +prop_executableFromShebang8 = executableFromShebang "/usr/bin/env --split-string foo=bar bash -x" == "bash" +prop_executableFromShebang9 = executableFromShebang "/usr/bin/env foo=bar dash" == "dash" +prop_executableFromShebang10 = executableFromShebang "/bin/busybox sh" == "busybox sh" +prop_executableFromShebang11 = executableFromShebang "/bin/busybox ash" == "busybox ash" + +-- Get the shell executable from a string like '/usr/bin/env bash' +executableFromShebang :: String -> String +executableFromShebang = shellFor + where + re = mkRegex "/env +(-S|--split-string=?)? *(.*)" + shellFor s | s `matches` re = + case matchRegex re s of + Just [flag, shell] -> fromEnvArgs (words shell) + _ -> "" + shellFor sb = + case words sb of + [] -> "" + [x] -> basename x + (first:second:args) | basename first == "busybox" -> + case basename second of + "sh" -> "busybox sh" + "ash" -> "busybox ash" + x -> x + (first:args) | basename first == "env" -> + fromEnvArgs args + (first:_) -> basename first + + fromEnvArgs args = fromMaybe "" $ find (notElem '=') $ skipFlags args + basename s = reverse . takeWhile (/= '/') . reverse $ s + skipFlags = dropWhile ("-" `isPrefixOf`) + + +-- Determining if a name is a variable +isVariableStartChar x = x == '_' || isAsciiLower x || isAsciiUpper x +isVariableChar x = isVariableStartChar x || isDigit x +isSpecialVariableChar = (`elem` "*@#?-$!") +variableNameRegex = mkRegex "[_a-zA-Z][_a-zA-Z0-9]*" + +prop_isVariableName1 = isVariableName "_fo123" +prop_isVariableName2 = not $ isVariableName "4" +prop_isVariableName3 = not $ isVariableName "test: " +isVariableName (x:r) = isVariableStartChar x && all isVariableChar r +isVariableName _ = False + + +-- Get the variable name from an expansion like ${var:-foo} +prop_getBracedReference1 = getBracedReference "foo" == "foo" +prop_getBracedReference2 = getBracedReference "#foo" == "foo" +prop_getBracedReference3 = getBracedReference "#" == "#" +prop_getBracedReference4 = getBracedReference "##" == "#" +prop_getBracedReference5 = getBracedReference "#!" == "!" +prop_getBracedReference6 = getBracedReference "!#" == "#" +prop_getBracedReference7 = getBracedReference "!foo#?" == "foo" +prop_getBracedReference8 = getBracedReference "foo-bar" == "foo" +prop_getBracedReference9 = getBracedReference "foo:-bar" == "foo" +prop_getBracedReference10 = getBracedReference "foo: -1" == "foo" +prop_getBracedReference11 = getBracedReference "!os*" == "" +prop_getBracedReference11b = getBracedReference "!os@" == "" +prop_getBracedReference12 = getBracedReference "!os?bar**" == "" +prop_getBracedReference13 = getBracedReference "foo[bar]" == "foo" +getBracedReference s = fromMaybe s $ + nameExpansion s `mplus` takeName noPrefix `mplus` getSpecial noPrefix `mplus` getSpecial s + where + noPrefix = dropPrefix s + dropPrefix (c:rest) | c `elem` "!#" = rest + dropPrefix cs = cs + takeName s = do + let name = takeWhile isVariableChar s + guard . not $ null name + return name + getSpecial (c:_) | isSpecialVariableChar c = return [c] + getSpecial _ = fail "empty or not special" + + nameExpansion ('!':next:rest) = do -- e.g. ${!foo*bar*} + guard $ isVariableChar next -- e.g. ${!@} + first <- find (not . isVariableChar) rest + guard $ first `elem` "*?@" + return "" + nameExpansion _ = Nothing + +-- Get the variable modifier like /a/b in ${var/a/b} +prop_getBracedModifier1 = getBracedModifier "foo:bar:baz" == ":bar:baz" +prop_getBracedModifier2 = getBracedModifier "!var:-foo" == ":-foo" +prop_getBracedModifier3 = getBracedModifier "foo[bar]" == "[bar]" +prop_getBracedModifier4 = getBracedModifier "foo[@]@Q" == "[@]@Q" +prop_getBracedModifier5 = getBracedModifier "@@Q" == "@Q" +getBracedModifier s = headOrDefault "" $ do + let var = getBracedReference s + a <- dropModifier s + dropPrefix var a + where + dropPrefix [] t = return t + dropPrefix (a:b) (c:d) | a == c = dropPrefix b d + dropPrefix _ _ = [] + + dropModifier (c:rest) | c `elem` "#!" = [rest, c:rest] + dropModifier x = [x] + +-- Get the variables from indices like ["x", "y"] in ${var[x+y+1]} +prop_getIndexReferences1 = getIndexReferences "var[x+y+1]" == ["x", "y"] +getIndexReferences s = fromMaybe [] $ do + index:_ <- matchRegex re s + return $ matchAllStrings variableNameRegex index + where + re = mkRegex "(\\[.*\\])" + +prop_getOffsetReferences1 = getOffsetReferences ":bar" == ["bar"] +prop_getOffsetReferences2 = getOffsetReferences ":bar:baz" == ["bar", "baz"] +prop_getOffsetReferences3 = getOffsetReferences "[foo]:bar" == ["bar"] +prop_getOffsetReferences4 = getOffsetReferences "[foo]:bar:baz" == ["bar", "baz"] +getOffsetReferences mods = fromMaybe [] $ do +-- if mods start with [, then drop until ] + _:offsets:_ <- matchRegex re mods + return $ matchAllStrings variableNameRegex offsets + where + re = mkRegex "^(\\[.+\\])? *:([^-=?+].*)" + + +-- Returns whether a token is a parameter expansion without any modifiers. +-- True for $var ${var} $1 $# +-- False for ${#var} ${var[x]} ${var:-0} +isUnmodifiedParameterExpansion t = + case t of + T_DollarBraced _ False _ -> True + T_DollarBraced _ _ list -> + let str = concat $ oversimplify list + in getBracedReference str == str + _ -> False + +-- Return the referenced variable if (and only if) it's an unmodified parameter expansion. +getUnmodifiedParameterExpansion t = + case t of + T_DollarBraced _ _ list -> do + let str = concat $ oversimplify list + guard $ getBracedReference str == str + return str + _ -> Nothing + +--- A list of the element and all its parents up to the root node. +getPath tree = NE.unfoldr $ \t -> (t, Map.lookup (getId t) tree) + +isClosingFileOp op = + case op of + T_IoDuplicate _ (T_GREATAND _) "-" -> True + T_IoDuplicate _ (T_LESSAND _) "-" -> True + _ -> False + +getEnableDirectives root = + case root of + T_Annotation _ list _ -> [s | EnableComment s <- list] + _ -> [] + +getExtendedAnalysisDirective :: Token -> Maybe Bool +getExtendedAnalysisDirective root = + case root of + T_Annotation _ list _ -> listToMaybe $ [s | ExtendedAnalysis s <- list] + _ -> Nothing + +return [] +runTests = $quickCheckAll diff --git a/src/ShellCheck/Analyzer.hs b/src/ShellCheck/Analyzer.hs index ffbc4e5..53717ed 100644 --- a/src/ShellCheck/Analyzer.hs +++ b/src/ShellCheck/Analyzer.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2022 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -17,7 +17,7 @@ You should have received a copy of the GNU General Public License along with this program. If not, see . -} -module ShellCheck.Analyzer (analyzeScript) where +module ShellCheck.Analyzer (analyzeScript, ShellCheck.Analyzer.optionalChecks) where import ShellCheck.Analytics import ShellCheck.AnalyzerLib @@ -25,6 +25,8 @@ import ShellCheck.Interface import Data.List import Data.Monoid import qualified ShellCheck.Checks.Commands +import qualified ShellCheck.Checks.ControlFlow +import qualified ShellCheck.Checks.Custom import qualified ShellCheck.Checks.ShellSupport @@ -33,13 +35,21 @@ analyzeScript :: AnalysisSpec -> AnalysisResult analyzeScript spec = newAnalysisResult { arComments = filterByAnnotation spec params . nub $ - runAnalytics spec - ++ runChecker params (checkers params) + runChecker params (checkers spec params) } where params = makeParameters spec -checkers params = mconcat $ map ($ params) [ - ShellCheck.Checks.Commands.checker, +checkers spec params = mconcat $ map ($ params) [ + ShellCheck.Analytics.checker spec, + ShellCheck.Checks.Commands.checker spec, + ShellCheck.Checks.ControlFlow.checker spec, + ShellCheck.Checks.Custom.checker, ShellCheck.Checks.ShellSupport.checker ] + +optionalChecks = mconcat $ [ + ShellCheck.Analytics.optionalChecks, + ShellCheck.Checks.Commands.optionalChecks, + ShellCheck.Checks.ControlFlow.optionalChecks + ] diff --git a/src/ShellCheck/CFG.hs b/src/ShellCheck/CFG.hs new file mode 100644 index 0000000..57aaf4b --- /dev/null +++ b/src/ShellCheck/CFG.hs @@ -0,0 +1,1316 @@ +{- + Copyright 2022 Vidar Holen + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{-# LANGUAGE TemplateHaskell #-} +{-# LANGUAGE DeriveAnyClass, DeriveGeneric #-} + +-- Constructs a Control Flow Graph from an AST +module ShellCheck.CFG ( + CFNode (..), + CFEdge (..), + CFEffect (..), + CFStringPart (..), + CFVariableProp (..), + CFGResult (..), + CFValue (..), + CFGraph, + CFGParameters (..), + IdTagged (..), + Scope (..), + buildGraph + , ShellCheck.CFG.runTests -- STRIP + ) + where + +import GHC.Generics (Generic) +import ShellCheck.AST +import ShellCheck.ASTLib +import ShellCheck.Data +import ShellCheck.Interface +import ShellCheck.Prelude +import ShellCheck.Regex +import Control.DeepSeq +import Control.Monad +import Control.Monad.Identity +import Data.Array.Unboxed +import Data.Array.ST +import Data.List hiding (map) +import qualified Data.List.NonEmpty as NE +import Data.Maybe +import qualified Data.Map as M +import qualified Data.Set as S +import Control.Monad.RWS.Lazy +import Data.Graph.Inductive.Graph +import Data.Graph.Inductive.Query.DFS +import Data.Graph.Inductive.Basic +import Data.Graph.Inductive.Query.Dominators +import Data.Graph.Inductive.PatriciaTree as G +import Debug.Trace -- STRIP + +import Test.QuickCheck.All (forAllProperties) +import Test.QuickCheck.Test (quickCheckWithResult, stdArgs, maxSuccess) + + +-- Our basic Graph type +type CFGraph = G.Gr CFNode CFEdge + +-- Node labels in a Control Flow Graph +data CFNode = + -- A no-op node for structural purposes + CFStructuralNode + -- A no-op for graph inspection purposes + | CFEntryPoint String + -- Drop current prefix assignments + | CFDropPrefixAssignments + -- A node with a certain effect on program state + | CFApplyEffects [IdTagged CFEffect] + -- The execution of a command or function by literal string if possible + | CFExecuteCommand (Maybe String) + -- Execute a subshell. These are represented by disjoint graphs just like + -- functions, but they don't require any form of name resolution + | CFExecuteSubshell String Node Node + -- Assignment of $? + | CFSetExitCode Id + -- The virtual 'exit' at the natural end of a subshell + | CFImpliedExit + -- An exit statement resolvable at CFG build time + | CFResolvedExit + -- An exit statement only resolvable at DFA time + | CFUnresolvedExit + -- An unreachable node, serving as the unconnected end point of a range + | CFUnreachable + -- Assignment of $! + | CFSetBackgroundPid Id + deriving (Eq, Ord, Show, Generic, NFData) + +-- Edge labels in a Control Flow Graph +data CFEdge = + CFEErrExit + -- Regular control flow edge + | CFEFlow + -- An edge that a human might think exists (e.g. from a backgrounded process to its parent) + | CFEFalseFlow + -- An edge followed on exit + | CFEExit + deriving (Eq, Ord, Show, Generic, NFData) + +-- Actions we track +data CFEffect = + CFSetProps (Maybe Scope) String (S.Set CFVariableProp) + | CFUnsetProps (Maybe Scope) String (S.Set CFVariableProp) + | CFReadVariable String + | CFWriteVariable String CFValue + | CFWriteGlobal String CFValue + | CFWriteLocal String CFValue + | CFWritePrefix String CFValue + | CFDefineFunction String Id Node Node + | CFUndefine String + | CFUndefineVariable String + | CFUndefineFunction String + | CFUndefineNameref String + -- Usage implies that this is an array (e.g. it's expanded with index) + | CFHintArray String + -- Operation implies that the variable will be defined (e.g. [ -z "$var" ]) + | CFHintDefined String + deriving (Eq, Ord, Show, Generic, NFData) + +data IdTagged a = IdTagged Id a + deriving (Eq, Ord, Show, Generic, NFData) + +-- Where a variable's value comes from +data CFValue = + -- The special 'uninitialized' value + CFValueUninitialized + -- An arbitrary array value + | CFValueArray + -- An arbitrary string value + | CFValueString + -- An arbitrary integer + | CFValueInteger + -- Token 'Id' concatenates and assigns the given parts + | CFValueComputed Id [CFStringPart] + deriving (Eq, Ord, Show, Generic, NFData) + +-- Simplified computed strings +data CFStringPart = + -- A known literal string value, like 'foo' + CFStringLiteral String + -- The contents of a variable, like $foo (may not be a string) + | CFStringVariable String + -- An value that is unknown but an integer + | CFStringInteger + -- An unknown string value, for things we can't handle + | CFStringUnknown + deriving (Eq, Ord, Show, Generic, NFData) + +-- The properties of a variable +data CFVariableProp = CFVPExport | CFVPArray | CFVPAssociative | CFVPInteger + deriving (Eq, Ord, Show, Generic, NFData) + +-- Options when generating CFG +data CFGParameters = CFGParameters { + -- Whether the last element in a pipeline runs in the current shell + cfLastpipe :: Bool, + -- Whether all elements in a pipeline count towards the exit status + cfPipefail :: Bool +} + +data CFGResult = CFGResult { + -- The graph itself + cfGraph :: CFGraph, + -- Map from Id to nominal start&end node (i.e. assuming normal execution without exits) + cfIdToRange :: M.Map Id (Node, Node), + -- A set of all nodes belonging to an Id, recursively + cfIdToNodes :: M.Map Id (S.Set Node), + -- An array (from,to) saying whether 'from' postdominates 'to' + cfPostDominators :: Array Node [Node] +} + deriving (Show) + +buildGraph :: CFGParameters -> Token -> CFGResult +buildGraph params root = + let + (nextNode, base) = execRWS (buildRoot root) (newCFContext params) 0 + (nodes, edges, mapping, association) = +-- renumberTopologically $ + removeUnnecessaryStructuralNodes + base + + idToRange = M.fromList mapping + isRealEdge (from, to, edge) = case edge of CFEFlow -> True; CFEExit -> True; _ -> False + onlyRealEdges = filter isRealEdge edges + (_, mainExit) = fromJust $ M.lookup (getId root) idToRange + + result = CFGResult { + cfGraph = mkGraph nodes edges, + cfIdToRange = idToRange, + cfIdToNodes = M.fromListWith S.union $ map (\(id, n) -> (id, S.singleton n)) association, + cfPostDominators = findPostDominators mainExit $ mkGraph nodes onlyRealEdges + } + in + result + +remapGraph :: M.Map Node Node -> CFW -> CFW +remapGraph remap (nodes, edges, mapping, assoc) = + ( + map (remapNode remap) nodes, + map (remapEdge remap) edges, + map (\(id, (a,b)) -> (id, (remapHelper remap a, remapHelper remap b))) mapping, + map (\(id, n) -> (id, remapHelper remap n)) assoc + ) + +prop_testRenumbering = + let + s = CFStructuralNode + before = ( + [(1,s), (3,s), (4, s), (8,s)], + [(1,3,CFEFlow), (3,4, CFEFlow), (4,8,CFEFlow)], + [(Id 0, (3,4))], + [(Id 1, 3), (Id 2, 4)] + ) + after = ( + [(0,s), (1,s), (2,s), (3,s)], + [(0,1,CFEFlow), (1,2, CFEFlow), (2,3,CFEFlow)], + [(Id 0, (1,2))], + [(Id 1, 1), (Id 2, 2)] + ) + in after == renumberGraph before + +-- Renumber the graph for prettiness, so there are no gaps in node numbers +renumberGraph :: CFW -> CFW +renumberGraph g@(nodes, edges, mapping, assoc) = + let renumbering = M.fromList (flip zip [0..] $ sort $ map fst nodes) + in remapGraph renumbering g + +prop_testRenumberTopologically = + let + s = CFStructuralNode + before = ( + [(4,s), (2,s), (3, s)], + [(4,2,CFEFlow), (2,3, CFEFlow)], + [(Id 0, (4,2))], + [] + ) + after = ( + [(0,s), (1,s), (2,s)], + [(0,1,CFEFlow), (1,2, CFEFlow)], + [(Id 0, (0,1))], + [] + ) + in after == renumberTopologically before + +-- Renumber the graph in topological order +renumberTopologically g@(nodes, edges, mapping, assoc) = + let renumbering = M.fromList (flip zip [0..] $ topsort (mkGraph nodes edges :: CFGraph)) + in remapGraph renumbering g + +prop_testRemoveStructural = + let + s = CFStructuralNode + before = ( + [(1,s), (2,s), (3, s), (4,s)], + [(1,2,CFEFlow), (2,3, CFEFlow), (3,4,CFEFlow)], + [(Id 0, (2,3))], + [(Id 0, 3)] + ) + after = ( + [(1,s), (2,s), (4,s)], + [(1,2,CFEFlow), (2,4,CFEFlow)], + [(Id 0, (2,2))], + [(Id 0, 2)] + ) + in after == removeUnnecessaryStructuralNodes before + +-- Collapse structural nodes that just form long chains like x->x->x. +-- This way we can generate them with abandon, without making DFA slower. +-- +-- Note in particular that we can't remove a structural node x in +-- foo -> x -> bar , because then the pre/post-condition for tokens +-- previously pointing to x would be wrong. +removeUnnecessaryStructuralNodes (nodes, edges, mapping, association) = + remapGraph recursiveRemapping + ( + filter (\(n, _) -> n `M.notMember` recursiveRemapping) nodes, + filter (`S.notMember` edgesToCollapse) edges, + mapping, + association + ) + where + regularEdges = filter isRegularEdge edges + inDegree = counter $ map (\(from,to,_) -> from) regularEdges + outDegree = counter $ map (\(from,to,_) -> to) regularEdges + structuralNodes = S.fromList [node | (node, CFStructuralNode) <- nodes] + candidateNodes = S.filter isLinear structuralNodes + edgesToCollapse = S.fromList $ filter filterEdges regularEdges + + remapping :: M.Map Node Node + remapping = M.fromList $ map orderEdge $ S.toList edgesToCollapse + recursiveRemapping = M.mapWithKey (\c _ -> recursiveLookup remapping c) remapping + + filterEdges (a,b,_) = + a `S.member` candidateNodes && b `S.member` candidateNodes + + orderEdge (a,b,_) = if a < b then (b,a) else (a,b) + counter = M.fromListWith (+) . map (\key -> (key, 1)) + isRegularEdge (_, _, CFEFlow) = True + isRegularEdge _ = False + + recursiveLookup :: M.Map Node Node -> Node -> Node + recursiveLookup map node = + case M.lookup node map of + Nothing -> node + Just x -> recursiveLookup map x + + isLinear node = + M.findWithDefault 0 node inDegree == 1 + && M.findWithDefault 0 node outDegree == 1 + + +remapNode :: M.Map Node Node -> LNode CFNode -> LNode CFNode +remapNode m (node, label) = + (remapHelper m node, newLabel) + where + newLabel = case label of + CFApplyEffects effects -> CFApplyEffects (map (remapEffect m) effects) + CFExecuteSubshell s a b -> CFExecuteSubshell s (remapHelper m a) (remapHelper m b) + _ -> label + +remapEffect map old@(IdTagged id effect) = + case effect of + CFDefineFunction name id start end -> IdTagged id $ CFDefineFunction name id (remapHelper map start) (remapHelper map end) + _ -> old + +remapEdge :: M.Map Node Node -> LEdge CFEdge -> LEdge CFEdge +remapEdge map (from, to, label) = (remapHelper map from, remapHelper map to, label) +remapHelper map n = M.findWithDefault n n map + +data Range = Range Node Node + deriving (Eq, Show) + +data CFContext = CFContext { + cfIsCondition :: Bool, + cfIsFunction :: Bool, + cfLoopStack :: [(Node, Node)], + cfTokenStack :: [Id], + cfExitTarget :: Maybe Node, + cfReturnTarget :: Maybe Node, + cfParameters :: CFGParameters +} +newCFContext params = CFContext { + cfIsCondition = False, + cfIsFunction = False, + cfLoopStack = [], + cfTokenStack = [], + cfExitTarget = Nothing, + cfReturnTarget = Nothing, + cfParameters = params +} + +-- The monad we generate a graph in +type CFM a = RWS CFContext CFW Int a +type CFW = ([LNode CFNode], [LEdge CFEdge], [(Id, (Node, Node))], [(Id, Node)]) + +newNode :: CFNode -> CFM Node +newNode label = do + n <- get + stack <- asks cfTokenStack + put (n+1) + tell ([(n, label)], [], [], map (\c -> (c, n)) stack) + return n + +newNodeRange :: CFNode -> CFM Range +-- newNodeRange label = nodeToRange <$> newNode label +newNodeRange label = nodeToRange <$> newNode label + +-- Build a disjoint piece of the graph and return a CFExecuteSubshell. The Id is used purely for debug naming. +subshell :: Id -> String -> CFM Range -> CFM Range +subshell id reason p = do + start <- newNode $ CFEntryPoint $ "Subshell " ++ show id ++ ": " ++ reason + end <- newNode CFStructuralNode + middle <- local (\c -> c { cfExitTarget = Just end, cfReturnTarget = Just end}) p + linkRanges [nodeToRange start, middle, nodeToRange end] + newNodeRange $ CFExecuteSubshell reason start end + + +withFunctionScope p = do + end <- newNode CFStructuralNode + body <- local (\c -> c { cfReturnTarget = Just end, cfIsFunction = True }) p + linkRanges [body, nodeToRange end] + +-- Anything that happens recursively in f will be attributed to this id +under :: Id -> CFM a -> CFM a +under id f = local (\c -> c { cfTokenStack = id:(cfTokenStack c) }) f + +nodeToRange :: Node -> Range +nodeToRange n = Range n n + +link :: Node -> Node -> CFEdge -> CFM () +link from to label = do + tell ([], [(from, to, label)], [], []) + +registerNode :: Id -> Range -> CFM () +registerNode id (Range start end) = tell ([], [], [(id, (start, end))], []) + +linkRange :: Range -> Range -> CFM Range +linkRange = linkRangeAs CFEFlow + +linkRangeAs :: CFEdge -> Range -> Range -> CFM Range +linkRangeAs label (Range start mid1) (Range mid2 end) = do + link mid1 mid2 label + return (Range start end) + +-- Like linkRange but without actually linking +spanRange :: Range -> Range -> Range +spanRange (Range start mid1) (Range mid2 end) = Range start end + +linkRanges :: [Range] -> CFM Range +linkRanges [] = error "Empty range" +linkRanges (first:rest) = foldM linkRange first rest + +sequentially :: [Token] -> CFM Range +sequentially list = do + first <- newStructuralNode + rest <- mapM build list + linkRanges (first:rest) + +withContext :: (CFContext -> CFContext) -> CFM a -> CFM a +withContext = local + +withReturn :: Range -> CFM a -> CFM a +withReturn _ p = p + +asCondition :: CFM Range -> CFM Range +asCondition = withContext (\c -> c { cfIsCondition = True }) + +newStructuralNode = newNodeRange CFStructuralNode + +buildRoot :: Token -> CFM Range +buildRoot t = under (getId t) $ do + entry <- newNodeRange $ CFEntryPoint "MAIN" + impliedExit <- newNode CFImpliedExit + end <- newNode CFStructuralNode + start <- local (\c -> c { cfExitTarget = Just end, cfReturnTarget = Just impliedExit}) $ build t + range <- linkRanges [entry, start, nodeToRange impliedExit, nodeToRange end] + registerNode (getId t) range + return range + +applySingle e = CFApplyEffects [e] + +-- Build the CFG. +build :: Token -> CFM Range +build t = do + range <- under (getId t) $ build' t + registerNode (getId t) range + return range + where + build' t = case t of + T_Annotation _ _ list -> build list + T_Script _ _ list -> do + sequentially list + + TA_Assignment id op var@(TA_Variable _ name indices) rhs -> do + -- value first: (( var[x=1] = (x=2) )) runs x=1 last + value <- build rhs + subscript <- sequentially indices + read <- + if op == "=" + then none + -- This is += or something + else newNodeRange $ applySingle $ IdTagged id $ CFReadVariable name + + write <- newNodeRange $ applySingle $ IdTagged id $ CFWriteVariable name $ + if null indices + then CFValueInteger + else CFValueArray + + linkRanges [value, subscript, read, write] + + TA_Assignment id op lhs rhs -> do + -- This is likely an invalid assignment like (( 1 = 2 )), but it + -- could be e.g. x=y; (( $x = 3 )); echo $y, so expand both sides + -- without updating anything + sequentially [lhs, rhs] + + TA_Binary _ _ a b -> sequentially [a,b] + TA_Expansion _ list -> sequentially list + TA_Sequence _ list -> sequentially list + TA_Parenthesis _ t -> build t + + TA_Trinary _ cond a b -> do + condition <- build cond + ifthen <- build a + elsethen <- build b + end <- newStructuralNode + linkRanges [condition, ifthen, end] + linkRanges [condition, elsethen, end] + + TA_Variable id name indices -> do + subscript <- sequentially indices + hint <- + if null indices + then none + else nodeToRange <$> newNode (applySingle $ IdTagged id $ CFHintArray name) + read <- nodeToRange <$> newNode (applySingle $ IdTagged id $ CFReadVariable name) + linkRanges [subscript, hint, read] + + TA_Unary id op (TA_Variable _ name indices) | "--" `isInfixOf` op || "++" `isInfixOf` op -> do + subscript <- sequentially indices + read <- newNodeRange $ applySingle $ IdTagged id $ CFReadVariable name + write <- newNodeRange $ applySingle $ IdTagged id $ CFWriteVariable name $ + if null indices + then CFValueInteger + else CFValueArray + linkRanges [subscript, read, write] + TA_Unary _ _ arg -> build arg + + TC_And _ SingleBracket _ lhs rhs -> do + sequentially [lhs, rhs] + + TC_And _ DoubleBracket _ lhs rhs -> do + left <- build lhs + right <- build rhs + end <- newStructuralNode + -- complete + linkRanges [left, right, end] + -- short circuit + linkRange left end + + -- TODO: Handle integer ops + TC_Binary _ mode str lhs rhs -> do + left <- build lhs + right <- build rhs + linkRange left right + + TC_Empty {} -> newStructuralNode + + TC_Group _ _ t -> build t + + -- TODO: Mark as checked + TC_Nullary _ _ arg -> build arg + + TC_Or _ SingleBracket _ lhs rhs -> sequentially [lhs, rhs] + + TC_Or _ DoubleBracket _ lhs rhs -> do + left <- build lhs + right <- build rhs + end <- newStructuralNode + -- complete + linkRanges [left, right, end] + -- short circuit + linkRange left end + + -- TODO: Handle -v, -z, -n + TC_Unary _ _ op arg -> do + build arg + + T_Arithmetic id root -> do + exe <- build root + status <- newNodeRange (CFSetExitCode id) + linkRange exe status + + T_AndIf _ lhs rhs -> do + left <- build lhs + right <- build rhs + end <- newStructuralNode + linkRange left right + linkRange right end + linkRange left end + + T_Array _ list -> sequentially list + + T_Assignment {} -> buildAssignment Nothing t + + T_Backgrounded id body -> do + start <- newStructuralNode + fork <- subshell id "backgrounding '&'" $ build body + pid <- newNodeRange $ CFSetBackgroundPid id + status <- newNodeRange $ CFSetExitCode id + + linkRange start fork + -- Add a join from the fork to warn about variable changes + linkRangeAs CFEFalseFlow fork pid + linkRanges [start, pid, status] + + T_Backticked id body -> + subshell id "`..` expansion" $ sequentially body + + T_Banged id cmd -> do + main <- build cmd + status <- newNodeRange (CFSetExitCode id) + linkRange main status + + T_BatsTest id _ body -> do + -- These are technically set by the 'run' command, but we'll just define them + -- up front to avoid figuring out which commands named "run" belong to Bats. + status <- newNodeRange $ applySingle $ IdTagged id $ CFWriteVariable "status" CFValueInteger + output <- newNodeRange $ applySingle $ IdTagged id $ CFWriteVariable "output" CFValueString + main <- build body + linkRanges [status, output, main] + + T_BraceExpansion _ list -> sequentially list + + T_BraceGroup id body -> + sequentially body + + T_CaseExpression id t [] -> build t + + T_CaseExpression id t list@(hd:tl) -> do + start <- newStructuralNode + token <- build t + branches <- mapM buildBranch (hd NE.:| tl) + end <- newStructuralNode + + let neighbors = zip (NE.toList branches) $ NE.tail branches + let (_, firstCond, _) = NE.head branches + let (_, lastCond, lastBody) = NE.last branches + + linkRange start token + linkRange token firstCond + mapM_ (uncurry $ linkBranch end) neighbors + linkRange lastBody end + + unless (any hasCatchAll list) $ + -- There's no *) branch, so assume we can fall through + void $ linkRange token end + + return $ spanRange start end + + where + -- for a | b | c, evaluate each in turn and allow short circuiting + buildCond list = do + start <- newStructuralNode + conds <- mapM build list + end <- newStructuralNode + linkRanges (start:conds) + mapM_ (`linkRange` end) conds + return $ spanRange start end + + buildBranch (typ, cond, body) = do + c <- buildCond cond + b <- sequentially body + linkRange c b + return (typ, c, b) + + linkBranch end (typ, cond, body) (_, nextCond, nextBody) = do + -- Failure case + linkRange cond nextCond + -- After body + case typ of + CaseBreak -> linkRange body end + CaseFallThrough -> linkRange body nextBody + CaseContinue -> linkRange body nextCond + + -- Find a *) if any + + hasCatchAll (_,cond,_) = any isCatchAll cond + isCatchAll c = fromMaybe False $ do + pg <- wordToExactPseudoGlob c + return $ pg `pseudoGlobIsSuperSetof` [PGMany] + + T_Condition id _ op -> do + cond <- build op + status <- newNodeRange $ CFSetExitCode id + linkRange cond status + + T_CoProc id maybeNameToken t -> do + -- If unspecified, "COPROC". If not a constant string, Nothing. + let maybeName = case maybeNameToken of + Just x -> getLiteralString x + Nothing -> Just "COPROC" + + let parentNode = case maybeName of + Just str -> applySingle $ IdTagged id $ CFWriteVariable str CFValueArray + Nothing -> CFStructuralNode + + start <- newStructuralNode + parent <- newNodeRange parentNode + child <- subshell id "coproc" $ build t + end <- newNodeRange $ CFSetExitCode id + + linkRange start parent + linkRange start child + linkRange parent end + linkRangeAs CFEFalseFlow child end + + return $ spanRange start end + T_CoProcBody _ t -> build t + + T_DollarArithmetic _ arith -> build arith + T_DollarDoubleQuoted _ list -> sequentially list + T_DollarSingleQuoted _ _ -> none + T_DollarBracket _ t -> build t + + T_DollarBraced id _ t -> do + let str = concat $ oversimplify t + let modifier = getBracedModifier str + let reference = getBracedReference str + let indices = getIndexReferences str + let offsets = getOffsetReferences str + vals <- build t + others <- mapM (\x -> nodeToRange <$> newNode (applySingle $ IdTagged id $ CFReadVariable x)) (indices ++ offsets) + deps <- linkRanges (vals:others) + read <- nodeToRange <$> newNode (applySingle $ IdTagged id $ CFReadVariable reference) + totalRead <- linkRange deps read + + if any (`isPrefixOf` modifier) ["=", ":="] + then do + optionalAssign <- newNodeRange (applySingle $ IdTagged id $ CFWriteVariable reference CFValueString) + result <- newStructuralNode + linkRange optionalAssign result + linkRange totalRead result + else return totalRead + + T_DoubleQuoted _ list -> sequentially list + + T_DollarExpansion id body -> + subshell id "$(..) expansion" $ sequentially body + + T_Extglob _ _ list -> sequentially list + + T_FdRedirect id ('{':identifier) op -> do + let name = takeWhile (/= '}') identifier + expression <- build op + rw <- newNodeRange $ + if isClosingFileOp op + then applySingle $ IdTagged id $ CFReadVariable name + else applySingle $ IdTagged id $ CFWriteVariable name CFValueInteger + + linkRange expression rw + + + T_FdRedirect _ name t -> do + build t + + T_ForArithmetic _ initT condT incT bodyT -> do + init <- build initT + cond <- build condT + body <- sequentially bodyT + inc <- build incT + end <- newStructuralNode + + -- Forward edges + linkRanges [init, cond, body, inc] + linkRange cond end + -- Backward edge + linkRange inc cond + return $ spanRange init end + + T_ForIn id name words body -> forInHelper id name words body + + -- For functions we generate an unlinked subgraph, and mention that in its definition node + T_Function id _ _ name body -> do + range <- local (\c -> c { cfExitTarget = Nothing }) $ do + entry <- newNodeRange $ CFEntryPoint $ "function " ++ name + f <- withFunctionScope $ build body + linkRange entry f + let (Range entry exit) = range + definition <- newNodeRange (applySingle $ IdTagged id $ CFDefineFunction name id entry exit) + exe <- newNodeRange (CFSetExitCode id) + linkRange definition exe + + T_Glob {} -> none + + T_HereString _ t -> build t + T_HereDoc _ _ _ _ list -> sequentially list + + T_IfExpression id ifs elses -> do + start <- newStructuralNode + branches <- doBranches start ifs elses [] + end <- newStructuralNode + mapM_ (`linkRange` end) branches + return $ spanRange start end + where + doBranches start ((conds, thens):rest) elses result = do + cond <- asCondition $ sequentially conds + action <- sequentially thens + linkRange start cond + linkRange cond action + doBranches cond rest elses (action:result) + doBranches start [] elses result = do + rest <- + if null elses + then newNodeRange (CFSetExitCode id) + else sequentially elses + linkRange start rest + return (rest:result) + + T_Include _ t -> build t + + T_IndexedElement _ indicesT valueT -> do + indices <- sequentially indicesT + value <- build valueT + linkRange indices value + + T_IoDuplicate _ op _ -> build op + + T_IoFile _ op t -> do + exp <- build t + doesntDoMuch <- build op + linkRange exp doesntDoMuch + + T_Literal {} -> none + + T_NormalWord _ list -> sequentially list + + T_OrIf _ lhs rhs -> do + left <- build lhs + right <- build rhs + end <- newStructuralNode + linkRange left right + linkRange right end + linkRange left end + + T_Pipeline _ _ [cmd] -> build cmd + T_Pipeline id _ cmds -> do + start <- newStructuralNode + hasLastpipe <- reader $ cfLastpipe . cfParameters + (leading, last) <- buildPipe hasLastpipe cmds + -- Ideally we'd let this exit code be that of the last command in the pipeline but ok + end <- newNodeRange $ CFSetExitCode id + + mapM_ (linkRange start) leading + mapM_ (\c -> linkRangeAs CFEFalseFlow c end) leading + linkRanges $ [start] ++ last ++ [end] + where + buildPipe True [x] = do + last <- build x + return ([], [last]) + buildPipe lp (first:rest) = do + this <- subshell id "pipeline" $ build first + (leading, last) <- buildPipe lp rest + return (this:leading, last) + buildPipe _ [] = return ([], []) + + T_ProcSub id op cmds -> do + start <- newStructuralNode + body <- subshell id (op ++ "() process substitution") $ sequentially cmds + end <- newStructuralNode + + linkRange start body + linkRangeAs CFEFalseFlow body end + linkRange start end + + T_Redirecting _ redirs cmd -> do + -- For simple commands, this is the other way around in bash + -- We do it in this order for comound commands like { x=name; } > "$x" + redir <- sequentially redirs + body <- build cmd + linkRange redir body + + T_SelectIn id name words body -> forInHelper id name words body + + T_SimpleCommand id vars [] -> do + -- Vars can also be empty, as in the command "> foo" + assignments <- sequentially vars + status <- newNodeRange (CFSetExitCode id) + linkRange assignments status + + T_SimpleCommand id vars (cmd:args) -> + handleCommand t vars (cmd NE.:| args) $ getUnquotedLiteral cmd + + T_SingleQuoted _ _ -> none + + T_SourceCommand _ originalCommand inlinedSource -> do + cmd <- build originalCommand + end <- newStructuralNode + inline <- withReturn end $ build inlinedSource + linkRange cmd inline + linkRange inline end + return $ spanRange cmd inline + + T_Subshell id body -> do + main <- subshell id "explicit (..) subshell" $ sequentially body + status <- newNodeRange (CFSetExitCode id) + linkRange main status + + T_UntilExpression id cond body -> whileHelper id cond body + T_WhileExpression id cond body -> whileHelper id cond body + + T_CLOBBER _ -> none + T_GREATAND _ -> none + T_LESSAND _ -> none + T_LESSGREAT _ -> none + T_DGREAT _ -> none + T_Greater _ -> none + T_Less _ -> none + T_ParamSubSpecialChar _ _ -> none + + x -> do + error ("Unimplemented: " ++ show x) -- STRIP + none + +-- Still in `where` clause + forInHelper id name words body = do + entry <- newStructuralNode + expansion <- sequentially words + assignmentChoice <- newStructuralNode + assignments <- + if null words || any willSplit words + then (:[]) <$> (newNodeRange $ applySingle $ IdTagged id $ CFWriteVariable name CFValueString) + else mapM (\t -> newNodeRange $ applySingle $ IdTagged id $ CFWriteVariable name $ CFValueComputed (getId t) $ tokenToParts t) words + body <- sequentially body + exit <- newStructuralNode + -- Forward edges + linkRanges [entry, expansion, assignmentChoice] + mapM_ (\t -> linkRanges [assignmentChoice, t, body]) assignments + linkRange body exit + linkRange expansion exit + -- Backward edge + linkRange body assignmentChoice + return $ spanRange entry exit + + whileHelper id cond body = do + condRange <- asCondition $ sequentially cond + bodyRange <- sequentially body + end <- newNodeRange (CFSetExitCode id) + + linkRange condRange bodyRange + linkRange bodyRange condRange + linkRange condRange end + + +handleCommand cmd vars args literalCmd = do + -- TODO: Handle assignments in declaring commands + + case literalCmd of + Just "exit" -> regularExpansion vars (NE.toList args) $ handleExit + Just "return" -> regularExpansion vars (NE.toList args) $ handleReturn + Just "unset" -> regularExpansionWithStatus vars args $ handleUnset args + + Just "declare" -> handleDeclare args + Just "local" -> handleDeclare args + Just "typeset" -> handleDeclare args + + Just "printf" -> regularExpansionWithStatus vars args $ handlePrintf args + Just "wait" -> regularExpansionWithStatus vars args $ handleWait args + + Just "mapfile" -> regularExpansionWithStatus vars args $ handleMapfile args + Just "readarray" -> regularExpansionWithStatus vars args $ handleMapfile args + + Just "read" -> regularExpansionWithStatus vars args $ handleRead args + + Just "DEFINE_boolean" -> regularExpansionWithStatus vars args $ handleDEFINE args + Just "DEFINE_float" -> regularExpansionWithStatus vars args $ handleDEFINE args + Just "DEFINE_integer" -> regularExpansionWithStatus vars args $ handleDEFINE args + Just "DEFINE_string" -> regularExpansionWithStatus vars args $ handleDEFINE args + + -- This will mostly behave like 'command' but ok + Just "builtin" -> + case args of + _ NE.:| [] -> regular + (_ NE.:| newcmd:newargs) -> + handleCommand newcmd vars (newcmd NE.:| newargs) $ getLiteralString newcmd + Just "command" -> + case args of + _ NE.:| [] -> regular + (_ NE.:| newcmd:newargs) -> + handleOthers (getId newcmd) vars (newcmd NE.:| newargs) $ getLiteralString newcmd + _ -> regular + + where + regular = handleOthers (getId cmd) vars args literalCmd + handleExit = do + exitNode <- reader cfExitTarget + case exitNode of + Just target -> do + exit <- newNode CFResolvedExit + link exit target CFEExit + unreachable <- newNode CFUnreachable + return $ Range exit unreachable + Nothing -> do + exit <- newNode CFUnresolvedExit + unreachable <- newNode CFUnreachable + return $ Range exit unreachable + + handleReturn = do + returnTarget <- reader cfReturnTarget + case returnTarget of + Nothing -> error $ pleaseReport "missing return target" + Just target -> do + ret <- newNode CFStructuralNode + link ret target CFEFlow + unreachable <- newNode CFUnreachable + return $ Range ret unreachable + + handleUnset (cmd NE.:| args) = do + case () of + _ | "n" `elem` flagNames -> unsetWith CFUndefineNameref + _ | "v" `elem` flagNames -> unsetWith CFUndefineVariable + _ | "f" `elem` flagNames -> unsetWith CFUndefineFunction + _ -> unsetWith CFUndefine + where + pairs :: [(String, Token)] -- [(Flag string, token)] e.g. [("-f", t), ("", myfunc)] + pairs = map (\(str, (flag, val)) -> (str, flag)) $ fromMaybe (map (\c -> ("", (c,c))) args) $ getGnuOpts "vfn" args + (names, flags) = partition (null . fst) pairs + flagNames = map fst flags + literalNames :: [(Token, String)] -- Literal names to unset, e.g. [(myfuncToken, "myfunc")] + literalNames = mapMaybe (\(_, t) -> (,) t <$> getLiteralString t) names + -- Apply a constructor like CFUndefineVariable to each literalName, and tag with its id + unsetWith c = newNodeRange $ CFApplyEffects $ map (\(token, name) -> IdTagged (getId token) $ c name) literalNames + + + variableAssignRegex = mkRegex "^([_a-zA-Z][_a-zA-Z0-9]*)=" + + handleDeclare (cmd NE.:| args) = do + isFunc <- asks cfIsFunction + -- This is a bit of a kludge: we don't have great support for things like + -- 'declare -i x=$x' so do one round with declare x=$x, followed by declare -i x + let (evaluated, assignments, added, removed) = mconcat $ map (toEffects isFunc) args + before <- sequentially $ evaluated + assignments <- newNodeRange $ CFApplyEffects assignments + addedProps <- if null added then newStructuralNode else newNodeRange $ CFApplyEffects added + removedProps <- if null removed then newStructuralNode else newNodeRange $ CFApplyEffects removed + result <- newNodeRange $ CFSetExitCode (getId cmd) + linkRanges [before, assignments, addedProps, removedProps, result] + where + opts = map fst $ getGenericOpts args + array = "a" `elem` opts || associative + associative = "A" `elem` opts + integer = "i" `elem` opts + func = "f" `elem` opts || "F" `elem` opts + global = "g" `elem` opts + export = "x" `elem` opts + writer isFunc = + case () of + _ | global -> CFWriteGlobal + _ | isFunc -> CFWriteLocal + _ -> CFWriteVariable + + scope isFunc = + case () of + _ | global -> Just GlobalScope + _ | isFunc -> Just LocalScope + _ -> Nothing + + addedProps = S.fromList $ concat $ [ + [ CFVPArray | array ], + [ CFVPInteger | integer ], + [ CFVPExport | export ], + [ CFVPAssociative | associative ] + ] + + removedProps = S.fromList $ concat $ [ + -- Array property can't be unset + [ CFVPInteger | 'i' `elem` unsetOptions ], + [ CFVPExport | 'e' `elem` unsetOptions ] + ] + + toEffects isFunc (T_Assignment id mode var idx t) = + let + pre = idx ++ [t] + val = [ IdTagged id $ (writer isFunc) var $ CFValueComputed (getId t) $ [ CFStringVariable var | mode == Append ] ++ tokenToParts t ] + added = [ IdTagged id $ CFSetProps (scope isFunc) var addedProps | not $ S.null addedProps ] + removed = [ IdTagged id $ CFUnsetProps (scope isFunc) var addedProps | not $ S.null removedProps ] + in + (pre, val, added, removed) + + toEffects isFunc t = + let + id = getId t + pre = [t] + literal = getLiteralStringDef "\0" t + isKnown = '\0' `notElem` literal + match = fmap head $ variableAssignRegex `matchRegex` literal + name = fromMaybe literal match + + asLiteral = + IdTagged id $ (writer isFunc) name $ + CFValueComputed (getId t) [ CFStringLiteral $ drop 1 $ dropWhile (/= '=') $ literal ] + asUnknown = + IdTagged id $ (writer isFunc) name $ + CFValueString + + added = [ IdTagged id $ CFSetProps (scope isFunc) name addedProps ] + removed = [ IdTagged id $ CFUnsetProps (scope isFunc) name removedProps ] + + in + case () of + _ | not (isVariableName name) -> (pre, [], [], []) + _ | isJust match && isKnown -> (pre, [asLiteral], added, removed) + _ | isJust match -> (pre, [asUnknown], added, removed) + -- e.g. declare -i x + _ -> (pre, [], added, removed) + + -- find "ia" from `define +i +a` + unsetOptions :: String + unsetOptions = + let + strings = mapMaybe getLiteralString args + plusses = filter ("+" `isPrefixOf`) strings + in + concatMap (drop 1) plusses + + handlePrintf (cmd NE.:| args) = + newNodeRange $ CFApplyEffects $ maybeToList findVar + where + findVar = do + flags <- getBsdOpts "v:" args + (flag, arg) <- lookup "v" flags + name <- getLiteralString arg + return $ IdTagged (getId arg) $ CFWriteVariable name CFValueString + + handleWait (cmd NE.:| args) = + newNodeRange $ CFApplyEffects $ maybeToList findVar + where + findVar = do + let flags = getGenericOpts args + (flag, arg) <- lookup "p" flags + name <- getLiteralString arg + return $ IdTagged (getId arg) $ CFWriteVariable name CFValueInteger + + handleMapfile (cmd NE.:| args) = + newNodeRange $ CFApplyEffects [findVar] + where + findVar = + let (id, name) = fromMaybe (getId cmd, "MAPFILE") $ getFromArg `mplus` getFromFallback + in IdTagged id $ CFWriteVariable name CFValueArray + + getFromArg = do + flags <- getGnuOpts flagsForMapfile args + (_, arg) <- lookup "" flags + name <- getLiteralString arg + return (getId arg, name) + + getFromFallback = + listToMaybe $ mapMaybe getIfVar $ reverse args + getIfVar c = do + name <- getLiteralString c + guard $ isVariableName name + return (getId c, name) + + handleRead (cmd NE.:| args) = newNodeRange $ CFApplyEffects main + where + main = fromMaybe fallback $ do + flags <- getGnuOpts flagsForRead args + return $ fromMaybe (withFields flags) $ withArray flags + + withArray :: [(String, (Token, Token))] -> Maybe [IdTagged CFEffect] + withArray flags = do + (_, token) <- lookup "a" flags + return $ fromMaybe [] $ do + name <- getLiteralString token + return [ IdTagged (getId token) $ CFWriteVariable name CFValueArray ] + + withFields flags = mapMaybe getAssignment flags + + getAssignment :: (String, (Token, Token)) -> Maybe (IdTagged CFEffect) + getAssignment f = do + ("", (t, _)) <- return f + name <- getLiteralString t + return $ IdTagged (getId t) $ CFWriteVariable name CFValueString + + fallback = + let + names = reverse $ map fromJust $ takeWhile isJust $ map (\c -> sequence (getId c, getLiteralString c)) $ reverse args + namesOrDefault = if null names then [(getId cmd, "REPLY")] else names + hasDashA = any (== "a") $ map fst $ getGenericOpts args + value = if hasDashA then CFValueArray else CFValueString + in + map (\(id, name) -> IdTagged id $ CFWriteVariable name value) namesOrDefault + + handleDEFINE (cmd NE.:| args) = + newNodeRange $ CFApplyEffects $ maybeToList findVar + where + findVar = do + name <- listToMaybe $ drop 1 args + str <- getLiteralString name + guard $ isVariableName str + return $ IdTagged (getId name) $ CFWriteVariable str CFValueString + + handleOthers id vars args cmd = + regularExpansion vars (NE.toList args) $ do + exe <- newNodeRange $ CFExecuteCommand cmd + status <- newNodeRange $ CFSetExitCode id + linkRange exe status + + regularExpansion vars args p = do + args <- sequentially args + assignments <- mapM (buildAssignment (Just PrefixScope)) vars + exe <- p + dropAssignments <- + if null vars + then + return [] + else do + drop <- newNodeRange CFDropPrefixAssignments + return [drop] + + linkRanges $ [args] ++ assignments ++ [exe] ++ dropAssignments + + regularExpansionWithStatus vars args@(cmd NE.:| _) p = do + initial <- regularExpansion vars (NE.toList args) p + status <- newNodeRange $ CFSetExitCode (getId cmd) + linkRange initial status + + +none = newStructuralNode + +data Scope = GlobalScope | LocalScope | PrefixScope + deriving (Eq, Ord, Show, Generic, NFData) + +buildAssignment scope t = do + op <- case t of + T_Assignment id mode var indices value -> do + expand <- build value + index <- sequentially indices + read <- case mode of + Append -> newNodeRange (applySingle $ IdTagged id $ CFReadVariable var) + Assign -> none + let valueType = if null indices then f id value else CFValueArray + let scoper = + case scope of + Just PrefixScope -> CFWritePrefix + Just LocalScope -> CFWriteLocal + Just GlobalScope -> CFWriteGlobal + Nothing -> CFWriteVariable + write <- newNodeRange $ applySingle $ IdTagged id $ scoper var valueType + linkRanges [expand, index, read, write] + where + f :: Id -> Token -> CFValue + f id t@T_NormalWord {} = CFValueComputed id $ [CFStringVariable var | mode == Append] ++ tokenToParts t + f id t@(T_Literal _ str) = CFValueComputed id $ [CFStringVariable var | mode == Append] ++ tokenToParts t + f _ T_Array {} = CFValueArray + + registerNode (getId t) op + return op + + +tokenToParts t = + case t of + T_NormalWord _ list -> concatMap tokenToParts list + T_DoubleQuoted _ list -> concatMap tokenToParts list + T_SingleQuoted _ str -> [ CFStringLiteral str ] + T_Literal _ str -> [ CFStringLiteral str ] + T_DollarArithmetic {} -> [ CFStringInteger ] + T_DollarBracket {} -> [ CFStringInteger ] + T_DollarBraced _ _ list | isUnmodifiedParameterExpansion t -> [ CFStringVariable (getBracedReference $ concat $ oversimplify list) ] + -- Check if getLiteralString can handle it, if not it's unknown + _ -> [maybe CFStringUnknown CFStringLiteral $ getLiteralString t] + + +-- Like & but well defined when the node already exists +safeUpdate ctx@(_,node,_,_) graph = ctx & (delNode node graph) + +-- Change all subshell invocations to instead link directly to their contents. +-- This is used for producing dominator trees. +inlineSubshells :: CFGraph -> CFGraph +inlineSubshells graph = relinkedGraph + where + subshells = ufold find [] graph + find (incoming, node, label, outgoing) acc = + case label of + CFExecuteSubshell _ start end -> (node, label, start, end, incoming, outgoing):acc + _ -> acc + + relinkedGraph = foldl' relink graph subshells + relink graph (node, label, start, end, incoming, outgoing) = + let + -- Link CFExecuteSubshell to the CFEntryPoint + subshellToStart = (incoming, node, label, [(CFEFlow, start)]) + -- Link the subshell exit to the + endToNexts = (endIncoming, endNode, endLabel, outgoing) + (endIncoming, endNode, endLabel, _) = context graph end + in + subshellToStart `safeUpdate` (endToNexts `safeUpdate` graph) + +findEntryNodes :: CFGraph -> [Node] +findEntryNodes graph = ufold find [] graph + where + find (incoming, node, label, _) list = + case label of + CFEntryPoint {} | null incoming -> node:list + _ -> list + +findDominators main graph = asSetMap + where + inlined = inlineSubshells graph + entryNodes = main : findEntryNodes graph + asLists = concatMap (dom inlined) entryNodes + asSetMap = M.fromList $ map (\(node, list) -> (node, S.fromList list)) asLists + +findTerminalNodes :: CFGraph -> [Node] +findTerminalNodes graph = ufold find [] graph + where + find (_, node, label, _) list = + case label of + CFUnresolvedExit -> node:list + CFApplyEffects effects -> f effects list + _ -> list + + f [] list = list + f (IdTagged _ (CFDefineFunction _ id start end):rest) list = f rest (end:list) + f (_:rest) list = f rest list + +findPostDominators :: Node -> CFGraph -> Array Node [Node] +findPostDominators mainexit graph = asArray + where + inlined = inlineSubshells graph + terminals = findTerminalNodes inlined + (incoming, _, label, outgoing) = context graph mainexit + withExitEdges = (incoming ++ map (\c -> (CFEFlow, c)) terminals, mainexit, label, outgoing) `safeUpdate` inlined + reversed = grev withExitEdges + postDoms = dom reversed mainexit + (_, maxNode) = nodeRange graph + -- Holes in the array cause "Exception: (Array.!): undefined array element" while + -- inspecting/debugging, so fill the array first and then update. + initializedArray = listArray (0, maxNode) $ repeat [] + asArray = initializedArray // postDoms + +return [] +runTests = $( [| $(forAllProperties) (quickCheckWithResult (stdArgs { maxSuccess = 1 }) ) |]) diff --git a/src/ShellCheck/CFGAnalysis.hs b/src/ShellCheck/CFGAnalysis.hs new file mode 100644 index 0000000..cf982e0 --- /dev/null +++ b/src/ShellCheck/CFGAnalysis.hs @@ -0,0 +1,1439 @@ +{- + Copyright 2022 Vidar Holen + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{-# LANGUAGE TemplateHaskell #-} +{-# LANGUAGE RankNTypes #-} +{-# LANGUAGE DeriveAnyClass, DeriveGeneric #-} +{-# LANGUAGE CPP #-} + +{- + Data Flow Analysis on a Control Flow Graph. + + This module implements a pretty standard iterative Data Flow Analysis. + For an overview of the process, see Wikipedia. + + Since shell scripts rely heavily on global variables, this DFA includes + tracking the value of globals across calls. Each function invocation is + treated as a separate DFA problem, and a caching mechanism (hopefully) + avoids any exponential explosions. + + To do efficient DFA join operations (or merges, as the code calls them), + some of the data structures have an integer version attached. On update, + the version is changed. If two states have the same version number, + a merge is skipped on the grounds that they are identical. It is easy + to unintentionally forget to update/invalidate the version number, + and bugs will ensure. + + For performance reasons, the entire code runs in plain ST, with a manual + context object Ctx being passed around. It relies heavily on mutable + STRefs. However, this turned out to be literally thousands of times faster + than my several attempts using RWST, so it can't be helped. +-} + +module ShellCheck.CFGAnalysis ( + analyzeControlFlow + ,CFGParameters (..) + ,CFGAnalysis (..) + ,ProgramState (..) + ,VariableState (..) + ,VariableValue (..) + ,VariableProperties + ,SpaceStatus (..) + ,NumericalStatus (..) + ,getIncomingState + ,getOutgoingState + ,doesPostDominate + ,variableMayBeDeclaredInteger + ,variableMayBeAssignedInteger + ,ShellCheck.CFGAnalysis.runTests -- STRIP + ) where + +import Control.DeepSeq +import Control.Monad +import Control.Monad.ST +import Data.Array.Unboxed +import Data.Char +import Data.Graph.Inductive.Graph +import Data.Graph.Inductive.Query.DFS +import Data.List hiding (map) +import Data.Maybe +import Data.STRef +import Debug.Trace -- STRIP +import GHC.Generics (Generic) +import qualified Data.Map as M +import qualified Data.Set as S +import qualified ShellCheck.Data as Data +import ShellCheck.AST +import ShellCheck.CFG +import ShellCheck.Prelude + +import Test.QuickCheck + + +-- The number of iterations for DFA to stabilize +iterationCount = 1000000 +-- There have been multiple bugs where bad caching caused oscillations. +-- As a precaution, disable caching if there's this many iterations left. +fallbackThreshold = 10000 +-- The number of cache entries to keep per node +cacheEntries = 10 + +logVerbose log = do + -- traceShowM log + return () +logInfo log = do + -- traceShowM log + return () + +-- The result of the data flow analysis +data CFGAnalysis = CFGAnalysis { + graph :: CFGraph, + tokenToRange :: M.Map Id (Node, Node), + tokenToNodes :: M.Map Id (S.Set Node), + postDominators :: Array Node [Node], + nodeToData :: M.Map Node (ProgramState, ProgramState) +} deriving (Show) + +-- The program state we expose externally +data ProgramState = ProgramState { + -- internalState :: InternalState, -- For debugging + variablesInScope :: M.Map String VariableState, + exitCodes :: S.Set Id, + stateIsReachable :: Bool +} deriving (Show, Eq, Generic, NFData) + +internalToExternal :: InternalState -> ProgramState +internalToExternal s = + ProgramState { + -- Censor the literal value to avoid introducing dependencies on it. It's just for debugging. + variablesInScope = M.map censor flatVars, + -- internalState = s, -- For debugging + exitCodes = fromMaybe S.empty $ sExitCodes s, + stateIsReachable = fromMaybe True $ sIsReachable s + } + where + censor s = s { + variableValue = (variableValue s) { + literalValue = Nothing + } + } + flatVars = M.unions $ map mapStorage [sPrefixValues s, sLocalValues s, sGlobalValues s] + +-- Conveniently get the state before a token id +getIncomingState :: CFGAnalysis -> Id -> Maybe ProgramState +getIncomingState analysis id = do + (start,end) <- M.lookup id $ tokenToRange analysis + fst <$> M.lookup start (nodeToData analysis) + +-- Conveniently get the state after a token id +getOutgoingState :: CFGAnalysis -> Id -> Maybe ProgramState +getOutgoingState analysis id = do + (start,end) <- M.lookup id $ tokenToRange analysis + snd <$> M.lookup end (nodeToData analysis) + +-- Conveniently determine whether one node postdominates another, +-- i.e. whether 'target' always unconditionally runs after 'base'. +doesPostDominate :: CFGAnalysis -> Id -> Id -> Bool +doesPostDominate analysis target base = fromMaybe False $ do + (_, baseEnd) <- M.lookup base $ tokenToRange analysis + (targetStart, _) <- M.lookup target $ tokenToRange analysis + return $ targetStart `elem` (postDominators analysis ! baseEnd) + +-- See if any execution path results in the variable containing a state +variableMayHaveState :: ProgramState -> String -> CFVariableProp -> Maybe Bool +variableMayHaveState state var property = do + value <- M.lookup var $ variablesInScope state + return $ any (S.member property) $ variableProperties value + +-- See if any execution path declares the variable an integer (declare -i). +variableMayBeDeclaredInteger state var = variableMayHaveState state var CFVPInteger + +-- See if any execution path suggests the variable may contain an integer value +variableMayBeAssignedInteger state var = do + value <- M.lookup var $ variablesInScope state + return $ (numericalStatus $ variableValue value) >= NumericalStatusMaybe + +getDataForNode analysis node = M.lookup node $ nodeToData analysis + +-- The current state of data flow at a point in the program, potentially as a diff +data InternalState = InternalState { + sVersion :: Integer, + sGlobalValues :: VersionedMap String VariableState, + sLocalValues :: VersionedMap String VariableState, + sPrefixValues :: VersionedMap String VariableState, + sFunctionTargets :: VersionedMap String FunctionValue, + sExitCodes :: Maybe (S.Set Id), + sIsReachable :: Maybe Bool +} deriving (Show, Generic, NFData) + +newInternalState = InternalState { + sVersion = 0, + sGlobalValues = vmEmpty, + sLocalValues = vmEmpty, + sPrefixValues = vmEmpty, + sFunctionTargets = vmEmpty, + sExitCodes = Nothing, + sIsReachable = Nothing +} + +unreachableState = modified newInternalState { + sIsReachable = Just False +} + +-- The default state we assume we get from the environment +createEnvironmentState :: InternalState +createEnvironmentState = do + foldl' (flip ($)) newInternalState $ concat [ + addVars Data.internalVariables unknownVariableState, + addVars Data.variablesWithoutSpaces spacelessVariableState, + addVars Data.specialIntegerVariables integerVariableState + ] + where + addVars names val = map (\name -> insertGlobal name val) names + spacelessVariableState = unknownVariableState { + variableValue = VariableValue { + literalValue = Nothing, + spaceStatus = SpaceStatusClean, + numericalStatus = NumericalStatusUnknown + } + } + integerVariableState = unknownVariableState { + variableValue = unknownIntegerValue + } + + +modified s = s { sVersion = -1 } + +insertGlobal :: String -> VariableState -> InternalState -> InternalState +insertGlobal name value state = modified state { + sGlobalValues = vmInsert name value $ sGlobalValues state +} + +insertLocal :: String -> VariableState -> InternalState -> InternalState +insertLocal name value state = modified state { + sLocalValues = vmInsert name value $ sLocalValues state +} + +insertPrefix :: String -> VariableState -> InternalState -> InternalState +insertPrefix name value state = modified state { + sPrefixValues = vmInsert name value $ sPrefixValues state +} + +insertFunction :: String -> FunctionValue -> InternalState -> InternalState +insertFunction name value state = modified state { + sFunctionTargets = vmInsert name value $ sFunctionTargets state +} + +addProperties :: S.Set CFVariableProp -> VariableState -> VariableState +addProperties props state = state { + variableProperties = S.map (S.union props) $ variableProperties state +} + +removeProperties :: S.Set CFVariableProp -> VariableState -> VariableState +removeProperties props state = state { + variableProperties = S.map (\s -> S.difference s props) $ variableProperties state +} + +setExitCode id = setExitCodes (S.singleton id) +setExitCodes set state = modified state { + sExitCodes = Just $ set +} + +-- Dependencies on values, e.g. "if there is a global variable named 'foo' without spaces" +-- This is used to see if the DFA of a function would result in the same state, so anything +-- that affects DFA must be tracked. +data StateDependency = + -- Complete variable state + DepState Scope String VariableState + -- Only variable properties (we need properties but not values for x=1) + | DepProperties Scope String VariableProperties + -- Function definition + | DepFunction String (S.Set FunctionDefinition) + -- Whether invoking the node would result in recursion (i.e., is the function on the stack?) + | DepIsRecursive Node Bool + -- The set of commands that could have provided the exit code $? + | DepExitCodes (S.Set Id) + deriving (Show, Eq, Ord, Generic, NFData) + +-- A function definition, or lack thereof +data FunctionDefinition = FunctionUnknown | FunctionDefinition String Node Node + deriving (Show, Eq, Ord, Generic, NFData) + +-- The Set of places a command name can point (it's a Set to handle conditionally defined functions) +type FunctionValue = S.Set FunctionDefinition + +-- Create an InternalState that fulfills the given dependencies +depsToState :: S.Set StateDependency -> InternalState +depsToState set = foldl insert newInternalState $ S.toList set + where + insert :: InternalState -> StateDependency -> InternalState + insert state dep = + case dep of + DepFunction name val -> insertFunction name val state + DepState scope name val -> insertIn True scope name val state + -- State includes properties and more, so don't overwrite a state with properties + DepProperties scope name props -> insertIn False scope name unknownVariableState { variableProperties = props } state + DepIsRecursive _ _ -> state + DepExitCodes s -> setExitCodes s state + + insertIn overwrite scope name val state = + let + (mapToCheck, inserter) = + case scope of + PrefixScope -> (sPrefixValues, insertPrefix) + LocalScope -> (sLocalValues, insertLocal) + GlobalScope -> (sGlobalValues, insertGlobal) + + alreadyExists = isJust $ vmLookup name $ mapToCheck state + in + if overwrite || not alreadyExists + then inserter name val state + else state + +unknownFunctionValue = S.singleton FunctionUnknown + +-- The information about the value of a single variable +data VariableValue = VariableValue { + literalValue :: Maybe String, -- TODO: For debugging. Remove me. + spaceStatus :: SpaceStatus, + numericalStatus :: NumericalStatus +} + deriving (Show, Eq, Ord, Generic, NFData) + +data VariableState = VariableState { + variableValue :: VariableValue, + variableProperties :: VariableProperties +} + deriving (Show, Eq, Ord, Generic, NFData) + +-- Whether or not the value needs quoting (has spaces/globs), or we don't know +data SpaceStatus = SpaceStatusEmpty | SpaceStatusClean | SpaceStatusDirty deriving (Show, Eq, Ord, Generic, NFData) +-- +-- Whether or not the value needs quoting (has spaces/globs), or we don't know +data NumericalStatus = NumericalStatusUnknown | NumericalStatusEmpty | NumericalStatusMaybe | NumericalStatusDefinitely deriving (Show, Eq, Ord, Generic, NFData) + +-- The set of possible sets of properties for this variable +type VariableProperties = S.Set (S.Set CFVariableProp) + +defaultProperties = S.singleton S.empty + +unknownVariableState = VariableState { + variableValue = unknownVariableValue, + variableProperties = defaultProperties +} + +unknownVariableValue = VariableValue { + literalValue = Nothing, + spaceStatus = SpaceStatusDirty, + numericalStatus = NumericalStatusUnknown +} + +emptyVariableValue = unknownVariableValue { + literalValue = Just "", + spaceStatus = SpaceStatusEmpty, + numericalStatus = NumericalStatusEmpty +} + +unsetVariableState = VariableState { + variableValue = emptyVariableValue, + variableProperties = defaultProperties +} + +mergeVariableState a b = VariableState { + variableValue = mergeVariableValue (variableValue a) (variableValue b), + variableProperties = S.union (variableProperties a) (variableProperties b) +} + +mergeVariableValue a b = VariableValue { + literalValue = if literalValue a == literalValue b then literalValue a else Nothing, + spaceStatus = mergeSpaceStatus (spaceStatus a) (spaceStatus b), + numericalStatus = mergeNumericalStatus (numericalStatus a) (numericalStatus b) +} + +mergeSpaceStatus a b = + case (a,b) of + (SpaceStatusEmpty, y) -> y + (x, SpaceStatusEmpty) -> x + (SpaceStatusClean, SpaceStatusClean) -> SpaceStatusClean + _ -> SpaceStatusDirty + +mergeNumericalStatus a b = + case (a,b) of + (NumericalStatusDefinitely, NumericalStatusDefinitely) -> NumericalStatusDefinitely + (NumericalStatusDefinitely, _) -> NumericalStatusMaybe + (_, NumericalStatusDefinitely) -> NumericalStatusMaybe + (NumericalStatusMaybe, _) -> NumericalStatusMaybe + (_, NumericalStatusMaybe) -> NumericalStatusMaybe + (NumericalStatusEmpty, NumericalStatusEmpty) -> NumericalStatusEmpty + _ -> NumericalStatusUnknown + +-- A VersionedMap is a Map that keeps an additional integer version to quickly determine if it has changed. +-- * Version -1 means it's unknown (possibly and presumably changed) +-- * Version 0 means it's empty +-- * Version N means it's equal to any other map with Version N (this is required but not enforced) +data VersionedMap k v = VersionedMap { + mapVersion :: Integer, + mapStorage :: M.Map k v +} + deriving (Generic, NFData) + +-- This makes states more readable but inhibits copy-paste +instance (Show k, Show v) => Show (VersionedMap k v) where + show m = (if mapVersion m >= 0 then "V" ++ show (mapVersion m) else "U") ++ " " ++ show (mapStorage m) + +instance Eq InternalState where + (==) a b = stateIsQuickEqual a b || stateIsSlowEqual a b + +instance (Eq k, Eq v) => Eq (VersionedMap k v) where + (==) a b = vmIsQuickEqual a b || mapStorage a == mapStorage b + +instance (Ord k, Ord v) => Ord (VersionedMap k v) where + compare a b = + if vmIsQuickEqual a b + then EQ + else mapStorage a `compare` mapStorage b + + +-- A context with STRefs manually passed around to function. +-- This is done because it was dramatically much faster than any RWS type stack +data Ctx s = Ctx { + -- The current node + cNode :: STRef s Node, + -- The current input state + cInput :: STRef s InternalState, + -- The current output state + cOutput :: STRef s InternalState, + + -- The current functions/subshells stack + cStack :: [StackEntry s], + -- The input graph + cGraph :: CFGraph, + -- An incrementing counter to version maps + cCounter :: STRef s Integer, + -- A cache of input state dependencies to output effects + cCache :: STRef s (M.Map Node [(S.Set StateDependency, InternalState)]), + -- Whether the cache is enabled (see fallbackThreshold) + cEnableCache :: STRef s Bool, + -- The states resulting from data flows per invocation path + cInvocations :: STRef s (M.Map [Node] (S.Set StateDependency, M.Map Node (InternalState, InternalState))) +} + +-- Whenever a function (or subshell) is invoked, a value like this is pushed onto the stack +data StackEntry s = StackEntry { + -- The entry point of this stack entry for the purpose of detecting recursion + entryPoint :: Node, + -- Whether this is a function call (as opposed to a subshell) + isFunctionCall :: Bool, + -- The node where this entry point was invoked + callSite :: Node, + -- A mutable set of dependencies we fetched from here or higher in the stack + dependencies :: STRef s (S.Set StateDependency), + -- The original input state for this stack entry + stackState :: InternalState +} + deriving (Eq, Generic, NFData) + +#if MIN_VERSION_deepseq(1,4,2) +-- Our deepseq already has a STRef instance +#else +-- Older deepseq (for GHC < 8) lacks this instance +instance NFData (STRef s a) where + rnf = (`seq` ()) +#endif + +-- Overwrite a base state with the contents of a diff state +-- This is unrelated to join/merge. +patchState :: InternalState -> InternalState -> InternalState +patchState base diff = + case () of + _ | sVersion diff == 0 -> base + _ | sVersion base == 0 -> diff + _ | stateIsQuickEqual base diff -> diff + _ -> + InternalState { + sVersion = -1, + sGlobalValues = vmPatch (sGlobalValues base) (sGlobalValues diff), + sLocalValues = vmPatch (sLocalValues base) (sLocalValues diff), + sPrefixValues = vmPatch (sPrefixValues base) (sPrefixValues diff), + sFunctionTargets = vmPatch (sFunctionTargets base) (sFunctionTargets diff), + sExitCodes = sExitCodes diff `mplus` sExitCodes base, + sIsReachable = sIsReachable diff `mplus` sIsReachable base + } + +patchOutputM ctx diff = do + let cOut = cOutput ctx + oldState <- readSTRef cOut + let newState = patchState oldState diff + writeSTRef cOut newState + +-- Merge (aka Join) two states. This is monadic because it requires looking up +-- values from the current context. For example: +-- +-- f() { +-- foo || x=2 +-- HERE # This merge requires looking up the value of $x in the parent frame +-- } +-- x=1 +-- f +mergeState :: forall s. Ctx s -> InternalState -> InternalState -> ST s InternalState +mergeState ctx a b = do + -- Kludge: we want `readVariable` & friends not to read from an intermediate state, + -- so temporarily set a blank input. + let cin = cInput ctx + old <- readSTRef cin + writeSTRef cin newInternalState + x <- merge a b + writeSTRef cin old + return x + + where + + merge a b = + case () of + _ | sIsReachable a == Just True && sIsReachable b == Just False + || sIsReachable a == Just False && sIsReachable b == Just True -> + error $ pleaseReport "Unexpected merge of reachable and unreachable state" + _ | sIsReachable a == Just False && sIsReachable b == Just False -> + return unreachableState + _ | sVersion a >= 0 && sVersion b >= 0 && sVersion a == sVersion b -> return a + _ -> do + globals <- mergeMaps ctx mergeVariableState readGlobal (sGlobalValues a) (sGlobalValues b) + locals <- mergeMaps ctx mergeVariableState readVariable (sLocalValues a) (sLocalValues b) + prefix <- mergeMaps ctx mergeVariableState readVariable (sPrefixValues a) (sPrefixValues b) + funcs <- mergeMaps ctx S.union readFunction (sFunctionTargets a) (sFunctionTargets b) + exitCodes <- mergeMaybes ctx S.union readExitCodes (sExitCodes a) (sExitCodes b) + return $ InternalState { + sVersion = -1, + sGlobalValues = globals, + sLocalValues = locals, + sPrefixValues = prefix, + sFunctionTargets = funcs, + sExitCodes = exitCodes, + sIsReachable = liftM2 (&&) (sIsReachable a) (sIsReachable b) + } + +-- Merge a number of states, or return a default if there are no states +-- (it can't fold from newInternalState because this would be equivalent of adding a new input edge). +mergeStates :: forall s. Ctx s -> InternalState -> [InternalState] -> ST s InternalState +mergeStates ctx def list = + case list of + [] -> return def + (first:rest) -> foldM (mergeState ctx) first rest + +-- Merge two maps, key by key. If both maps have a key, the 'merger' is used. +-- If only one has the key, the 'reader' is used to fetch a second, and the two are merged as above. +mergeMaps :: (Ord k) => forall s. + Ctx s -> + (v -> v -> v) -> + (Ctx s -> k -> ST s v) -> + (VersionedMap k v) -> + (VersionedMap k v) -> + ST s (VersionedMap k v) +mergeMaps ctx merger reader a b = + if vmIsQuickEqual a b + then return a + else do + new <- M.fromDistinctAscList <$> reverse <$> f [] (M.toAscList $ mapStorage a) (M.toAscList $ mapStorage b) + vmFromMap ctx new + where + f l [] [] = return l + f l [] b = f l b [] + f l ((k,v):rest1) [] = do + other <- reader ctx k + f ((k, merger v other):l) rest1 [] + f l l1@((k1, v1):rest1) l2@((k2, v2):rest2) = + case k1 `compare` k2 of + EQ -> + f ((k1, merger v1 v2):l) rest1 rest2 + LT -> do + nv2 <- reader ctx k1 + f ((k1, merger v1 nv2):l) rest1 l2 + GT -> do + nv1 <- reader ctx k2 + f ((k2, merger nv1 v2):l) l1 rest2 + +-- Merge two Maybes, like mergeMaps for a single element +mergeMaybes ctx merger reader a b = + case (a, b) of + (Nothing, Nothing) -> return Nothing + (Just v1, Nothing) -> single v1 + (Nothing, Just v2) -> single v2 + (Just v1, Just v2) -> return $ Just $ merger v1 v2 + where + single val = do + result <- merger val <$> reader ctx + return $ Just result + +vmFromMap ctx map = return $ VersionedMap { + mapVersion = -1, + mapStorage = map +} + +-- Give a VersionedMap a version if it does not already have one. +versionMap ctx map = + if mapVersion map >= 0 + then return map + else do + v <- nextVersion ctx + return map { + mapVersion = v + } + +-- Give an InternalState a version if it does not already have one. +versionState ctx state = + if sVersion state >= 0 + then return state + else do + self <- nextVersion ctx + ssGlobalValues <- versionMap ctx $ sGlobalValues state + ssLocalValues <- versionMap ctx $ sLocalValues state + ssFunctionTargets <- versionMap ctx $ sFunctionTargets state + return state { + sVersion = self, + sGlobalValues = ssGlobalValues, + sLocalValues = ssLocalValues, + sFunctionTargets = ssFunctionTargets + } + +-- Like 'not null' but for 2+ elements +is2plus :: [a] -> Bool +is2plus l = case l of + _:_:_ -> True + _ -> False + +-- Use versions to see if two states are trivially identical +stateIsQuickEqual a b = + let + va = sVersion a + vb = sVersion b + in + va >= 0 && vb >= 0 && va == vb + +-- A manual slow path 'Eq' (it's not derived because it's part of the custom Eq instance) +stateIsSlowEqual a b = + check sGlobalValues + && check sLocalValues + && check sPrefixValues + && check sFunctionTargets + && check sIsReachable + where + check f = f a == f b + +-- Check if two VersionedMaps are trivially equal +vmIsQuickEqual :: VersionedMap k v -> VersionedMap k v -> Bool +vmIsQuickEqual a b = + let + va = mapVersion a + vb = mapVersion b + in + va >= 0 && vb >= 0 && va == vb + +-- A new, empty VersionedMap +vmEmpty = VersionedMap { + mapVersion = 0, + mapStorage = M.empty +} + +-- Map.null for VersionedMaps +vmNull :: VersionedMap k v -> Bool +vmNull m = mapVersion m == 0 || (M.null $ mapStorage m) + +-- Map.lookup for VersionedMaps +vmLookup name map = M.lookup name $ mapStorage map + +-- Map.insert for VersionedMaps +vmInsert key val map = VersionedMap { + mapVersion = -1, + mapStorage = M.insert key val $ mapStorage map +} + +-- Overwrite all keys in the first map with values from the second +vmPatch :: (Ord k) => VersionedMap k v -> VersionedMap k v -> VersionedMap k v +vmPatch base diff = + case () of + _ | mapVersion base == 0 -> diff + _ | mapVersion diff == 0 -> base + _ | vmIsQuickEqual base diff -> diff + _ -> VersionedMap { + mapVersion = -1, + mapStorage = M.union (mapStorage diff) (mapStorage base) + } + +-- Set a variable. This includes properties. Applies it to the appropriate scope. +writeVariable :: forall s. Ctx s -> String -> VariableState -> ST s () +writeVariable ctx name val = do + typ <- readVariableScope ctx name + case typ of + GlobalScope -> writeGlobal ctx name val + LocalScope -> writeLocal ctx name val + -- Prefixed variables actually become local variables in the invoked function + PrefixScope -> writeLocal ctx name val + +writeGlobal ctx name val = do + modifySTRef (cOutput ctx) $ insertGlobal name val + +writeLocal ctx name val = do + modifySTRef (cOutput ctx) $ insertLocal name val + +writePrefix ctx name val = do + modifySTRef (cOutput ctx) $ insertPrefix name val + +updateVariableValue ctx name val = do + (props, scope) <- readVariablePropertiesWithScope ctx name + let f = case scope of + GlobalScope -> writeGlobal + LocalScope -> writeLocal + PrefixScope -> writeLocal -- Updates become local + f ctx name $ VariableState { variableValue = val, variableProperties = props } + +updateGlobalValue ctx name val = do + props <- readGlobalProperties ctx name + writeGlobal ctx name VariableState { variableValue = val, variableProperties = props } + +updateLocalValue ctx name val = do + props <- readLocalProperties ctx name + writeLocal ctx name VariableState { variableValue = val, variableProperties = props } + +updatePrefixValue ctx name val = do + -- Prefix variables don't inherit properties + writePrefix ctx name VariableState { variableValue = val, variableProperties = defaultProperties } + + +-- Look up a variable value, and also return its scope +readVariableWithScope :: forall s. Ctx s -> String -> ST s (VariableState, Scope) +readVariableWithScope ctx name = lookupStack get dep def ctx name + where + def = (unknownVariableState, GlobalScope) + get = getVariableWithScope + dep k (val, scope) = DepState scope k val + +-- Look up the variable's properties. This can be done independently to avoid incurring a dependency on the value. +readVariablePropertiesWithScope :: forall s. Ctx s -> String -> ST s (VariableProperties, Scope) +readVariablePropertiesWithScope ctx name = lookupStack get dep def ctx name + where + def = (defaultProperties, GlobalScope) + get s k = do + (val, scope) <- getVariableWithScope s k + return (variableProperties val, scope) + dep k (val, scope) = DepProperties scope k val + +readVariableScope ctx name = snd <$> readVariablePropertiesWithScope ctx name + +getVariableWithScope :: InternalState -> String -> Maybe (VariableState, Scope) +getVariableWithScope s name = + case (vmLookup name $ sPrefixValues s, vmLookup name $ sLocalValues s, vmLookup name $ sGlobalValues s) of + (Just var, _, _) -> return (var, PrefixScope) + (_, Just var, _) -> return (var, LocalScope) + (_, _, Just var) -> return (var, GlobalScope) + _ -> Nothing + +undefineFunction ctx name = + writeFunction ctx name $ FunctionUnknown + +undefineVariable ctx name = + writeVariable ctx name $ unsetVariableState + +readVariable ctx name = fst <$> readVariableWithScope ctx name +readVariableProperties ctx name = fst <$> readVariablePropertiesWithScope ctx name + +readGlobal ctx name = lookupStack get dep def ctx name + where + def = unknownVariableState -- could come from the environment + get s name = vmLookup name $ sGlobalValues s + dep k v = DepState GlobalScope k v + + +readGlobalProperties ctx name = lookupStack get dep def ctx name + where + def = defaultProperties + get s name = variableProperties <$> (vmLookup name $ sGlobalValues s) + -- This dependency will fail to match if it's shadowed by a local variable, + -- such as in x=1; f() { local -i x; declare -ag x; } because we'll look at + -- x and find it to be local and not global. FIXME? + dep k v = DepProperties GlobalScope k v + +readLocal ctx name = lookupStackUntilFunction get dep def ctx name + where + def = unsetVariableState -- can't come from the environment + get s name = vmLookup name $ sLocalValues s + dep k v = DepState LocalScope k v + +-- We only want to look up the local properties of the current function, +-- though preferably even if we're in a subshell. FIXME? +readLocalProperties ctx name = fst <$> lookupStackUntilFunction get dep def ctx name + where + def = (defaultProperties, LocalScope) + with tag f = do + val <- variableProperties <$> f + return (val, tag) + + get s name = (with LocalScope $ vmLookup name $ sLocalValues s) `mplus` (with PrefixScope $ vmLookup name $ sPrefixValues s) + dep k (val, scope) = DepProperties scope k val + +readFunction ctx name = lookupStack get dep def ctx name + where + def = unknownFunctionValue + get s name = vmLookup name $ sFunctionTargets s + dep k v = DepFunction k v + +writeFunction ctx name val = do + modifySTRef (cOutput ctx) $ insertFunction name $ S.singleton val + +readExitCodes ctx = lookupStack get dep def ctx () + where + get s () = sExitCodes s + def = S.empty + dep () v = DepExitCodes v + +-- Look up each state on the stack until a value is found (or the default is used), +-- then add this value as a StateDependency. +lookupStack' :: forall s k v. + -- Whether to stop at function boundaries + Bool + -- A function that maybe finds a value from a state + -> (InternalState -> k -> Maybe v) + -- A function that creates a dependency on what was found + -> (k -> v -> StateDependency) + -- A default value, if the value can't be found anywhere + -> v + -- Context + -> Ctx s + -- The key to look up + -> k + -- Returning the result + -> ST s v +lookupStack' functionOnly get dep def ctx key = do + top <- readSTRef $ cInput ctx + case get top key of + Just v -> return v + Nothing -> f (cStack ctx) + where + f [] = return def + f (s:_) | functionOnly && isFunctionCall s = return def + f (s:rest) = do + -- Go up the stack until we find the value, and add + -- a dependency on each state (including where it was found) + res <- maybe (f rest) return (get (stackState s) key) + modifySTRef (dependencies s) $ S.insert $ dep key res + return res + +lookupStack = lookupStack' False +lookupStackUntilFunction = lookupStack' True + +-- Like lookupStack but without adding dependencies +peekStack get def ctx key = do + top <- readSTRef $ cInput ctx + case get top key of + Just v -> return v + Nothing -> f (cStack ctx) + where + f [] = return def + f (s:rest) = + case get (stackState s) key of + Just v -> return v + Nothing -> f rest + +-- Check if the current context fulfills a StateDependency if entering `entry` +fulfillsDependency ctx entry dep = + case dep of + DepState scope name val -> (== (val, scope)) <$> peek scope ctx name + DepProperties scope name props -> do + (state, s) <- peek scope ctx name + return $ scope == s && variableProperties state == props + DepFunction name val -> (== val) <$> peekFunc ctx name + -- Hack. Since we haven't pushed the soon-to-be invoked function on the stack, + -- it won't be found by the normal check. + DepIsRecursive node val | node == entry -> return True + DepIsRecursive node val -> return $ val == any (\f -> entryPoint f == node) (cStack ctx) + DepExitCodes val -> (== val) <$> peekStack (\s k -> sExitCodes s) S.empty ctx () + -- _ -> error $ "Unknown dep " ++ show dep + where + peek scope = peekStack getVariableWithScope $ if scope == GlobalScope then (unknownVariableState, GlobalScope) else (unsetVariableState, LocalScope) + peekFunc = peekStack (\state name -> vmLookup name $ sFunctionTargets state) unknownFunctionValue + +-- Check if the current context fulfills all StateDependencies +fulfillsDependencies ctx entry deps = + f $ S.toList deps + where + f [] = return True + f (dep:rest) = do + res <- fulfillsDependency ctx entry dep + if res + then f rest + else return False + +-- Create a brand new Ctx given a Control Flow Graph (CFG) +newCtx g = do + c <- newSTRef 1 + input <- newSTRef undefined + output <- newSTRef undefined + node <- newSTRef undefined + cache <- newSTRef M.empty + enableCache <- newSTRef True + invocations <- newSTRef M.empty + return $ Ctx { + cCounter = c, + cInput = input, + cOutput = output, + cNode = node, + cCache = cache, + cEnableCache = enableCache, + cStack = [], + cInvocations = invocations, + cGraph = g + } + +-- The next incrementing version for VersionedMaps +nextVersion ctx = do + let ctr = cCounter ctx + n <- readSTRef ctr + writeSTRef ctr $! n+1 + return n + +-- Create a new StackEntry +newStackEntry ctx point isCall = do + deps <- newSTRef S.empty + state <- readSTRef $ cOutput ctx + callsite <- readSTRef $ cNode ctx + return $ StackEntry { + entryPoint = point, + isFunctionCall = isCall, + callSite = callsite, + dependencies = deps, + stackState = state + } + +-- Call a function with a new stack entry on the stack +withNewStackFrame ctx node isCall f = do + newEntry <- newStackEntry ctx node isCall + newInput <- newSTRef newInternalState + newOutput <- newSTRef newInternalState + newNode <- newSTRef node + let newCtx = ctx { + cInput = newInput, + cOutput = newOutput, + cNode = newNode, + cStack = newEntry : cStack ctx + } + x <- f newCtx + + {- + deps <- readSTRef $ dependencies newEntry + selfcheck <- fulfillsDependencies newCtx deps + unless selfcheck $ error $ pleaseReport $ "Unmet stack dependencies on " ++ show (node, deps) + -} + + return (x, newEntry) + +-- Check if invoking this function would be a recursive loop +-- (i.e. we already have the function on the stack) +wouldBeRecursive ctx node = f (cStack ctx) + where + f [] = return False + f (s:rest) = do + res <- + if entryPoint s == node + then return True + else f rest + modifySTRef (dependencies s) $ S.insert $ DepIsRecursive node res + return res + +-- The main DFA 'transfer' function, applying the effects of a node to the output state +transfer ctx label = + --traceShow ("Transferring", label) $ + case label of + CFStructuralNode -> return () + CFEntryPoint _ -> return () + CFImpliedExit -> return () + CFResolvedExit {} -> return () + + CFExecuteCommand cmd -> transferCommand ctx cmd + CFExecuteSubshell reason entry exit -> transferSubshell ctx reason entry exit + CFApplyEffects effects -> mapM_ (\(IdTagged _ f) -> transferEffect ctx f) effects + CFSetExitCode id -> transferExitCode ctx id + + CFUnresolvedExit -> patchOutputM ctx unreachableState + CFUnreachable -> patchOutputM ctx unreachableState + + -- TODO + CFSetBackgroundPid _ -> return () + CFDropPrefixAssignments {} -> + modifySTRef (cOutput ctx) $ \c -> modified c { sPrefixValues = vmEmpty } +-- _ -> error $ "Unknown " ++ show label + + +-- Transfer the effects of a subshell invocation. This is similar to a function call +-- to allow easily discarding the effects (otherwise the InternalState would have +-- to represent subshell depth, while this way it can simply use the function stack). +transferSubshell ctx reason entry exit = do + let cout = cOutput ctx + initial <- readSTRef cout + runCached ctx entry (f entry exit) + res <- readSTRef cout + -- Clear subshell changes. TODO: track this to warn about modifications. + writeSTRef cout $ initial { + sExitCodes = sExitCodes res + } + where + f entry exit ctx = do + (states, frame) <- withNewStackFrame ctx entry False (flip dataflow $ entry) + let (_, res) = fromMaybe (error $ pleaseReport "Subshell has no exit") $ M.lookup exit states + deps <- readSTRef $ dependencies frame + registerFlowResult ctx entry states deps + return (deps, res) + +-- Transfer the effects of executing a command, i.e. the merged union of all possible function definitions. +transferCommand ctx Nothing = return () +transferCommand ctx (Just name) = do + targets <- readFunction ctx name + logVerbose ("Transferring ",name,targets) + transferMultiple ctx $ map (flip transferFunctionValue) $ S.toList targets + +-- Transfer a set of function definitions and merge the output states. +transferMultiple ctx funcs = do + logVerbose ("Transferring set of ", length funcs) + original <- readSTRef out + branches <- mapM (apply ctx original) funcs + merged <- mergeStates ctx original branches + let patched = patchState original merged + writeSTRef out patched + where + out = cOutput ctx + apply ctx original f = do + writeSTRef out original + f ctx + readSTRef out + +-- Transfer the effects of a single function definition. +transferFunctionValue ctx funcVal = + case funcVal of + FunctionUnknown -> return () + FunctionDefinition name entry exit -> do + isRecursive <- wouldBeRecursive ctx entry + if isRecursive + then return () -- TODO: Find a better strategy for recursion + else runCached ctx entry (f name entry exit) + where + f name entry exit ctx = do + (states, frame) <- withNewStackFrame ctx entry True (flip dataflow $ entry) + deps <- readSTRef $ dependencies frame + let res = + case M.lookup exit states of + Just (input, output) -> do + -- Discard local variables. TODO: track&retain variables declared local in previous scopes? + modified output { sLocalValues = vmEmpty } + Nothing -> do + -- e.g. f() { exit; } + unreachableState + registerFlowResult ctx entry states deps + return (deps, res) + +transferExitCode ctx id = do + modifySTRef (cOutput ctx) $ setExitCode id + +-- Register/save the result of a dataflow of a function. +-- At the end, all the different values from different flows are merged together. +registerFlowResult ctx entry states deps = do + -- This function is called in the context of a CFExecuteCommand and not its invoked function, + -- so manually add the current node to the stack. + current <- readSTRef $ cNode ctx + let parents = map callSite $ cStack ctx + -- A unique path to this flow context. The specific value doesn't matter, as long as it's + -- unique per invocation of the function. This is required so that 'x=1; f; x=2; f' won't + -- overwrite each other. + let path = entry : current : parents + modifySTRef (cInvocations ctx) $ M.insert path (deps, states) + + +-- Look up a node in the cache and see if the dependencies of any entries are matched. +-- In that case, reuse the previous result instead of doing a new data flow. +runCached :: forall s. Ctx s -> Node -> (Ctx s -> ST s (S.Set StateDependency, InternalState)) -> ST s () +runCached ctx node f = do + cache <- getCache ctx node + case cache of + Just v -> do + logInfo ("Running cached", node) + -- do { (deps, diff) <- f ctx; unless (v == diff) $ traceShowM ("Cache FAILED to match actual result", node, deps, diff); } + patchOutputM ctx v + + Nothing -> do + logInfo ("Cache failed", node) + (deps, diff) <- f ctx + modifySTRef (cCache ctx) (M.insertWith (\_ old -> (deps, diff):(take cacheEntries old)) node [(deps,diff)]) + logVerbose ("Recomputed cache for", node, deps) + -- do { f <- fulfillsDependencies ctx node deps; unless (f) $ traceShowM ("New dependencies FAILED to match", node, deps); } + patchOutputM ctx diff + +-- Get a cached version whose dependencies are currently fulfilled, if any. +getCache :: forall s. Ctx s -> Node -> ST s (Maybe InternalState) +getCache ctx node = do + cache <- readSTRef $ cCache ctx + enable <- readSTRef $ cEnableCache ctx + logVerbose ("Cache for", node, "length", length $ M.findWithDefault [] node cache, M.lookup node cache) + if enable + then f $ M.findWithDefault [] node cache + else return Nothing + where + f [] = return Nothing + f ((deps, value):rest) = do + match <- fulfillsDependencies ctx node deps + if match + then return $ Just value + else f rest + +-- Transfer a single CFEffect to the output state. +transferEffect ctx effect = + case effect of + CFReadVariable name -> + case name of + "?" -> void $ readExitCodes ctx + _ -> void $ readVariable ctx name + CFWriteVariable name value -> do + val <- cfValueToVariableValue ctx value + updateVariableValue ctx name val + CFWriteGlobal name value -> do + val <- cfValueToVariableValue ctx value + updateGlobalValue ctx name val + CFWriteLocal name value -> do + val <- cfValueToVariableValue ctx value + updateLocalValue ctx name val + CFWritePrefix name value -> do + val <- cfValueToVariableValue ctx value + updatePrefixValue ctx name val + + CFSetProps scope name props -> + case scope of + Nothing -> do + state <- readVariable ctx name + writeVariable ctx name $ addProperties props state + Just GlobalScope -> do + state <- readGlobal ctx name + writeGlobal ctx name $ addProperties props state + Just LocalScope -> do + out <- readSTRef (cOutput ctx) + state <- readLocal ctx name + writeLocal ctx name $ addProperties props state + Just PrefixScope -> do + -- Prefix values become local + state <- readLocal ctx name + writeLocal ctx name $ addProperties props state + + CFUnsetProps scope name props -> + case scope of + Nothing -> do + state <- readVariable ctx name + writeVariable ctx name $ removeProperties props state + Just GlobalScope -> do + state <- readGlobal ctx name + writeGlobal ctx name $ removeProperties props state + Just LocalScope -> do + out <- readSTRef (cOutput ctx) + state <- readLocal ctx name + writeLocal ctx name $ removeProperties props state + Just PrefixScope -> do + -- Prefix values become local + state <- readLocal ctx name + writeLocal ctx name $ removeProperties props state + + + CFUndefineVariable name -> undefineVariable ctx name + CFUndefineFunction name -> undefineFunction ctx name + CFUndefine name -> do + -- This should really just unset one or the other + undefineVariable ctx name + undefineFunction ctx name + CFDefineFunction name id entry exit -> + writeFunction ctx name $ FunctionDefinition name entry exit + + -- TODO + CFUndefineNameref name -> undefineVariable ctx name + CFHintArray name -> return () + CFHintDefined name -> return () +-- _ -> error $ "Unknown effect " ++ show effect + + +-- Transfer the CFG's idea of a value into our VariableState +cfValueToVariableValue ctx val = + case val of + CFValueArray -> return unknownVariableValue -- TODO: Track array status + CFValueComputed _ parts -> foldM f emptyVariableValue parts + CFValueInteger -> return unknownIntegerValue + CFValueString -> return unknownVariableValue + CFValueUninitialized -> return emptyVariableValue +-- _ -> error $ "Unknown value: " ++ show val + where + f val part = do + next <- computeValue ctx part + return $ val `appendVariableValue` next + +-- A value can be computed from 0 or more parts, such as x="literal$y$z" +computeValue ctx part = + case part of + CFStringLiteral str -> return $ literalToVariableValue str + CFStringInteger -> return unknownIntegerValue + CFStringUnknown -> return unknownVariableValue + CFStringVariable name -> variableStateToValue <$> readVariable ctx name + where + variableStateToValue state = + case () of + _ | all (CFVPInteger `S.member`) $ variableProperties state -> unknownIntegerValue + _ -> variableValue state + +-- Append two VariableValues as if with z="$x$y" +appendVariableValue :: VariableValue -> VariableValue -> VariableValue +appendVariableValue a b = + unknownVariableValue { + literalValue = liftM2 (++) (literalValue a) (literalValue b), + spaceStatus = appendSpaceStatus (spaceStatus a) (spaceStatus b), + numericalStatus = appendNumericalStatus (numericalStatus a) (numericalStatus b) + } + +appendSpaceStatus a b = + case (a,b) of + (SpaceStatusEmpty, _) -> b + (_, SpaceStatusEmpty) -> a + (SpaceStatusClean, SpaceStatusClean) -> a + _ ->SpaceStatusDirty + +appendNumericalStatus a b = + case (a,b) of + (NumericalStatusEmpty, x) -> x + (x, NumericalStatusEmpty) -> x + (NumericalStatusDefinitely, NumericalStatusDefinitely) -> NumericalStatusDefinitely + (NumericalStatusUnknown, _) -> NumericalStatusUnknown + (_, NumericalStatusUnknown) -> NumericalStatusUnknown + _ -> NumericalStatusMaybe + +unknownIntegerValue = unknownVariableValue { + literalValue = Nothing, + spaceStatus = SpaceStatusClean, + numericalStatus = NumericalStatusDefinitely +} + +literalToVariableValue str = unknownVariableValue { + literalValue = Just str, + spaceStatus = literalToSpaceStatus str, + numericalStatus = literalToNumericalStatus str +} + +withoutChanges ctx f = do + let inp = cInput ctx + let out = cOutput ctx + prevInput <- readSTRef inp + prevOutput <- readSTRef out + res <- f + writeSTRef inp prevInput + writeSTRef out prevOutput + return res + +-- Get the SpaceStatus for a literal string, i.e. if it needs quoting +literalToSpaceStatus str = + case str of + "" -> SpaceStatusEmpty + _ | all (`notElem` " \t\n*?[") str -> SpaceStatusClean + _ -> SpaceStatusDirty + +-- Get the NumericalStatus for a literal string, i.e. whether it's an integer +literalToNumericalStatus str = + case str of + "" -> NumericalStatusEmpty + '-':rest -> if isNumeric rest then NumericalStatusDefinitely else NumericalStatusUnknown + rest -> if isNumeric rest then NumericalStatusDefinitely else NumericalStatusUnknown + where + isNumeric = all isDigit + +type StateMap = M.Map Node (InternalState, InternalState) + +-- Classic, iterative Data Flow Analysis. See Wikipedia for a description of the process. +dataflow :: forall s. Ctx s -> Node -> ST s StateMap +dataflow ctx entry = do + pending <- newSTRef $ S.singleton entry + states <- newSTRef $ M.empty + -- Should probably be done via a stack frame instead + withoutChanges ctx $ + f iterationCount pending states + readSTRef states + where + graph = cGraph ctx + f 0 _ _ = error $ pleaseReport "DFA did not reach fix point" + f n pending states = do + ps <- readSTRef pending + + when (n == fallbackThreshold) $ do + -- This should never happen, but has historically been due to caching bugs. + -- Try disabling the cache and continuing. + logInfo "DFA is not stabilizing! Disabling cache." + writeSTRef (cEnableCache ctx) False + + if S.null ps + then return () + else do + let (next, rest) = S.deleteFindMin ps + nexts <- process states next + writeSTRef pending $ S.union (S.fromList nexts) rest + f (n-1) pending states + + process states node = do + stateMap <- readSTRef states + let inputs = filter (\c -> sIsReachable c /= Just False) $ mapMaybe (\c -> fmap snd $ M.lookup c stateMap) incoming + input <- + case incoming of + [] -> return newInternalState + _ -> + case inputs of + [] -> return unreachableState + (x:rest) -> foldM (mergeState ctx) x rest + writeSTRef (cInput ctx) $ input + writeSTRef (cOutput ctx) $ input + writeSTRef (cNode ctx) $ node + transfer ctx label + newOutput <- readSTRef $ cOutput ctx + result <- + if is2plus outgoing + then + -- Version the state because we split and will probably merge later + versionState ctx newOutput + else return newOutput + writeSTRef states $ M.insert node (input, result) stateMap + case M.lookup node stateMap of + Nothing -> return outgoing + Just (oldInput, oldOutput) -> + if oldOutput == result + then return [] + else return outgoing + where + (incomingL, _, label, outgoingL) = context graph $ node + incoming = map snd $ filter isRegular $ incomingL + outgoing = map snd outgoingL + isRegular = ((== CFEFlow) . fst) + +runRoot ctx env entry exit = do + writeSTRef (cInput ctx) $ env + writeSTRef (cOutput ctx) $ env + writeSTRef (cNode ctx) $ entry + (states, frame) <- withNewStackFrame ctx entry False $ \c -> dataflow c entry + deps <- readSTRef $ dependencies frame + registerFlowResult ctx entry states deps + -- Return the final state, used to invoke functions that were declared but not invoked + return $ snd $ fromMaybe (error $ pleaseReport "Missing exit state") $ M.lookup exit states + + +analyzeControlFlow :: CFGParameters -> Token -> CFGAnalysis +analyzeControlFlow params t = + let + cfg = buildGraph params t + (entry, exit) = M.findWithDefault (error $ pleaseReport "Missing root") (getId t) (cfIdToRange cfg) + in + runST $ f cfg entry exit + where + f cfg entry exit = do + let env = createEnvironmentState + ctx <- newCtx $ cfGraph cfg + -- Do a dataflow analysis starting on the root node + exitState <- runRoot ctx env entry exit + + -- All nodes we've touched + invocations <- readSTRef $ cInvocations ctx + let invokedNodes = M.fromSet (const ()) $ S.unions $ map (M.keysSet . snd) $ M.elems invocations + + -- Invoke all functions that were declared but not invoked + -- This is so that we still get warnings for dead code + -- (it's probably not actually dead, just used by a script that sources ours) + let declaredFunctions = getFunctionTargets exitState + let uninvoked = M.difference declaredFunctions invokedNodes + + let stragglerInput = + (env `patchState` exitState) { + -- We don't want `die() { exit $?; }; echo "Sourced"` to assume $? is always echo + sExitCodes = Nothing + } + + analyzeStragglers ctx stragglerInput uninvoked + + -- Now round up all the states from all data flows + -- (FIXME: this excludes functions that were defined in straggling functions) + invocations <- readSTRef $ cInvocations ctx + invokedStates <- flattenByNode ctx $ groupByNode $ M.map addDeps invocations + + -- Fill in the map with unreachable states for anything we didn't get to + let baseStates = M.fromDistinctAscList $ map (\c -> (c, (unreachableState, unreachableState))) $ uncurry enumFromTo $ nodeRange $ cfGraph cfg + let allStates = M.union invokedStates baseStates + + -- Convert to external states + let nodeToData = M.map (\(a,b) -> (internalToExternal a, internalToExternal b)) allStates + + return $ nodeToData `deepseq` CFGAnalysis { + graph = cfGraph cfg, + tokenToRange = cfIdToRange cfg, + tokenToNodes = cfIdToNodes cfg, + nodeToData = nodeToData, + postDominators = cfPostDominators cfg + } + + + -- Include the dependencies in the state of each function, e.g. if it depends on `x=foo` then add that. + addDeps :: (S.Set StateDependency, M.Map Node (InternalState, InternalState)) -> M.Map Node (InternalState, InternalState) + addDeps (deps, m) = let base = depsToState deps in M.map (\(a,b) -> (base `patchState` a, base `patchState` b)) m + + -- Collect all the states that each node has resulted in. + groupByNode :: forall k v. M.Map k (M.Map Node v) -> M.Map Node [v] + groupByNode pathMap = M.fromListWith (++) $ map (\(k,v) -> (k,[v])) $ concatMap M.toList $ M.elems pathMap + + -- Merge all the pre/post states for each node. This would have been a foldM if Map had one. + flattenByNode ctx m = M.fromDistinctAscList <$> (mapM (mergePair ctx) $ M.toList m) + + mergeAllStates ctx pairs = + let + (pres, posts) = unzip pairs + in do + pre <- mergeStates ctx (error $ pleaseReport "Null node states") pres + post <- mergeStates ctx (error $ pleaseReport "Null node states") posts + return (pre, post) + + mergePair ctx (node, list) = do + merged <- mergeAllStates ctx list + return (node, merged) + + -- Get the all the functions defined in an InternalState + getFunctionTargets :: InternalState -> M.Map Node FunctionDefinition + getFunctionTargets state = + let + declaredFuncs = S.unions $ M.elems $ mapStorage $ sFunctionTargets state + getFunc d = + case d of + FunctionDefinition _ entry _ -> Just (entry, d) + _ -> Nothing + funcs = mapMaybe getFunc $ S.toList declaredFuncs + in + M.fromList funcs + + +analyzeStragglers ctx state stragglers = do + mapM_ f $ M.elems stragglers + where + f def@(FunctionDefinition name entry exit) = do + writeSTRef (cInput ctx) state + writeSTRef (cOutput ctx) state + writeSTRef (cNode ctx) entry + transferFunctionValue ctx def + + + +return [] +runTests = $quickCheckAll diff --git a/src/ShellCheck/Checks/Commands.hs b/src/ShellCheck/Checks/Commands.hs index 2c8b00a..c37a67d 100644 --- a/src/ShellCheck/Checks/Commands.hs +++ b/src/ShellCheck/Checks/Commands.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2022 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -19,27 +19,38 @@ -} {-# LANGUAGE TemplateHaskell #-} {-# LANGUAGE FlexibleContexts #-} +{-# LANGUAGE MultiWayIf #-} +{-# LANGUAGE PatternGuards #-} -- This module contains checks that examine specific commands by name. -module ShellCheck.Checks.Commands (checker , ShellCheck.Checks.Commands.runTests) where +module ShellCheck.Checks.Commands (checker, optionalChecks, ShellCheck.Checks.Commands.runTests) where import ShellCheck.AST import ShellCheck.ASTLib import ShellCheck.AnalyzerLib +import ShellCheck.CFG +import qualified ShellCheck.CFGAnalysis as CF import ShellCheck.Data import ShellCheck.Interface import ShellCheck.Parser +import ShellCheck.Prelude import ShellCheck.Regex import Control.Monad import Control.Monad.RWS import Data.Char +import Data.Functor.Identity +import qualified Data.Graph.Inductive.Graph as G import Data.List import Data.Maybe -import qualified Data.Map.Strict as Map +import qualified Data.List.NonEmpty as NE +import qualified Data.Map.Strict as M +import qualified Data.Set as S import Test.QuickCheck.All (forAllProperties) import Test.QuickCheck.Test (quickCheckWithResult, stdArgs, maxSuccess) +import Debug.Trace -- STRIP + data CommandName = Exactly String | Basename String deriving (Eq, Ord) @@ -51,16 +62,15 @@ verify :: CommandCheck -> String -> Bool verify f s = producesComments (getChecker [f]) s == Just True verifyNot f s = producesComments (getChecker [f]) s == Just False -arguments (T_SimpleCommand _ _ (cmd:args)) = args - commandChecks :: [CommandCheck] commandChecks = [ checkTr ,checkFindNameGlob - ,checkNeedlessExpr + ,checkExpr ,checkGrepRe ,checkTrapQuotes ,checkReturn + ,checkExit ,checkFindExecWithSingleArgument ,checkUnusedEchoEscapes ,checkInjectableFindSh @@ -89,28 +99,99 @@ commandChecks = [ ,checkMvArguments, checkCpArguments, checkLnArguments ,checkFindRedirections ,checkReadExpansions - ,checkWhich ,checkSudoRedirect ,checkSudoArgs + ,checkSourceArgs + ,checkChmodDashr + ,checkXargsDashi + ,checkUnquotedEchoSpaces + ,checkEvalArray ] + ++ map checkArgComparison ("alias" : declaringCommands) + ++ map checkMaskedReturns declaringCommands + ++ map checkMultipleDeclaring declaringCommands + ++ map checkBackreferencingDeclaration declaringCommands -buildCommandMap :: [CommandCheck] -> Map.Map CommandName (Token -> Analysis) -buildCommandMap = foldl' addCheck Map.empty + +optionalChecks = map fst optionalCommandChecks +optionalCommandChecks :: [(CheckDescription, CommandCheck)] +optionalCommandChecks = [ + (newCheckDescription { + cdName = "deprecate-which", + cdDescription = "Suggest 'command -v' instead of 'which'", + cdPositive = "which javac", + cdNegative = "command -v javac" + }, checkWhich) + ] +optionalCheckMap = M.fromList $ map (\(desc, check) -> (cdName desc, check)) optionalCommandChecks + +prop_verifyOptionalExamples = all check optionalCommandChecks + where + check (desc, check) = + verify check (cdPositive desc) + && verifyNot check (cdNegative desc) + +-- Run a check against the getopt parser. If it fails, the lists are empty. +checkGetOpts str flags args f = + flags == actualFlags && args == actualArgs + where + toTokens = map (T_Literal (Id 0)) . words + opts = fromMaybe [] $ f (toTokens str) + actualFlags = filter (not . null) $ map fst opts + actualArgs = [onlyLiteralString x | ("", (_, x)) <- opts] + +-- Short options +prop_checkGetOptsS1 = checkGetOpts "-f x" ["f"] [] $ getOpts (True, True) "f:" [] +prop_checkGetOptsS2 = checkGetOpts "-fx" ["f"] [] $ getOpts (True, True) "f:" [] +prop_checkGetOptsS3 = checkGetOpts "-f -x" ["f", "x"] [] $ getOpts (True, True) "fx" [] +prop_checkGetOptsS4 = checkGetOpts "-f -x" ["f"] [] $ getOpts (True, True) "f:" [] +prop_checkGetOptsS5 = checkGetOpts "-fx" [] [] $ getOpts (True, True) "fx:" [] + +prop_checkGenericOptsS1 = checkGetOpts "-f x" ["f"] [] $ return . getGenericOpts +prop_checkGenericOptsS2 = checkGetOpts "-abc x" ["a", "b", "c"] [] $ return . getGenericOpts +prop_checkGenericOptsS3 = checkGetOpts "-abc -x" ["a", "b", "c", "x"] [] $ return . getGenericOpts +prop_checkGenericOptsS4 = checkGetOpts "-x" ["x"] [] $ return . getGenericOpts + +-- Long options +prop_checkGetOptsL1 = checkGetOpts "--foo=bar baz" ["foo"] ["baz"] $ getOpts (True, False) "" [("foo", True)] +prop_checkGetOptsL2 = checkGetOpts "--foo bar baz" ["foo"] ["baz"] $ getOpts (True, False) "" [("foo", True)] +prop_checkGetOptsL3 = checkGetOpts "--foo baz" ["foo"] ["baz"] $ getOpts (True, True) "" [] +prop_checkGetOptsL4 = checkGetOpts "--foo baz" [] [] $ getOpts (True, False) "" [] + +prop_checkGenericOptsL1 = checkGetOpts "--foo=bar" ["foo"] [] $ return . getGenericOpts +prop_checkGenericOptsL2 = checkGetOpts "--foo bar" ["foo"] ["bar"] $ return . getGenericOpts +prop_checkGenericOptsL3 = checkGetOpts "-x --foo" ["x", "foo"] [] $ return . getGenericOpts + +-- Know when to terminate +prop_checkGetOptsT1 = checkGetOpts "-a x -b" ["a", "b"] ["x"] $ getOpts (True, True) "ab" [] +prop_checkGetOptsT2 = checkGetOpts "-a x -b" ["a"] ["x","-b"] $ getOpts (False, True) "ab" [] +prop_checkGetOptsT3 = checkGetOpts "-a -- -b" ["a"] ["-b"] $ getOpts (True, True) "ab" [] +prop_checkGetOptsT4 = checkGetOpts "-a -- -b" ["a", "b"] [] $ getOpts (True, True) "a:b" [] + +prop_checkGenericOptsT1 = checkGetOpts "-x -- -y" ["x"] ["-y"] $ return . getGenericOpts +prop_checkGenericOptsT2 = checkGetOpts "-xy --" ["x", "y"] [] $ return . getGenericOpts + + +buildCommandMap :: [CommandCheck] -> M.Map CommandName (Token -> Analysis) +buildCommandMap = foldl' addCheck M.empty where addCheck map (CommandCheck name function) = - Map.insertWith composeAnalyzers name function map + M.insertWith composeAnalyzers name function map -checkCommand :: Map.Map CommandName (Token -> Analysis) -> Token -> Analysis -checkCommand map t@(T_SimpleCommand id _ (cmd:rest)) = fromMaybe (return ()) $ do +checkCommand :: M.Map CommandName (Token -> Analysis) -> Token -> Analysis +checkCommand map t@(T_SimpleCommand id cmdPrefix (cmd:rest)) = sequence_ $ do name <- getLiteralString cmd return $ - if '/' `elem` name - then - Map.findWithDefault nullCheck (Basename $ basename name) map t - else do - Map.findWithDefault nullCheck (Exactly name) map t - Map.findWithDefault nullCheck (Basename name) map t + if | '/' `elem` name -> + M.findWithDefault nullCheck (Basename $ basename name) map t + | name == "builtin", (h:_) <- rest -> + let t' = T_SimpleCommand id cmdPrefix rest + selectedBuiltin = onlyLiteralString h + in M.findWithDefault nullCheck (Exactly selectedBuiltin) map t' + | otherwise -> do + M.findWithDefault nullCheck (Exactly name) map t + M.findWithDefault nullCheck (Basename name) map t where basename = reverse . takeWhile (/= '/') . reverse @@ -125,23 +206,29 @@ getChecker list = Checker { map = buildCommandMap list -checker :: Parameters -> Checker -checker params = getChecker commandChecks +checker :: AnalysisSpec -> Parameters -> Checker +checker spec params = getChecker $ commandChecks ++ optionals + where + keys = asOptionalChecks spec + optionals = + if "all" `elem` keys + then map snd optionalCommandChecks + else mapMaybe (\x -> M.lookup x optionalCheckMap) keys prop_checkTr1 = verify checkTr "tr [a-f] [A-F]" prop_checkTr2 = verify checkTr "tr 'a-z' 'A-Z'" -prop_checkTr2a= verify checkTr "tr '[a-z]' '[A-Z]'" +prop_checkTr2a = verify checkTr "tr '[a-z]' '[A-Z]'" prop_checkTr3 = verifyNot checkTr "tr -d '[:lower:]'" -prop_checkTr3a= verifyNot checkTr "tr -d '[:upper:]'" -prop_checkTr3b= verifyNot checkTr "tr -d '|/_[:upper:]'" +prop_checkTr3a = verifyNot checkTr "tr -d '[:upper:]'" +prop_checkTr3b = verifyNot checkTr "tr -d '|/_[:upper:]'" prop_checkTr4 = verifyNot checkTr "ls [a-z]" prop_checkTr5 = verify checkTr "tr foo bar" prop_checkTr6 = verify checkTr "tr 'hello' 'world'" prop_checkTr8 = verifyNot checkTr "tr aeiou _____" prop_checkTr9 = verifyNot checkTr "a-z n-za-m" -prop_checkTr10= verifyNot checkTr "tr --squeeze-repeats rl lr" -prop_checkTr11= verifyNot checkTr "tr abc '[d*]'" -prop_checkTr12= verifyNot checkTr "tr '[=e=]' 'e'" +prop_checkTr10 = verifyNot checkTr "tr --squeeze-repeats rl lr" +prop_checkTr11 = verifyNot checkTr "tr abc '[d*]'" +prop_checkTr12 = verifyNot checkTr "tr '[=e=]' 'e'" checkTr = CommandCheck (Basename "tr") (mapM_ f . arguments) where f w | isGlob w = -- The user will go [ab] -> '[ab]' -> 'ab'. Fixme? @@ -166,30 +253,83 @@ prop_checkFindNameGlob1 = verify checkFindNameGlob "find / -name *.php" prop_checkFindNameGlob2 = verify checkFindNameGlob "find / -type f -ipath *(foo)" prop_checkFindNameGlob3 = verifyNot checkFindNameGlob "find * -name '*.php'" checkFindNameGlob = CommandCheck (Basename "find") (f . arguments) where - acceptsGlob (Just s) = s `elem` [ "-ilname", "-iname", "-ipath", "-iregex", "-iwholename", "-lname", "-name", "-path", "-regex", "-wholename" ] - acceptsGlob _ = False + acceptsGlob s = s `elem` [ "-ilname", "-iname", "-ipath", "-iregex", "-iwholename", "-lname", "-name", "-path", "-regex", "-wholename" ] f [] = return () - f [x] = return () - f (a:b:r) = do - when (acceptsGlob (getLiteralString a) && isGlob b) $ do - let (Just s) = getLiteralString a + f (x:xs) = foldr g (const $ return ()) xs x + g b acc a = do + forM_ (getLiteralString a) $ \s -> when (acceptsGlob s && isGlob b) $ warn (getId b) 2061 $ "Quote the parameter to " ++ s ++ " so the shell won't interpret it." - f (b:r) + acc b -prop_checkNeedlessExpr = verify checkNeedlessExpr "foo=$(expr 3 + 2)" -prop_checkNeedlessExpr2 = verify checkNeedlessExpr "foo=`echo \\`expr 3 + 2\\``" -prop_checkNeedlessExpr3 = verifyNot checkNeedlessExpr "foo=$(expr foo : regex)" -prop_checkNeedlessExpr4 = verifyNot checkNeedlessExpr "foo=$(expr foo \\< regex)" -checkNeedlessExpr = CommandCheck (Basename "expr") f where - f t = +prop_checkExpr = verify checkExpr "foo=$(expr 3 + 2)" +prop_checkExpr2 = verify checkExpr "foo=`echo \\`expr 3 + 2\\``" +prop_checkExpr3 = verifyNot checkExpr "foo=$(expr foo : regex)" +prop_checkExpr4 = verifyNot checkExpr "foo=$(expr foo \\< regex)" +prop_checkExpr5 = verify checkExpr "# shellcheck disable=SC2003\nexpr match foo bar" +prop_checkExpr6 = verify checkExpr "# shellcheck disable=SC2003\nexpr foo : fo*" +prop_checkExpr7 = verify checkExpr "# shellcheck disable=SC2003\nexpr 5 -3" +prop_checkExpr8 = verifyNot checkExpr "# shellcheck disable=SC2003\nexpr \"$@\"" +prop_checkExpr9 = verifyNot checkExpr "# shellcheck disable=SC2003\nexpr 5 $rest" +prop_checkExpr10 = verify checkExpr "# shellcheck disable=SC2003\nexpr length \"$var\"" +prop_checkExpr11 = verify checkExpr "# shellcheck disable=SC2003\nexpr foo > bar" +prop_checkExpr12 = verify checkExpr "# shellcheck disable=SC2003\nexpr 1 | 2" +prop_checkExpr13 = verify checkExpr "# shellcheck disable=SC2003\nexpr 1 * 2" +prop_checkExpr14 = verify checkExpr "# shellcheck disable=SC2003\nexpr \"$x\" >= \"$y\"" + +checkExpr = CommandCheck (Basename "expr") f where + f t = do when (all (`notElem` exceptions) (words $ arguments t)) $ style (getId $ getCommandTokenOrThis t) 2003 "expr is antiquated. Consider rewriting this using $((..)), ${} or [[ ]]." + + case arguments t of + [lhs, op, rhs] -> do + checkOp lhs + case getWordParts op of + [T_Glob _ "*"] -> + err (getId op) 2304 + "* must be escaped to multiply: \\*. Modern $((x * y)) avoids this issue." + [T_Literal _ ":"] | isGlob rhs -> + warn (getId rhs) 2305 + "Quote regex argument to expr to avoid it expanding as a glob." + _ -> return () + + [single] | not (willSplit single) -> + warn (getId single) 2307 + "'expr' expects 3+ arguments but sees 1. Make sure each operator/operand is a separate argument, and escape <>&|." + + [first, second] | + onlyLiteralString first /= "length" + && not (willSplit first || willSplit second) -> do + checkOp first + warn (getId t) 2307 + "'expr' expects 3+ arguments, but sees 2. Make sure each operator/operand is a separate argument, and escape <>&|." + + (first:rest) -> do + checkOp first + forM_ rest $ \t -> + -- We already find 95%+ of multiplication and regex earlier, so don't bother classifying this further. + when (isGlob t) $ warn (getId t) 2306 "Escape glob characters in arguments to expr to avoid pathname expansion." + + _ -> return () + -- These operators are hard to replicate in POSIX - exceptions = [ ":", "<", ">", "<=", ">=" ] + exceptions = [ ":", "<", ">", "<=", ">=", + -- We can offer better suggestions for these + "match", "length", "substr", "index"] words = mapMaybe getLiteralString + checkOp side = + case getLiteralString side of + Just "match" -> msg "'expr match' has unspecified results. Prefer 'expr str : regex'." + Just "length" -> msg "'expr length' has unspecified results. Prefer ${#var}." + Just "substr" -> msg "'expr substr' has unspecified results. Prefer 'cut' or ${var#???}." + Just "index" -> msg "'expr index' has unspecified results. Prefer x=${var%%[chars]*}; $((${#x}+1))." + _ -> return () + where + msg = info (getId side) 2308 + prop_checkGrepRe1 = verify checkGrepRe "cat foo | grep *.mp3" prop_checkGrepRe2 = verify checkGrepRe "grep -Ev cow*test *.mp3" @@ -200,23 +340,30 @@ prop_checkGrepRe6 = verifyNot checkGrepRe "grep foo \\*.mp3" prop_checkGrepRe7 = verify checkGrepRe "grep *foo* file" prop_checkGrepRe8 = verify checkGrepRe "ls | grep foo*.jpg" prop_checkGrepRe9 = verifyNot checkGrepRe "grep '[0-9]*' file" -prop_checkGrepRe10= verifyNot checkGrepRe "grep '^aa*' file" -prop_checkGrepRe11= verifyNot checkGrepRe "grep --include=*.png foo" -prop_checkGrepRe12= verifyNot checkGrepRe "grep -F 'Foo*' file" -prop_checkGrepRe13= verifyNot checkGrepRe "grep -- -foo bar*" -prop_checkGrepRe14= verifyNot checkGrepRe "grep -e -foo bar*" -prop_checkGrepRe15= verifyNot checkGrepRe "grep --regex -foo bar*" +prop_checkGrepRe10 = verifyNot checkGrepRe "grep '^aa*' file" +prop_checkGrepRe11 = verifyNot checkGrepRe "grep --include=*.png foo" +prop_checkGrepRe12 = verifyNot checkGrepRe "grep -F 'Foo*' file" +prop_checkGrepRe13 = verifyNot checkGrepRe "grep -- -foo bar*" +prop_checkGrepRe14 = verifyNot checkGrepRe "grep -e -foo bar*" +prop_checkGrepRe15 = verifyNot checkGrepRe "grep --regex -foo bar*" +prop_checkGrepRe16 = verifyNot checkGrepRe "grep --include 'Foo*' file" +prop_checkGrepRe17 = verifyNot checkGrepRe "grep --exclude 'Foo*' file" +prop_checkGrepRe18 = verifyNot checkGrepRe "grep --exclude-dir 'Foo*' file" +prop_checkGrepRe19 = verify checkGrepRe "grep -- 'Foo*' file" +prop_checkGrepRe20 = verifyNot checkGrepRe "grep --fixed-strings 'Foo*' file" +prop_checkGrepRe21 = verifyNot checkGrepRe "grep -o 'x*' file" +prop_checkGrepRe22 = verifyNot checkGrepRe "grep --only-matching 'x*' file" +prop_checkGrepRe23 = verifyNot checkGrepRe "grep '.*' file" checkGrepRe = CommandCheck (Basename "grep") check where check cmd = f cmd (arguments cmd) -- --regex=*(extglob) doesn't work. Fixme? - skippable (Just s) = not ("--regex=" `isPrefixOf` s) && "-" `isPrefixOf` s - skippable _ = False + skippable s = not ("--regex=" `isPrefixOf` s) && "-" `isPrefixOf` s f _ [] = return () f cmd (x:r) = - let str = getLiteralStringExt (const $ return "_") x + let str = getLiteralStringDef "_" x in - if str `elem` [Just "--", Just "-e", Just "--regex"] + if str `elem` ["--", "-e", "--regex"] then checkRE cmd r -- Regex is *after* this else if skippable str @@ -228,36 +375,33 @@ checkGrepRe = CommandCheck (Basename "grep") check where when (isGlob re) $ warn (getId re) 2062 "Quote the grep pattern so the shell won't interpret it." - unless (cmd `hasFlag` "F") $ do + unless (any (`elem` flags) grepGlobFlags) $ do let string = concat $ oversimplify re if isConfusedGlobRegex string then warn (getId re) 2063 "Grep uses regex, but this looks like a glob." - else potentially $ do + else sequence_ $ do char <- getSuspiciousRegexWildcard string return $ info (getId re) 2022 $ "Note that unlike globs, " ++ [char] ++ "* here matches '" ++ [char, char, char] ++ "' but not '" ++ wordStartingWith char ++ "'." + where + flags = map snd $ getAllFlags cmd + grepGlobFlags = ["fixed-strings", "F", "include", "exclude", "exclude-dir", "o", "only-matching"] wordStartingWith c = - head . filter ([c] `isPrefixOf`) $ candidates + headOrDefault (c:"test") . filter ([c] `isPrefixOf`) $ candidates where candidates = - sampleWords ++ map (\(x:r) -> toUpper x : r) sampleWords ++ [c:"test"] + sampleWords ++ map (\(x:r) -> toUpper x : r) sampleWords - getSuspiciousRegexWildcard str = - if not $ str `matches` contra - then do - match <- matchRegex suspicious str - str <- match !!! 0 - str !!! 0 - else - fail "looks good" - where - suspicious = mkRegex "([A-Za-z1-9])\\*" - contra = mkRegex "[^a-zA-Z1-9]\\*|[][^$+\\\\]" + getSuspiciousRegexWildcard str = case matchRegex suspicious str of + Just [[c]] | not (str `matches` contra) -> Just c + _ -> fail "looks good" + suspicious = mkRegex "([A-Za-z1-9])\\*" + contra = mkRegex "[^a-zA-Z1-9]\\*|[][^$+\\\\]" prop_checkTrapQuotes1 = verify checkTrapQuotes "trap \"echo $num\" INT" -prop_checkTrapQuotes1a= verify checkTrapQuotes "trap \"echo `ls`\" INT" +prop_checkTrapQuotes1a = verify checkTrapQuotes "trap \"echo `ls`\" INT" prop_checkTrapQuotes2 = verifyNot checkTrapQuotes "trap 'echo $num' INT" prop_checkTrapQuotes3 = verify checkTrapQuotes "trap \"echo $((1+num))\" EXIT DEBUG" checkTrapQuotes = CommandCheck (Exactly "trap") (f . arguments) where @@ -268,7 +412,7 @@ checkTrapQuotes = CommandCheck (Exactly "trap") (f . arguments) where warning id = warn id 2064 "Use single quotes, otherwise this expands now rather than when signalled." checkExpansions (T_DollarExpansion id _) = warning id checkExpansions (T_Backticked id _) = warning id - checkExpansions (T_DollarBraced id _) = warning id + checkExpansions (T_DollarBraced id _ _) = warning id checkExpansions (T_DollarArithmetic id _) = warning id checkExpansions _ = return () @@ -280,21 +424,34 @@ prop_checkReturn4 = verifyNot checkReturn "return $((a|b))" prop_checkReturn5 = verify checkReturn "return -1" prop_checkReturn6 = verify checkReturn "return 1000" prop_checkReturn7 = verify checkReturn "return 'hello world'" -checkReturn = CommandCheck (Exactly "return") (f . arguments) +checkReturn = CommandCheck (Exactly "return") (returnOrExit + (\c -> err c 2151 "Only one integer 0-255 can be returned. Use stdout for other data.") + (\c -> err c 2152 "Can only return 0-255. Other data should be written to stdout.")) + +prop_checkExit1 = verifyNot checkExit "exit" +prop_checkExit2 = verifyNot checkExit "exit 1" +prop_checkExit3 = verifyNot checkExit "exit $var" +prop_checkExit4 = verifyNot checkExit "exit $((a|b))" +prop_checkExit5 = verify checkExit "exit -1" +prop_checkExit6 = verify checkExit "exit 1000" +prop_checkExit7 = verify checkExit "exit 'hello world'" +checkExit = CommandCheck (Exactly "exit") (returnOrExit + (\c -> err c 2241 "The exit status can only be one integer 0-255. Use stdout for other data.") + (\c -> err c 2242 "Can only exit with status 0-255. Other data should be written to stdout/stderr.")) + +returnOrExit multi invalid = (f . arguments) where f (first:second:_) = - err (getId second) 2151 - "Only one integer 0-255 can be returned. Use stdout for other data." + multi (getId first) f [value] = when (isInvalid $ literal value) $ - err (getId value) 2152 - "Can only return 0-255. Other data should be written to stdout." + invalid (getId value) f _ = return () - isInvalid s = s == "" || any (not . isDigit) s || length s > 5 + isInvalid s = null s || any (not . isDigit) s || length s > 5 || let value = (read s :: Integer) in value > 255 - literal token = fromJust $ getLiteralStringExt lit token + literal token = runIdentity $ getLiteralStringExt lit token lit (T_DollarBraced {}) = return "0" lit (T_DollarArithmetic {}) = return "0" lit (T_DollarExpansion {}) = return "0" @@ -311,7 +468,7 @@ checkFindExecWithSingleArgument = CommandCheck (Basename "find") (f . arguments) check (exec:arg:term:_) = do execS <- getLiteralString exec termS <- getLiteralString term - cmdS <- getLiteralStringExt (const $ return " ") arg + let cmdS = getLiteralStringDef " " arg guard $ execS `elem` ["-exec", "-execdir"] && termS `elem` [";", "+"] guard $ cmdS `matches` commandRegex @@ -325,9 +482,16 @@ prop_checkUnusedEchoEscapes2 = verifyNot checkUnusedEchoEscapes "echo -e 'foi\\n prop_checkUnusedEchoEscapes3 = verify checkUnusedEchoEscapes "echo \"n:\\t42\"" prop_checkUnusedEchoEscapes4 = verifyNot checkUnusedEchoEscapes "echo lol" prop_checkUnusedEchoEscapes5 = verifyNot checkUnusedEchoEscapes "echo -n -e '\n'" +prop_checkUnusedEchoEscapes6 = verify checkUnusedEchoEscapes "echo '\\506'" +prop_checkUnusedEchoEscapes7 = verify checkUnusedEchoEscapes "echo '\\5a'" +prop_checkUnusedEchoEscapes8 = verifyNot checkUnusedEchoEscapes "echo '\\8a'" +prop_checkUnusedEchoEscapes9 = verifyNot checkUnusedEchoEscapes "echo '\\d5a'" +prop_checkUnusedEchoEscapes10 = verify checkUnusedEchoEscapes "echo '\\x4a'" +prop_checkUnusedEchoEscapes11 = verify checkUnusedEchoEscapes "echo '\\xat'" +prop_checkUnusedEchoEscapes12 = verifyNot checkUnusedEchoEscapes "echo '\\xth'" checkUnusedEchoEscapes = CommandCheck (Basename "echo") f where - hasEscapes = mkRegex "\\\\[rnt]" + hasEscapes = mkRegex "\\\\([rntabefv\\']|[0-7]{1,3}|x([0-9]|[A-F]|[a-f]){1,2})" f cmd = whenShell [Sh, Bash, Ksh] $ unless (cmd `hasFlag` "e") $ @@ -407,10 +571,10 @@ prop_checkMkdirDashPM20 = verifyNot checkMkdirDashPM "mkdir -p -m 0755 .././bin" prop_checkMkdirDashPM21 = verifyNot checkMkdirDashPM "mkdir -p -m 0755 ../../bin" checkMkdirDashPM = CommandCheck (Basename "mkdir") check where - check t = potentially $ do + check t = sequence_ $ do let flags = getAllFlags t - dashP <- find ((\f -> f == "p" || f == "parents") . snd) flags - dashM <- find ((\f -> f == "m" || f == "mode") . snd) flags + dashP <- find (\(_,f) -> f == "p" || f == "parents") flags + dashM <- find (\(_,f) -> f == "m" || f == "mode") flags -- mkdir -pm 0700 dir is fine, so is ../dir, but dir/subdir is not. guard $ any couldHaveSubdirs (drop 1 $ arguments t) return $ warn (getId $ fst dashM) 2174 "When used with -p, -m only applies to the deepest directory." @@ -430,10 +594,10 @@ prop_checkNonportableSignals7 = verifyNot checkNonportableSignals "trap 'stop' i checkNonportableSignals = CommandCheck (Exactly "trap") (f . arguments) where f args = case args of - first:rest -> unless (isFlag first) $ mapM_ check rest + first:rest | not $ isFlag first -> mapM_ check rest _ -> return () - check param = potentially $ do + check param = sequence_ $ do str <- getLiteralString param let id = getId param return $ sequence_ $ mapMaybe (\f -> f id str) [ @@ -462,14 +626,14 @@ prop_checkInteractiveSu4 = verifyNot checkInteractiveSu "su root < script" checkInteractiveSu = CommandCheck (Basename "su") f where f cmd = when (length (arguments cmd) <= 1) $ do - path <- pathTo cmd + path <- getPathM cmd when (all undirected path) $ info (getId cmd) 2117 "To run commands as another user, use su -c or sudo." - undirected (T_Pipeline _ _ l) = length l <= 1 + undirected (T_Pipeline _ _ (_:_:_)) = False -- This should really just be modifications to stdin, but meh - undirected (T_Redirecting _ list _) = null list + undirected (T_Redirecting _ (_:_) _) = False undirected _ = True @@ -486,9 +650,8 @@ checkSshCommandString = CommandCheck (Basename "ssh") (f . arguments) ([], hostport:r@(_:_)) -> checkArg $ last r _ -> return () checkArg (T_NormalWord _ [T_DoubleQuoted id parts]) = - case filter (not . isConstant) parts of - [] -> return () - (x:_) -> info (getId x) 2029 + forM_ (find (not . isConstant) parts) $ + \x -> info (getId x) 2029 "Note that, unescaped, this expands on the client side." checkArg _ = return () @@ -502,61 +665,97 @@ prop_checkPrintfVar6 = verify checkPrintfVar "printf foo bar baz" prop_checkPrintfVar7 = verify checkPrintfVar "printf -- foo bar baz" prop_checkPrintfVar8 = verifyNot checkPrintfVar "printf '%s %s %s' \"${var[@]}\"" prop_checkPrintfVar9 = verifyNot checkPrintfVar "printf '%s %s %s\\n' *.png" -prop_checkPrintfVar10= verifyNot checkPrintfVar "printf '%s %s %s' foo bar baz" -prop_checkPrintfVar11= verifyNot checkPrintfVar "printf '%(%s%s)T' -1" -prop_checkPrintfVar12= verify checkPrintfVar "printf '%s %s\\n' 1 2 3" -prop_checkPrintfVar13= verifyNot checkPrintfVar "printf '%s %s\\n' 1 2 3 4" -prop_checkPrintfVar14= verify checkPrintfVar "printf '%*s\\n' 1" -prop_checkPrintfVar15= verifyNot checkPrintfVar "printf '%*s\\n' 1 2" -prop_checkPrintfVar16= verifyNot checkPrintfVar "printf $'string'" -prop_checkPrintfVar17= verify checkPrintfVar "printf '%-*s\\n' 1" -prop_checkPrintfVar18= verifyNot checkPrintfVar "printf '%-*s\\n' 1 2" +prop_checkPrintfVar10 = verifyNot checkPrintfVar "printf '%s %s %s' foo bar baz" +prop_checkPrintfVar11 = verifyNot checkPrintfVar "printf '%(%s%s)T' -1" +prop_checkPrintfVar12 = verify checkPrintfVar "printf '%s %s\\n' 1 2 3" +prop_checkPrintfVar13 = verifyNot checkPrintfVar "printf '%s %s\\n' 1 2 3 4" +prop_checkPrintfVar14 = verify checkPrintfVar "printf '%*s\\n' 1" +prop_checkPrintfVar15 = verifyNot checkPrintfVar "printf '%*s\\n' 1 2" +prop_checkPrintfVar16 = verifyNot checkPrintfVar "printf $'string'" +prop_checkPrintfVar17 = verify checkPrintfVar "printf '%-*s\\n' 1" +prop_checkPrintfVar18 = verifyNot checkPrintfVar "printf '%-*s\\n' 1 2" +prop_checkPrintfVar19 = verifyNot checkPrintfVar "printf '%(%s)T'" +prop_checkPrintfVar20 = verifyNot checkPrintfVar "printf '%d %(%s)T' 42" +prop_checkPrintfVar21 = verify checkPrintfVar "printf '%d %(%s)T'" +prop_checkPrintfVar22 = verify checkPrintfVar "printf '%s\n%s' foo" + checkPrintfVar = CommandCheck (Exactly "printf") (f . arguments) where f (doubledash:rest) | getLiteralString doubledash == Just "--" = f rest f (dashv:var:rest) | getLiteralString dashv == Just "-v" = f rest f (format:params) = check format params f _ = return () - countFormats string = - case string of - '%':'%':rest -> countFormats rest - '%':'(':rest -> 1 + countFormats (dropWhile (/= ')') rest) - '%':rest -> regexBasedCountFormats rest + countFormats (dropWhile (/= '%') rest) - _:rest -> countFormats rest - [] -> 0 - - regexBasedCountFormats rest = - maybe 1 (foldl (\acc group -> acc + (if group == "*" then 1 else 0)) 1) (matchRegex re rest) - where - -- constructed based on specifications in "man printf" - re = mkRegex "#?-?\\+? ?0?(\\*|\\d*).?(\\d*|\\*)[diouxXfFeEgGaAcsb]" - -- \____ _____/\___ ____/ \____ ____/\________ ________/ - -- V V V V - -- flags field width precision format character - -- field width and precision can be specified with a '*' instead of a digit, - -- in which case printf will accept one more argument for each '*' used check format more = do - fromMaybe (return ()) $ do + sequence_ $ do string <- getLiteralString format - let vars = countFormats string + let formats = getPrintfFormats string + let formatCount = length formats + let argCount = length more + let pluraliseIfMany word n = if n > 1 then word ++ "s" else word - return $ do - when (vars == 0 && more /= []) $ + return $ if + | argCount == 0 && formatCount == 0 -> + return () -- This is fine + | formatCount == 0 && argCount > 0 -> err (getId format) 2182 "This printf format string has no variables. Other arguments are ignored." - - when (vars > 0 - && ((length more) `mod` vars /= 0 || null more) - && all (not . mayBecomeMultipleArgs) more) $ + | any mayBecomeMultipleArgs more -> + return () -- We don't know so trust the user + | argCount < formatCount && onlyTrailingTs formats argCount -> + return () -- Allow trailing %()Ts since they use the current time + | argCount > 0 && argCount `mod` formatCount == 0 -> + return () -- Great: a suitable number of arguments + | otherwise -> warn (getId format) 2183 $ - "This format string has " ++ show vars ++ " variables, but is passed " ++ show (length more) ++ " arguments." - + "This format string has " ++ show formatCount ++ " " ++ pluraliseIfMany "variable" formatCount ++ + ", but is passed " ++ show argCount ++ pluraliseIfMany " argument" argCount ++ "." unless ('%' `elem` concat (oversimplify format) || isLiteral format) $ info (getId format) 2059 - "Don't use variables in the printf format string. Use printf \"..%s..\" \"$foo\"." + "Don't use variables in the printf format string. Use printf '..%s..' \"$foo\"." + where + onlyTrailingTs format argCount = + all (== 'T') $ drop argCount format +prop_checkGetPrintfFormats1 = getPrintfFormats "%s" == "s" +prop_checkGetPrintfFormats2 = getPrintfFormats "%0*s" == "*s" +prop_checkGetPrintfFormats3 = getPrintfFormats "%(%s)T" == "T" +prop_checkGetPrintfFormats4 = getPrintfFormats "%d%%%(%s)T" == "dT" +prop_checkGetPrintfFormats5 = getPrintfFormats "%bPassed: %d, %bFailed: %d%b, Skipped: %d, %bErrored: %d%b\\n" == "bdbdbdbdb" +prop_checkGetPrintfFormats6 = getPrintfFormats "%s%s" == "ss" +prop_checkGetPrintfFormats7 = getPrintfFormats "%s\n%s" == "ss" +getPrintfFormats = getFormats + where + -- Get the arguments in the string as a string of type characters, + -- e.g. "Hello %s" -> "s" and "%(%s)T %0*d\n" -> "T*d" + getFormats :: String -> String + getFormats string = + case string of + '%':'%':rest -> getFormats rest + '%':'(':rest -> + case dropWhile (/= ')') rest of + ')':c:trailing -> c : getFormats trailing + _ -> "" + '%':rest -> regexBasedGetFormats rest + _:rest -> getFormats rest + [] -> "" + + regexBasedGetFormats rest = + case matchRegex re rest of + Just [width, precision, typ, rest, _] -> + (if width == "*" then "*" else "") ++ + (if precision == "*" then "*" else "") ++ + typ ++ getFormats rest + Nothing -> take 1 rest ++ getFormats rest + where + -- constructed based on specifications in "man printf" + re = mkRegex "#?-?\\+? ?0?(\\*|\\d*)\\.?(\\d*|\\*)([diouxXfFeEgGaAcsbq])((\n|.)*)" + -- \____ _____/\___ ____/ \____ ____/\_________ _________/ \______ / + -- V V V V V + -- flags field width precision format character rest + -- field width and precision can be specified with an '*' instead of a digit, + -- in which case printf will accept one more argument for each '*' used prop_checkUuoeCmd1 = verify checkUuoeCmd "echo $(date)" @@ -579,17 +778,12 @@ prop_checkSetAssignment5 = verifyNot checkSetAssignment "set 'a=5'" prop_checkSetAssignment6 = verifyNot checkSetAssignment "set" checkSetAssignment = CommandCheck (Exactly "set") (f . arguments) where - f (var:value:rest) = - let str = literal var in - when (isVariableName str || isAssignment str) $ - msg (getId var) - f (var:_) = - when (isAssignment $ literal var) $ - msg (getId var) + f (var:rest) + | (not (null rest) && isVariableName str) || isAssignment str = + warn (getId var) 2121 "To assign a variable, use just 'var=value', no 'set ..'." + where str = literal var f _ = return () - msg id = warn id 2121 "To assign a variable, use just 'var=value', no 'set ..'." - isAssignment str = '=' `elem` str literal (T_NormalWord _ l) = concatMap literal l literal (T_Literal _ str) = str @@ -602,9 +796,8 @@ prop_checkExportedExpansions3 = verifyNot checkExportedExpansions "export foo" prop_checkExportedExpansions4 = verifyNot checkExportedExpansions "export ${foo?}" checkExportedExpansions = CommandCheck (Exactly "export") (mapM_ check . arguments) where - check t = potentially $ do - var <- getSingleUnmodifiedVariable t - let name = bracedString var + check t = sequence_ $ do + name <- getSingleUnmodifiedBracedString t return . warn (getId t) 2163 $ "This does not export '" ++ name ++ "'. Remove $/${} for that, or use ${var?} to quiet." @@ -616,30 +809,43 @@ prop_checkReadExpansions5 = verify checkReadExpansions "read \"$var\"" prop_checkReadExpansions6 = verify checkReadExpansions "read -a $var" prop_checkReadExpansions7 = verifyNot checkReadExpansions "read $1" prop_checkReadExpansions8 = verifyNot checkReadExpansions "read ${var?}" +prop_checkReadExpansions9 = verify checkReadExpansions "read arr[val]" checkReadExpansions = CommandCheck (Exactly "read") check where - options = getGnuOpts "sreu:n:N:i:p:a:" + options = getGnuOpts flagsForRead getVars cmd = fromMaybe [] $ do - opts <- options cmd - return . map snd $ filter (\(x,_) -> x == "" || x == "a") opts + opts <- options $ arguments cmd + return [y | (x,(_, y)) <- opts, null x || x == "a"] - check cmd = mapM_ warning $ getVars cmd - warning t = potentially $ do - var <- getSingleUnmodifiedVariable t - let name = bracedString var + check cmd = do + mapM_ dollarWarning $ getVars cmd + mapM_ arrayWarning $ arguments cmd + + dollarWarning t = sequence_ $ do + name <- getSingleUnmodifiedBracedString t guard $ isVariableName name -- e.g. not $1 return . warn (getId t) 2229 $ "This does not read '" ++ name ++ "'. Remove $/${} for that, or use ${var?} to quiet." + arrayWarning word = + when (any isUnquotedBracket $ getWordParts word) $ + warn (getId word) 2313 $ + "Quote array indices to avoid them expanding as globs." + + isUnquotedBracket t = + case t of + T_Glob _ ('[':_) -> True + _ -> False + -- Return the single variable expansion that makes up this word, if any. -- e.g. $foo -> $foo, "$foo"'' -> $foo , "hello $name" -> Nothing -getSingleUnmodifiedVariable :: Token -> Maybe Token -getSingleUnmodifiedVariable word = +getSingleUnmodifiedBracedString :: Token -> Maybe String +getSingleUnmodifiedBracedString word = case getWordParts word of - [t@(T_DollarBraced {})] -> - let contents = bracedString t + [T_DollarBraced _ _ l] -> + let contents = concat $ oversimplify l name = getBracedReference contents - in guard (contents == name) >> return t + in guard (contents == name) >> return contents _ -> Nothing prop_checkAliasesUsesArgs1 = verify checkAliasesUsesArgs "alias a='cp $1 /a'" @@ -650,7 +856,7 @@ checkAliasesUsesArgs = CommandCheck (Exactly "alias") (f . arguments) re = mkRegex "\\$\\{?[0-9*@]" f = mapM_ checkArg checkArg arg = - let string = fromJust $ getLiteralStringExt (const $ return "_") arg in + let string = getLiteralStringDef "_" arg in when ('=' `elem` string && string `matches` re) $ err (getId arg) 2142 "Aliases can't use positional parameters. Use a function." @@ -663,13 +869,16 @@ checkAliasesExpandEarly = CommandCheck (Exactly "alias") (f . arguments) where f = mapM_ checkArg checkArg arg | '=' `elem` concat (oversimplify arg) = - forM_ (take 1 $ filter (not . isLiteral) $ getWordParts arg) $ + forM_ (find (not . isLiteral) $ getWordParts arg) $ \x -> warn (getId x) 2139 "This expands when defined, not when used. Consider escaping." checkArg _ = return () prop_checkUnsetGlobs1 = verify checkUnsetGlobs "unset foo[1]" prop_checkUnsetGlobs2 = verifyNot checkUnsetGlobs "unset foo" +prop_checkUnsetGlobs3 = verify checkUnsetGlobs "unset foo[$i]" +prop_checkUnsetGlobs4 = verify checkUnsetGlobs "unset foo[x${i}y]" +prop_checkUnsetGlobs5 = verifyNot checkUnsetGlobs "unset foo][" checkUnsetGlobs = CommandCheck (Exactly "unset") (mapM_ check . arguments) where check arg = @@ -696,7 +905,7 @@ checkFindWithoutPath = CommandCheck (Basename "find") f -- path. We assume that all the pre-path flags are single characters from a -- list of GNU and macOS flags. hasPath (first:rest) = - let flag = fromJust $ getLiteralStringExt (const $ return "___") first in + let flag = getLiteralStringDef "___" first in not ("-" `isPrefixOf` flag) || isLeadingFlag flag && hasPath rest hasPath [] = False isLeadingFlag flag = length flag <= 2 || all (`elem` leadingFlagChars) flag @@ -722,7 +931,7 @@ prop_checkTimedCommand2 = verify checkTimedCommand "#!/bin/dash\ntime ( foo; bar prop_checkTimedCommand3 = verifyNot checkTimedCommand "#!/bin/sh\ntime sleep 1" checkTimedCommand = CommandCheck (Exactly "time") f where f (T_SimpleCommand _ _ (c:args@(_:_))) = - whenShell [Sh, Dash] $ do + whenShell [Sh, Dash, BusyboxSh] $ do let cmd = last args -- "time" is parsed with a command as argument when (isPiped cmd) $ warn (getId c) 2176 "'time' is undefined for pipelines. time single stage or bash -c instead." @@ -746,11 +955,27 @@ checkTimedCommand = CommandCheck (Exactly "time") f where prop_checkLocalScope1 = verify checkLocalScope "local foo=3" prop_checkLocalScope2 = verifyNot checkLocalScope "f() { local foo=3; }" checkLocalScope = CommandCheck (Exactly "local") $ \t -> - whenShell [Bash, Dash] $ do -- Ksh allows it, Sh doesn't support local + whenShell [Bash, Dash, BusyboxSh] $ do -- Ksh allows it, Sh doesn't support local path <- getPathM t - unless (any isFunction path) $ + unless (any isFunctionLike path) $ err (getId $ getCommandTokenOrThis t) 2168 "'local' is only valid in functions." +prop_checkMultipleDeclaring1 = verify (checkMultipleDeclaring "local") "q() { local readonly var=1; }" +prop_checkMultipleDeclaring2 = verifyNot (checkMultipleDeclaring "local") "q() { local var=1; }" +prop_checkMultipleDeclaring3 = verify (checkMultipleDeclaring "readonly") "readonly local foo=5" +prop_checkMultipleDeclaring4 = verify (checkMultipleDeclaring "export") "export readonly foo=5" +prop_checkMultipleDeclaring5 = verifyNot (checkMultipleDeclaring "local") "f() { local -r foo=5; }" +prop_checkMultipleDeclaring6 = verifyNot (checkMultipleDeclaring "declare") "declare -rx foo=5" +prop_checkMultipleDeclaring7 = verifyNot (checkMultipleDeclaring "readonly") "readonly 'local' foo=5" +checkMultipleDeclaring cmd = CommandCheck (Exactly cmd) (mapM_ check . arguments) + where + check t = sequence_ $ do + lit <- getUnquotedLiteral t + guard $ lit `elem` declaringCommands + return $ err (getId $ getCommandTokenOrThis t) 2316 $ + "This applies " ++ cmd ++ " to the variable named " ++ lit ++ + ", which is probably not what you want. Use a separate command or the appropriate `declare` options instead." + prop_checkDeprecatedTempfile1 = verify checkDeprecatedTempfile "var=$(tempfile)" prop_checkDeprecatedTempfile2 = verifyNot checkDeprecatedTempfile "tempfile=$(mktemp)" checkDeprecatedTempfile = CommandCheck (Basename "tempfile") $ @@ -769,42 +994,56 @@ prop_checkWhileGetoptsCase2 = verify checkWhileGetoptsCase "while getopts 'a:' x prop_checkWhileGetoptsCase3 = verifyNot checkWhileGetoptsCase "while getopts 'a:b' x; do case $x in a) foo;; b) bar;; *) :;esac; done" prop_checkWhileGetoptsCase4 = verifyNot checkWhileGetoptsCase "while getopts 'a:123' x; do case $x in a) foo;; [0-9]) bar;; esac; done" prop_checkWhileGetoptsCase5 = verifyNot checkWhileGetoptsCase "while getopts 'a:' x; do case $x in a) foo;; \\?) bar;; *) baz;; esac; done" +prop_checkWhileGetoptsCase6 = verifyNot checkWhileGetoptsCase "while getopts 'a:b' x; do case $y in a) foo;; esac; done" +prop_checkWhileGetoptsCase7 = verifyNot checkWhileGetoptsCase "while getopts 'a:b' x; do case x$x in xa) foo;; xb) foo;; esac; done" +prop_checkWhileGetoptsCase8 = verifyNot checkWhileGetoptsCase "while getopts 'a:b' x; do x=a; case $x in a) foo;; esac; done" checkWhileGetoptsCase = CommandCheck (Exactly "getopts") f where f :: Token -> Analysis - f t@(T_SimpleCommand _ _ (cmd:arg1:_)) = do + f t@(T_SimpleCommand _ _ (cmd:arg1:name:_)) = do path <- getPathM t - potentially $ do + params <- ask + sequence_ $ do options <- getLiteralString arg1 - (T_WhileExpression _ _ body) <- findFirst whileLoop path - caseCmd <- mapMaybe findCase body !!! 0 - return $ check (getId arg1) (map (:[]) $ filter (/= ':') options) caseCmd + getoptsVar <- getLiteralString name + (T_WhileExpression _ _ body) <- findFirst whileLoop (NE.toList path) + T_CaseExpression id var list <- mapMaybe findCase body !!! 0 + + -- Make sure getopts name and case variable matches + [T_DollarBraced _ _ bracedWord] <- return $ getWordParts var + [T_Literal _ caseVar] <- return $ getWordParts bracedWord + guard $ caseVar == getoptsVar + + -- Make sure the variable isn't modified + guard . not $ modifiesVariable params (T_BraceGroup (Id 0) body) getoptsVar + + return $ check (getId arg1) (map (:[]) $ filter (/= ':') options) id list f _ = return () - check :: Id -> [String] -> Token -> Analysis - check optId opts (T_CaseExpression id _ list) = do - unless (Nothing `Map.member` handledMap) $ do - mapM_ (warnUnhandled optId id) $ catMaybes $ Map.keys notHandled + check :: Id -> [String] -> Id -> [(CaseType, [Token], [Token])] -> Analysis + check optId opts id list = do + unless (Nothing `M.member` handledMap) $ do + mapM_ (warnUnhandled optId id) $ catMaybes $ M.keys notHandled - unless (any (`Map.member` handledMap) [Just "*",Just "?"]) $ + unless (any (`M.member` handledMap) [Just "*",Just "?"]) $ warn id 2220 "Invalid flags are not handled. Add a *) case." - mapM_ warnRedundant $ Map.toList notRequested + mapM_ warnRedundant $ M.toList notRequested where - handledMap = Map.fromList (concatMap getHandledStrings list) - requestedMap = Map.fromList $ map (\x -> (Just x, ())) opts + handledMap = M.fromList (concatMap getHandledStrings list) + requestedMap = M.fromList $ map (\x -> (Just x, ())) opts - notHandled = Map.difference requestedMap handledMap - notRequested = Map.difference handledMap requestedMap + notHandled = M.difference requestedMap handledMap + notRequested = M.difference handledMap requestedMap warnUnhandled optId caseId str = - warn caseId 2213 $ "getopts specified -" ++ str ++ ", but it's not handled by this 'case'." + warn caseId 2213 $ "getopts specified -" ++ (e4m str) ++ ", but it's not handled by this 'case'." - warnRedundant (key, expr) = potentially $ do - str <- key - guard $ str `notElem` ["*", ":", "?"] - return $ warn (getId expr) 2214 "This case is not specified by getopts." + warnRedundant (Just str, expr) + | str `notElem` ["*", ":", "?"] = + warn (getId expr) 2214 "This case is not specified by getopts." + warnRedundant _ = return () getHandledStrings (_, globs, _) = map (\x -> (literal x, x)) globs @@ -815,7 +1054,7 @@ checkWhileGetoptsCase = CommandCheck (Exactly "getopts") f fromGlob t = case t of - T_Glob _ ('[':c:']':[]) -> return [c] + T_Glob _ ['[', c, ']'] -> return [c] T_Glob _ "*" -> return "*" T_Glob _ "?" -> return "?" _ -> Nothing @@ -840,17 +1079,17 @@ prop_checkCatastrophicRm4 = verify checkCatastrophicRm "rm -fr /home/$(whoami)/* prop_checkCatastrophicRm5 = verifyNot checkCatastrophicRm "rm -r /home/${USER:-thing}/*" prop_checkCatastrophicRm6 = verify checkCatastrophicRm "rm --recursive /etc/*$config*" prop_checkCatastrophicRm8 = verify checkCatastrophicRm "rm -rf /home" -prop_checkCatastrophicRm10= verifyNot checkCatastrophicRm "rm -r \"${DIR}\"/{.gitignore,.gitattributes,ci}" -prop_checkCatastrophicRm11= verify checkCatastrophicRm "rm -r /{bin,sbin}/$exec" -prop_checkCatastrophicRm12= verify checkCatastrophicRm "rm -r /{{usr,},{bin,sbin}}/$exec" -prop_checkCatastrophicRm13= verifyNot checkCatastrophicRm "rm -r /{{a,b},{c,d}}/$exec" +prop_checkCatastrophicRm10 = verifyNot checkCatastrophicRm "rm -r \"${DIR}\"/{.gitignore,.gitattributes,ci}" +prop_checkCatastrophicRm11 = verify checkCatastrophicRm "rm -r /{bin,sbin}/$exec" +prop_checkCatastrophicRm12 = verify checkCatastrophicRm "rm -r /{{usr,},{bin,sbin}}/$exec" +prop_checkCatastrophicRm13 = verifyNot checkCatastrophicRm "rm -r /{{a,b},{c,d}}/$exec" prop_checkCatastrophicRmA = verify checkCatastrophicRm "rm -rf /usr /lib/nvidia-current/xorg/xorg" prop_checkCatastrophicRmB = verify checkCatastrophicRm "rm -rf \"$STEAMROOT/\"*" checkCatastrophicRm = CommandCheck (Basename "rm") $ \t -> when (isRecursive t) $ mapM_ (mapM_ checkWord . braceExpand) $ arguments t where - isRecursive = any (`elem` ["r", "R", "recursive"]) . map snd . getAllFlags + isRecursive = any ((`elem` ["r", "R", "recursive"]) . snd) . getAllFlags checkWord token = case getLiteralString token of @@ -860,7 +1099,7 @@ checkCatastrophicRm = CommandCheck (Basename "rm") $ \t -> Nothing -> checkWord' token - checkWord' token = fromMaybe (return ()) $ do + checkWord' token = sequence_ $ do filename <- getPotentialPath token let path = fixPath filename return . when (path `elem` importantPaths) $ @@ -873,7 +1112,7 @@ checkCatastrophicRm = CommandCheck (Basename "rm") $ \t -> getPotentialPath = getLiteralStringExt f where f (T_Glob _ str) = return str - f (T_DollarBraced _ word) = + f (T_DollarBraced _ _ word) = let var = onlyLiteralString word in -- This shouldn't handle non-colon cases. if any (`isInfixOf` var) [":?", ":-", ":="] @@ -882,9 +1121,9 @@ checkCatastrophicRm = CommandCheck (Basename "rm") $ \t -> f _ = return "" stripTrailing c = reverse . dropWhile (== c) . reverse - skipRepeating c (a:b:rest) | a == b && b == c = skipRepeating c (b:rest) - skipRepeating c (a:r) = a:skipRepeating c r - skipRepeating _ [] = [] + skipRepeating c = foldr go [] + where + go a r = a : case r of b:rest | b == c && a == b -> rest; _ -> r paths = [ "", "/bin", "/etc", "/home", "/mnt", "/usr", "/usr/share", "/usr/local", @@ -910,10 +1149,9 @@ missingDestination handler token = do _ -> return () where args = getAllFlags token - params = map fst $ filter (\(_,x) -> x == "") args + params = [x | (x,"") <- args] hasTarget = - any (\x -> x /= "" && x `isPrefixOf` "target-directory") $ - map snd args + any (\(_,x) -> x /= "" && x `isPrefixOf` "target-directory") args prop_checkMvArguments1 = verify checkMvArguments "mv 'foo bar'" prop_checkMvArguments2 = verifyNot checkMvArguments "mv foo bar" @@ -956,7 +1194,7 @@ checkFindRedirections = CommandCheck (Basename "find") f prop_checkWhich = verify checkWhich "which '.+'" checkWhich = CommandCheck (Basename "which") $ - \t -> info (getId $ getCommandTokenOrThis t) 2230 "which is non-standard. Use builtin 'command -v' instead." + \t -> info (getId $ getCommandTokenOrThis t) 2230 "'which' is non-standard. Use builtin 'command -v' instead." prop_checkSudoRedirect1 = verify checkSudoRedirect "sudo echo 3 > /proc/file" prop_checkSudoRedirect2 = verify checkSudoRedirect "sudo cmd < input" @@ -973,7 +1211,7 @@ checkSudoRedirect = CommandCheck (Basename "sudo") f Just (T_Redirecting _ redirs _) -> mapM_ warnAbout redirs warnAbout (T_FdRedirect _ s (T_IoFile id op file)) - | (s == "" || s == "&") && not (special file) = + | (null s || s == "&") && not (special file) = case op of T_Less _ -> info (getId op) 2024 @@ -997,10 +1235,9 @@ prop_checkSudoArgs6 = verifyNot checkSudoArgs "sudo -n -u export ls" prop_checkSudoArgs7 = verifyNot checkSudoArgs "sudo docker export foo" checkSudoArgs = CommandCheck (Basename "sudo") f where - f t = potentially $ do - opts <- parseOpts t - let nonFlags = map snd $ filter (\(flag, _) -> flag == "") opts - commandArg <- nonFlags !!! 0 + f t = sequence_ $ do + opts <- parseOpts $ arguments t + (_,(commandArg, _)) <- find (null . fst) opts command <- getLiteralString commandArg guard $ command `elem` builtins return $ warn (getId t) 2232 $ "Can't use sudo with builtins like " ++ command ++ ". Did you want sudo sh -c .. instead?" @@ -1008,5 +1245,228 @@ checkSudoArgs = CommandCheck (Basename "sudo") f -- This mess is why ShellCheck prefers not to know. parseOpts = getBsdOpts "vAknSbEHPa:g:h:p:u:c:T:r:" +prop_checkSourceArgs1 = verify checkSourceArgs "#!/bin/sh\n. script arg" +prop_checkSourceArgs2 = verifyNot checkSourceArgs "#!/bin/sh\n. script" +prop_checkSourceArgs3 = verifyNot checkSourceArgs "#!/bin/bash\n. script arg" +checkSourceArgs = CommandCheck (Exactly ".") f + where + f t = whenShell [Sh, Dash] $ + case arguments t of + (file:arg1:_) -> warn (getId arg1) 2240 $ + "The dot command does not support arguments in sh/dash. Set them as variables." + _ -> return () + +prop_checkChmodDashr1 = verify checkChmodDashr "chmod -r 0755 dir" +prop_checkChmodDashr2 = verifyNot checkChmodDashr "chmod -R 0755 dir" +prop_checkChmodDashr3 = verifyNot checkChmodDashr "chmod a-r dir" +checkChmodDashr = CommandCheck (Basename "chmod") f + where + f t = mapM_ check $ arguments t + check t = sequence_ $ do + flag <- getLiteralString t + guard $ flag == "-r" + return $ warn (getId t) 2253 "Use -R to recurse, or explicitly a-r to remove read permissions." + +prop_checkXargsDashi1 = verify checkXargsDashi "xargs -i{} echo {}" +prop_checkXargsDashi2 = verifyNot checkXargsDashi "xargs -I{} echo {}" +prop_checkXargsDashi3 = verifyNot checkXargsDashi "xargs sed -i -e foo" +prop_checkXargsDashi4 = verify checkXargsDashi "xargs -e sed -i foo" +prop_checkXargsDashi5 = verifyNot checkXargsDashi "xargs -x sed -i foo" +checkXargsDashi = CommandCheck (Basename "xargs") f + where + f t = sequence_ $ do + opts <- parseOpts $ arguments t + (option, value) <- lookup "i" opts + return $ info (getId option) 2267 "GNU xargs -i is deprecated in favor of -I{}" + parseOpts = getBsdOpts "0oprtxadR:S:J:L:l:n:P:s:e:E:i:I:" + + +prop_checkArgComparison1 = verify (checkArgComparison "declare") "declare a = b" +prop_checkArgComparison2 = verify (checkArgComparison "declare") "declare a =b" +prop_checkArgComparison3 = verifyNot (checkArgComparison "declare") "declare a=b" +prop_checkArgComparison4 = verify (checkArgComparison "export") "export a +=b" +prop_checkArgComparison7 = verifyNot (checkArgComparison "declare") "declare -a +i foo" +prop_checkArgComparison8 = verify (checkArgComparison "let") "let x = 0" +prop_checkArgComparison9 = verify (checkArgComparison "alias") "alias x =0" +-- This mirrors checkSecondArgIsComparison but for arguments to local/readonly/declare/export +checkArgComparison cmd = CommandCheck (Exactly cmd) wordsWithEqual + where + wordsWithEqual t = mapM_ check $ arguments t + check arg = do + sequence_ $ do + str <- getLeadingUnquotedString arg + case str of + '=':_ -> + return $ err (headId arg) 2290 $ + "Remove spaces around = to assign." + '+':'=':_ -> + return $ err (headId arg) 2290 $ + "Remove spaces around += to append." + _ -> Nothing + + -- 'let' is parsed as a sequence of arithmetic expansions, + -- so we want the additional warning for "x=" + when (cmd == "let") $ sequence_ $ do + token <- getTrailingUnquotedLiteral arg + str <- getLiteralString token + guard $ "=" `isSuffixOf` str + return $ err (getId token) 2290 $ + "Remove spaces around = to assign." + + headId t = + case t of + T_NormalWord _ (x:_) -> getId x + _ -> getId t + + +prop_checkMaskedReturns1 = verify (checkMaskedReturns "local") "f() { local a=$(false); }" +prop_checkMaskedReturns2 = verify (checkMaskedReturns "declare") "declare a=$(false)" +prop_checkMaskedReturns3 = verify (checkMaskedReturns "declare") "declare a=\"`false`\"" +prop_checkMaskedReturns4 = verify (checkMaskedReturns "readonly") "readonly a=$(false)" +prop_checkMaskedReturns5 = verify (checkMaskedReturns "readonly") "readonly a=\"`false`\"" +prop_checkMaskedReturns6 = verifyNot (checkMaskedReturns "declare") "declare a; a=$(false)" +prop_checkMaskedReturns7 = verifyNot (checkMaskedReturns "local") "f() { local -r a=$(false); }" +prop_checkMaskedReturns8 = verifyNot (checkMaskedReturns "readonly") "a=$(false); readonly a" +prop_checkMaskedReturns9 = verify (checkMaskedReturns "typeset") "#!/bin/ksh\n f() { typeset -r x=$(false); }" +prop_checkMaskedReturns10 = verifyNot (checkMaskedReturns "typeset") "#!/bin/ksh\n function f { typeset -r x=$(false); }" +prop_checkMaskedReturns11 = verifyNot (checkMaskedReturns "typeset") "#!/bin/bash\n f() { typeset -r x=$(false); }" +prop_checkMaskedReturns12 = verify (checkMaskedReturns "typeset") "typeset -r x=$(false);" +prop_checkMaskedReturns13 = verify (checkMaskedReturns "typeset") "f() { typeset -g x=$(false); }" +prop_checkMaskedReturns14 = verify (checkMaskedReturns "declare") "declare x=${ false; }" +prop_checkMaskedReturns15 = verify (checkMaskedReturns "declare") "f() { declare x=$(false); }" +checkMaskedReturns str = CommandCheck (Exactly str) checkCmd + where + checkCmd t = do + path <- getPathM t + shell <- asks shellType + sequence_ $ do + name <- getCommandName t + + let flags = map snd (getAllFlags t) + let hasDashR = "r" `elem` flags + let hasDashG = "g" `elem` flags + let isInScopedFunction = any (isScopedFunction shell) path + + let isLocal = not hasDashG && isLocalInFunction name && isInScopedFunction + let isReadOnly = name == "readonly" || hasDashR + + -- Don't warn about local variables that are declared readonly, + -- because the workaround `local x; x=$(false); local -r x;` is annoying + guard . not $ isLocal && isReadOnly + + return $ mapM_ checkArgs $ arguments t + + checkArgs (T_Assignment id _ _ _ word) | any hasReturn $ getWordParts word = + warn id 2155 "Declare and assign separately to avoid masking return values." + checkArgs _ = return () + + isLocalInFunction = (`elem` ["local", "declare", "typeset"]) + isScopedFunction shell t = + case t of + T_BatsTest {} -> True + -- In ksh, only functions declared with 'function' have their own scope + T_Function _ (FunctionKeyword hasFunction) _ _ _ -> shell /= Ksh || hasFunction + _ -> False + + hasReturn t = case t of + T_Backticked {} -> True + T_DollarExpansion {} -> True + T_DollarBraceCommandExpansion {} -> True + _ -> False + + +prop_checkUnquotedEchoSpaces1 = verify checkUnquotedEchoSpaces "echo foo bar" +prop_checkUnquotedEchoSpaces2 = verifyNot checkUnquotedEchoSpaces "echo foo" +prop_checkUnquotedEchoSpaces3 = verifyNot checkUnquotedEchoSpaces "echo foo bar" +prop_checkUnquotedEchoSpaces4 = verifyNot checkUnquotedEchoSpaces "echo 'foo bar'" +prop_checkUnquotedEchoSpaces5 = verifyNot checkUnquotedEchoSpaces "echo a > myfile.txt b" +prop_checkUnquotedEchoSpaces6 = verifyNot checkUnquotedEchoSpaces " echo foo\\\n bar" +checkUnquotedEchoSpaces = CommandCheck (Basename "echo") check + where + check t = do + let args = arguments t + m <- asks tokenPositions + redir <- getClosestCommandM t + sequence_ $ do + let positions = mapMaybe (\c -> M.lookup (getId c) m) args + let pairs = zip positions (drop 1 positions) + (T_Redirecting _ redirTokens _) <- redir + let redirPositions = mapMaybe (\c -> fst <$> M.lookup (getId c) m) redirTokens + guard $ any (hasSpacesBetween redirPositions) pairs + return $ info (getId t) 2291 "Quote repeated spaces to avoid them collapsing into one." + + hasSpacesBetween redirs ((a,b), (c,d)) = + posLine a == posLine d + && ((posColumn c) - (posColumn b)) >= 4 + && not (any (\x -> b < x && x < c) redirs) + + +prop_checkEvalArray1 = verify checkEvalArray "eval $@" +prop_checkEvalArray2 = verify checkEvalArray "eval \"${args[@]}\"" +prop_checkEvalArray3 = verify checkEvalArray "eval \"${args[@]@Q}\"" +prop_checkEvalArray4 = verifyNot checkEvalArray "eval \"${args[*]@Q}\"" +prop_checkEvalArray5 = verifyNot checkEvalArray "eval \"$*\"" +checkEvalArray = CommandCheck (Exactly "eval") (mapM_ check . concatMap getWordParts . arguments) + where + check t = + when (isArrayExpansion t) $ + if isEscaped t + then style (getId t) 2293 "When eval'ing @Q-quoted words, use * rather than @ as the index." + else warn (getId t) 2294 "eval negates the benefit of arrays. Drop eval to preserve whitespace/symbols (or eval as string)." + + isEscaped q = + case q of + -- Match ${arr[@]@Q} and ${@@Q} and such + T_DollarBraced _ _ l -> 'Q' `elem` getBracedModifier (concat $ oversimplify l) + _ -> False + + +prop_checkBackreferencingDeclaration1 = verify (checkBackreferencingDeclaration "declare") "declare x=1 y=foo$x" +prop_checkBackreferencingDeclaration2 = verify (checkBackreferencingDeclaration "readonly") "readonly x=1 y=$((1+x))" +prop_checkBackreferencingDeclaration3 = verify (checkBackreferencingDeclaration "local") "local x=1 y=$(echo $x)" +prop_checkBackreferencingDeclaration4 = verify (checkBackreferencingDeclaration "local") "local x=1 y[$x]=z" +prop_checkBackreferencingDeclaration5 = verify (checkBackreferencingDeclaration "declare") "declare x=var $x=1" +prop_checkBackreferencingDeclaration6 = verify (checkBackreferencingDeclaration "declare") "declare x=var $x=1" +prop_checkBackreferencingDeclaration7 = verify (checkBackreferencingDeclaration "declare") "declare x=var $k=$x" +checkBackreferencingDeclaration cmd = CommandCheck (Exactly cmd) check + where + check t = do + maybeCfga <- asks cfgAnalysis + mapM_ (\cfga -> foldM_ (perArg cfga) M.empty $ arguments t) maybeCfga + + perArg cfga leftArgs t = + case t of + T_Assignment id _ name idx t -> do + warnIfBackreferencing cfga leftArgs $ t:idx + return $ M.insert name id leftArgs + t -> do + warnIfBackreferencing cfga leftArgs [t] + return leftArgs + + warnIfBackreferencing cfga backrefs l = do + references <- findReferences cfga l + let reused = M.intersection backrefs references + mapM msg $ M.toList reused + + msg (name, id) = warn id 2318 $ "This assignment is used again in this '" ++ cmd ++ "', but won't have taken effect. Use two '" ++ cmd ++ "'s." + + findReferences cfga list = do + let graph = CF.graph cfga + let nodesMap = CF.tokenToNodes cfga + let nodes = S.unions $ map (\id -> M.findWithDefault S.empty id nodesMap) $ map getId $ list + let labels = mapMaybe (G.lab graph) $ S.toList nodes + let references = M.fromList $ concatMap refFromLabel labels + return references + + refFromLabel lab = + case lab of + CFApplyEffects effects -> mapMaybe refFromEffect effects + _ -> [] + refFromEffect e = + case e of + IdTagged id (CFReadVariable name) -> return (name, id) + _ -> Nothing + + return [] runTests = $( [| $(forAllProperties) (quickCheckWithResult (stdArgs { maxSuccess = 1 }) ) |]) diff --git a/src/ShellCheck/Checks/ControlFlow.hs b/src/ShellCheck/Checks/ControlFlow.hs new file mode 100644 index 0000000..9f63141 --- /dev/null +++ b/src/ShellCheck/Checks/ControlFlow.hs @@ -0,0 +1,101 @@ +{- + Copyright 2022 Vidar Holen + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{-# LANGUAGE TemplateHaskell #-} + +-- Checks that run on the Control Flow Graph (as opposed to the AST) +-- This is scaffolding for a work in progress. + +module ShellCheck.Checks.ControlFlow (checker, optionalChecks, ShellCheck.Checks.ControlFlow.runTests) where + +import ShellCheck.AST +import ShellCheck.ASTLib +import ShellCheck.CFG hiding (cfgAnalysis) +import ShellCheck.CFGAnalysis +import ShellCheck.AnalyzerLib +import ShellCheck.Data +import ShellCheck.Interface + +import Control.Monad +import Control.Monad.Reader +import Data.Graph.Inductive.Graph +import qualified Data.Map as M +import qualified Data.Set as S +import Data.List +import Data.Maybe + +import Test.QuickCheck.All (forAllProperties) +import Test.QuickCheck.Test (quickCheckWithResult, stdArgs, maxSuccess) + + +optionalChecks :: [CheckDescription] +optionalChecks = [] + +-- A check that runs on the entire graph +type ControlFlowCheck = Analysis +-- A check invoked once per node, with its (pre,post) data +type ControlFlowNodeCheck = LNode CFNode -> (ProgramState, ProgramState) -> Analysis +-- A check invoked once per effect, with its node's (pre,post) data +type ControlFlowEffectCheck = IdTagged CFEffect -> Node -> (ProgramState, ProgramState) -> Analysis + + +checker :: AnalysisSpec -> Parameters -> Checker +checker spec params = Checker { + perScript = const $ sequence_ controlFlowChecks, + perToken = const $ return () +} + +controlFlowChecks :: [ControlFlowCheck] +controlFlowChecks = [ + runNodeChecks controlFlowNodeChecks + ] + +controlFlowNodeChecks :: [ControlFlowNodeCheck] +controlFlowNodeChecks = [ + runEffectChecks controlFlowEffectChecks + ] + +controlFlowEffectChecks :: [ControlFlowEffectCheck] +controlFlowEffectChecks = [ + ] + +runNodeChecks :: [ControlFlowNodeCheck] -> ControlFlowCheck +runNodeChecks perNode = do + cfg <- asks cfgAnalysis + mapM_ runOnAll cfg + where + getData datas n@(node, label) = do + (pre, post) <- M.lookup node datas + return (n, (pre, post)) + + runOn :: (LNode CFNode, (ProgramState, ProgramState)) -> Analysis + runOn (node, prepost) = mapM_ (\c -> c node prepost) perNode + runOnAll cfg = mapM_ runOn $ mapMaybe (getData $ nodeToData cfg) $ labNodes (graph cfg) + +runEffectChecks :: [ControlFlowEffectCheck] -> ControlFlowNodeCheck +runEffectChecks list = checkNode + where + checkNode (node, label) prepost = + case label of + CFApplyEffects effects -> mapM_ (\effect -> mapM_ (\c -> c effect node prepost) list) effects + _ -> return () + + +return [] +runTests = $( [| $(forAllProperties) (quickCheckWithResult (stdArgs { maxSuccess = 1 }) ) |]) diff --git a/src/ShellCheck/Checks/Custom.hs b/src/ShellCheck/Checks/Custom.hs new file mode 100644 index 0000000..17e9c9e --- /dev/null +++ b/src/ShellCheck/Checks/Custom.hs @@ -0,0 +1,21 @@ +{- + This empty file is provided for ease of patching in site specific checks. + However, there are no guarantees regarding compatibility between versions. +-} + +{-# LANGUAGE TemplateHaskell #-} +module ShellCheck.Checks.Custom (checker, ShellCheck.Checks.Custom.runTests) where + +import ShellCheck.AnalyzerLib +import Test.QuickCheck + +checker :: Parameters -> Checker +checker params = Checker { + perScript = const $ return (), + perToken = const $ return () + } + +prop_CustomTestsWork = True + +return [] +runTests = $quickCheckAll diff --git a/src/ShellCheck/Checks/ShellSupport.hs b/src/ShellCheck/Checks/ShellSupport.hs index 1bc425f..b664879 100644 --- a/src/ShellCheck/Checks/ShellSupport.hs +++ b/src/ShellCheck/Checks/ShellSupport.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2016 Vidar Holen + Copyright 2012-2020 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -19,20 +19,24 @@ -} {-# LANGUAGE TemplateHaskell #-} {-# LANGUAGE FlexibleContexts #-} +{-# LANGUAGE ViewPatterns #-} module ShellCheck.Checks.ShellSupport (checker , ShellCheck.Checks.ShellSupport.runTests) where import ShellCheck.AST import ShellCheck.ASTLib import ShellCheck.AnalyzerLib import ShellCheck.Interface +import ShellCheck.Prelude import ShellCheck.Regex import Control.Monad import Control.Monad.RWS import Data.Char +import Data.Functor.Identity import Data.List import Data.Maybe import qualified Data.Map as Map +import qualified Data.Set as Set import Test.QuickCheck.All (forAllProperties) import Test.QuickCheck.Test (quickCheckWithResult, stdArgs, maxSuccess) @@ -57,6 +61,9 @@ checks = [ ,checkBraceExpansionVars ,checkMultiDimensionalArrays ,checkPS1Assignments + ,checkMultipleBangs + ,checkBangAfterPipe + ,checkNegatedUnaryOps ] testChecker (ForShell _ t) = @@ -70,243 +77,416 @@ verifyNot c s = producesComments (testChecker c) s == Just False prop_checkForDecimals1 = verify checkForDecimals "((3.14*c))" prop_checkForDecimals2 = verify checkForDecimals "foo[1.2]=bar" prop_checkForDecimals3 = verifyNot checkForDecimals "declare -A foo; foo[1.2]=bar" -checkForDecimals = ForShell [Sh, Dash, Bash] f +checkForDecimals = ForShell [Sh, Dash, BusyboxSh, Bash] f where - f t@(TA_Expansion id _) = potentially $ do - str <- getLiteralString t - first <- str !!! 0 - guard $ isDigit first && '.' `elem` str + f t@(TA_Expansion id _) = sequence_ $ do + first:rest <- getLiteralString t + guard $ isDigit first && '.' `elem` rest return $ err id 2079 "(( )) doesn't support decimals. Use bc or awk." f _ = return () prop_checkBashisms = verify checkBashisms "while read a; do :; done < <(a)" -prop_checkBashisms2 = verify checkBashisms "[ foo -nt bar ]" +prop_checkBashisms2 = verifyNot checkBashisms "[ foo -nt bar ]" prop_checkBashisms3 = verify checkBashisms "echo $((i++))" prop_checkBashisms4 = verify checkBashisms "rm !(*.hs)" prop_checkBashisms5 = verify checkBashisms "source file" prop_checkBashisms6 = verify checkBashisms "[ \"$a\" == 42 ]" +prop_checkBashisms6b = verify checkBashisms "test \"$a\" == 42" +prop_checkBashisms6c = verify checkBashisms "[ foo =~ bar ]" +prop_checkBashisms6d = verify checkBashisms "test foo =~ bar" prop_checkBashisms7 = verify checkBashisms "echo ${var[1]}" prop_checkBashisms8 = verify checkBashisms "echo ${!var[@]}" prop_checkBashisms9 = verify checkBashisms "echo ${!var*}" -prop_checkBashisms10= verify checkBashisms "echo ${var:4:12}" -prop_checkBashisms11= verifyNot checkBashisms "echo ${var:-4}" -prop_checkBashisms12= verify checkBashisms "echo ${var//foo/bar}" -prop_checkBashisms13= verify checkBashisms "exec -c env" -prop_checkBashisms14= verify checkBashisms "echo -n \"Foo: \"" -prop_checkBashisms15= verify checkBashisms "let n++" -prop_checkBashisms16= verify checkBashisms "echo $RANDOM" -prop_checkBashisms17= verify checkBashisms "echo $((RANDOM%6+1))" -prop_checkBashisms18= verify checkBashisms "foo &> /dev/null" -prop_checkBashisms19= verify checkBashisms "foo > file*.txt" -prop_checkBashisms20= verify checkBashisms "read -ra foo" -prop_checkBashisms21= verify checkBashisms "[ -a foo ]" -prop_checkBashisms22= verifyNot checkBashisms "[ foo -a bar ]" -prop_checkBashisms23= verify checkBashisms "trap mything ERR INT" -prop_checkBashisms24= verifyNot checkBashisms "trap mything INT TERM" -prop_checkBashisms25= verify checkBashisms "cat < /dev/tcp/host/123" -prop_checkBashisms26= verify checkBashisms "trap mything ERR SIGTERM" -prop_checkBashisms27= verify checkBashisms "echo *[^0-9]*" -prop_checkBashisms28= verify checkBashisms "exec {n}>&2" -prop_checkBashisms29= verify checkBashisms "echo ${!var}" -prop_checkBashisms30= verify checkBashisms "printf -v '%s' \"$1\"" -prop_checkBashisms31= verify checkBashisms "printf '%q' \"$1\"" -prop_checkBashisms32= verifyNot checkBashisms "#!/bin/dash\n[ foo -nt bar ]" -prop_checkBashisms33= verify checkBashisms "#!/bin/sh\necho -n foo" -prop_checkBashisms34= verifyNot checkBashisms "#!/bin/dash\necho -n foo" -prop_checkBashisms35= verifyNot checkBashisms "#!/bin/dash\nlocal foo" -prop_checkBashisms36= verifyNot checkBashisms "#!/bin/dash\nread -p foo -r bar" -prop_checkBashisms37= verifyNot checkBashisms "HOSTNAME=foo; echo $HOSTNAME" -prop_checkBashisms38= verify checkBashisms "RANDOM=9; echo $RANDOM" -prop_checkBashisms39= verify checkBashisms "foo-bar() { true; }" -prop_checkBashisms40= verify checkBashisms "echo $(/dev/null" -prop_checkBashisms48= verifyNot checkBashisms "#!/bin/sh\necho $LINENO" -prop_checkBashisms49= verify checkBashisms "#!/bin/dash\necho $MACHTYPE" -prop_checkBashisms50= verify checkBashisms "#!/bin/sh\ncmd >& file" -prop_checkBashisms51= verifyNot checkBashisms "#!/bin/sh\ncmd 2>&1" -prop_checkBashisms52= verifyNot checkBashisms "#!/bin/sh\ncmd >&2" -prop_checkBashisms53= verifyNot checkBashisms "#!/bin/sh\nprintf -- -f\n" -prop_checkBashisms54= verify checkBashisms "#!/bin/sh\nfoo+=bar" -prop_checkBashisms55= verify checkBashisms "#!/bin/sh\necho ${@%foo}" -prop_checkBashisms56= verifyNot checkBashisms "#!/bin/sh\necho ${##}" -checkBashisms = ForShell [Sh, Dash] $ \t -> do +prop_checkBashisms10 = verify checkBashisms "echo ${var:4:12}" +prop_checkBashisms11 = verifyNot checkBashisms "echo ${var:-4}" +prop_checkBashisms12 = verify checkBashisms "echo ${var//foo/bar}" +prop_checkBashisms13 = verify checkBashisms "exec -c env" +prop_checkBashisms14 = verify checkBashisms "echo -n \"Foo: \"" +prop_checkBashisms15 = verify checkBashisms "let n++" +prop_checkBashisms16 = verify checkBashisms "echo $RANDOM" +prop_checkBashisms17 = verify checkBashisms "echo $((RANDOM%6+1))" +prop_checkBashisms18 = verify checkBashisms "foo &> /dev/null" +prop_checkBashisms19 = verify checkBashisms "foo > file*.txt" +prop_checkBashisms20 = verify checkBashisms "read -ra foo" +prop_checkBashisms21 = verify checkBashisms "[ -a foo ]" +prop_checkBashisms21b = verify checkBashisms "test -a foo" +prop_checkBashisms22 = verifyNot checkBashisms "[ foo -a bar ]" +prop_checkBashisms23 = verify checkBashisms "trap mything ERR INT" +prop_checkBashisms24 = verifyNot checkBashisms "trap mything INT TERM" +prop_checkBashisms25 = verify checkBashisms "cat < /dev/tcp/host/123" +prop_checkBashisms26 = verify checkBashisms "trap mything ERR SIGTERM" +prop_checkBashisms27 = verify checkBashisms "echo *[^0-9]*" +prop_checkBashisms28 = verify checkBashisms "exec {n}>&2" +prop_checkBashisms29 = verify checkBashisms "echo ${!var}" +prop_checkBashisms30 = verify checkBashisms "printf -v '%s' \"$1\"" +prop_checkBashisms31 = verify checkBashisms "printf '%q' \"$1\"" +prop_checkBashisms32 = verifyNot checkBashisms "#!/bin/dash\n[ foo -nt bar ]" +prop_checkBashisms33 = verify checkBashisms "#!/bin/sh\necho -n foo" +prop_checkBashisms34 = verifyNot checkBashisms "#!/bin/dash\necho -n foo" +prop_checkBashisms35 = verifyNot checkBashisms "#!/bin/dash\nlocal foo" +prop_checkBashisms36 = verifyNot checkBashisms "#!/bin/dash\nread -p foo -r bar" +prop_checkBashisms37 = verifyNot checkBashisms "HOSTNAME=foo; echo $HOSTNAME" +prop_checkBashisms38 = verify checkBashisms "RANDOM=9; echo $RANDOM" +prop_checkBashisms39 = verify checkBashisms "foo-bar() { true; }" +prop_checkBashisms40 = verify checkBashisms "echo $(/dev/null" +prop_checkBashisms48 = verifyNot checkBashisms "#!/bin/sh\necho $LINENO" +prop_checkBashisms49 = verify checkBashisms "#!/bin/dash\necho $MACHTYPE" +prop_checkBashisms50 = verify checkBashisms "#!/bin/sh\ncmd >& file" +prop_checkBashisms51 = verifyNot checkBashisms "#!/bin/sh\ncmd 2>&1" +prop_checkBashisms52 = verifyNot checkBashisms "#!/bin/sh\ncmd >&2" +prop_checkBashisms52b = verifyNot checkBashisms "#!/bin/sh\ncmd >& $var" +prop_checkBashisms52c = verify checkBashisms "#!/bin/sh\ncmd >& $dir/$var" +prop_checkBashisms53 = verifyNot checkBashisms "#!/bin/sh\nprintf -- -f\n" +prop_checkBashisms54 = verify checkBashisms "#!/bin/sh\nfoo+=bar" +prop_checkBashisms55 = verify checkBashisms "#!/bin/sh\necho ${@%foo}" +prop_checkBashisms56 = verifyNot checkBashisms "#!/bin/sh\necho ${##}" +prop_checkBashisms57 = verifyNot checkBashisms "#!/bin/dash\nulimit -c 0" +prop_checkBashisms58 = verify checkBashisms "#!/bin/sh\nulimit -c 0" +prop_checkBashisms59 = verify checkBashisms "#!/bin/sh\njobs -s" +prop_checkBashisms60 = verifyNot checkBashisms "#!/bin/sh\njobs -p" +prop_checkBashisms61 = verifyNot checkBashisms "#!/bin/sh\njobs -lp" +prop_checkBashisms62 = verify checkBashisms "#!/bin/sh\nexport -f foo" +prop_checkBashisms63 = verifyNot checkBashisms "#!/bin/sh\nexport -p" +prop_checkBashisms64 = verify checkBashisms "#!/bin/sh\nreadonly -a" +prop_checkBashisms65 = verifyNot checkBashisms "#!/bin/sh\nreadonly -p" +prop_checkBashisms66 = verifyNot checkBashisms "#!/bin/sh\ncd -P ." +prop_checkBashisms67 = verify checkBashisms "#!/bin/sh\ncd -P -e ." +prop_checkBashisms68 = verify checkBashisms "#!/bin/sh\numask -p" +prop_checkBashisms69 = verifyNot checkBashisms "#!/bin/sh\numask -S" +prop_checkBashisms70 = verify checkBashisms "#!/bin/sh\ntrap -l" +prop_checkBashisms71 = verify checkBashisms "#!/bin/sh\ntype -a ls" +prop_checkBashisms72 = verifyNot checkBashisms "#!/bin/sh\ntype ls" +prop_checkBashisms73 = verify checkBashisms "#!/bin/sh\nunset -n namevar" +prop_checkBashisms74 = verifyNot checkBashisms "#!/bin/sh\nunset -f namevar" +prop_checkBashisms75 = verifyNot checkBashisms "#!/bin/sh\necho \"-n foo\"" +prop_checkBashisms76 = verifyNot checkBashisms "#!/bin/sh\necho \"-ne foo\"" +prop_checkBashisms77 = verifyNot checkBashisms "#!/bin/sh\necho -Q foo" +prop_checkBashisms78 = verify checkBashisms "#!/bin/sh\necho -ne foo" +prop_checkBashisms79 = verify checkBashisms "#!/bin/sh\nhash -l" +prop_checkBashisms80 = verifyNot checkBashisms "#!/bin/sh\nhash -r" +prop_checkBashisms81 = verifyNot checkBashisms "#!/bin/dash\nhash -v" +prop_checkBashisms82 = verifyNot checkBashisms "#!/bin/sh\nset -v +o allexport -o errexit -C" +prop_checkBashisms83 = verifyNot checkBashisms "#!/bin/sh\nset --" +prop_checkBashisms84 = verify checkBashisms "#!/bin/sh\nset -o pipefail" +prop_checkBashisms85 = verify checkBashisms "#!/bin/sh\nset -B" +prop_checkBashisms86 = verifyNot checkBashisms "#!/bin/dash\nset -o emacs" +prop_checkBashisms87 = verify checkBashisms "#!/bin/sh\nset -o emacs" +prop_checkBashisms88 = verifyNot checkBashisms "#!/bin/sh\nset -- wget -o foo 'https://some.url'" +prop_checkBashisms89 = verifyNot checkBashisms "#!/bin/sh\nopts=$-\nset -\"$opts\"" +prop_checkBashisms90 = verifyNot checkBashisms "#!/bin/sh\nset -o \"$opt\"" +prop_checkBashisms91 = verify checkBashisms "#!/bin/sh\nwait -n" +prop_checkBashisms92 = verify checkBashisms "#!/bin/sh\necho $((16#FF))" +prop_checkBashisms93 = verify checkBashisms "#!/bin/sh\necho $(( 10#$(date +%m) ))" +prop_checkBashisms94 = verify checkBashisms "#!/bin/sh\n[ -v var ]" +prop_checkBashisms95 = verify checkBashisms "#!/bin/sh\necho $_" +prop_checkBashisms96 = verifyNot checkBashisms "#!/bin/dash\necho $_" +prop_checkBashisms97 = verify checkBashisms "#!/bin/sh\necho ${var,}" +prop_checkBashisms98 = verify checkBashisms "#!/bin/sh\necho ${var^^}" +prop_checkBashisms99 = verify checkBashisms "#!/bin/dash\necho [^f]oo" +prop_checkBashisms100 = verify checkBashisms "read -r" +prop_checkBashisms101 = verify checkBashisms "read" +prop_checkBashisms102 = verifyNot checkBashisms "read -r foo" +prop_checkBashisms103 = verifyNot checkBashisms "read foo" +prop_checkBashisms104 = verifyNot checkBashisms "read ''" +prop_checkBashisms105 = verifyNot checkBashisms "#!/bin/busybox sh\nset -o pipefail" +prop_checkBashisms106 = verifyNot checkBashisms "#!/bin/busybox sh\nx=x\n[[ \"$x\" = \"$x\" ]]" +prop_checkBashisms107 = verifyNot checkBashisms "#!/bin/busybox sh\nx=x\n[ \"$x\" == \"$x\" ]" +prop_checkBashisms108 = verifyNot checkBashisms "#!/bin/busybox sh\necho magic &> /dev/null" +prop_checkBashisms109 = verifyNot checkBashisms "#!/bin/busybox sh\ntrap stop EXIT SIGTERM" +prop_checkBashisms110 = verifyNot checkBashisms "#!/bin/busybox sh\nsource /dev/null" +prop_checkBashisms111 = verify checkBashisms "#!/bin/dash\nx='test'\n${x:0:3}" -- SC3057 +prop_checkBashisms112 = verifyNot checkBashisms "#!/bin/busybox sh\nx='test'\n${x:0:3}" -- SC3057 +prop_checkBashisms113 = verify checkBashisms "#!/bin/dash\nx='test'\n${x/st/xt}" -- SC3060 +prop_checkBashisms114 = verifyNot checkBashisms "#!/bin/busybox sh\nx='test'\n${x/st/xt}" -- SC3060 +prop_checkBashisms115 = verify checkBashisms "#!/bin/busybox sh\nx='test'\n${!x}" -- SC3053 +prop_checkBashisms116 = verify checkBashisms "#!/bin/busybox sh\nx='test'\n${x[1]}" -- SC3054 +prop_checkBashisms117 = verify checkBashisms "#!/bin/busybox sh\nx='test'\n${!x[@]}" -- SC3055 +prop_checkBashisms118 = verify checkBashisms "#!/bin/busybox sh\nxyz=1\n${!x*}" -- SC3056 +prop_checkBashisms119 = verify checkBashisms "#!/bin/busybox sh\nx='test'\n${x^^[t]}" -- SC3059 +prop_checkBashisms120 = verify checkBashisms "#!/bin/sh\n[ x == y ]" +prop_checkBashisms121 = verifyNot checkBashisms "#!/bin/sh\n# shellcheck shell=busybox\n[ x == y ]" +prop_checkBashisms122 = verify checkBashisms "#!/bin/dash\n$'a'" +prop_checkBashisms123 = verifyNot checkBashisms "#!/bin/busybox sh\n$'a'" +prop_checkBashisms124 = verify checkBashisms "#!/bin/dash\ntype -p test" +prop_checkBashisms125 = verifyNot checkBashisms "#!/bin/busybox sh\ntype -p test" +prop_checkBashisms126 = verifyNot checkBashisms "#!/bin/busybox sh\nread -p foo -r bar" +prop_checkBashisms127 = verifyNot checkBashisms "#!/bin/busybox sh\necho -ne foo" +prop_checkBashisms128 = verify checkBashisms "#!/bin/dash\ntype -p test" +prop_checkBashisms129 = verify checkBashisms "#!/bin/sh\n[ -k /tmp ]" +prop_checkBashisms130 = verifyNot checkBashisms "#!/bin/dash\ntest -k /tmp" +prop_checkBashisms131 = verify checkBashisms "#!/bin/sh\n[ -o errexit ]" +checkBashisms = ForShell [Sh, Dash, BusyboxSh] $ \t -> do params <- ask kludge params t where -- This code was copy-pasted from Analytics where params was a variable kludge params = bashism where - isDash = shellType params == Dash - warnMsg id s = + isBusyboxSh = shellType params == BusyboxSh + isDash = shellType params == Dash || isBusyboxSh + warnMsg id code s = if isDash - then warn id 2169 $ "In dash, " ++ s ++ " not supported." - else warn id 2039 $ "In POSIX sh, " ++ s ++ " undefined." + then err id code $ "In dash, " ++ s ++ " not supported." + else warn id code $ "In POSIX sh, " ++ s ++ " undefined." + asStr = getLiteralString + + bashism (T_ProcSub id _ _) = warnMsg id 3001 "process substitution is" + bashism (T_Extglob id _ _) = warnMsg id 3002 "extglob is" + bashism (T_DollarSingleQuoted id _) = + unless isBusyboxSh $ warnMsg id 3003 "$'..' is" + bashism (T_DollarDoubleQuoted id _) = warnMsg id 3004 "$\"..\" is" + bashism (T_ForArithmetic id _ _ _ _) = warnMsg id 3005 "arithmetic for loops are" + bashism (T_Arithmetic id _) = warnMsg id 3006 "standalone ((..)) is" + bashism (T_DollarBracket id _) = warnMsg id 3007 "$[..] in place of $((..)) is" + bashism (T_SelectIn id _ _ _) = warnMsg id 3008 "select loops are" + bashism (T_BraceExpansion id _) = warnMsg id 3009 "brace expansion is" + bashism (T_Condition id DoubleBracket _) = + unless isBusyboxSh $ warnMsg id 3010 "[[ ]] is" + bashism (T_HereString id _) = warnMsg id 3011 "here-strings are" + + bashism (TC_Binary id _ op _ _) = + checkTestOp bashismBinaryTestFlags op id + bashism (T_SimpleCommand id _ [asStr -> Just "test", lhs, asStr -> Just op, rhs]) = + checkTestOp bashismBinaryTestFlags op id + bashism (TC_Unary id _ op _) = + checkTestOp bashismUnaryTestFlags op id + bashism (T_SimpleCommand id _ [asStr -> Just "test", asStr -> Just op, _]) = + checkTestOp bashismUnaryTestFlags op id - bashism (T_ProcSub id _ _) = warnMsg id "process substitution is" - bashism (T_Extglob id _ _) = warnMsg id "extglob is" - bashism (T_DollarSingleQuoted id _) = warnMsg id "$'..' is" - bashism (T_DollarDoubleQuoted id _) = warnMsg id "$\"..\" is" - bashism (T_ForArithmetic id _ _ _ _) = warnMsg id "arithmetic for loops are" - bashism (T_Arithmetic id _) = warnMsg id "standalone ((..)) is" - bashism (T_DollarBracket id _) = warnMsg id "$[..] in place of $((..)) is" - bashism (T_SelectIn id _ _ _) = warnMsg id "select loops are" - bashism (T_BraceExpansion id _) = warnMsg id "brace expansion is" - bashism (T_Condition id DoubleBracket _) = warnMsg id "[[ ]] is" - bashism (T_HereString id _) = warnMsg id "here-strings are" - bashism (TC_Binary id SingleBracket op _ _) - | op `elem` [ "<", ">", "\\<", "\\>", "<=", ">=", "\\<=", "\\>="] = - unless isDash $ warnMsg id $ "lexicographical " ++ op ++ " is" - bashism (TC_Binary id SingleBracket op _ _) - | op `elem` [ "-nt", "-ef" ] = - unless isDash $ warnMsg id $ op ++ " is" - bashism (TC_Binary id SingleBracket "==" _ _) = - warnMsg id "== in place of = is" - bashism (TC_Binary id SingleBracket "=~" _ _) = - warnMsg id "=~ regex matching is" - bashism (TC_Unary id _ "-a" _) = - warnMsg id "unary -a in place of -e is" bashism (TA_Unary id op _) | op `elem` [ "|++", "|--", "++|", "--|"] = - warnMsg id $ filter (/= '|') op ++ " is" - bashism (TA_Binary id "**" _ _) = warnMsg id "exponentials are" - bashism (T_FdRedirect id "&" (T_IoFile _ (T_Greater _) _)) = warnMsg id "&> is" - bashism (T_FdRedirect id "" (T_IoFile _ (T_GREATAND _) _)) = warnMsg id ">& is" - bashism (T_FdRedirect id ('{':_) _) = warnMsg id "named file descriptors are" + warnMsg id 3018 $ filter (/= '|') op ++ " is" + bashism (TA_Binary id "**" _ _) = warnMsg id 3019 "exponentials are" + bashism (T_FdRedirect id "&" (T_IoFile _ (T_Greater _) _)) = + unless isBusyboxSh $ warnMsg id 3020 "&> is" + bashism (T_FdRedirect id "" (T_IoFile _ (T_GREATAND _) file)) = + unless (all isDigit $ onlyLiteralString file) $ warnMsg id 3021 ">& filename (as opposed to >& fd) is" + bashism (T_FdRedirect id ('{':_) _) = warnMsg id 3022 "named file descriptors are" bashism (T_FdRedirect id num _) - | all isDigit num && length num > 1 = warnMsg id "FDs outside 0-9 are" + | all isDigit num && length num > 1 = warnMsg id 3023 "FDs outside 0-9 are" bashism (T_Assignment id Append _ _ _) = - warnMsg id "+= is" + warnMsg id 3024 "+= is" bashism (T_IoFile id _ word) | isNetworked = - warnMsg id "/dev/{tcp,udp} is" + warnMsg id 3025 "/dev/{tcp,udp} is" where file = onlyLiteralString word isNetworked = any (`isPrefixOf` file) ["/dev/tcp", "/dev/udp"] bashism (T_Glob id str) | "[^" `isInfixOf` str = - warnMsg id "^ in place of ! in glob bracket expressions is" + warnMsg id 3026 "^ in place of ! in glob bracket expressions is" bashism t@(TA_Variable id str _) | isBashVariable str = - warnMsg id $ str ++ " is" + warnMsg id 3028 $ str ++ " is" - bashism t@(T_DollarBraced id token) = do - mapM_ check expansion + bashism t@(T_DollarBraced id _ token) = do + unless isBusyboxSh $ mapM_ check simpleExpansions + mapM_ check advancedExpansions when (isBashVariable var) $ - warnMsg id $ var ++ " is" + warnMsg id 3028 $ var ++ " is" where - str = bracedString t + str = concat $ oversimplify token var = getBracedReference str - check (regex, feature) = - when (isJust $ matchRegex regex str) $ warnMsg id feature + check (regex, code, feature) = + when (isJust $ matchRegex regex str) $ warnMsg id code feature bashism t@(T_Pipe id "|&") = - warnMsg id "|& in place of 2>&1 | is" + warnMsg id 3029 "|& in place of 2>&1 | is" bashism (T_Array id _) = - warnMsg id "arrays are" + warnMsg id 3030 "arrays are" bashism (T_IoFile id _ t) | isGlob t = - warnMsg id "redirecting to/from globs is" + warnMsg id 3031 "redirecting to/from globs is" bashism (T_CoProc id _ _) = - warnMsg id "coproc is" + warnMsg id 3032 "coproc is" bashism (T_Function id _ _ str _) | not (isVariableName str) = - warnMsg id "naming functions outside [a-zA-Z_][a-zA-Z0-9_]* is" + warnMsg id 3033 "naming functions outside [a-zA-Z_][a-zA-Z0-9_]* is" bashism (T_DollarExpansion id [x]) | isOnlyRedirection x = - warnMsg id "$( [(Id, String)] + getLiteralArgs = foldr go [] + where + go first rest = case getLiteralString first of + Just str -> (getId first, str) : rest + Nothing -> [] + + -- Check a flag-option pair (such as -o errexit) + checkOptions (flag@(fid,flag') : opt@(oid,opt') : rest) + | flag' `matches` oFlagRegex = do + when (opt' `notElem` longOptions) $ + warnMsg oid 3040 $ "set option " <> opt' <> " is" + checkFlags (flag:rest) + | otherwise = checkFlags (flag:opt:rest) + checkOptions (flag:rest) = checkFlags (flag:rest) + checkOptions _ = return () + + -- Check that each option in a sequence of flags + -- (such as -aveo) is valid + checkFlags (flag@(fid, flag'):rest) + | startsOption flag' = do + unless (flag' `matches` validFlagsRegex) $ + forM_ (tail flag') $ \letter -> + when (letter `notElem` optionsSet) $ + warnMsg fid 3041 $ "set flag " <> ('-':letter:" is") + checkOptions rest + | beginsWithDoubleDash flag' = do + warnMsg fid 3042 $ "set flag " <> flag' <> " is" + checkOptions rest + -- Either a word that doesn't start with a dash, or simply '--', + -- so stop checking. + | otherwise = return () + checkFlags [] = return () + + options = "abCefhmnuvxo" + optionsSet = Set.fromList options + startsOption = (`matches` mkRegex "^(\\+|-[^-])") + oFlagRegex = mkRegex $ "^[-+][" <> options <> "]*o$" + validFlagsRegex = mkRegex $ "^[-+]([" <> options <> "]+o?|o)$" + beginsWithDoubleDash = (`matches` mkRegex "^--.+$") + longOptions = Set.fromList + [ "allexport", "errexit", "ignoreeof", "monitor", "noclobber" + , "noexec", "noglob", "nolog", "notify" , "nounset", "verbose" + , "vi", "xtrace" ] bashism t@(T_SimpleCommand id _ (cmd:rest)) = let name = fromMaybe "" $ getCommandName t flags = getLeadingFlags t in do + when (name == "local" && not isDash) $ + -- This is so commonly accepted that we'll make it a special case + warnMsg id 3043 $ "'local' is" when (name `elem` unsupportedCommands) $ - warnMsg id $ "'" ++ name ++ "' is" - potentially $ do - allowed <- Map.lookup name allowedFlags - (word, flag) <- listToMaybe $ - filter (\x -> (not . null . snd $ x) && snd x `notElem` allowed) flags - return . warnMsg (getId word) $ name ++ " -" ++ flag ++ " is" + warnMsg id 3044 $ "'" ++ name ++ "' is" + sequence_ $ do + allowed' <- Map.lookup name allowedFlags + allowed <- allowed' + (word, flag) <- find + (\x -> (not . null . snd $ x) && snd x `notElem` allowed) flags + return . warnMsg (getId word) 3045 $ name ++ " -" ++ flag ++ " is" - when (name == "source") $ warnMsg id "'source' in place of '.' is" + when (name == "source" && not isBusyboxSh) $ + warnMsg id 3046 "'source' in place of '.' is" when (name == "trap") $ let - check token = potentially $ do + check token = sequence_ $ do str <- getLiteralString token let upper = map toUpper str return $ do when (upper `elem` ["ERR", "DEBUG", "RETURN"]) $ - warnMsg (getId token) $ "trapping " ++ str ++ " is" - when ("SIG" `isPrefixOf` upper) $ - warnMsg (getId token) + warnMsg (getId token) 3047 $ "trapping " ++ str ++ " is" + when (not isBusyboxSh && "SIG" `isPrefixOf` upper) $ + warnMsg (getId token) 3048 "prefixing signal names with 'SIG' is" when (not isDash && upper /= str) $ - warnMsg (getId token) + warnMsg (getId token) 3049 "using lower/mixed case for signal names is" in mapM_ check (drop 1 rest) - when (name == "printf") $ potentially $ do + when (name == "printf") $ sequence_ $ do format <- rest !!! 0 -- flags are covered by allowedFlags let literal = onlyLiteralString format guard $ "%q" `isInfixOf` literal - return $ warnMsg (getId format) "printf %q is" + return $ warnMsg (getId format) 3050 "printf %q is" + + when (name == "read" && all isFlag rest) $ + warnMsg (getId cmd) 3061 "read without a variable is" where unsupportedCommands = [ "let", "caller", "builtin", "complete", "compgen", "declare", "dirs", "disown", "enable", "mapfile", "readarray", "pushd", "popd", "shopt", "suspend", "typeset" - ] ++ if not isDash then ["local"] else [] - allowedFlags = Map.fromList [ - ("exec", []), - ("export", ["-p"]), - ("printf", []), - ("read", if isDash then ["r", "p"] else ["r"]), - ("ulimit", ["f"]) ] - bashism t@(T_SourceCommand id src _) = - let name = fromMaybe "" $ getCommandName src - in do - when (name == "source") $ warnMsg id "'source' in place of '.' is" + allowedFlags = Map.fromList [ + ("cd", Just ["L", "P"]), + ("exec", Just []), + ("export", Just ["p"]), + ("hash", Just $ if isDash then ["r", "v"] else ["r"]), + ("jobs", Just ["l", "p"]), + ("printf", Just []), + ("read", Just $ if isDash || isBusyboxSh then ["r", "p"] else ["r"]), + ("readonly", Just ["p"]), + ("trap", Just []), + ("type", Just $ if isBusyboxSh then ["p"] else []), + ("ulimit", if isDash then Nothing else Just ["f"]), + ("umask", Just ["S"]), + ("unset", Just ["f", "v"]), + ("wait", Just []) + ] + bashism t@(T_SourceCommand id src _) + | getCommandName src == Just "source" = + unless isBusyboxSh $ + warnMsg id 3051 "'source' in place of '.' is" + bashism (TA_Expansion _ (T_Literal id str : _)) + | str `matches` radix = warnMsg id 3052 "arithmetic base conversion is" + where + radix = mkRegex "^[0-9]+#" bashism _ = return () varChars="_0-9a-zA-Z" - expansion = let re = mkRegex in [ - (re $ "^![" ++ varChars ++ "]", "indirect expansion is"), - (re $ "^[" ++ varChars ++ "]+\\[.*\\]$", "array references are"), - (re $ "^![" ++ varChars ++ "]+\\[[*@]]$", "array key expansion is"), - (re $ "^![" ++ varChars ++ "]+[*@]$", "name matching prefixes are"), - (re $ "^[" ++ varChars ++ "*@]+:[^-=?+]", "string indexing is"), - (re $ "^([*@][%#]|#[@*])", "string operations on $@/$* are"), - (re $ "^[" ++ varChars ++ "*@]+(\\[.*\\])?/", "string replacement is") + advancedExpansions = let re = mkRegex in [ + (re $ "^![" ++ varChars ++ "]", 3053, "indirect expansion is"), + (re $ "^[" ++ varChars ++ "]+\\[.*\\]$", 3054, "array references are"), + (re $ "^![" ++ varChars ++ "]+\\[[*@]]$", 3055, "array key expansion is"), + (re $ "^![" ++ varChars ++ "]+[*@]$", 3056, "name matching prefixes are"), + (re $ "^[" ++ varChars ++ "*@]+(\\[.*\\])?[,^]", 3059, "case modification is") + ] + simpleExpansions = let re = mkRegex in [ + (re $ "^[" ++ varChars ++ "*@]+:[^-=?+]", 3057, "string indexing is"), + (re $ "^([*@][%#]|#[@*])", 3058, "string operations on $@/$* are"), + (re $ "^[" ++ varChars ++ "*@]+(\\[.*\\])?/", 3060, "string replacement is") ] bashVars = [ + -- This list deliberately excludes $BASH_VERSION as it's often used + -- for shell identification. "OSTYPE", "MACHTYPE", "HOSTTYPE", "HOSTNAME", - "DIRSTACK", "EUID", "UID", "SHLVL", "PIPESTATUS", "SHELLOPTS" + "DIRSTACK", "EUID", "UID", "SHLVL", "PIPESTATUS", "SHELLOPTS", + "_", "BASHOPTS", "BASHPID", "BASH_ALIASES", "BASH_ARGC", + "BASH_ARGV", "BASH_ARGV0", "BASH_CMDS", "BASH_COMMAND", + "BASH_EXECUTION_STRING", "BASH_LINENO", "BASH_REMATCH", "BASH_SOURCE", + "BASH_SUBSHELL", "BASH_VERSINFO", "EPOCHREALTIME", "EPOCHSECONDS", + "FUNCNAME", "GROUPS", "MACHTYPE", "MAPFILE" ] bashDynamicVars = [ "RANDOM", "SECONDS" ] - dashVars = [ ] + dashVars = [ "_" ] isBashVariable var = (var `elem` bashDynamicVars || var `elem` bashVars && not (isAssigned var)) @@ -317,32 +497,89 @@ checkBashisms = ForShell [Sh, Dash] $ \t -> do Assignment (_, _, name, _) -> name == var _ -> False + checkTestOp table op id = sequence_ $ do + (code, shells, msg) <- Map.lookup op table + guard . not $ shellType params `elem` shells + return $ warnMsg id code (msg op) + + +buildTestFlagMap list = Map.fromList $ concatMap (\(x,y) -> map (\c -> (c,y)) x) list +bashismBinaryTestFlags = buildTestFlagMap [ + -- ([list of applicable flags], + -- (error code, exempt shells, message builder :: String -> String)), + -- + -- Distinct error codes allow the wiki to give more helpful, targeted + -- information. + (["<", ">", "\\<", "\\>", "<=", ">=", "\\<=", "\\>="], + (3012, [Dash, BusyboxSh], \op -> "lexicographical " ++ op ++ " is")), + (["=="], + (3014, [BusyboxSh], \op -> op ++ " in place of = is")), + (["=~"], + (3015, [], \op -> op ++ " regex matching is")), + + ([], (0,[],const "")) + ] +bashismUnaryTestFlags = buildTestFlagMap [ + (["-v"], + (3016, [], \op -> "test " ++ op ++ " (in place of [ -n \"${var+x}\" ]) is")), + (["-a"], + (3017, [], \op -> "unary " ++ op ++ " in place of -e is")), + (["-o"], + (3062, [], \op -> "test " ++ op ++ " to check options is")), + (["-R"], + (3063, [], \op -> "test " ++ op ++ " and namerefs in general are")), + (["-N"], + (3064, [], \op -> "test " ++ op ++ " is")), + (["-k"], + (3065, [Dash, BusyboxSh], \op -> "test " ++ op ++ " is")), + (["-G"], + (3066, [Dash, BusyboxSh], \op -> "test " ++ op ++ " is")), + (["-O"], + (3067, [Dash, BusyboxSh], \op -> "test " ++ op ++ " is")), + + ([], (0,[],const "")) + ] + + prop_checkEchoSed1 = verify checkEchoSed "FOO=$(echo \"$cow\" | sed 's/foo/bar/g')" +prop_checkEchoSed1b = verify checkEchoSed "FOO=$(sed 's/foo/bar/g' <<< \"$cow\")" prop_checkEchoSed2 = verify checkEchoSed "rm $(echo $cow | sed -e 's,foo,bar,')" +prop_checkEchoSed2b = verify checkEchoSed "rm $(sed -e 's,foo,bar,' <<< $cow)" checkEchoSed = ForShell [Bash, Ksh] f where + f (T_Redirecting id lefts r) = + when (any redirectHereString lefts) $ + checkSed id rcmd + where + redirectHereString :: Token -> Bool + redirectHereString t = case t of + (T_FdRedirect _ _ T_HereString{}) -> True + _ -> False + rcmd = oversimplify r + f (T_Pipeline id _ [a, b]) = when (acmd == ["echo", "${VAR}"]) $ - case bcmd of - ["sed", v] -> checkIn v - ["sed", "-e", v] -> checkIn v - _ -> return () + checkSed id bcmd where - -- This should have used backreferences, but TDFA doesn't support them - sedRe = mkRegex "^s(.)([^\n]*)g?$" - isSimpleSed s = fromMaybe False $ do - [first,rest] <- matchRegex sedRe s - let delimiters = filter (== head first) rest - guard $ length delimiters == 2 - return True - acmd = oversimplify a bcmd = oversimplify b - checkIn s = - when (isSimpleSed s) $ - style id 2001 "See if you can use ${variable//search/replace} instead." + f _ = return () + checkSed id ["sed", v] = checkIn id v + checkSed id ["sed", "-e", v] = checkIn id v + checkSed _ _ = return () + + -- This should have used backreferences, but TDFA doesn't support them + sedRe = mkRegex "^s(.)([^\n]*)g?$" + isSimpleSed s = isJust $ do + [h:_,rest] <- matchRegex sedRe s + let delimiters = filter (== h) rest + guard $ length delimiters == 2 + checkIn id s = + when (isSimpleSed s) $ + style id 2001 "See if you can use ${variable//search/replace} instead." + prop_checkBraceExpansionVars1 = verify checkBraceExpansionVars "echo {1..$n}" prop_checkBraceExpansionVars2 = verifyNot checkBraceExpansionVars "echo {1,3,$n}" @@ -365,11 +602,11 @@ checkBraceExpansionVars = ForShell [Bash] f T_DollarBraced {} -> return "$" T_DollarExpansion {} -> return "$" T_DollarArithmetic {} -> return "$" - otherwise -> return "-" - toString t = fromJust $ getLiteralStringExt literalExt t + _ -> return "-" + toString t = runIdentity $ getLiteralStringExt literalExt t isEvaled t = do cmd <- getClosestCommandM t - return $ isJust cmd && fromJust cmd `isUnqualifiedCommand` "eval" + return $ maybe False (`isUnqualifiedCommand` "eval") cmd prop_checkMultiDimensionalArrays1 = verify checkMultiDimensionalArrays "foo[a][b]=3" @@ -384,20 +621,20 @@ checkMultiDimensionalArrays = ForShell [Bash] f case token of T_Assignment _ _ name (first:second:_) _ -> about second T_IndexedElement _ (first:second:_) _ -> about second - T_DollarBraced {} -> - when (isMultiDim token) $ about token + T_DollarBraced _ _ l -> + when (isMultiDim l) $ about token _ -> return () about t = warn (getId t) 2180 "Bash does not support multidimensional arrays. Use 1D or associative arrays." re = mkRegex "^\\[.*\\]\\[.*\\]" -- Fixme, this matches ${foo:- [][]} and such as well - isMultiDim t = getBracedModifier (bracedString t) `matches` re + isMultiDim l = getBracedModifier (concat $ oversimplify l) `matches` re prop_checkPS11 = verify checkPS1Assignments "PS1='\\033[1;35m\\$ '" -prop_checkPS11a= verify checkPS1Assignments "export PS1='\\033[1;35m\\$ '" +prop_checkPS11a = verify checkPS1Assignments "export PS1='\\033[1;35m\\$ '" prop_checkPSf2 = verify checkPS1Assignments "PS1='\\h \\e[0m\\$ '" prop_checkPS13 = verify checkPS1Assignments "PS1=$'\\x1b[c '" prop_checkPS14 = verify checkPS1Assignments "PS1=$'\\e[3m; '" -prop_checkPS14a= verify checkPS1Assignments "export PS1=$'\\e[3m; '" +prop_checkPS14a = verify checkPS1Assignments "export PS1=$'\\e[3m; '" prop_checkPS15 = verifyNot checkPS1Assignments "PS1='\\[\\033[1;35m\\]\\$ '" prop_checkPS16 = verifyNot checkPS1Assignments "PS1='\\[\\e1m\\e[1m\\]\\$ '" prop_checkPS17 = verifyNot checkPS1Assignments "PS1='e033x1B'" @@ -419,5 +656,46 @@ checkPS1Assignments = ForShell [Bash] f escapeRegex = mkRegex "\\\\x1[Bb]|\\\\e|\x1B|\\\\033" +prop_checkMultipleBangs1 = verify checkMultipleBangs "! ! true" +prop_checkMultipleBangs2 = verifyNot checkMultipleBangs "! true" +checkMultipleBangs = ForShell [Dash, BusyboxSh, Sh] f + where + f token = case token of + T_Banged id (T_Banged _ _) -> + err id 2325 "Multiple ! in front of pipelines are a bash/ksh extension. Use only 0 or 1." + _ -> return () + + +prop_checkBangAfterPipe1 = verify checkBangAfterPipe "true | ! true" +prop_checkBangAfterPipe2 = verifyNot checkBangAfterPipe "true | ( ! true )" +prop_checkBangAfterPipe3 = verifyNot checkBangAfterPipe "! ! true | true" +checkBangAfterPipe = ForShell [Dash, BusyboxSh, Sh, Bash] f + where + f token = case token of + T_Pipeline _ _ cmds -> mapM_ check cmds + _ -> return () + + check token = case token of + T_Banged id _ -> + err id 2326 "! is not allowed in the middle of pipelines. Use command group as in cmd | { ! cmd; } if necessary." + _ -> return () + + +prop_checkNegatedUnaryOps1 = verify checkNegatedUnaryOps "[ ! -o braceexpand ]" +prop_checkNegatedUnaryOps2 = verifyNot checkNegatedUnaryOps "[ -o braceexpand ]" +prop_checkNegatedUnaryOps3 = verifyNot checkNegatedUnaryOps "[[ ! -o braceexpand ]]" +prop_checkNegatedUnaryOps4 = verifyNot checkNegatedUnaryOps "! [ -o braceexpand ]" +prop_checkNegatedUnaryOps5 = verify checkNegatedUnaryOps "[ ! -a file ]" +checkNegatedUnaryOps = ForShell [Bash] f + where + f token = case token of + TC_Unary id SingleBracket "!" (TC_Unary _ _ op _) | op `elem` ["-a", "-o"] -> + err id 2332 $ msg op + _ -> return () + + msg "-o" = "[ ! -o opt ] is always true because -o becomes logical OR. Use [[ ]] or ! [ -o opt ]." + msg "-a" = "[ ! -a file ] is always true because -a becomes logical AND. Use -e instead." + msg _ = pleaseReport "unhandled negated unary message" + return [] runTests = $( [| $(forAllProperties) (quickCheckWithResult (stdArgs { maxSuccess = 1 }) ) |]) diff --git a/src/ShellCheck/Data.hs b/src/ShellCheck/Data.hs index 8ce0026..688d0d7 100644 --- a/src/ShellCheck/Data.hs +++ b/src/ShellCheck/Data.hs @@ -2,9 +2,27 @@ module ShellCheck.Data where import ShellCheck.Interface import Data.Version (showVersion) -import Paths_ShellCheck (version) -shellcheckVersion = showVersion version + +{- +If you are here because you saw an error about Paths_ShellCheck in this file, +simply comment out the import below and define the version as a constant string. + +Instead of: + + import Paths_ShellCheck (version) + shellcheckVersion = showVersion version + +Use: + + -- import Paths_ShellCheck (version) + shellcheckVersion = "kludge" + +-} + +import Paths_ShellCheck (version) +shellcheckVersion = showVersion version -- VERSIONSTRING + internalVariables = [ -- Generic @@ -12,23 +30,27 @@ internalVariables = [ -- Bash "BASH", "BASHOPTS", "BASHPID", "BASH_ALIASES", "BASH_ARGC", - "BASH_ARGV", "BASH_CMDS", "BASH_COMMAND", "BASH_EXECUTION_STRING", - "BASH_LINENO", "BASH_REMATCH", "BASH_SOURCE", "BASH_SUBSHELL", - "BASH_VERSINFO", "BASH_VERSION", "COMP_CWORD", "COMP_KEY", - "COMP_LINE", "COMP_POINT", "COMP_TYPE", "COMP_WORDBREAKS", - "COMP_WORDS", "COPROC", "DIRSTACK", "EUID", "FUNCNAME", "GROUPS", - "HISTCMD", "HOSTNAME", "HOSTTYPE", "LINENO", "MACHTYPE", "MAPFILE", - "OLDPWD", "OPTARG", "OPTIND", "OSTYPE", "PIPESTATUS", "PPID", "PWD", - "RANDOM", "READLINE_LINE", "READLINE_POINT", "REPLY", "SECONDS", - "SHELLOPTS", "SHLVL", "UID", "BASH_ENV", "BASH_XTRACEFD", "CDPATH", - "COLUMNS", "COMPREPLY", "EMACS", "ENV", "FCEDIT", "FIGNORE", + "BASH_ARGV", "BASH_ARGV0", "BASH_CMDS", "BASH_COMMAND", + "BASH_EXECUTION_STRING", "BASH_LINENO", "BASH_LOADABLES_PATH", + "BASH_REMATCH", "BASH_SOURCE", "BASH_SUBSHELL", "BASH_VERSINFO", + "BASH_VERSION", "COMP_CWORD", "COMP_KEY", "COMP_LINE", "COMP_POINT", + "COMP_TYPE", "COMP_WORDBREAKS", "COMP_WORDS", "COPROC", "DIRSTACK", + "EPOCHREALTIME", "EPOCHSECONDS", "EUID", "FUNCNAME", "GROUPS", "HISTCMD", + "HOSTNAME", "HOSTTYPE", "LINENO", "MACHTYPE", "MAPFILE", "OLDPWD", + "OPTARG", "OPTIND", "OSTYPE", "PIPESTATUS", "PPID", "PWD", "RANDOM", + "READLINE_ARGUMENT", "READLINE_LINE", "READLINE_MARK", "READLINE_POINT", + "REPLY", "SECONDS", "SHELLOPTS", "SHLVL", "SRANDOM", "UID", "BASH_COMPAT", + "BASH_ENV", "BASH_XTRACEFD", "CDPATH", "CHILD_MAX", "COLUMNS", + "COMPREPLY", "EMACS", "ENV", "EXECIGNORE", "FCEDIT", "FIGNORE", "FUNCNEST", "GLOBIGNORE", "HISTCONTROL", "HISTFILE", "HISTFILESIZE", "HISTIGNORE", "HISTSIZE", "HISTTIMEFORMAT", "HOME", "HOSTFILE", "IFS", - "IGNOREEOF", "INPUTRC", "LANG", "LC_ALL", "LC_COLLATE", "LC_CTYPE", - "LC_MESSAGES", "LC_MONETARY", "LC_NUMERIC", "LC_TIME", "LINES", "MAIL", - "MAILCHECK", "MAILPATH", "OPTERR", "PATH", "POSIXLY_CORRECT", - "PROMPT_COMMAND", "PROMPT_DIRTRIM", "PS1", "PS2", "PS3", "PS4", "SHELL", - "TIMEFORMAT", "TMOUT", "TMPDIR", "auto_resume", "histchars", "COPROC", + "IGNOREEOF", "INPUTRC", "INSIDE_EMACS", "LANG", "LC_ALL", "LC_COLLATE", + "LC_CTYPE", "LC_MESSAGES", "LC_MONETARY", "LC_NUMERIC", "LC_TIME", + "LINES", "MAIL", "MAILCHECK", "MAILPATH", "OPTERR", "PATH", + "POSIXLY_CORRECT", "PROMPT_COMMAND", "PROMPT_DIRTRIM", "PS0", "PS1", + "PS2", "PS3", "PS4", "SHELL", "TIMEFORMAT", "TMOUT", "TMPDIR", + "BASH_MONOSECONDS", "BASH_TRAPSIG", "GLOBSORT", + "auto_resume", "histchars", -- Other "USER", "TZ", "TERM", "LOGNAME", "LD_LIBRARY_PATH", "LANGUAGE", "DISPLAY", @@ -36,13 +58,37 @@ internalVariables = [ -- Ksh , ".sh.version" + + -- shflags + , "FLAGS_ARGC", "FLAGS_ARGV", "FLAGS_ERROR", "FLAGS_FALSE", "FLAGS_HELP", + "FLAGS_PARENT", "FLAGS_RESERVED", "FLAGS_TRUE", "FLAGS_VERSION", + "flags_error", "flags_return" + + -- Bats + ,"stderr", "stderr_lines" ] -variablesWithoutSpaces = [ - "$", "-", "?", "!", "#", - "BASHPID", "BASH_ARGC", "BASH_LINENO", "BASH_SUBSHELL", "EUID", "LINENO", - "OPTIND", "PPID", "RANDOM", "SECONDS", "SHELLOPTS", "SHLVL", "UID", - "COLUMNS", "HISTFILESIZE", "HISTSIZE", "LINES" +specialIntegerVariables = [ + "$", "?", "!", "#" + ] + +specialVariablesWithoutSpaces = "-" : specialIntegerVariables + +variablesWithoutSpaces = specialVariablesWithoutSpaces ++ [ + "BASHPID", "BASH_ARGC", "BASH_LINENO", "BASH_SUBSHELL", "EUID", + "EPOCHREALTIME", "EPOCHSECONDS", "LINENO", "OPTIND", "PPID", "RANDOM", + "READLINE_ARGUMENT", "READLINE_MARK", "READLINE_POINT", "SECONDS", + "SHELLOPTS", "SHLVL", "SRANDOM", "UID", "COLUMNS", "HISTFILESIZE", + "HISTSIZE", "LINES", "BASH_MONOSECONDS", "BASH_TRAPSIG" + + -- shflags + , "FLAGS_ERROR", "FLAGS_FALSE", "FLAGS_TRUE" + ] + +specialVariables = specialVariablesWithoutSpaces ++ ["@", "*"] + +unbracedVariables = specialVariables ++ [ + "0", "1", "2", "3", "4", "5", "6", "7", "8", "9" ] arrayVariables = [ @@ -79,10 +125,10 @@ commonCommands = [ nonReadingCommands = [ "alias", "basename", "bg", "cal", "cd", "chgrp", "chmod", "chown", - "cp", "du", "echo", "export", "false", "fg", "fuser", "getconf", + "cp", "du", "echo", "export", "fg", "fuser", "getconf", "getopt", "getopts", "ipcrm", "ipcs", "jobs", "kill", "ln", "ls", "locale", "mv", "printf", "ps", "pwd", "renice", "rm", "rmdir", - "set", "sleep", "touch", "trap", "true", "ulimit", "unalias", "uname" + "set", "sleep", "touch", "trap", "ulimit", "unalias", "uname" ] sampleWords = [ @@ -98,6 +144,10 @@ binaryTestOps = [ "-gt", "-ge", "=~", ">", "<", "=", "\\<", "\\>", "\\<=", "\\>=" ] +arithmeticBinaryTestOps = [ + "-eq", "-ne", "-lt", "-le", "-gt", "-ge" + ] + unaryTestOps = [ "!", "-a", "-b", "-c", "-d", "-e", "-f", "-g", "-h", "-L", "-k", "-p", "-r", "-s", "-S", "-t", "-u", "-w", "-x", "-O", "-G", "-N", "-z", "-n", @@ -109,9 +159,19 @@ shellForExecutable name = case name of "sh" -> return Sh "bash" -> return Bash + "bats" -> return Bash + "busybox" -> return BusyboxSh -- Used for directives and --shell=busybox + "busybox sh" -> return BusyboxSh + "busybox ash" -> return BusyboxSh "dash" -> return Dash "ash" -> return Dash -- There's also a warning for this. "ksh" -> return Ksh "ksh88" -> return Ksh "ksh93" -> return Ksh - otherwise -> Nothing + "oksh" -> return Ksh + _ -> Nothing + +flagsForRead = "sreu:n:N:i:p:a:t:" +flagsForMapfile = "d:n:O:s:u:C:c:t" + +declaringCommands = ["local", "declare", "export", "readonly", "typeset", "let"] diff --git a/src/ShellCheck/Debug.hs b/src/ShellCheck/Debug.hs new file mode 100644 index 0000000..b6015e5 --- /dev/null +++ b/src/ShellCheck/Debug.hs @@ -0,0 +1,313 @@ +{- + +This file contains useful functions for debugging and developing ShellCheck. + +To invoke them interactively, run: + + cabal repl + +At the ghci prompt, enter: + + :load ShellCheck.Debug + +You can now invoke the functions. Here are some examples: + + shellcheckString "echo $1" + stringToAst "(( x+1 ))" + stringToCfg "if foo; then bar; else baz; fi" + writeFile "/tmp/test.dot" $ stringToCfgViz "while foo; do bar; done" + +The latter file can be rendered to png with GraphViz: + + dot -Tpng /tmp/test.dot > /tmp/test.png + +To run all unit tests in a module: + + ShellCheck.Parser.runTests + ShellCheck.Analytics.runTests + +To run a specific test: + + :load ShellCheck.Analytics + prop_checkUuoc3 + +If you make code changes, reload in seconds at any time with: + + :r + +=========================================================================== + +Crash course in printf debugging in Haskell: + + import Debug.Trace + + greet 0 = return () + -- Print when a function is invoked + greet n | trace ("calling greet " ++ show n) False = undefined + greet n = do + putStrLn "Enter name" + name <- getLine + -- Print at some point in any monadic function + traceM $ "user entered " ++ name + putStrLn $ "Hello " ++ name + -- Print a value before passing it on + greet $ traceShowId (n - 1) + + +=========================================================================== + +If you want to invoke `ghci` directly, such as on `shellcheck.hs`, to +debug all of ShellCheck including I/O, you may see an error like this: + + src/ShellCheck/Data.hs:5:1: error: + Could not load module ‘Paths_ShellCheck’ + it is a hidden module in the package ‘ShellCheck-0.8.0’ + +This can easily be circumvented by running `./setgitversion` or manually +editing src/ShellCheck/Data.hs to replace the auto-deduced version number +with a constant string as indicated. + +Afterwards, you can run the ShellCheck tool, as if from the shell, with: + + $ ghci shellcheck.hs + ghci> runMain ["-x", "file.sh"] + +-} + +module ShellCheck.Debug () where + +import ShellCheck.Analyzer +import ShellCheck.AST +import ShellCheck.CFG +import ShellCheck.Checker +import ShellCheck.CFGAnalysis as CF +import ShellCheck.Interface +import ShellCheck.Parser +import ShellCheck.Prelude + +import Control.Monad +import Control.Monad.Identity +import Control.Monad.RWS +import Control.Monad.Writer +import Data.Graph.Inductive.Graph as G +import Data.List +import Data.Maybe +import qualified Data.Map as M +import qualified Data.Set as S + + +-- Run all of ShellCheck (minus output formatters) +shellcheckString :: String -> CheckResult +shellcheckString scriptString = + runIdentity $ checkScript dummySystemInterface checkSpec + where + checkSpec :: CheckSpec + checkSpec = emptyCheckSpec { + csScript = scriptString + } + +dummySystemInterface :: SystemInterface Identity +dummySystemInterface = mockedSystemInterface [ + -- A tiny, fake filesystem for sourced files + ("lib/mylib1.sh", "foo=$(cat $1 | wc -l)"), + ("lib/mylib2.sh", "bar=42") + ] + +-- Parameters used when generating Control Flow Graphs +cfgParams :: CFGParameters +cfgParams = CFGParameters { + cfLastpipe = False, + cfPipefail = False +} + +-- An example script to play with +exampleScript :: String +exampleScript = unlines [ + "#!/bin/sh", + "count=0", + "for file in *", + "do", + " (( count++ ))", + "done", + "echo $count" + ] + +-- Parse the script string into ShellCheck's ParseResult +parseScriptString :: String -> ParseResult +parseScriptString scriptString = + runIdentity $ parseScript dummySystemInterface parseSpec + where + parseSpec :: ParseSpec + parseSpec = newParseSpec { + psFilename = "myscript", + psScript = scriptString + } + + +-- Parse the script string into an Abstract Syntax Tree +stringToAst :: String -> Token +stringToAst scriptString = + case maybeRoot of + Just root -> root + Nothing -> error $ "Script failed to parse: " ++ show parserWarnings + where + parseResult :: ParseResult + parseResult = parseScriptString scriptString + + maybeRoot :: Maybe Token + maybeRoot = prRoot parseResult + + parserWarnings :: [PositionedComment] + parserWarnings = prComments parseResult + + +astToCfgResult :: Token -> CFGResult +astToCfgResult = buildGraph cfgParams + +astToDfa :: Token -> CFGAnalysis +astToDfa = analyzeControlFlow cfgParams + +astToCfg :: Token -> CFGraph +astToCfg = cfGraph . astToCfgResult + +stringToCfg :: String -> CFGraph +stringToCfg = astToCfg . stringToAst + +stringToDfa :: String -> CFGAnalysis +stringToDfa = astToDfa . stringToAst + +cfgToGraphViz :: CFGraph -> String +cfgToGraphViz = cfgToGraphVizWith show + +stringToCfgViz :: String -> String +stringToCfgViz = cfgToGraphViz . stringToCfg + +stringToDfaViz :: String -> String +stringToDfaViz = dfaToGraphViz . stringToDfa + +-- Dump a Control Flow Graph as GraphViz with extended information +stringToDetailedCfgViz :: String -> String +stringToDetailedCfgViz scriptString = cfgToGraphVizWith nodeLabel graph + where + ast :: Token + ast = stringToAst scriptString + + cfgResult :: CFGResult + cfgResult = astToCfgResult ast + + graph :: CFGraph + graph = cfGraph cfgResult + + idToToken :: M.Map Id Token + idToToken = M.fromList $ execWriter $ doAnalysis (\c -> tell [(getId c, c)]) ast + + idToNode :: M.Map Id (Node, Node) + idToNode = cfIdToRange cfgResult + + nodeToStartIds :: M.Map Node (S.Set Id) + nodeToStartIds = + M.fromListWith S.union $ + map (\(id, (start, _)) -> (start, S.singleton id)) $ + M.toList idToNode + + nodeToEndIds :: M.Map Node (S.Set Id) + nodeToEndIds = + M.fromListWith S.union $ + map (\(id, (_, end)) -> (end, S.singleton id)) $ + M.toList idToNode + + formatId :: Id -> String + formatId id = fromMaybe ("Unknown " ++ show id) $ do + (OuterToken _ token) <- M.lookup id idToToken + firstWord <- words (show token) !!! 0 + -- Strip off "Inner_" + (_ : tokenName) <- return $ dropWhile (/= '_') firstWord + return $ tokenName ++ " " ++ show id + + formatGroup :: S.Set Id -> String + formatGroup set = intercalate ", " $ map formatId $ S.toList set + + nodeLabel (node, label) = unlines [ + show node ++ ". " ++ show label, + "Begin: " ++ formatGroup (M.findWithDefault S.empty node nodeToStartIds), + "End: " ++ formatGroup (M.findWithDefault S.empty node nodeToEndIds) + ] + + +-- Dump a Control Flow Graph with Data Flow Analysis as GraphViz +dfaToGraphViz :: CF.CFGAnalysis -> String +dfaToGraphViz analysis = cfgToGraphVizWith label $ CF.graph analysis + where + label (node, label) = + let + desc = show node ++ ". " ++ show label + in + fromMaybe ("No DFA available\n\n" ++ desc) $ do + (pre, post) <- M.lookup node $ CF.nodeToData analysis + return $ unlines [ + "Precondition: " ++ show pre, + "", + desc, + "", + "Postcondition: " ++ show post + ] + + +-- Dump an Control Flow Graph to GraphViz with a given node formatter +cfgToGraphVizWith :: (LNode CFNode -> String) -> CFGraph -> String +cfgToGraphVizWith nodeLabel graph = concat [ + "digraph {\n", + concatMap dumpNode (labNodes graph), + concatMap dumpLink (labEdges graph), + tagVizEntries graph, + "}\n" + ] + where + dumpNode l@(node, label) = show node ++ " [label=" ++ quoteViz (nodeLabel l) ++ "]\n" + dumpLink (from, to, typ) = show from ++ " -> " ++ show to ++ " [style=" ++ quoteViz (edgeStyle typ) ++ "]\n" + edgeStyle CFEFlow = "solid" + edgeStyle CFEExit = "bold" + edgeStyle CFEFalseFlow = "dotted" + +quoteViz str = "\"" ++ escapeViz str ++ "\"" +escapeViz [] = [] +escapeViz (c:rest) = + case c of + '\"' -> '\\' : '\"' : escapeViz rest + '\n' -> '\\' : 'l' : escapeViz rest + '\\' -> '\\' : '\\' : escapeViz rest + _ -> c : escapeViz rest + + +-- Dump an Abstract Syntax Tree (or branch thereof) to GraphViz format +astToGraphViz :: Token -> String +astToGraphViz token = concat [ + "digraph {\n", + formatTree token, + "}\n" + ] + where + formatTree :: Token -> String + formatTree t = snd $ execRWS (doStackAnalysis push pop t) () [] + + push :: Token -> RWS () String [Int] () + push (OuterToken (Id n) inner) = do + stack <- get + put (n : stack) + case stack of + [] -> return () + (top:_) -> tell $ show top ++ " -> " ++ show n ++ "\n" + tell $ show n ++ " [label=" ++ quoteViz (show n ++ ": " ++ take 32 (show inner)) ++ "]\n" + + pop :: Token -> RWS () String [Int] () + pop _ = modify tail + + +-- For each entry point, set the rank so that they'll align in the graph +tagVizEntries :: CFGraph -> String +tagVizEntries graph = "{ rank=same " ++ rank ++ " }" + where + entries = mapMaybe find $ labNodes graph + find (node, CFEntryPoint name) = return (node, name) + find _ = Nothing + rank = unwords $ map (\(c, _) -> show c) entries diff --git a/src/ShellCheck/Fixer.hs b/src/ShellCheck/Fixer.hs new file mode 100644 index 0000000..0d3c8f4 --- /dev/null +++ b/src/ShellCheck/Fixer.hs @@ -0,0 +1,412 @@ +{- + Copyright 2018-2019 Vidar Holen, Ng Zhi An + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} + +{-# LANGUAGE TemplateHaskell #-} +module ShellCheck.Fixer (applyFix, removeTabStops, mapPositions, Ranged(..), runTests) where + +import ShellCheck.Interface +import ShellCheck.Prelude +import Control.Monad +import Control.Monad.State +import Data.Array +import Data.List +import Data.Semigroup +import GHC.Exts (sortWith) +import Test.QuickCheck + +-- The Ranged class is used for types that has a start and end position. +class Ranged a where + start :: a -> Position + end :: a -> Position + overlap :: a -> a -> Bool + overlap x y = + xEnd > yStart && yEnd > xStart + where + yStart = start y + yEnd = end y + xStart = start x + xEnd = end x + -- Set a new start and end position on a Ranged + setRange :: (Position, Position) -> a -> a + +-- Tests auto-verify that overlap commutes +assertOverlap x y = overlap x y && overlap y x +assertNoOverlap x y = not (overlap x y) && not (overlap y x) + +prop_overlap_contiguous = assertNoOverlap + (tFromStart 10 12 "foo" 1) + (tFromStart 12 14 "bar" 2) + +prop_overlap_adjacent_zerowidth = assertNoOverlap + (tFromStart 3 3 "foo" 1) + (tFromStart 3 3 "bar" 2) + +prop_overlap_enclosed = assertOverlap + (tFromStart 3 5 "foo" 1) + (tFromStart 1 10 "bar" 2) + +prop_overlap_partial = assertOverlap + (tFromStart 1 5 "foo" 1) + (tFromStart 3 7 "bar" 2) + + +instance Ranged PositionedComment where + start = pcStartPos + end = pcEndPos + setRange (s, e) pc = pc { + pcStartPos = s, + pcEndPos = e + } + +instance Ranged Replacement where + start = repStartPos + end = repEndPos + setRange (s, e) r = r { + repStartPos = s, + repEndPos = e + } + +-- The Monoid instance for Fix merges fixes that do not conflict. +-- TODO: Make an efficient 'mconcat' +instance Monoid Fix where + mempty = newFix + mappend = (<>) + mconcat = foldl mappend mempty -- fold left to right since <> discards right on overlap + +instance Semigroup Fix where + f1 <> f2 = + -- FIXME: This might need to also discard adjacent zero-width ranges for + -- when two fixes change the same AST node, e.g. `foo` -> "$(foo)" + if or [ r2 `overlap` r1 | r1 <- fixReplacements f1, r2 <- fixReplacements f2 ] + then f1 + else newFix { + fixReplacements = fixReplacements f1 ++ fixReplacements f2 + } + +-- Conveniently apply a transformation to positions in a Fix +mapPositions :: (Position -> Position) -> Fix -> Fix +mapPositions f = adjustFix + where + adjustReplacement rep = + rep { + repStartPos = f $ repStartPos rep, + repEndPos = f $ repEndPos rep + } + adjustFix fix = + fix { + fixReplacements = map adjustReplacement $ fixReplacements fix + } + +-- Rewrite a Ranged from a tabstop of 8 to 1 +removeTabStops :: Ranged a => a -> Array Int String -> a +removeTabStops range ls = + let startColumn = realignColumn lineNo colNo range + endColumn = realignColumn endLineNo endColNo range + startPosition = (start range) { posColumn = startColumn } + endPosition = (end range) { posColumn = endColumn } in + setRange (startPosition, endPosition) range + where + realignColumn lineNo colNo c = + if lineNo c > 0 && lineNo c <= fromIntegral (length ls) + then real (ls ! fromIntegral (lineNo c)) 0 0 (colNo c) + else colNo c + real _ r v target | target <= v = r + -- hit this case at the end of line, and if we don't hit the target + -- return real + (target - v) + real [] r v target = r + (target - v) + real ('\t':rest) r v target = real rest (r+1) (v + 8 - (v `mod` 8)) target + real (_:rest) r v target = real rest (r+1) (v+1) target + lineNo = posLine . start + endLineNo = posLine . end + colNo = posColumn . start + endColNo = posColumn . end + + +-- A replacement that spans multiple line is applied by: +-- 1. merging the affected lines into a single string using `unlines` +-- 2. apply the replacement as if it only spanned a single line +-- The tricky part is adjusting the end column of the replacement +-- (the end line doesn't matter because there is only one line) +-- +-- aaS <--- start of replacement (row 1 column 3) +-- bbbb +-- cEc +-- \------- end of replacement (row 3 column 2) +-- +-- a flattened string will look like: +-- +-- "aaS\nbbbb\ncEc\n" +-- +-- The column of E has to be adjusted by: +-- 1. lengths of lines to be replaced, except the end row itself +-- 2. end column of the replacement +-- 3. number of '\n' by `unlines` +multiToSingleLine :: [Fix] -> Array Int String -> ([Fix], String) +multiToSingleLine fixes lines = + (map (mapPositions adjust) fixes, unlines $ elems lines) + where + -- A prefix sum tree from line number to column shift. + -- FIXME: The tree will be totally unbalanced. + shiftTree :: PSTree Int + shiftTree = + foldl (\t (n,s) -> addPSValue (n+1) (length s + 1) t) newPSTree $ + assocs lines + singleString = unlines $ elems lines + adjust pos = + pos { + posLine = 1, + posColumn = (posColumn pos) + + (fromIntegral $ getPrefixSum (fromIntegral $ posLine pos) shiftTree) + } + +-- Apply a fix and return resulting lines. +-- The number of lines can increase or decrease with no obvious mapping back, so +-- the function does not return an array. +applyFix :: Fix -> Array Int String -> [String] +applyFix fix fileLines = + let + untabbed = fix { + fixReplacements = + map (\c -> removeTabStops c fileLines) $ + fixReplacements fix + } + (adjustedFixes, singleLine) = multiToSingleLine [untabbed] fileLines + in + lines . runFixer $ applyFixes2 adjustedFixes singleLine + + +-- start and end comes from pos, which is 1 based +prop_doReplace1 = doReplace 0 0 "1234" "A" == "A1234" -- technically not valid +prop_doReplace2 = doReplace 1 1 "1234" "A" == "A1234" +prop_doReplace3 = doReplace 1 2 "1234" "A" == "A234" +prop_doReplace4 = doReplace 3 3 "1234" "A" == "12A34" +prop_doReplace5 = doReplace 4 4 "1234" "A" == "123A4" +prop_doReplace6 = doReplace 5 5 "1234" "A" == "1234A" +doReplace start end o r = + let si = fromIntegral (start-1) + ei = fromIntegral (end-1) + (x, xs) = splitAt si o + z = drop (ei - si) xs + in + x ++ r ++ z + +-- Fail if the 'expected' string is not result when applying 'fixes' to 'original'. +testFixes :: String -> String -> [Fix] -> Bool +testFixes expected original fixes = + actual == expected + where + actual = runFixer (applyFixes2 fixes original) + + +-- A Fixer allows doing repeated modifications of a string where each +-- replacement automatically accounts for shifts from previous ones. +type Fixer a = State (PSTree Int) a + +-- Apply a single replacement using its indices into the original string. +-- It does not handle multiple lines, all line indices must be 1. +applyReplacement2 :: Replacement -> String -> Fixer String +applyReplacement2 rep string = do + tree <- get + let transform pos = pos + getPrefixSum pos tree + let originalPos = (repStartPos rep, repEndPos rep) + (oldStart, oldEnd) = tmap (fromInteger . posColumn) originalPos + (newStart, newEnd) = tmap transform (oldStart, oldEnd) + + let (l1, l2) = tmap posLine originalPos in + when (l1 /= 1 || l2 /= 1) $ + error $ pleaseReport "bad cross-line fix" + + let replacer = repString rep + let shift = (length replacer) - (oldEnd - oldStart) + let insertionPoint = + case repInsertionPoint rep of + InsertBefore -> oldStart + InsertAfter -> oldEnd+1 + put $ addPSValue insertionPoint shift tree + + return $ doReplace newStart newEnd string replacer + where + tmap f (a,b) = (f a, f b) + +-- Apply a list of Replacements in the correct order +applyReplacements2 :: [Replacement] -> String -> Fixer String +applyReplacements2 reps str = + foldM (flip applyReplacement2) str $ + reverse $ sortWith repPrecedence reps + +-- Apply all fixes with replacements in the correct order +applyFixes2 :: [Fix] -> String -> Fixer String +applyFixes2 fixes = applyReplacements2 (concatMap fixReplacements fixes) + +-- Get the final value of a Fixer. +runFixer :: Fixer a -> a +runFixer f = evalState f newPSTree + + + +-- A Prefix Sum Tree that lets you look up the sum of values at and below an index. +-- It's implemented essentially as a Fenwick tree without the bit-based balancing. +-- The last Num is the sum of the left branch plus current element. +data PSTree n = PSBranch n (PSTree n) (PSTree n) n | PSLeaf + deriving (Show) + +newPSTree :: Num n => PSTree n +newPSTree = PSLeaf + +-- Get the sum of values whose keys are <= 'target' +getPrefixSum :: (Ord n, Num n) => n -> PSTree n -> n +getPrefixSum = f 0 + where + f sum _ PSLeaf = sum + f sum target (PSBranch pivot left right cumulative) = + case target `compare` pivot of + LT -> f sum target left + GT -> f (sum+cumulative) target right + EQ -> sum+cumulative + +-- Add a value to the Prefix Sum tree at the given index. +-- Values accumulate: addPSValue 42 2 . addPSValue 42 3 == addPSValue 42 5 +addPSValue :: (Ord n, Num n) => n -> n -> PSTree n -> PSTree n +addPSValue key value tree = if value == 0 then tree else f tree + where + f PSLeaf = PSBranch key PSLeaf PSLeaf value + f (PSBranch pivot left right sum) = + case key `compare` pivot of + LT -> PSBranch pivot (f left) right (sum + value) + GT -> PSBranch pivot left (f right) sum + EQ -> PSBranch pivot left right (sum + value) + +prop_pstreeSumsCorrectly kvs targets = + let + -- Trivial O(n * m) implementation + dumbPrefixSums :: [(Int, Int)] -> [Int] -> [Int] + dumbPrefixSums kvs targets = + let prefixSum target = sum [v | (k,v) <- kvs, k <= target] + in map prefixSum targets + -- PSTree O(n * log m) implementation + smartPrefixSums :: [(Int, Int)] -> [Int] -> [Int] + smartPrefixSums kvs targets = + let tree = foldl (\tree (pos, shift) -> addPSValue pos shift tree) PSLeaf kvs + in map (\x -> getPrefixSum x tree) targets + in smartPrefixSums kvs targets == dumbPrefixSums kvs targets + + +-- Semi-convenient functions for constructing tests. +testFix :: [Replacement] -> Fix +testFix list = newFix { + fixReplacements = list + } + +tFromStart :: Int -> Int -> String -> Int -> Replacement +tFromStart start end repl order = + newReplacement { + repStartPos = newPosition { + posLine = 1, + posColumn = fromIntegral start + }, + repEndPos = newPosition { + posLine = 1, + posColumn = fromIntegral end + }, + repString = repl, + repPrecedence = order, + repInsertionPoint = InsertAfter + } + +tFromEnd start end repl order = + (tFromStart start end repl order) { + repInsertionPoint = InsertBefore + } + +prop_simpleFix1 = testFixes "hello world" "hell world" [ + testFix [ + tFromEnd 5 5 "o" 1 + ]] + +prop_anchorsLeft = testFixes "-->foobar<--" "--><--" [ + testFix [ + tFromStart 4 4 "foo" 1, + tFromStart 4 4 "bar" 2 + ]] + +prop_anchorsRight = testFixes "-->foobar<--" "--><--" [ + testFix [ + tFromEnd 4 4 "bar" 1, + tFromEnd 4 4 "foo" 2 + ]] + +prop_anchorsBoth1 = testFixes "-->foobar<--" "--><--" [ + testFix [ + tFromStart 4 4 "bar" 2, + tFromEnd 4 4 "foo" 1 + ]] + +prop_anchorsBoth2 = testFixes "-->foobar<--" "--><--" [ + testFix [ + tFromEnd 4 4 "foo" 2, + tFromStart 4 4 "bar" 1 + ]] + +prop_composeFixes1 = testFixes "cd \"$1\" || exit" "cd $1" [ + testFix [ + tFromStart 4 4 "\"" 10, + tFromEnd 6 6 "\"" 10 + ], + testFix [ + tFromEnd 6 6 " || exit" 5 + ]] + +prop_composeFixes2 = testFixes "$(\"$1\")" "`$1`" [ + testFix [ + tFromStart 1 2 "$(" 5, + tFromEnd 4 5 ")" 5 + ], + testFix [ + tFromStart 2 2 "\"" 10, + tFromEnd 4 4 "\"" 10 + ]] + +prop_composeFixes3 = testFixes "(x)[x]" "xx" [ + testFix [ + tFromStart 1 1 "(" 4, + tFromEnd 2 2 ")" 3, + tFromStart 2 2 "[" 2, + tFromEnd 3 3 "]" 1 + ]] + +prop_composeFixes4 = testFixes "(x)[x]" "xx" [ + testFix [ + tFromStart 1 1 "(" 4, + tFromStart 2 2 "[" 3, + tFromEnd 2 2 ")" 2, + tFromEnd 3 3 "]" 1 + ]] + +prop_composeFixes5 = testFixes "\"$(x)\"" "`x`" [ + testFix [ + tFromStart 1 2 "$(" 2, + tFromEnd 3 4 ")" 2, + tFromStart 1 1 "\"" 1, + tFromEnd 4 4 "\"" 1 + ]] + + +return [] +runTests = $quickCheckAll diff --git a/src/ShellCheck/Formatter/CheckStyle.hs b/src/ShellCheck/Formatter/CheckStyle.hs index b3f2074..3f898c3 100644 --- a/src/ShellCheck/Formatter/CheckStyle.hs +++ b/src/ShellCheck/Formatter/CheckStyle.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2019 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -24,8 +24,8 @@ import ShellCheck.Formatter.Format import Data.Char import Data.List -import GHC.Exts import System.IO +import qualified Data.List.NonEmpty as NE format :: IO Formatter format = return Formatter { @@ -45,12 +45,12 @@ outputResults cr sys = else mapM_ outputGroup fileGroups where comments = crComments cr - fileGroups = groupWith sourceFile comments + fileGroups = NE.groupWith sourceFile comments outputGroup group = do - let filename = sourceFile (head group) - result <- (siReadFile sys) filename + let filename = sourceFile (NE.head group) + result <- siReadFile sys (Just True) filename let contents = either (const "") id result - outputFile filename contents group + outputFile filename contents (NE.toList group) outputFile filename contents warnings = do let comments = makeNonVirtual warnings contents @@ -88,7 +88,7 @@ outputError file error = putStrLn $ concat [ attr s v = concat [ s, "='", escape v, "' " ] escape = concatMap escape' escape' c = if isOk c then [c] else "&#" ++ show (ord c) ++ ";" -isOk x = any ($x) [isAsciiUpper, isAsciiLower, isDigit, (`elem` " ./")] +isOk x = any ($ x) [isAsciiUpper, isAsciiLower, isDigit, (`elem` " ./")] severity "error" = "error" severity "warning" = "warning" diff --git a/src/ShellCheck/Formatter/Diff.hs b/src/ShellCheck/Formatter/Diff.hs new file mode 100644 index 0000000..15d00d7 --- /dev/null +++ b/src/ShellCheck/Formatter/Diff.hs @@ -0,0 +1,254 @@ +{- + Copyright 2019 Vidar 'koala_man' Holen + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{-# LANGUAGE TemplateHaskell #-} +module ShellCheck.Formatter.Diff (format, ShellCheck.Formatter.Diff.runTests) where + +import ShellCheck.Interface +import ShellCheck.Fixer +import ShellCheck.Formatter.Format + +import Control.Monad +import Data.Algorithm.Diff +import Data.Array +import Data.IORef +import Data.List +import qualified Data.Monoid as Monoid +import Data.Maybe +import qualified Data.Map as M +import GHC.Exts (sortWith) +import System.IO +import System.FilePath + +import Test.QuickCheck + +format :: FormatterOptions -> IO Formatter +format options = do + foundIssues <- newIORef False + reportedIssues <- newIORef False + shouldColor <- shouldOutputColor (foColorOption options) + let color = if shouldColor then colorize else nocolor + return Formatter { + header = return (), + footer = checkFooter foundIssues reportedIssues color, + onFailure = reportFailure color, + onResult = reportResult foundIssues reportedIssues color + } + + +contextSize = 3 +red = 31 +green = 32 +yellow = 33 +cyan = 36 +bold = 1 + +nocolor n = id +colorize n s = (ansi n) ++ s ++ (ansi 0) +ansi n = "\x1B[" ++ show n ++ "m" + +printErr :: ColorFunc -> String -> IO () +printErr color = hPutStrLn stderr . color bold . color red +reportFailure color file msg = printErr color $ file ++ ": " ++ msg + +checkFooter foundIssues reportedIssues color = do + found <- readIORef foundIssues + output <- readIORef reportedIssues + when (found && not output) $ + printErr color "Issues were detected, but none were auto-fixable. Use another format to see them." + +type ColorFunc = (Int -> String -> String) +data LFStatus = LinefeedMissing | LinefeedOk +data DiffDoc a = DiffDoc String LFStatus [DiffRegion a] +data DiffRegion a = DiffRegion (Int, Int) (Int, Int) [Diff a] + +reportResult :: (IORef Bool) -> (IORef Bool) -> ColorFunc -> CheckResult -> SystemInterface IO -> IO () +reportResult foundIssues reportedIssues color result sys = do + let comments = crComments result + unless (null comments) $ writeIORef foundIssues True + let suggestedFixes = mapMaybe pcFix comments + let fixmap = buildFixMap suggestedFixes + mapM_ output $ M.toList fixmap + where + output (name, fix) = do + file <- siReadFile sys (Just True) name + case file of + Right contents -> do + putStrLn $ formatDoc color $ makeDiff name contents fix + writeIORef reportedIssues True + Left msg -> reportFailure color name msg + +hasTrailingLinefeed str = + case str of + [] -> True + _ -> last str == '\n' + +coversLastLine regions = + case regions of + [] -> False + _ -> (fst $ last regions) + +-- TODO: Factor this out into a unified diff library because we're doing a lot +-- of the heavy lifting anyways. +makeDiff :: String -> String -> Fix -> DiffDoc String +makeDiff name contents fix = do + let hunks = groupDiff $ computeDiff contents fix + let lf = if coversLastLine hunks && not (hasTrailingLinefeed contents) + then LinefeedMissing + else LinefeedOk + DiffDoc name lf $ findRegions hunks + +computeDiff :: String -> Fix -> [Diff String] +computeDiff contents fix = + let old = lines contents + array = listArray (1, fromIntegral $ (length old)) old + new = applyFix fix array + in getDiff old new + +-- Group changes into hunks +groupDiff :: [Diff a] -> [(Bool, [Diff a])] +groupDiff = filter (\(_, l) -> not (null l)) . hunt [] + where + -- Churn through 'Both's until we find a difference + hunt current [] = [(False, reverse current)] + hunt current (x@Both {}:rest) = hunt (x:current) rest + hunt current list = + let (context, previous) = splitAt contextSize current + in (False, reverse previous) : gather context 0 list + + -- Pick out differences until we find a run of Both's + gather current n [] = + let (extras, patch) = splitAt (max 0 $ n - contextSize) current + in [(True, reverse patch), (False, reverse extras)] + + gather current n list@(Both {}:_) | n == contextSize*2 = + let (context, previous) = splitAt contextSize current + in (True, reverse previous) : hunt context list + + gather current n (x@Both {}:rest) = gather (x:current) (n+1) rest + gather current n (x:rest) = gather (x:current) 0 rest + +-- Get line numbers for hunks +findRegions :: [(Bool, [Diff String])] -> [DiffRegion String] +findRegions = find' 1 1 + where + find' _ _ [] = [] + find' left right ((output, run):rest) = + let (dl, dr) = countDelta run + remainder = find' (left+dl) (right+dr) rest + in + if output + then DiffRegion (left, dl) (right, dr) run : remainder + else remainder + +-- Get left/right line counts for a hunk +countDelta :: [Diff a] -> (Int, Int) +countDelta = count' 0 0 + where + count' left right [] = (left, right) + count' left right (x:rest) = + case x of + Both {} -> count' (left+1) (right+1) rest + First {} -> count' (left+1) right rest + Second {} -> count' left (right+1) rest + +formatRegion :: ColorFunc -> LFStatus -> DiffRegion String -> String +formatRegion color lf (DiffRegion left right diffs) = + let header = color cyan ("@@ -" ++ (tup left) ++ " +" ++ (tup right) ++" @@") + in + unlines $ header : reverse (getStrings lf (reverse diffs)) + where + noLF = "\\ No newline at end of file" + + getStrings LinefeedOk list = map format list + getStrings LinefeedMissing list@((Both _ _):_) = noLF : map format list + getStrings LinefeedMissing list@((First _):_) = noLF : map format list + getStrings LinefeedMissing (last:rest) = format last : getStrings LinefeedMissing rest + + tup (a,b) = (show a) ++ "," ++ (show b) + format (Both x _) = ' ':x + format (First x) = color red $ '-':x + format (Second x) = color green $ '+':x + +splitLast [] = ([], []) +splitLast x = + let (last, rest) = splitAt 1 $ reverse x + in (reverse rest, last) + +formatDoc color (DiffDoc name lf regions) = + let (most, last) = splitLast regions + in + (color bold $ "--- " ++ ("a" name)) ++ "\n" ++ + (color bold $ "+++ " ++ ("b" name)) ++ "\n" ++ + concatMap (formatRegion color LinefeedOk) most ++ + concatMap (formatRegion color lf) last + +-- Create a Map from filename to Fix +buildFixMap :: [Fix] -> M.Map String Fix +buildFixMap fixes = perFile + where + splitFixes = splitFixByFile $ mconcat fixes + perFile = groupByMap (posFile . repStartPos . head . fixReplacements) splitFixes + +splitFixByFile :: Fix -> [Fix] +splitFixByFile fix = map makeFix $ groupBy sameFile (fixReplacements fix) + where + sameFile rep1 rep2 = (posFile $ repStartPos rep1) == (posFile $ repStartPos rep2) + makeFix reps = newFix { fixReplacements = reps } + +groupByMap :: (Ord k, Monoid v) => (v -> k) -> [v] -> M.Map k v +groupByMap f = M.fromListWith Monoid.mappend . map (\x -> (f x, x)) + +-- For building unit tests +b n = Both n n +l = First +r = Second + +prop_identifiesProperContext = groupDiff [b 1, b 2, b 3, b 4, l 5, b 6, b 7, b 8, b 9] == + [(False, [b 1]), -- Omitted + (True, [b 2, b 3, b 4, l 5, b 6, b 7, b 8]), -- A change with three lines of context + (False, [b 9])] -- Omitted + +prop_includesContextFromStartIfNecessary = groupDiff [b 4, l 5, b 6, b 7, b 8, b 9] == + [ -- Nothing omitted + (True, [b 4, l 5, b 6, b 7, b 8]), -- A change with three lines of context + (False, [b 9])] -- Omitted + +prop_includesContextUntilEndIfNecessary = groupDiff [b 4, l 5] == + [ -- Nothing omitted + (True, [b 4, l 5]) + ] -- Nothing Omitted + +prop_splitsIntoMultipleHunks = groupDiff [l 1, b 1, b 2, b 3, b 4, b 5, b 6, b 7, r 8] == + [ -- Nothing omitted + (True, [l 1, b 1, b 2, b 3]), + (False, [b 4]), + (True, [b 5, b 6, b 7, r 8]) + ] -- Nothing Omitted + +prop_splitsIntoMultipleHunksUnlessTouching = groupDiff [l 1, b 1, b 2, b 3, b 4, b 5, b 6, r 7] == + [ + (True, [l 1, b 1, b 2, b 3, b 4, b 5, b 6, r 7]) + ] + +prop_countDeltasWorks = countDelta [b 1, l 2, r 3, r 4, b 5] == (3,4) +prop_countDeltasWorks2 = countDelta [] == (0,0) + +return [] +runTests = $quickCheckAll diff --git a/src/ShellCheck/Formatter/Format.hs b/src/ShellCheck/Formatter/Format.hs index 5e46713..53b59a4 100644 --- a/src/ShellCheck/Formatter/Format.hs +++ b/src/ShellCheck/Formatter/Format.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2019 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -21,6 +21,14 @@ module ShellCheck.Formatter.Format where import ShellCheck.Data import ShellCheck.Interface +import ShellCheck.Fixer + +import Control.Monad +import Data.Array +import Data.List +import System.IO +import System.Info +import System.Environment -- A formatter that carries along an arbitrary piece of data data Formatter = Formatter { @@ -50,21 +58,25 @@ severityText pc = makeNonVirtual comments contents = map fix comments where - ls = lines contents - fix c = c { - pcStartPos = (pcStartPos c) { - posColumn = realignColumn lineNo colNo c - } - , pcEndPos = (pcEndPos c) { - posColumn = realignColumn endLineNo endColNo c - } + list = lines contents + arr = listArray (1, length list) list + untabbedFix f = newFix { + fixReplacements = map (\r -> removeTabStops r arr) (fixReplacements f) } - realignColumn lineNo colNo c = - if lineNo c > 0 && lineNo c <= fromIntegral (length ls) - then real (ls !! fromIntegral (lineNo c - 1)) 0 0 (colNo c) - else colNo c - real _ r v target | target <= v = r - real [] r v _ = r -- should never happen - real ('\t':rest) r v target = - real rest (r+1) (v + 8 - (v `mod` 8)) target - real (_:rest) r v target = real rest (r+1) (v+1) target + fix c = (removeTabStops c arr) { + pcFix = fmap untabbedFix (pcFix c) + } + + +shouldOutputColor :: ColorOption -> IO Bool +shouldOutputColor colorOption = + case colorOption of + ColorAlways -> return True + ColorNever -> return False + ColorAuto -> do + isTerminal <- hIsTerminalDevice stdout + term <- lookupEnv "TERM" + let windows = "mingw" `isPrefixOf` os + let dumbTerm = term `elem` [Just "dumb", Just "", Nothing] + let isUsableTty = isTerminal && not windows && not dumbTerm + return isUsableTty diff --git a/src/ShellCheck/Formatter/GCC.hs b/src/ShellCheck/Formatter/GCC.hs index b8a0bb0..b921753 100644 --- a/src/ShellCheck/Formatter/GCC.hs +++ b/src/ShellCheck/Formatter/GCC.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2019 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -23,8 +23,8 @@ import ShellCheck.Interface import ShellCheck.Formatter.Format import Data.List -import GHC.Exts import System.IO +import qualified Data.List.NonEmpty as NE format :: IO Formatter format = return Formatter { @@ -39,13 +39,13 @@ outputError file error = hPutStrLn stderr $ file ++ ": " ++ error outputAll cr sys = mapM_ f groups where comments = crComments cr - groups = groupWith sourceFile comments - f :: [PositionedComment] -> IO () + groups = NE.groupWith sourceFile comments + f :: NE.NonEmpty PositionedComment -> IO () f group = do - let filename = sourceFile (head group) - result <- (siReadFile sys) filename + let filename = sourceFile (NE.head group) + result <- siReadFile sys (Just True) filename let contents = either (const "") id result - outputResult filename contents group + outputResult filename contents (NE.toList group) outputResult filename contents warnings = do let comments = makeNonVirtual warnings contents diff --git a/src/ShellCheck/Formatter/JSON1.hs b/src/ShellCheck/Formatter/JSON1.hs new file mode 100644 index 0000000..b4dbe35 --- /dev/null +++ b/src/ShellCheck/Formatter/JSON1.hs @@ -0,0 +1,128 @@ +{-# LANGUAGE OverloadedStrings #-} +{- + Copyright 2012-2019 Vidar Holen + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +module ShellCheck.Formatter.JSON1 (format) where + +import ShellCheck.Interface +import ShellCheck.Formatter.Format + +import Control.DeepSeq +import Data.Aeson +import Data.IORef +import Data.Monoid +import System.IO +import qualified Data.ByteString.Lazy.Char8 as BL +import qualified Data.List.NonEmpty as NE + +format :: IO Formatter +format = do + ref <- newIORef [] + return Formatter { + header = return (), + onResult = collectResult ref, + onFailure = outputError, + footer = finish ref + } + +data Json1Output = Json1Output { + comments :: [PositionedComment] + } + +instance ToJSON Json1Output where + toJSON result = object [ + "comments" .= comments result + ] + toEncoding result = pairs ( + "comments" .= comments result + ) + +instance ToJSON Replacement where + toJSON replacement = + let start = repStartPos replacement + end = repEndPos replacement + str = repString replacement in + object [ + "precedence" .= repPrecedence replacement, + "insertionPoint" .= + case repInsertionPoint replacement of + InsertBefore -> "beforeStart" :: String + InsertAfter -> "afterEnd", + "line" .= posLine start, + "column" .= posColumn start, + "endLine" .= posLine end, + "endColumn" .= posColumn end, + "replacement" .= str + ] + +instance ToJSON PositionedComment where + toJSON comment = + let start = pcStartPos comment + end = pcEndPos comment + c = pcComment comment in + object [ + "file" .= posFile start, + "line" .= posLine start, + "endLine" .= posLine end, + "column" .= posColumn start, + "endColumn" .= posColumn end, + "level" .= severityText comment, + "code" .= cCode c, + "message" .= cMessage c, + "fix" .= pcFix comment + ] + + toEncoding comment = + let start = pcStartPos comment + end = pcEndPos comment + c = pcComment comment in + pairs ( + "file" .= posFile start + <> "line" .= posLine start + <> "endLine" .= posLine end + <> "column" .= posColumn start + <> "endColumn" .= posColumn end + <> "level" .= severityText comment + <> "code" .= cCode c + <> "message" .= cMessage c + <> "fix" .= pcFix comment + ) + +instance ToJSON Fix where + toJSON fix = object [ + "replacements" .= fixReplacements fix + ] + +outputError file msg = hPutStrLn stderr $ file ++ ": " ++ msg + +collectResult ref cr sys = mapM_ f groups + where + comments = crComments cr + groups = NE.groupWith sourceFile comments + f :: NE.NonEmpty PositionedComment -> IO () + f group = do + let filename = sourceFile (NE.head group) + result <- siReadFile sys (Just True) filename + let contents = either (const "") id result + let comments' = makeNonVirtual comments contents + deepseq comments' $ modifyIORef ref (\x -> comments' ++ x) + +finish ref = do + list <- readIORef ref + BL.putStrLn $ encode $ Json1Output { comments = list } diff --git a/src/ShellCheck/Formatter/Quiet.hs b/src/ShellCheck/Formatter/Quiet.hs new file mode 100644 index 0000000..b7e0ee9 --- /dev/null +++ b/src/ShellCheck/Formatter/Quiet.hs @@ -0,0 +1,36 @@ +{- + Copyright 2019 Austin Voecks + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +module ShellCheck.Formatter.Quiet (format) where + +import ShellCheck.Interface +import ShellCheck.Formatter.Format + +import Control.Monad +import Data.IORef +import System.Exit + +format :: FormatterOptions -> IO Formatter +format options = + return Formatter { + header = return (), + footer = return (), + onFailure = \ _ _ -> exitFailure, + onResult = \ result _ -> unless (null $ crComments result) exitFailure + } diff --git a/src/ShellCheck/Parser.hs b/src/ShellCheck/Parser.hs index d656ed9..84c3ce4 100644 --- a/src/ShellCheck/Parser.hs +++ b/src/ShellCheck/Parser.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2022 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -24,9 +24,10 @@ module ShellCheck.Parser (parseScript, runTests) where import ShellCheck.AST -import ShellCheck.ASTLib +import ShellCheck.ASTLib hiding (runTests) import ShellCheck.Data import ShellCheck.Interface +import ShellCheck.Prelude import Control.Applicative ((<*), (*>)) import Control.Monad @@ -34,10 +35,9 @@ import Control.Monad.Identity import Control.Monad.Trans import Data.Char import Data.Functor -import Data.List (isPrefixOf, isInfixOf, isSuffixOf, partition, sortBy, intercalate, nub) +import Data.List (isPrefixOf, isInfixOf, isSuffixOf, partition, sortBy, intercalate, nub, find) import Data.Maybe import Data.Monoid -import Debug.Trace import GHC.Exts (sortWith) import Prelude hiding (readList) import System.IO @@ -46,7 +46,8 @@ import Text.Parsec.Error import Text.Parsec.Pos import qualified Control.Monad.Reader as Mr import qualified Control.Monad.State as Ms -import qualified Data.Map as Map +import qualified Data.List.NonEmpty as NE +import qualified Data.Map.Strict as Map import Test.QuickCheck.All (quickCheckAll) @@ -66,10 +67,10 @@ doubleQuote = char '"' variableStart = upper <|> lower <|> oneOf "_" variableChars = upper <|> lower <|> digit <|> oneOf "_" -- Chars to allow in function names -functionChars = variableChars <|> oneOf ":+?-./^@" +functionChars = variableChars <|> oneOf ":+?-./^@," -- Chars to allow in functions using the 'function' keyword extendedFunctionChars = functionChars <|> oneOf "[]*=!" -specialVariable = oneOf "@*#?-$!" +specialVariable = oneOf (concat specialVariables) paramSubSpecialChars = oneOf "/:+-=%" quotableChars = "|&;<>()\\ '\t\n\r\xA0" ++ doubleQuotableChars quotable = almostSpace <|> oneOf quotableChars @@ -87,11 +88,23 @@ extglobStart = oneOf extglobStartChars unicodeDoubleQuotes = "\x201C\x201D\x2033\x2036" unicodeSingleQuotes = "\x2018\x2019" -prop_spacing = isOk spacing " \\\n # Comment" +prop_spacing1 = isOk spacing " \\\n # Comment" +prop_spacing2 = isOk spacing "# We can continue lines with \\" +prop_spacing3 = isWarning spacing " \\\n # --verbose=true \\" spacing = do - x <- many (many1 linewhitespace <|> try (string "\\\n" >> return "")) + x <- many (many1 linewhitespace <|> continuation) optional readComment return $ concat x + where + continuation = do + try (string "\\\n") + -- The line was continued. Warn if this next line is a comment with a trailing \ + whitespace <- many linewhitespace + optional $ do + x <- readComment + when ("\\" `isSuffixOf` x) $ + parseProblem ErrorC 1143 "This backslash is part of a comment and does not continue the line." + return whitespace spacing1 = do spacing <- spacing @@ -113,6 +126,7 @@ allspacing = do allspacingOrFail = do s <- allspacing when (null s) $ fail "Expected whitespace" + return s readUnicodeQuote = do start <- startSpan @@ -122,22 +136,17 @@ readUnicodeQuote = do return $ T_Literal id [c] carriageReturn = do - parseNote ErrorC 1017 "Literal carriage return. Run script through tr -d '\\r' ." + pos <- getPosition char '\r' + parseProblemAt pos ErrorC 1017 "Literal carriage return. Run script through tr -d '\\r' ." + return '\r' -almostSpace = - choice [ - check '\xA0' "unicode non-breaking space", - check '\x200B' "unicode zerowidth space" - ] - where - check c name = do - parseNote ErrorC 1018 $ "This is a " ++ name ++ ". Delete and retype it." - char c +almostSpace = do + parseNote ErrorC 1018 $ "This is a unicode space. Delete and retype it." + oneOf "\xA0\x2002\x2003\x2004\x2005\x2006\x2007\x2008\x2009\x200B\x202F" return ' ' --------- Message/position annotation on top of user state -data Note = Note Id Severity Code String deriving (Show, Eq) data ParseNote = ParseNote SourcePos SourcePos Severity Code String deriving (Show, Eq) data Context = ContextName SourcePos String @@ -146,7 +155,7 @@ data Context = deriving (Show) data HereDocContext = - HereDocPending Token [Context] -- on linefeed, read this T_HereDoc + HereDocPending Id Dashed Quoted String [Context] -- on linefeed, read this T_HereDoc deriving (Show) data UserState = UserState { @@ -165,10 +174,6 @@ initialUserState = UserState { } codeForParseNote (ParseNote _ _ _ code _) = code -noteToParseNote map (Note id severity code message) = - ParseNote pos pos severity code message - where - pos = fromJust $ Map.lookup id map getLastId = lastId <$> getState @@ -190,17 +195,17 @@ getNextIdSpanningTokens startTok endTok = do -- Get an ID starting from the first token of the list, and ending after the last getNextIdSpanningTokenList list = - if null list - then do + case list of + [] -> do pos <- getPosition getNextIdBetween pos pos - else - getNextIdSpanningTokens (head list) (last list) + (h:_) -> + getNextIdSpanningTokens h (last list) -- Get the span covered by an id getSpanForId :: Monad m => Id -> SCParser m (SourcePos, SourcePos) getSpanForId id = - Map.findWithDefault (error "Internal error: no position for id. Please report!") id <$> + Map.findWithDefault (error $ pleaseReport "no parser span for id") id <$> getMap -- Create a new id with the same span as an existing one @@ -213,8 +218,13 @@ startSpan = IncompleteInterval <$> getPosition endSpan (IncompleteInterval start) = do endPos <- getPosition - id <- getNextIdBetween start endPos - return id + getNextIdBetween start endPos + +getSpanPositionsFor m = do + start <- getPosition + m + end <- getPosition + return (start, end) addToHereDocMap id list = do state <- getState @@ -223,12 +233,12 @@ addToHereDocMap id list = do hereDocMap = Map.insert id list map } -addPendingHereDoc t = do +addPendingHereDoc id d q str = do state <- getState context <- getCurrentContexts let docs = pendingHereDocs state putState $ state { - pendingHereDocs = HereDocPending t context : docs + pendingHereDocs = HereDocPending id d q str context : docs } popPendingHereDocs = do @@ -250,19 +260,38 @@ addParseNote n = do parseNotes = n : parseNotes state } +ignoreProblemsOf p = do + systemState <- lift . lift $ Ms.get + p <* (lift . lift . Ms.put $ systemState) + shouldIgnoreCode code = do context <- getCurrentContexts checkSourced <- Mr.asks checkSourced - return $ any (disabling checkSourced) context + return $ any (contextItemDisablesCode checkSourced code) context + +-- Does this item on the context stack disable warnings for 'code'? +contextItemDisablesCode :: Bool -> Integer -> Context -> Bool +contextItemDisablesCode alsoCheckSourced code = disabling alsoCheckSourced where disabling checkSourced item = case item of ContextAnnotation list -> any disabling' list ContextSource _ -> not $ checkSourced _ -> False - disabling' (DisableComment n) = code == n + disabling' (DisableComment n m) = code >= n && code < m disabling' _ = False + + +getCurrentAnnotations includeSource = + concatMap get . takeWhile (not . isBoundary) <$> getCurrentContexts + where + get (ContextAnnotation list) = list + get _ = [] + isBoundary (ContextSource _) = not includeSource + isBoundary _ = False + + shouldFollow file = do context <- getCurrentContexts if any isThisFile context @@ -306,6 +335,8 @@ initialSystemState = SystemState { data Environment m = Environment { systemInterface :: SystemInterface m, checkSourced :: Bool, + ignoreRC :: Bool, + currentFilename :: String, shellTypeOverride :: Maybe Shell } @@ -314,16 +345,15 @@ parseProblem level code msg = do parseProblemAt pos level code msg setCurrentContexts c = Ms.modify (\state -> state { contextStack = c }) -getCurrentContexts = contextStack <$> Ms.get +getCurrentContexts = Ms.gets contextStack popContext = do v <- getCurrentContexts - if not $ null v - then do - let (a:r) = v + case v of + (a:r) -> do setCurrentContexts r return $ Just a - else + [] -> return Nothing pushContext c = do @@ -363,16 +393,15 @@ parseNoteAtId id c l a = do parseNoteAtWithEnd start end c l a = addParseNote $ ParseNote start end c l a --------- Convenient combinators -thenSkip main follow = do - r <- main - optional follow - return r +thenSkip main follow = main <* optional follow unexpecting s p = try $ (try p >> fail ("Unexpected " ++ s)) <|> return () notFollowedBy2 = unexpecting "" +isFollowedBy p = (lookAhead . try $ p $> True) <|> return False + reluctantlyTill p end = (lookAhead (void (try end) <|> eof) >> return []) <|> do x <- p @@ -410,7 +439,7 @@ acceptButWarn parser level code note = parsecBracket before after op = do val <- before - (op val <* optional (after val)) <|> (after val *> fail "") + op val `thenSkip` after val <|> (after val *> fail "") swapContext contexts p = parsecBracket (getCurrentContexts <* setCurrentContexts contexts) @@ -423,13 +452,14 @@ called s p = do pos <- getPosition withContext (ContextName pos s) p -withAnnotations anns = - withContext (ContextAnnotation anns) +withAnnotations anns p = + if null anns then p else withContext (ContextAnnotation anns) p readConditionContents single = readCondContents `attempting` lookAhead (do pos <- getPosition s <- readVariableName + spacing1 when (s `elem` commonCommands) $ parseProblemAt pos WarningC 1014 "Use 'if cmd; then ..' to check exit code, or 'if [[ $(cmd) == .. ]]' to check output.") @@ -521,7 +551,7 @@ readConditionContents single = notFollowedBy2 (try (spacing >> string "]")) x <- readNormalWord pos <- getPosition - when (endedWith "]" x && notArrayIndex x) $ do + when (notArrayIndex x && endedWith "]" x && not (x `containsLiteral` "[")) $ do parseProblemAt pos ErrorC 1020 $ "You need a space before the " ++ (if single then "]" else "]]") ++ "." fail "Missing space before ]" @@ -537,6 +567,7 @@ readConditionContents single = endedWith _ _ = False notArrayIndex (T_NormalWord id s@(_:T_Literal _ t:_)) = t /= "[" notArrayIndex _ = True + containsLiteral x s = s `isInfixOf` onlyLiteralString x readCondAndOp = readAndOrOp TC_And "&&" False <|> readAndOrOp TC_And "-a" True @@ -575,9 +606,9 @@ readConditionContents single = return $ TC_Nullary id typ x ) - checkTrailingOp x = fromMaybe (return ()) $ do + checkTrailingOp x = sequence_ $ do (T_Literal id str) <- getTrailingUnquotedLiteral x - trailingOp <- listToMaybe (filter (`isSuffixOf` str) binaryTestOps) + trailingOp <- find (`isSuffixOf` str) binaryTestOps return $ parseProblemAtId id ErrorC 1108 $ "You need a space before and after the " ++ trailingOp ++ " ." @@ -682,20 +713,20 @@ prop_a6 = isOk readArithmeticContents " 1 | 2 ||3|4" prop_a7 = isOk readArithmeticContents "3*2**10" prop_a8 = isOk readArithmeticContents "3" prop_a9 = isOk readArithmeticContents "a^!-b" -prop_a10= isOk readArithmeticContents "! $?" -prop_a11= isOk readArithmeticContents "10#08 * 16#f" -prop_a12= isOk readArithmeticContents "\"$((3+2))\" + '37'" -prop_a13= isOk readArithmeticContents "foo[9*y+x]++" -prop_a14= isOk readArithmeticContents "1+`echo 2`" -prop_a15= isOk readArithmeticContents "foo[`echo foo | sed s/foo/4/g` * 3] + 4" -prop_a16= isOk readArithmeticContents "$foo$bar" -prop_a17= isOk readArithmeticContents "i<(0+(1+1))" -prop_a18= isOk readArithmeticContents "a?b:c" -prop_a19= isOk readArithmeticContents "\\\n3 +\\\n 2" -prop_a20= isOk readArithmeticContents "a ? b ? c : d : e" -prop_a21= isOk readArithmeticContents "a ? b : c ? d : e" -prop_a22= isOk readArithmeticContents "!!a" -prop_a23= isOk readArithmeticContents "~0" +prop_a10 = isOk readArithmeticContents "! $?" +prop_a11 = isOk readArithmeticContents "10#08 * 16#f" +prop_a12 = isOk readArithmeticContents "\"$((3+2))\" + '37'" +prop_a13 = isOk readArithmeticContents "foo[9*y+x]++" +prop_a14 = isOk readArithmeticContents "1+`echo 2`" +prop_a15 = isOk readArithmeticContents "foo[`echo foo | sed s/foo/4/g` * 3] + 4" +prop_a16 = isOk readArithmeticContents "$foo$bar" +prop_a17 = isOk readArithmeticContents "i<(0+(1+1))" +prop_a18 = isOk readArithmeticContents "a?b:c" +prop_a19 = isOk readArithmeticContents "\\\n3 +\\\n 2" +prop_a20 = isOk readArithmeticContents "a ? b ? c : d : e" +prop_a21 = isOk readArithmeticContents "a ? b : c ? d : e" +prop_a22 = isOk readArithmeticContents "!!a" +prop_a23 = isOk readArithmeticContents "~0" readArithmeticContents :: Monad m => SCParser m Token readArithmeticContents = readSequence @@ -784,11 +815,13 @@ readArithmeticContents = return $ TA_Expansion id pieces readGroup = do + start <- startSpan char '(' s <- readSequence char ')' + id <- endSpan start spacing - return s + return $ TA_Parenthesis id s readArithTerm = readGroup <|> readVariable <|> readExpansion @@ -888,8 +921,8 @@ prop_readCondition7 = isOk readCondition "[[ ${line} =~ ^[[:space:]]*# ]]" prop_readCondition8 = isOk readCondition "[[ $l =~ ogg|flac ]]" prop_readCondition9 = isOk readCondition "[ foo -a -f bar ]" prop_readCondition10 = isOk readCondition "[[\na == b\n||\nc == d ]]" -prop_readCondition10a= isOk readCondition "[[\na == b ||\nc == d ]]" -prop_readCondition10b= isOk readCondition "[[ a == b\n||\nc == d ]]" +prop_readCondition10a = isOk readCondition "[[\na == b ||\nc == d ]]" +prop_readCondition10b = isOk readCondition "[[ a == b\n||\nc == d ]]" prop_readCondition11 = isOk readCondition "[[ a == b ||\n c == d ]]" prop_readCondition12 = isWarning readCondition "[ a == b \n -o c == d ]" prop_readCondition13 = isOk readCondition "[[ foo =~ ^fo{1,3}$ ]]" @@ -903,6 +936,12 @@ prop_readCondition20 = isOk readCondition "[[ echo_rc -eq 0 ]]" prop_readCondition21 = isOk readCondition "[[ $1 =~ ^(a\\ b)$ ]]" prop_readCondition22 = isOk readCondition "[[ $1 =~ \\.a\\.(\\.b\\.)\\.c\\. ]]" prop_readCondition23 = isOk readCondition "[[ -v arr[$var] ]]" +prop_readCondition25 = isOk readCondition "[[ lex.yy.c -ot program.l ]]" +prop_readCondition26 = isOk readScript "[[ foo ]]\\\n && bar" +prop_readCondition27 = not $ isOk readConditionCommand "[[ x ]] foo" +prop_readCondition28 = isOk readCondition "[[ x = [\"$1\"] ]]" +prop_readCondition29 = isOk readCondition "[[ x = [*] ]]" + readCondition = called "test expression" $ do opos <- getPosition start <- startSpan @@ -929,10 +968,9 @@ readCondition = called "test expression" $ do cpos <- getPosition close <- try (string "]]") <|> string "]" <|> fail "Expected test to end here (don't wrap commands in []/[[]])" id <- endSpan start - when (open == "[[" && close /= "]]") $ parseProblemAt cpos ErrorC 1033 "Did you mean ]] ?" - when (open == "[" && close /= "]" ) $ parseProblemAt opos ErrorC 1034 "Did you mean [[ ?" + when (open == "[[" && close /= "]]") $ parseProblemAt cpos ErrorC 1033 "Test expression was opened with double [[ but closed with single ]. Make sure they match." + when (open == "[" && close /= "]" ) $ parseProblemAt opos ErrorC 1034 "Test expression was opened with single [ but closed with double ]]. Make sure they match." spacing - many readCmdWord -- Read and throw away remainders to get then/do warnings. Fixme? return $ T_Condition id typ condition readAnnotationPrefix = do @@ -946,42 +984,99 @@ prop_readAnnotation3 = isOk readAnnotation "# shellcheck disable=SC1234 source=/ prop_readAnnotation4 = isWarning readAnnotation "# shellcheck cats=dogs disable=SC1234\n" prop_readAnnotation5 = isOk readAnnotation "# shellcheck disable=SC2002 # All cats are precious\n" prop_readAnnotation6 = isOk readAnnotation "# shellcheck disable=SC1234 # shellcheck foo=bar\n" +prop_readAnnotation7 = isOk readAnnotation "# shellcheck disable=SC1000,SC2000-SC3000,SC1001\n" +prop_readAnnotation8 = isOk readAnnotation "# shellcheck disable=all\n" +prop_readAnnotation9 = isOk readAnnotation "# shellcheck source='foo bar' source-path=\"baz etc\"\n" +prop_readAnnotation10 = isOk readAnnotation "# shellcheck disable='SC1234,SC2345' enable=\"foo\" shell='bash'\n" +prop_readAnnotation11 = isOk (readAnnotationWithoutPrefix False) "external-sources='true'" + readAnnotation = called "shellcheck directive" $ do try readAnnotationPrefix many1 linewhitespace + readAnnotationWithoutPrefix True + +readAnnotationWithoutPrefix sandboxed = do values <- many1 readKey optional readAnyComment - void linefeed <|> do + void linefeed <|> eof <|> do parseNote ErrorC 1125 "Invalid key=value pair? Ignoring the rest of this directive starting here." many (noneOf "\n") void linefeed <|> eof many linewhitespace return $ concat values where + plainOrQuoted p = quoted p <|> p + quoted p = do + c <- oneOf "'\"" + start <- getPosition + str <- many1 $ noneOf (c:"\n") + char c <|> fail "Missing terminating quote for directive." + subParse start p str readKey = do keyPos <- getPosition - key <- many1 letter + key <- many1 (letter <|> char '-') char '=' <|> fail "Expected '=' after directive key" annotations <- case key of - "disable" -> readCode `sepBy` char ',' + "disable" -> plainOrQuoted $ readElement `sepBy` char ',' where + readElement = readRange <|> readAll + readAll = do + string "all" + return $ DisableComment 0 1000000 + readRange = do + from <- readCode + to <- choice [ char '-' *> readCode, return $ from+1 ] + return $ DisableComment from to readCode = do optional $ string "SC" int <- many1 digit - return $ DisableComment (read int) + return $ read int + + "enable" -> plainOrQuoted $ readName `sepBy` char ',' + where + readName = EnableComment <$> many1 (letter <|> char '-') "source" -> do - filename <- many1 $ noneOf " \n" + filename <- quoted (many1 anyChar) <|> (many1 $ noneOf " \n") return [SourceOverride filename] + "source-path" -> do + dirname <- quoted (many1 anyChar) <|> (many1 $ noneOf " \n") + return [SourcePath dirname] + "shell" -> do pos <- getPosition - shell <- many1 $ noneOf " \n" + shell <- quoted (many1 anyChar) <|> (many1 $ noneOf " \n") when (isNothing $ shellForExecutable shell) $ parseNoteAt pos ErrorC 1103 "This shell type is unknown. Use e.g. sh or bash." return [ShellOverride shell] + "extended-analysis" -> do + pos <- getPosition + value <- plainOrQuoted $ many1 letter + case value of + "true" -> return [ExtendedAnalysis True] + "false" -> return [ExtendedAnalysis False] + _ -> do + parseNoteAt pos ErrorC 1146 "Unknown extended-analysis value. Expected true/false." + return [] + + "external-sources" -> do + pos <- getPosition + value <- plainOrQuoted $ many1 letter + case value of + "true" -> + if sandboxed + then do + parseNoteAt pos ErrorC 1144 "external-sources can only be enabled in .shellcheckrc, not in individual files." + return [] + else return [ExternalSources True] + "false" -> return [ExternalSources False] + _ -> do + parseNoteAt pos ErrorC 1145 "Unknown external-sources value. Expected true/false." + return [] + _ -> do parseNoteAt keyPos WarningC 1107 "This directive is unknown. It will be ignored." anyChar `reluctantlyTill` whitespace @@ -998,6 +1093,7 @@ readComment = do unexpecting "shellcheck annotation" readAnnotationPrefix readAnyComment +prop_readAnyComment = isOk readAnyComment "# Comment" readAnyComment = do char '#' many $ noneOf "\r\n" @@ -1014,14 +1110,16 @@ prop_readNormalWord9 = isOk readSubshell "(foo\\ ;\nbar)" prop_readNormalWord10 = isWarning readNormalWord "\x201Chello\x201D" prop_readNormalWord11 = isWarning readNormalWord "\x2018hello\x2019" prop_readNormalWord12 = isWarning readNormalWord "hello\x2018" -readNormalWord = readNormalishWord "" +readNormalWord = readNormalishWord "" ["do", "done", "then", "fi", "esac"] -readNormalishWord end = do +readPatternWord = readNormalishWord "" ["esac"] + +readNormalishWord end terms = do start <- startSpan pos <- getPosition x <- many1 (readNormalWordPart end) id <- endSpan start - checkPossibleTermination pos x + checkPossibleTermination pos x terms return $ T_NormalWord id x readIndexSpan = do @@ -1041,10 +1139,10 @@ readIndexSpan = do id <- endSpan start return $ T_Literal id str -checkPossibleTermination pos [T_Literal _ x] = - when (x `elem` ["do", "done", "then", "fi", "esac"]) $ +checkPossibleTermination pos [T_Literal _ x] terminators = + when (x `elem` terminators) $ parseProblemAt pos WarningC 1010 $ "Use semicolon or linefeed before '" ++ x ++ "' (or quote to make it literal)." -checkPossibleTermination _ _ = return () +checkPossibleTermination _ _ _ = return () readNormalWordPart end = do notFollowedBy2 $ oneOf end @@ -1101,7 +1199,7 @@ readDollarBracedPart = readSingleQuoted <|> readDoubleQuoted <|> readDollarBracedLiteral = do start <- startSpan - vars <- (readBraceEscaped <|> (anyChar >>= \x -> return [x])) `reluctantlyTill1` bracedQuotable + vars <- (readBraceEscaped <|> ((\x -> [x]) <$> anyChar)) `reluctantlyTill1` bracedQuotable id <- endSpan start return $ T_Literal id $ concat vars @@ -1200,7 +1298,7 @@ readBackTicked quoted = called "backtick expansion" $ do suggestForgotClosingQuote startPos endPos "backtick expansion" -- Result positions may be off due to escapes - result <- subParse subStart subParser (unEscape subString) + result <- subParse subStart (tryWithErrors subParser <|> return []) (unEscape subString) return $ T_Backticked id result where unEscape [] = [] @@ -1316,6 +1414,8 @@ prop_readGlob5 = isOk readGlob "[^[:alpha:]1-9]" prop_readGlob6 = isOk readGlob "[\\|]" prop_readGlob7 = isOk readGlob "[^[]" prop_readGlob8 = isOk readGlob "[*?]" +prop_readGlob9 = isOk readGlob "[!]^]" +prop_readGlob10 = isOk readGlob "[]]" readGlob = readExtglob <|> readSimple <|> readClass <|> readGlobbyLiteral where readSimple = do @@ -1323,22 +1423,25 @@ readGlob = readExtglob <|> readSimple <|> readClass <|> readGlobbyLiteral c <- oneOf "*?" id <- endSpan start return $ T_Glob id [c] - -- Doesn't handle weird things like [^]a] and [$foo]. fixme? readClass = try $ do start <- startSpan char '[' - s <- many1 (predefined <|> readNormalLiteralPart "]" <|> globchars) + negation <- charToString (oneOf "!^") <|> return "" + leadingBracket <- charToString (oneOf "]") <|> return "" + s <- many (predefined <|> readNormalLiteralPart "]" <|> globchars) + guard $ not (null leadingBracket) || not (null s) char ']' id <- endSpan start - return $ T_Glob id $ "[" ++ concat s ++ "]" + return $ T_Glob id $ "[" ++ concat (negation:leadingBracket:s) ++ "]" where - globchars = fmap return . oneOf $ "!$[" ++ extglobStartChars + globchars = charToString $ oneOf $ "![" ++ extglobStartChars predefined = do try $ string "[:" s <- many1 letter string ":]" return $ "[:" ++ s ++ ":]" + charToString = fmap return readGlobbyLiteral = do start <- startSpan c <- extglobStart <|> char '[' @@ -1361,6 +1464,8 @@ readNormalEscaped = called "escaped char" $ do do next <- quotable <|> oneOf "?*@!+[]{}.,~#" when (next == ' ') $ checkTrailingSpaces pos <|> return () + -- Check if this line is followed by a commented line with a trailing backslash + when (next == '\n') $ try . lookAhead $ void spacing return $ if next == '\n' then "" else [next] <|> do @@ -1428,7 +1533,6 @@ readSingleEscaped = do case x of '\'' -> parseProblemAt pos InfoC 1003 "Want to escape a single quote? echo 'This is how it'\\''s done'."; - '\n' -> parseProblemAt pos InfoC 1004 "This backslash+linefeed is literal. Break outside single quotes if you just want to break the line." _ -> return () return [s] @@ -1457,7 +1561,7 @@ readGenericLiteral endChars = do return $ concat strings readGenericLiteral1 endExp = do - strings <- (readGenericEscaped <|> (anyChar >>= \x -> return [x])) `reluctantlyTill1` endExp + strings <- (readGenericEscaped <|> ((\x -> [x]) <$> anyChar)) `reluctantlyTill1` endExp return $ concat strings readGenericEscaped = do @@ -1506,10 +1610,10 @@ ensureDollar = readNormalDollar = do ensureDollar - readDollarExp <|> readDollarDoubleQuote <|> readDollarSingleQuote <|> readDollarLonely + readDollarExp <|> readDollarDoubleQuote <|> readDollarSingleQuote <|> readDollarLonely False readDoubleQuotedDollar = do ensureDollar - readDollarExp <|> readDollarLonely + readDollarExp <|> readDollarLonely True prop_readDollarExpression1 = isOk readDollarExpression "$(((1) && 3))" @@ -1523,7 +1627,7 @@ readDollarExpression = do readDollarExp = arithmetic <|> readDollarExpansion <|> readDollarBracket <|> readDollarBraceCommandExpansion <|> readDollarBraced <|> readDollarVariable where arithmetic = readAmbiguous "$((" readDollarArithmetic readDollarExpansion (\pos -> - parseNoteAt pos WarningC 1102 "Shells disambiguate $(( differently or not at all. For $(command substition), add space after $( . For $((arithmetics)), fix parsing errors.") + parseNoteAt pos ErrorC 1102 "Shells disambiguate $(( differently or not at all. For $(command substitution), add space after $( . For $((arithmetics)), fix parsing errors.") prop_readDollarSingleQuote = isOk readDollarSingleQuote "$'foo\\\'lol'" readDollarSingleQuote = called "$'..' expression" $ do @@ -1572,6 +1676,7 @@ readArithmeticExpression = called "((..)) command" $ do c <- readArithmeticContents string "))" id <- endSpan start + spacing return (T_Arithmetic id c) -- If the next characters match prefix, try two different parsers and warn if the alternate parser had to be used @@ -1611,11 +1716,11 @@ readDollarBraced = called "parameter expansion" $ do word <- readDollarBracedWord char '}' id <- endSpan start - return $ T_DollarBraced id word + return $ T_DollarBraced id True word -prop_readDollarExpansion1= isOk readDollarExpansion "$(echo foo; ls\n)" -prop_readDollarExpansion2= isOk readDollarExpansion "$( )" -prop_readDollarExpansion3= isOk readDollarExpansion "$( command \n#comment \n)" +prop_readDollarExpansion1 = isOk readDollarExpansion "$(echo foo; ls\n)" +prop_readDollarExpansion2 = isOk readDollarExpansion "$( )" +prop_readDollarExpansion3 = isOk readDollarExpansion "$( command \n#comment \n)" readDollarExpansion = called "command expansion" $ do start <- startSpan try (string "$(") @@ -1638,7 +1743,7 @@ readDollarVariable = do let singleCharred p = do value <- wrapString ((:[]) <$> p) id <- endSpan start - return $ (T_DollarBraced id value) + return $ (T_DollarBraced id False value) let positional = do value <- singleCharred digit @@ -1651,7 +1756,7 @@ readDollarVariable = do let regular = do value <- wrapString readVariableName id <- endSpan start - return (T_DollarBraced id value) `attempting` do + return (T_DollarBraced id False value) `attempting` do lookAhead $ char '[' parseNoteAt pos ErrorC 1087 "Use braces when expanding arrays, e.g. ${array[idx]} (or ${var}[.. to quiet)." @@ -1671,12 +1776,32 @@ readVariableName = do rest <- many variableChars return (f:rest) -readDollarLonely = do + +prop_readDollarLonely1 = isWarning readNormalWord "\"$\"var" +prop_readDollarLonely2 = isWarning readNormalWord "\"$\"\"var\"" +prop_readDollarLonely3 = isOk readNormalWord "\"$\"$var" +prop_readDollarLonely4 = isOk readNormalWord "\"$\"*" +prop_readDollarLonely5 = isOk readNormalWord "$\"str\"" +readDollarLonely quoted = do start <- startSpan char '$' id <- endSpan start - n <- lookAhead (anyChar <|> (eof >> return '_')) + when quoted $ do + isHack <- quoteForEscape + when isHack $ + parseProblemAtId id StyleC 1135 + "Prefer escape over ending quote to make $ literal. Instead of \"It costs $\"5, use \"It costs \\$5\"." return $ T_Literal id "$" + where + quoteForEscape = option False $ try . lookAhead $ do + char '"' + -- Check for "foo $""bar" + optional $ char '"' + c <- anyVar + -- Don't trigger on [[ x == "$"* ]] or "$"$pattern + return $ c `notElem` "*$" + anyVar = variableStart <|> digit <|> specialVariable + prop_readHereDoc = isOk readScript "cat << foo\nlol\ncow\nfoo" prop_readHereDoc2 = isNotOk readScript "cat <<- EOF\n cow\n EOF" @@ -1687,16 +1812,19 @@ prop_readHereDoc6 = isOk readScript "cat << foo\\ bar\ncow\nfoo bar" prop_readHereDoc7 = isOk readScript "cat << foo\n\\$(f ())\nfoo" prop_readHereDoc8 = isOk readScript "cat <>bar\netc\nfoo" prop_readHereDoc9 = isOk readScript "if true; then cat << foo; fi\nbar\nfoo\n" -prop_readHereDoc10= isOk readScript "if true; then cat << foo << bar; fi\nfoo\nbar\n" -prop_readHereDoc11= isOk readScript "cat << foo $(\nfoo\n)lol\nfoo\n" -prop_readHereDoc12= isOk readScript "cat << foo|cat\nbar\nfoo" -prop_readHereDoc13= isOk readScript "cat <<'#!'\nHello World\n#!\necho Done" -prop_readHereDoc14= isWarning readScript "cat << foo\nbar\nfoo \n" -prop_readHereDoc15= isWarning readScript "cat < (Quoted, String) + unquote "" = (Unquoted, "") + unquote [c] = (Unquoted, [c]) + unquote s@(cl:tl) = + case reverse tl of + (cr:tr) | cr == cl && cl `elem` "\"'" -> (Quoted, reverse tr) + _ -> (if '\\' `elem` s then (Quoted, filter ((/=) '\\') s) else (Unquoted, s)) -- Fun fact: bash considers << foo"" quoted, but not << <("foo"). - -- Instead of replicating this, just read a token and strip quotes. readToken = do str <- readStringForParser readNormalWord - return (if any (`elem` quotes) str then Quoted else Unquoted, - filter (not . (`elem` quotes)) str) - + -- A here doc actually works with \r\n because the \r becomes part of the token + crstr <- (carriageReturn >> (return $ str ++ "\r")) <|> return str + return $ unquote crstr readPendingHereDocs = do docs <- popPendingHereDocs mapM_ readDoc docs where - readDoc (HereDocPending (T_HereDoc id dashed quoted endToken _) ctx) = + readDoc (HereDocPending id dashed quoted endToken ctx) = swapContext ctx $ do docStartPos <- getPosition @@ -1771,7 +1904,7 @@ readPendingHereDocs = do let thereIsNoTrailer = null trailingSpace && null trailer let leaderIsOk = null leadingSpace || dashed == Dashed && leadingSpacesAreTabs - let trailerStart = if null trailer then '\0' else head trailer + let trailerStart = case trailer of [] -> '\0'; (h:_) -> h let hasTrailingSpace = not $ null trailingSpace let hasTrailer = not $ null trailer let ppt = parseProblemAt trailerPos ErrorC @@ -1803,7 +1936,7 @@ readPendingHereDocs = do -- The end token is just a prefix skipLine | hasTrailer -> - error "ShellCheck bug, please report (here doc trailer)." + error $ pleaseReport "unexpected heredoc trailer" -- The following cases assume no trailing text: | dashed == Undashed && (not $ null leadingSpace) -> do @@ -1837,14 +1970,14 @@ readPendingHereDocs = do debugHereDoc tokenId endToken doc | endToken `isInfixOf` doc = let lookAt line = when (endToken `isInfixOf` line) $ - parseProblemAtId tokenId ErrorC 1042 ("Close matches include '" ++ line ++ "' (!= '" ++ endToken ++ "').") + parseProblemAtId tokenId ErrorC 1042 ("Close matches include '" ++ (e4m line) ++ "' (!= '" ++ (e4m endToken) ++ "').") in do - parseProblemAtId tokenId ErrorC 1041 ("Found '" ++ endToken ++ "' further down, but not on a separate line.") + parseProblemAtId tokenId ErrorC 1041 ("Found '" ++ (e4m endToken) ++ "' further down, but not on a separate line.") mapM_ lookAt (lines doc) | map toLower endToken `isInfixOf` map toLower doc = - parseProblemAtId tokenId ErrorC 1043 ("Found " ++ endToken ++ " further down, but with wrong casing.") + parseProblemAtId tokenId ErrorC 1043 ("Found " ++ (e4m endToken) ++ " further down, but with wrong casing.") | otherwise = - parseProblemAtId tokenId ErrorC 1044 ("Couldn't find end token `" ++ endToken ++ "' in the here document.") + parseProblemAtId tokenId ErrorC 1044 ("Couldn't find end token `" ++ (e4m endToken) ++ "' in the here document.") readFilename = readNormalWord @@ -1900,8 +2033,6 @@ readIoRedirect = do spacing return $ T_FdRedirect id n redir -readRedirectList = many1 readIoRedirect - prop_readHereString = isOk readHereString "<<< \"Hello $world\"" readHereString = called "here string" $ do start <- startSpan @@ -1911,13 +2042,16 @@ readHereString = called "here string" $ do word <- readNormalWord return $ T_HereString id word +prop_readNewlineList1 = isOk readScript "&> /dev/null echo foo" readNewlineList = many1 ((linefeed <|> carriageReturn) `thenSkip` spacing) <* checkBadBreak where checkBadBreak = optional $ do pos <- getPosition - try $ lookAhead (oneOf "|&") -- |, || or && - parseProblemAt pos ErrorC 1133 "Unexpected start of line. If breaking lines, |/||/&& should be at the end of the previous one." + try $ lookAhead (oneOf "|&") -- See if the next thing could be |, || or && + notFollowedBy2 (string "&>") -- Except &> or &>> which is valid + parseProblemAt pos ErrorC 1133 + "Unexpected start of line. If breaking lines, |/||/&& should be at the end of the previous one." readLineBreak = optional readNewlineList prop_readSeparator1 = isWarning readScript "a &; b" @@ -1975,6 +2109,7 @@ prop_readSimpleCommand4 = isOk readSimpleCommand "typeset -a foo=(lol)" prop_readSimpleCommand5 = isOk readSimpleCommand "time if true; then echo foo; fi" prop_readSimpleCommand6 = isOk readSimpleCommand "time -p ( ls -l; )" prop_readSimpleCommand7 = isOk readSimpleCommand "\\ls" +prop_readSimpleCommand7b = isOk readSimpleCommand "\\:" prop_readSimpleCommand8 = isWarning readSimpleCommand "// Lol" prop_readSimpleCommand9 = isWarning readSimpleCommand "/* Lolbert */" prop_readSimpleCommand10 = isWarning readSimpleCommand "/**** Lolbert */" @@ -1982,6 +2117,7 @@ prop_readSimpleCommand11 = isOk readSimpleCommand "/\\* foo" prop_readSimpleCommand12 = isWarning readSimpleCommand "elsif foo" prop_readSimpleCommand13 = isWarning readSimpleCommand "ElseIf foo" prop_readSimpleCommand14 = isWarning readSimpleCommand "elseif[$i==2]" +prop_readSimpleCommand15 = isWarning readSimpleCommand "trap 'foo\"bar' INT" readSimpleCommand = called "simple command" $ do prefix <- option [] readCmdPrefix skipAnnotationAndWarn @@ -1996,7 +2132,11 @@ readSimpleCommand = called "simple command" $ do Just cmd -> do validateCommand cmd - suffix <- option [] $ getParser readCmdSuffix cmd [ + -- We have to ignore possible parsing problems from the lookAhead parser + firstArgument <- ignoreProblemsOf . optionMaybe . try . lookAhead $ readCmdWord + suffix <- option [] $ getParser readCmdSuffix + -- If `export` or other modifier commands are called with `builtin` we have to look at the first argument + (if isCommand ["builtin"] cmd then fromMaybe cmd firstArgument else cmd) [ (["declare", "export", "local", "readonly", "typeset"], readModifierSuffix), (["time"], readTimeSuffix), (["let"], readLetSuffix), @@ -2007,9 +2147,12 @@ readSimpleCommand = called "simple command" $ do id2 <- getNewIdFor id1 let result = makeSimpleCommand id1 id2 prefix [cmd] suffix - if isCommand ["source", "."] cmd - then readSource result - else return result + case () of + _ | isCommand ["source", "."] cmd -> readSource result + _ | isCommand ["trap"] cmd -> do + syntaxCheckTrap result + return result + _ -> return result where isCommand strings (T_NormalWord _ [T_Literal _ s]) = s `elem` strings isCommand _ _ = False @@ -2019,10 +2162,6 @@ readSimpleCommand = called "simple command" $ do then action else getParser def cmd rest - cStyleComment cmd = - case cmd of - _ -> False - validateCommand cmd = case cmd of (T_NormalWord _ [T_Literal _ "//"]) -> commentWarning (getId cmd) @@ -2033,6 +2172,17 @@ readSimpleCommand = called "simple command" $ do parseProblemAtId (getId cmd) ErrorC 1131 "Use 'elif' to start another branch." _ -> return () + syntaxCheckTrap cmd = + case cmd of + (T_Redirecting _ _ (T_SimpleCommand _ _ (cmd:arg:_))) -> checkArg arg (getLiteralString arg) + _ -> return () + where + checkArg _ Nothing = return () + checkArg arg (Just ('-':_)) = return () + checkArg arg (Just str) = do + (start,end) <- getSpanForId (getId arg) + subParse start (tryWithErrors (readCompoundListOrEmpty >> verifyEof) <|> return ()) str + commentWarning id = parseProblemAtId id ErrorC 1127 "Was this intended as a comment? Use # in sh." @@ -2055,31 +2205,40 @@ readSimpleCommand = called "simple command" $ do readSource :: Monad m => Token -> SCParser m Token -readSource t@(T_Redirecting _ _ (T_SimpleCommand cmdId _ (cmd:file:_))) = do +readSource t@(T_Redirecting _ _ (T_SimpleCommand cmdId _ (cmd:file':rest'))) = do + let file = getFile file' rest' override <- getSourceOverride let literalFile = do - name <- override `mplus` getLiteralString file + name <- override `mplus` getLiteralString file `mplus` stripDynamicPrefix file -- Hack to avoid 'source ~/foo' trying to read from literal tilde guard . not $ "~/" `isPrefixOf` name return name case literalFile of Nothing -> do parseNoteAtId (getId file) WarningC 1090 - "Can't follow non-constant source. Use a directive to specify location." + "ShellCheck can't follow non-constant source. Use a directive to specify location." return t Just filename -> do proceed <- shouldFollow filename if not proceed then do + -- FIXME: This actually gets squashed without -a parseNoteAtId (getId file) InfoC 1093 "This file appears to be recursively sourced. Ignoring." return t else do sys <- Mr.asks systemInterface - input <- + (input, resolvedFile) <- if filename == "/dev/null" -- always allow /dev/null - then return (Right "") - else system $ siReadFile sys filename + then return (Right "", filename) + else do + allAnnotations <- getCurrentAnnotations True + currentScript <- Mr.asks currentFilename + let paths = mapMaybe getSourcePath allAnnotations + let externalSources = listToMaybe $ mapMaybe getExternalSources allAnnotations + resolved <- system $ siFindSource sys currentScript externalSources paths filename + contents <- system $ siReadFile sys externalSources resolved + return (contents, resolved) case input of Left err -> do parseNoteAtId (getId file) InfoC 1091 $ @@ -2090,7 +2249,7 @@ readSource t@(T_Redirecting _ _ (T_SimpleCommand cmdId _ (cmd:file:_))) = do id2 <- getNewIdFor cmdId let included = do - src <- subRead filename script + src <- subRead resolvedFile script return $ T_SourceCommand id1 t (T_Include id2 src) let failed = do @@ -2100,24 +2259,60 @@ readSource t@(T_Redirecting _ _ (T_SimpleCommand cmdId _ (cmd:file:_))) = do included <|> failed where + getFile :: Token -> [Token] -> Token + getFile file (next:rest) = + case getLiteralString file of + Just "--" -> next + x -> file + getFile file _ = file + + getSourcePath t = + case t of + SourcePath x -> Just x + _ -> Nothing + + getExternalSources t = + case t of + ExternalSources b -> Just b + _ -> Nothing + + -- If the word has a single expansion as the directory, try stripping it + -- This affects `$foo/bar` but not `${foo}-dir/bar` or `/foo/$file` + stripDynamicPrefix word = + case getWordParts word of + exp : rest | isStringExpansion exp -> do + str <- getLiteralString (T_NormalWord (Id 0) rest) + guard $ "/" `isPrefixOf` str + return $ "." ++ str + _ -> Nothing + subRead name script = withContext (ContextSource name) $ - inSeparateContext $ - subParse (initialPos name) readScript script + inSeparateContext $ do + oldState <- getState + setState $ oldState { pendingHereDocs = [] } + result <- subParse (initialPos name) (readScriptFile True) script + newState <- getState + setState $ newState { pendingHereDocs = pendingHereDocs oldState } + return result readSource t = return t prop_readPipeline = isOk readPipeline "! cat /etc/issue | grep -i ubuntu" prop_readPipeline2 = isWarning readPipeline "!cat /etc/issue | grep -i ubuntu" prop_readPipeline3 = isOk readPipeline "for f; do :; done|cat" +prop_readPipeline4 = isOk readPipeline "! ! true" +prop_readPipeline5 = isOk readPipeline "true | ! true" readPipeline = do unexpecting "keyword/token" readKeyword - do - (T_Bang id) <- g_Bang - pipe <- readPipeSequence - return $ T_Banged id pipe - <|> - readPipeSequence + readBanged readPipeSequence + +readBanged parser = do + pos <- getPosition + (T_Bang id) <- g_Bang + next <- readBanged parser + return $ T_Banged id next + <|> parser prop_readAndOr = isOk readAndOr "grep -i lol foo || exit 1" prop_readAndOr1 = isOk readAndOr "# shellcheck disable=1\nfoo" @@ -2133,7 +2328,7 @@ readAndOr = do parseProblemAt apos ErrorC 1123 "ShellCheck directives are only valid in front of complete compound commands, like 'if', not e.g. individual 'elif' branches." andOr <- withAnnotations annotations $ - chainr1 readPipeline $ do + chainl1 readPipeline $ do op <- g_AND_IF <|> g_OR_IF readLineBreak return $ case op of T_AND_IF id -> T_AndIf id @@ -2173,14 +2368,14 @@ readTerm = do readPipeSequence = do start <- startSpan - (cmds, pipes) <- sepBy1WithSeparators readCommand + (cmds, pipes) <- sepBy1WithSeparators (readBanged readCommand) (readPipe `thenSkip` (spacing >> readLineBreak)) id <- endSpan start spacing return $ T_Pipeline id pipes cmds where sepBy1WithSeparators p s = do - let elems = p >>= \x -> return ([x], []) + let elems = (\x -> ([x], [])) <$> p let seps = do separator <- s return $ \(a,b) (c,d) -> (a++c, b ++ d ++ [separator]) @@ -2197,15 +2392,20 @@ readPipe = do readCommand = choice [ readCompoundCommand, + readConditionCommand, readCoProc, readSimpleCommand ] readCmdName = do + -- If the command name is `!` then + optional . lookAhead . try $ do + char '!' + whitespace -- Ignore alias suppression optional . try $ do char '\\' - lookAhead $ variableChars + lookAhead $ variableChars <|> oneOf ":." readCmdWord readCmdWord = do @@ -2225,6 +2425,7 @@ prop_readIfClause2 = isWarning readIfClause "if false; then; echo oo; fi" prop_readIfClause3 = isWarning readIfClause "if false; then true; else; echo lol; fi" prop_readIfClause4 = isWarning readIfClause "if false; then true; else if true; then echo lol; fi; fi" prop_readIfClause5 = isOk readIfClause "if false; then true; else\nif true; then echo lol; fi; fi" +prop_readIfClause6 = isWarning readIfClause "if true\nthen\nDo the thing\nfi" readIfClause = called "if expression" $ do start <- startSpan pos <- getPosition @@ -2308,6 +2509,7 @@ readSubshell = called "explicit subshell" $ do allspacing char ')' <|> fail "Expected ) closing the subshell" id <- endSpan start + spacing return $ T_Subshell id list prop_readBraceGroup = isOk readBraceGroup "{ a; b | c | d; e; }" @@ -2328,8 +2530,33 @@ readBraceGroup = called "brace group" $ do parseProblem ErrorC 1056 "Expected a '}'. If you have one, try a ; or \\n in front of it." fail "Missing '}'" id <- endSpan start + spacing return $ T_BraceGroup id list +prop_readBatsTest1 = isOk readBatsTest "@test 'can parse' {\n true\n}" +prop_readBatsTest2 = isOk readBatsTest "@test random text !(@*$Y&! {\n true\n}" +prop_readBatsTest3 = isOk readBatsTest "@test foo { bar { baz {\n true\n}" +prop_readBatsTest4 = isNotOk readBatsTest "@test foo \n{\n true\n}" +readBatsTest = called "bats @test" $ do + start <- startSpan + try $ string "@test " + spacing + name <- readBatsName + spacing + test <- readBraceGroup + id <- endSpan start + return $ T_BatsTest id name test + where + readBatsName = do + line <- try . lookAhead $ many1 $ noneOf "\n" + let name = reverse $ f $ reverse line + string name + + -- We want everything before the last " {" in a string, so we find everything after "{ " in its reverse + f ('{':' ':rest) = dropWhile isSpace rest + f (a:rest) = f rest + f [] = "" + prop_readWhileClause = isOk readWhileClause "while [[ -e foo ]]; do sleep 1; done" readWhileClause = called "while loop" $ do start <- startSpan @@ -2357,7 +2584,7 @@ readDoGroup kwId = do parseProblem ErrorC 1058 "Expected 'do'." return "Expected 'do'" - acceptButWarn g_Semi ErrorC 1059 "No semicolons directly after 'do'." + acceptButWarn g_Semi ErrorC 1059 "Semicolon is not allowed directly after 'do'. You can just delete it." allspacing optional (do @@ -2369,10 +2596,16 @@ readDoGroup kwId = do parseProblemAtId (getId doKw) ErrorC 1061 "Couldn't find 'done' for this 'do'." parseProblem ErrorC 1062 "Expected 'done' matching previously mentioned 'do'." return "Expected 'done'" + + optional . lookAhead $ do + pos <- getPosition + try $ string "<(" + parseProblemAt pos ErrorC 1142 "Use 'done < <(cmd)' to redirect from process substitution (currently missing one '<')." return commands prop_readForClause = isOk readForClause "for f in *; do rm \"$f\"; done" +prop_readForClause1 = isOk readForClause "for f in *; { rm \"$f\"; }" prop_readForClause3 = isOk readForClause "for f; do foo; done" prop_readForClause4 = isOk readForClause "for((i=0; i<10; i++)); do echo $i; done" prop_readForClause5 = isOk readForClause "for ((i=0;i<10 && n>x;i++,--n))\ndo \necho $i\ndone" @@ -2380,9 +2613,9 @@ prop_readForClause6 = isOk readForClause "for ((;;))\ndo echo $i\ndone" prop_readForClause7 = isOk readForClause "for ((;;)) do echo $i\ndone" prop_readForClause8 = isOk readForClause "for ((;;)) ; do echo $i\ndone" prop_readForClause9 = isOk readForClause "for i do true; done" -prop_readForClause10= isOk readForClause "for ((;;)) { true; }" -prop_readForClause12= isWarning readForClause "for $a in *; do echo \"$a\"; done" -prop_readForClause13= isOk readForClause "for foo\nin\\\n bar\\\n baz\ndo true; done" +prop_readForClause10 = isOk readForClause "for ((;;)) { true; }" +prop_readForClause12 = isWarning readForClause "for $a in *; do echo \"$a\"; done" +prop_readForClause13 = isOk readForClause "for foo\nin\\\n bar\\\n baz\ndo true; done" readForClause = called "for loop" $ do pos <- getPosition (T_For id) <- g_For @@ -2390,19 +2623,31 @@ readForClause = called "for loop" $ do readArithmetic id <|> readRegular id where readArithmetic id = called "arithmetic for condition" $ do - try $ string "((" + readArithmeticDelimiter '(' "Missing second '(' to start arithmetic for ((;;)) loop" x <- readArithmeticContents char ';' >> spacing y <- readArithmeticContents char ';' >> spacing z <- readArithmeticContents spacing - string "))" + readArithmeticDelimiter ')' "Missing second ')' to terminate 'for ((;;))' loop condition" spacing optional $ readSequentialSep >> spacing group <- readBraced <|> readDoGroup id return $ T_ForArithmetic id x y z group + -- For c='(' read "((" and be lenient about spaces + readArithmeticDelimiter c msg = do + char c + startPos <- getPosition + sp <- spacing + endPos <- getPosition + char c <|> do + parseProblemAt startPos ErrorC 1137 msg + fail "" + unless (null sp) $ + parseProblemAtWithEnd startPos endPos ErrorC 1138 $ "Remove spaces between " ++ [c,c] ++ " in arithmetic for loop." + readBraced = do (T_BraceGroup _ list) <- readBraceGroup return list @@ -2412,7 +2657,7 @@ readForClause = called "for loop" $ do "Don't use $ on the iterator name in for loops." name <- readVariableName `thenSkip` allspacing values <- readInClause <|> (optional readSequentialSep >> return []) - group <- readDoGroup id + group <- readBraced <|> readDoGroup id return $ T_ForIn id name values group prop_readSelectClause1 = isOk readSelectClause "select foo in *; do echo $foo; done" @@ -2450,6 +2695,7 @@ prop_readCaseClause2 = isOk readCaseClause "case foo\n in * ) echo bar;; esac" prop_readCaseClause3 = isOk readCaseClause "case foo\n in * ) echo bar & ;; esac" prop_readCaseClause4 = isOk readCaseClause "case foo\n in *) echo bar ;& bar) foo; esac" prop_readCaseClause5 = isOk readCaseClause "case foo\n in *) echo bar;;& foo) baz;; esac" +prop_readCaseClause6 = isOk readCaseClause "case foo\n in if) :;; done) :;; esac" readCaseClause = called "case expression" $ do start <- startSpan g_Case @@ -2501,10 +2747,10 @@ prop_readFunctionDefinition6 = isOk readFunctionDefinition "?(){ foo; }" prop_readFunctionDefinition7 = isOk readFunctionDefinition "..(){ cd ..; }" prop_readFunctionDefinition8 = isOk readFunctionDefinition "foo() (ls)" prop_readFunctionDefinition9 = isOk readFunctionDefinition "function foo { true; }" -prop_readFunctionDefinition10= isOk readFunctionDefinition "function foo () { true; }" -prop_readFunctionDefinition11= isWarning readFunctionDefinition "function foo{\ntrue\n}" -prop_readFunctionDefinition12= isOk readFunctionDefinition "function []!() { true; }" -prop_readFunctionDefinition13= isOk readFunctionDefinition "@require(){ true; }" +prop_readFunctionDefinition10 = isOk readFunctionDefinition "function foo () { true; }" +prop_readFunctionDefinition11 = isWarning readFunctionDefinition "function foo{\ntrue\n}" +prop_readFunctionDefinition12 = isOk readFunctionDefinition "function []!() { true; }" +prop_readFunctionDefinition13 = isOk readFunctionDefinition "@require(){ true; }" readFunctionDefinition = called "function" $ do start <- startSpan functionSignature <- try readFunctionSignature @@ -2532,7 +2778,7 @@ readFunctionDefinition = called "function" $ do readWithoutFunction = try $ do name <- many1 functionChars - guard $ name /= "time" -- Interfers with time ( foo ) + guard $ name /= "time" -- Interferes with time ( foo ) spacing readParens return $ \id -> T_Function id (FunctionKeyword False) (FunctionParentheses True) name @@ -2549,17 +2795,29 @@ readFunctionDefinition = called "function" $ do prop_readCoProc1 = isOk readCoProc "coproc foo { echo bar; }" prop_readCoProc2 = isOk readCoProc "coproc { echo bar; }" prop_readCoProc3 = isOk readCoProc "coproc echo bar" +prop_readCoProc4 = isOk readCoProc "coproc a=b echo bar" +prop_readCoProc5 = isOk readCoProc "coproc 'foo' { echo bar; }" +prop_readCoProc6 = isOk readCoProc "coproc \"foo$$\" { echo bar; }" +prop_readCoProc7 = isOk readCoProc "coproc 'foo' ( echo bar )" +prop_readCoProc8 = isOk readCoProc "coproc \"foo$$\" while true; do true; done" readCoProc = called "coproc" $ do start <- startSpan try $ do string "coproc" - whitespace + spacing1 choice [ try $ readCompoundCoProc start, readSimpleCoProc start ] where readCompoundCoProc start = do - var <- optionMaybe $ - readVariableName `thenSkip` whitespace - body <- readBody readCompoundCommand + notFollowedBy2 readAssignmentWord + (var, body) <- choice [ + try $ do + body <- readBody readCompoundCommand + return (Nothing, body), + try $ do + var <- readNormalWord `thenSkip` spacing + body <- readBody readCompoundCommand + return (Just var, body) + ] id <- endSpan start return $ T_CoProc id var body readSimpleCoProc start = do @@ -2572,34 +2830,63 @@ readCoProc = called "coproc" $ do id <- endSpan start return $ T_CoProcBody id body +readPattern = (readPatternWord `thenSkip` spacing) `sepBy1` (char '|' `thenSkip` spacing) -readPattern = (readNormalWord `thenSkip` spacing) `sepBy1` (char '|' `thenSkip` spacing) +prop_readConditionCommand = isOk readConditionCommand "[[ x ]] > foo 2>&1" +readConditionCommand = do + cmd <- readCondition + redirs <- many readIoRedirect + id <- getNextIdSpanningTokenList (cmd:redirs) + + pos <- getPosition + hasDashAo <- isFollowedBy $ do + c <- choice $ try . string <$> ["-o", "-a", "or", "and"] + posEnd <- getPosition + parseProblemAtWithEnd pos posEnd ErrorC 1139 $ + "Use " ++ alt c ++ " instead of '" ++ c ++ "' between test commands." + + -- If the next word is a keyword, readNormalWord will trigger a warning + hasKeyword <- isFollowedBy readKeyword + hasWord <- isFollowedBy readNormalWord + + when (hasWord && not (hasKeyword || hasDashAo)) $ do + -- We have other words following, and no error has been emitted. + posEnd <- getPosition + parseProblemAtWithEnd pos posEnd ErrorC 1140 "Unexpected parameters after condition. Missing &&/||, or bad expression?" + + return $ T_Redirecting id redirs cmd + where + alt "or" = "||" + alt "-o" = "||" + alt "and" = "&&" + alt "-a" = "&&" + alt _ = "|| or &&" prop_readCompoundCommand = isOk readCompoundCommand "{ echo foo; }>/dev/null" readCompoundCommand = do cmd <- choice [ readBraceGroup, readAmbiguous "((" readArithmeticExpression readSubshell (\pos -> - parseNoteAt pos WarningC 1105 "Shells disambiguate (( differently or not at all. For subshell, add spaces around ( . For ((, fix parsing errors."), + parseNoteAt pos ErrorC 1105 "Shells disambiguate (( differently or not at all. For subshell, add spaces around ( . For ((, fix parsing errors."), readSubshell, - readCondition, readWhileClause, readUntilClause, readIfClause, readForClause, readSelectClause, readCaseClause, + readBatsTest, readFunctionDefinition ] - spacing redirs <- many readIoRedirect id <- getNextIdSpanningTokenList (cmd:redirs) - unless (null redirs) $ optional $ do - lookAhead $ try (spacing >> needsSeparator) - parseProblem WarningC 1013 "Bash requires ; or \\n here, after redirecting nested compound commands." + optional . lookAhead $ do + notFollowedBy2 $ choice [readKeyword, g_Lbrace] + pos <- getPosition + many1 readNormalWord + posEnd <- getPosition + parseProblemAtWithEnd pos posEnd ErrorC 1141 "Unexpected tokens after compound command. Bad redirection or missing ;/&&/||/|?" return $ T_Redirecting id redirs cmd - where - needsSeparator = choice [ g_Then, g_Else, g_Elif, g_Fi, g_Do, g_Done, g_Esac, g_Rbrace ] readCompoundList = readTerm @@ -2629,13 +2916,13 @@ readLetSuffix = many1 (readIoRedirect <|> try readLetExpression <|> readCmdWord) startPos <- getPosition expression <- readStringForParser readCmdWord let (unQuoted, newPos) = kludgeAwayQuotes expression startPos - subParse newPos readArithmeticContents unQuoted + subParse newPos (readArithmeticContents <* eof) unQuoted kludgeAwayQuotes :: String -> SourcePos -> (String, SourcePos) kludgeAwayQuotes s p = case s of - first:rest@(_:_) -> - let (last:backwards) = reverse rest + first:second:rest -> + let (last NE.:| backwards) = NE.reverse (second NE.:| rest) middle = reverse backwards in if first `elem` "'\"" && first == last @@ -2669,75 +2956,84 @@ readLiteralForParser parser = do prop_readAssignmentWord = isOk readAssignmentWord "a=42" prop_readAssignmentWord2 = isOk readAssignmentWord "b=(1 2 3)" -prop_readAssignmentWord3 = isWarning readAssignmentWord "$b = 13" -prop_readAssignmentWord4 = isWarning readAssignmentWord "b = $(lol)" prop_readAssignmentWord5 = isOk readAssignmentWord "b+=lol" -prop_readAssignmentWord6 = isWarning readAssignmentWord "b += (1 2 3)" prop_readAssignmentWord7 = isOk readAssignmentWord "a[3$n'']=42" prop_readAssignmentWord8 = isOk readAssignmentWord "a[4''$(cat foo)]=42" prop_readAssignmentWord9 = isOk readAssignmentWord "IFS= " -prop_readAssignmentWord9a= isOk readAssignmentWord "foo=" -prop_readAssignmentWord9b= isOk readAssignmentWord "foo= " -prop_readAssignmentWord9c= isOk readAssignmentWord "foo= #bar" -prop_readAssignmentWord10= isWarning readAssignmentWord "foo$n=42" -prop_readAssignmentWord11= isOk readAssignmentWord "foo=([a]=b [c] [d]= [e f )" -prop_readAssignmentWord12= isOk readAssignmentWord "a[b <<= 3 + c]='thing'" -prop_readAssignmentWord13= isOk readAssignmentWord "var=( (1 2) (3 4) )" -prop_readAssignmentWord14= isOk readAssignmentWord "var=( 1 [2]=(3 4) )" -prop_readAssignmentWord15= isOk readAssignmentWord "var=(1 [2]=(3 4))" +prop_readAssignmentWord9a = isOk readAssignmentWord "foo=" +prop_readAssignmentWord9b = isOk readAssignmentWord "foo= " +prop_readAssignmentWord9c = isOk readAssignmentWord "foo= #bar" +prop_readAssignmentWord11 = isOk readAssignmentWord "foo=([a]=b [c] [d]= [e f )" +prop_readAssignmentWord12 = isOk readAssignmentWord "a[b <<= 3 + c]='thing'" +prop_readAssignmentWord13 = isOk readAssignmentWord "var=( (1 2) (3 4) )" +prop_readAssignmentWord14 = isOk readAssignmentWord "var=( 1 [2]=(3 4) )" +prop_readAssignmentWord15 = isOk readAssignmentWord "var=(1 [2]=(3 4))" readAssignmentWord = readAssignmentWordExt True readWellFormedAssignment = readAssignmentWordExt False -readAssignmentWordExt lenient = try $ do - start <- startSpan - pos <- getPosition - when lenient $ - optional (char '$' >> parseNote ErrorC 1066 "Don't use $ on the left side of assignments.") - variable <- readVariableName - when lenient $ - optional (readNormalDollar >> parseNoteAt pos ErrorC - 1067 "For indirection, use (associative) arrays or 'read \"var$n\" <<< \"value\"'") - indices <- many readArrayIndex - hasLeftSpace <- fmap (not . null) spacing - pos <- getPosition - id <- endSpan start - op <- readAssignmentOp +readAssignmentWordExt lenient = called "variable assignment" $ do + -- Parse up to and including the = in a 'try' + (id, variable, op, indices) <- try $ do + start <- startSpan + pos <- getPosition + -- Check for a leading $ at parse time, to warn for $foo=(bar) which + -- would otherwise cause a parse failure so it can't be checked later. + leadingDollarPos <- + if lenient + then optionMaybe $ getSpanPositionsFor (char '$') + else return Nothing + variable <- readVariableName + indices <- many readArrayIndex + hasLeftSpace <- fmap (not . null) spacing + opStart <- getPosition + id <- endSpan start + op <- readAssignmentOp + opEnd <- getPosition + + when (isJust leadingDollarPos || hasLeftSpace) $ do + hasParen <- isFollowedBy (spacing >> char '(') + when hasParen $ + sequence_ $ do + (l, r) <- leadingDollarPos + return $ parseProblemAtWithEnd l r ErrorC 1066 "Don't use $ on the left side of assignments." + + -- Fail so that this is not parsed as an assignment. + fail "" + -- At this point we know for sure. + return (id, variable, op, indices) + + rightPosStart <- getPosition hasRightSpace <- fmap (not . null) spacing + rightPosEnd <- getPosition isEndOfCommand <- fmap isJust $ optionMaybe (try . lookAhead $ (void (oneOf "\r\n;&|)") <|> eof)) - if not hasLeftSpace && (hasRightSpace || isEndOfCommand) + + if hasRightSpace || isEndOfCommand then do - when (variable /= "IFS" && hasRightSpace && not isEndOfCommand) $ - parseNoteAt pos WarningC 1007 + when (variable /= "IFS" && hasRightSpace && not isEndOfCommand) $ do + parseProblemAtWithEnd rightPosStart rightPosEnd WarningC 1007 "Remove space after = if trying to assign a value (for empty string, use var='' ... )." value <- readEmptyLiteral return $ T_Assignment id op variable indices value else do - when (hasLeftSpace || hasRightSpace) $ - parseNoteAt pos ErrorC 1068 $ - "Don't put spaces around the " - ++ if op == Append - then "+= when appending." - else "= in assignments." + optional $ do + lookAhead $ char '=' + parseProblem ErrorC 1097 "Unexpected ==. For assignment, use =. For comparison, use [/[[. Or quote for literal string." + value <- readArray <|> readNormalWord spacing return $ T_Assignment id op variable indices value where readAssignmentOp = do - pos <- getPosition - unexpecting "" $ string "===" + -- This is probably some kind of ascii art border + unexpecting "===" (string "===") choice [ string "+=" >> return Append, - do - try (string "==") - parseProblemAt pos ErrorC 1097 - "Unexpected ==. For assignment, use =. For comparison, use [/[[." - return Assign, - string "=" >> return Assign ] - readEmptyLiteral = do - start <- startSpan - id <- endSpan start - return $ T_Literal id "" + +readEmptyLiteral = do + start <- startSpan + id <- endSpan start + return $ T_Literal id "" readArrayIndex = do start <- startSpan @@ -2795,6 +3091,7 @@ redirToken c t = try $ do tryWordToken s t = tryParseWordToken s t `thenSkip` spacing tryParseWordToken keyword t = try $ do + pos <- getPosition start <- startSpan str <- anycaseString keyword id <- endSpan start @@ -2810,9 +3107,10 @@ tryParseWordToken keyword t = try $ do _ -> return () lookAhead keywordSeparator - when (str /= keyword) $ - parseProblem ErrorC 1081 $ - "Scripts are case sensitive. Use '" ++ keyword ++ "', not '" ++ str ++ "'." + when (str /= keyword) $ do + parseProblemAt pos ErrorC 1081 $ + "Scripts are case sensitive. Use '" ++ keyword ++ "', not '" ++ str ++ "' (or quote if literal)." + fail "" return $ t id anycaseString = @@ -2885,12 +3183,14 @@ prop_readShebang5 = isWarning readShebang "\n#!/bin/sh" prop_readShebang6 = isWarning readShebang " # Copyright \n!#/bin/bash" prop_readShebang7 = isNotOk readShebang "# Copyright \nfoo\n#!/bin/bash" readShebang = do + start <- startSpan anyShebang <|> try readMissingBang <|> withHeader many linewhitespace str <- many $ noneOf "\r\n" + id <- endSpan start optional carriageReturn optional linefeed - return str + return $ T_Literal id str where anyShebang = choice $ map try [ readCorrect, @@ -2966,62 +3266,112 @@ verifyEof = eof <|> choice [ try (lookAhead p) action -prop_readScript1 = isOk readScriptFile "#!/bin/bash\necho hello world\n" -prop_readScript2 = isWarning readScriptFile "#!/bin/bash\r\necho hello world\n" -prop_readScript3 = isWarning readScriptFile "#!/bin/bash\necho hello\xA0world" -prop_readScript4 = isWarning readScriptFile "#!/usr/bin/perl\nfoo=(" -prop_readScript5 = isOk readScriptFile "#!/bin/bash\n#This is an empty script\n\n" -readScriptFile = do + +readConfigFile :: Monad m => FilePath -> SCParser m [Annotation] +readConfigFile filename = do + shouldIgnore <- Mr.asks ignoreRC + if shouldIgnore then return [] else read' filename + where + read' filename = do + sys <- Mr.asks systemInterface + contents <- system $ siGetConfig sys filename + case contents of + Nothing -> return [] + Just (file, str) -> readConfig file str + + readConfig filename contents = do + result <- lift $ runParserT readConfigKVs initialUserState filename contents + case result of + Right result -> + return result + + Left err -> do + parseProblem ErrorC 1134 $ errorFor filename err + return [] + + errorFor filename err = + let line = "line " ++ (show . sourceLine $ errorPos err) + suggestion = getStringFromParsec $ errorMessages err + in + "Failed to process " ++ (e4m filename) ++ ", " ++ line ++ ": " + ++ suggestion + +prop_readConfigKVs1 = isOk readConfigKVs "disable=1234" +prop_readConfigKVs2 = isOk readConfigKVs "# Comment\ndisable=1234 # Comment\n" +prop_readConfigKVs3 = isOk readConfigKVs "" +prop_readConfigKVs4 = isOk readConfigKVs "\n\n\n\n\t \n" +prop_readConfigKVs5 = isOk readConfigKVs "# shellcheck accepts annotation-like comments in rc files\ndisable=1234" +readConfigKVs = do + anySpacingOrComment + annotations <- many (readAnnotationWithoutPrefix False <* anySpacingOrComment) + eof + return $ concat annotations +anySpacingOrComment = + many (void allspacingOrFail <|> void readAnyComment) + +prop_readScript1 = isOk readScript "#!/bin/bash\necho hello world\n" +prop_readScript2 = isWarning readScript "#!/bin/bash\r\necho hello world\n" +prop_readScript3 = isWarning readScript "#!/bin/bash\necho hello\xA0world" +prop_readScript4 = isWarning readScript "#!/usr/bin/perl\nfoo=(" +prop_readScript5 = isOk readScript "#!/bin/bash\n#This is an empty script\n\n" +prop_readScript6 = isOk readScript "#!/usr/bin/env -S X=FOO bash\n#This is an empty script\n\n" +prop_readScript7 = isOk readScript "#!/bin/zsh\n# shellcheck disable=SC1071\nfor f (a b); echo $f\n" +readScriptFile sourced = do start <- startSpan pos <- getPosition - optional $ do - readUtf8Bom - parseProblem ErrorC 1082 - "This file has a UTF-8 BOM. Remove it with: LC_CTYPE=C sed '1s/^...//' < yourscript ." - sb <- option "" readShebang - allspacing - annotationStart <- startSpan - annotations <- readAnnotations - annotationId <- endSpan annotationStart - let shellAnnotationSpecified = - any (\x -> case x of ShellOverride {} -> True; _ -> False) annotations - shellFlagSpecified <- isJust <$> Mr.asks shellTypeOverride - let ignoreShebang = shellAnnotationSpecified || shellFlagSpecified + rcAnnotations <- if sourced + then return [] + else do + filename <- Mr.asks currentFilename + readConfigFile filename - unless ignoreShebang $ - verifyShebang pos (getShell sb) - if ignoreShebang || isValidShell (getShell sb) /= Just False - then do - commands <- withAnnotations annotations readCompoundListOrEmpty - id <- endSpan start - verifyEof - let script = T_Annotation annotationId annotations $ - T_Script id sb commands - reparseIndices script - else do - many anyChar - id <- endSpan start - return $ T_Script id sb [] + -- Put the rc annotations on the stack so that one can ignore e.g. SC1084 in .shellcheckrc + withAnnotations rcAnnotations $ do + hasBom <- wasIncluded readUtf8Bom + shebang <- readShebang <|> readEmptyLiteral + let (T_Literal _ shebangString) = shebang + allspacing + annotationStart <- startSpan + fileAnnotations <- readAnnotations + + -- Similarly put the filewide annotations on the stack to allow earlier suppression + withAnnotations fileAnnotations $ do + when (hasBom) $ + parseProblemAt pos ErrorC 1082 + "This file has a UTF-8 BOM. Remove it with: LC_CTYPE=C sed '1s/^...//' < yourscript ." + let annotations = fileAnnotations ++ rcAnnotations + annotationId <- endSpan annotationStart + let shellAnnotationSpecified = + any (\x -> case x of ShellOverride {} -> True; _ -> False) annotations + shellFlagSpecified <- isJust <$> Mr.asks shellTypeOverride + let ignoreShebang = shellAnnotationSpecified || shellFlagSpecified + + unless ignoreShebang $ + verifyShebang pos (executableFromShebang shebangString) + if ignoreShebang || isValidShell (executableFromShebang shebangString) /= Just False + then do + commands <- readCompoundListOrEmpty + id <- endSpan start + readPendingHereDocs + verifyEof + let script = T_Annotation annotationId annotations $ + T_Script id shebang commands + userstate <- getState + reparseIndices $ reattachHereDocs script (hereDocMap userstate) + else do + many anyChar + id <- endSpan start + return $ T_Script id shebang [] where - basename s = reverse . takeWhile (/= '/') . reverse $ s - getShell sb = - case words sb of - [] -> "" - [x] -> basename x - (first:second:_) -> - if basename first == "env" - then second - else basename first - verifyShebang pos s = do case isValidShell s of Just True -> return () - Just False -> parseProblemAt pos ErrorC 1071 "ShellCheck only supports sh/bash/dash/ksh scripts. Sorry!" - Nothing -> parseProblemAt pos InfoC 1008 "This shebang was unrecognized. Note that ShellCheck only handles sh/bash/dash/ksh." + Just False -> parseProblemAt pos ErrorC 1071 "ShellCheck only supports sh/bash/dash/ksh/'busybox sh' scripts. Sorry!" + Nothing -> parseProblemAt pos ErrorC 1008 "This shebang was unrecognized. ShellCheck only supports sh/bash/dash/ksh/'busybox sh'. Add a 'shell' directive to specify." isValidShell s = - let good = s == "" || any (`isPrefixOf` s) goodShells + let good = null s || any (`isPrefixOf` s) goodShells bad = any (`isPrefixOf` s) badShells in if good @@ -3034,15 +3384,20 @@ readScriptFile = do "sh", "ash", "dash", + "busybox sh", "bash", - "ksh" + "bats", + "ksh", + "oksh" ] badShells = [ "awk", "csh", "expect", + "fish", "perl", "python", + "python3", "ruby", "tcsh", "zsh" @@ -3050,7 +3405,7 @@ readScriptFile = do readUtf8Bom = called "Byte Order Mark" $ string "\xFEFF" -readScript = readScriptFile +readScript = readScriptFile False -- Interactively run a specific parser in ghci: -- debugParse readSimpleCommand "echo 'hello world'" @@ -3085,6 +3440,8 @@ testEnvironment = Environment { systemInterface = (mockedSystemInterface []), checkSourced = False, + currentFilename = "myscript", + ignoreRC = False, shellTypeOverride = Nothing } @@ -3093,23 +3450,22 @@ isOk p s = parsesCleanly p s == Just True -- The string parses with no wa isWarning p s = parsesCleanly p s == Just False -- The string parses with warnings isNotOk p s = parsesCleanly p s == Nothing -- The string does not parse -parsesCleanly parser string = runIdentity $ do - (res, sys) <- runParser testEnvironment - (parser >> eof >> getState) "-" string - case (res, sys) of - (Right userState, systemState) -> - return $ Just . null $ parseNotes userState ++ parseProblems systemState - (Left _, _) -> return Nothing +-- If the parser matches the string, return Right [ParseNotes+ParseProblems] +-- If it does not match the string, return Left [ParseProblems] +getParseOutput parser string = runIdentity $ do + (res, systemState) <- runParser testEnvironment + (parser >> eof >> getState) "-" string + return $ case res of + Right userState -> + Right $ parseNotes userState ++ parseProblems systemState + Left _ -> Left $ parseProblems systemState --- For printf debugging: print the value of an expression --- Example: return $ dump $ T_Literal id [c] -dump :: Show a => a -> a -dump x = trace (show x) x - --- Like above, but print a specific expression: --- Example: return $ dumps ("Returning: " ++ [c]) $ T_Literal id [c] -dumps :: Show x => x -> a -> a -dumps t = trace (show t) +-- If the parser matches the string, return Just whether it was clean (without emitting suggestions) +-- Otherwise, Nothing +parsesCleanly parser string = + case getParseOutput parser string of + Right list -> Just $ null list + Left _ -> Nothing parseWithNotes parser = do item <- parser @@ -3127,9 +3483,8 @@ makeErrorFor parsecError = pos = errorPos parsecError getStringFromParsec errors = - case map f errors of - r -> unwords (take 1 $ catMaybes $ reverse r) ++ - " Fix any mentioned problems and try again." + headOrDefault "" (mapMaybe f $ reverse errors) ++ + " Fix any mentioned problems and try again." where f err = case err of @@ -3160,34 +3515,36 @@ parseShell env name contents = do return newParseResult { prComments = map toPositionedComment $ nub $ parseNotes userstate ++ parseProblems state, prTokenPositions = Map.map startEndPosToPos (positionMap userstate), - prRoot = Just $ - reattachHereDocs script (hereDocMap userstate) + prRoot = Just script } - Left err -> + Left err -> do + let context = contextStack state return newParseResult { prComments = map toPositionedComment $ - notesForContext (contextStack state) - ++ [makeErrorFor err] + (filter (not . isIgnored context) $ + notesForContext context + ++ [makeErrorFor err]) ++ parseProblems state, prTokenPositions = Map.empty, prRoot = Nothing } - -notesForContext list = zipWith ($) [first, second] $ filter isName list where - isName (ContextName _ _) = True - isName _ = False - first (ContextName pos str) = ParseNote pos pos ErrorC 1073 $ + -- A final pass for ignoring parse errors after failed parsing + isIgnored stack note = any (contextItemDisablesCode False (codeForParseNote note)) stack + +notesForContext list = zipWith ($) [first, second] [(pos, str) | ContextName pos str <- list] + where + first (pos, str) = ParseNote pos pos ErrorC 1073 $ "Couldn't parse this " ++ str ++ ". Fix to allow more checks." - second (ContextName pos str) = ParseNote pos pos InfoC 1009 $ + second (pos, str) = ParseNote pos pos InfoC 1009 $ "The mentioned syntax error was in this " ++ str ++ "." -- Go over all T_UnparsedIndex and reparse them as either arithmetic or text -- depending on declare -A statements. -reparseIndices root = - analyze blank blank f root +reparseIndices root = process root where + process = analyze blank blank f associative = getAssociativeArrays root isAssociative s = s `elem` associative f (T_Assignment id mode name indices value) = do @@ -3212,8 +3569,9 @@ reparseIndices root = fixAssignmentIndex name word = case word of - T_UnparsedIndex id pos src -> - parsed name pos src + T_UnparsedIndex id pos src -> do + idx <- parsed name pos src + process idx -- Recursively parse for cases like x[y[z=1]]=1 _ -> return word parsed name pos src = @@ -3260,6 +3618,8 @@ parseScript sys spec = env = Environment { systemInterface = sys, checkSourced = psCheckSourced spec, + currentFilename = psFilename spec, + ignoreRC = psIgnoreRC spec, shellTypeOverride = psShellTypeOverride spec } @@ -3296,4 +3656,3 @@ tryWithErrors parser = do return [] runTests = $quickCheckAll - diff --git a/src/ShellCheck/Prelude.hs b/src/ShellCheck/Prelude.hs new file mode 100644 index 0000000..7610c46 --- /dev/null +++ b/src/ShellCheck/Prelude.hs @@ -0,0 +1,51 @@ +{- + Copyright 2022 Vidar Holen + + This file is part of ShellCheck. + https://www.shellcheck.net + + ShellCheck is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + ShellCheck is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} + +-- Generic basic utility functions +module ShellCheck.Prelude where + +import Data.Semigroup + + +-- Get element 0 or a default. Like `head` but safe. +headOrDefault _ (a:_) = a +headOrDefault def _ = def + +-- Get the last element or a default. Like `last` but safe. +lastOrDefault def [] = def +lastOrDefault _ list = last list + +--- Get element n of a list, or Nothing. Like `!!` but safe. +(!!!) list i = + case drop i list of + [] -> Nothing + (r:_) -> Just r + + +-- Like mconcat but for Semigroups +sconcat1 :: (Semigroup t) => [t] -> t +sconcat1 [x] = x +sconcat1 (x:xs) = x <> sconcat1 xs + +sconcatOrDefault def [] = def +sconcatOrDefault _ list = sconcat1 list + +-- For more actionable "impossible" errors +pleaseReport str = "ShellCheck internal error, please report: " ++ str diff --git a/src/ShellCheck/Regex.hs b/src/ShellCheck/Regex.hs index f975775..9367ee7 100644 --- a/src/ShellCheck/Regex.hs +++ b/src/ShellCheck/Regex.hs @@ -1,5 +1,5 @@ {- - Copyright 2012-2015 Vidar Holen + Copyright 2012-2019 Vidar Holen This file is part of ShellCheck. https://www.shellcheck.net @@ -30,7 +30,7 @@ import Text.Regex.TDFA -- Precompile the regex mkRegex :: String -> Regex mkRegex str = - let make :: RegexMaker Regex CompOption ExecOption String => String -> Regex + let make :: String -> Regex make = makeRegex in make str diff --git a/stack.yaml b/stack.yaml index d39cada..4cf5c74 100644 --- a/stack.yaml +++ b/stack.yaml @@ -2,7 +2,7 @@ # For more information, see: https://docs.haskellstack.org/en/stable/yaml_configuration/ # Specifies the GHC version and set of packages available (e.g., lts-3.5, nightly-2015-09-21, ghc-7.10.2) -resolver: lts-8.5 +resolver: lts-18.15 # Local packages, usually specified by relative directory name packages: diff --git a/striptests b/striptests index 8281f78..6e64607 100755 --- a/striptests +++ b/striptests @@ -29,6 +29,7 @@ detestify() { state = 0; } + /STRIP/ { next; } /LANGUAGE TemplateHaskell/ { next; } /^import.*Test\./ { next; } @@ -75,4 +76,3 @@ find . -name '.git' -prune -o -type f -name '*.hs' -print | do modify "$file" detestify done - diff --git a/test/buildtest b/test/buildtest index e3aa1eb..469539b 100755 --- a/test/buildtest +++ b/test/buildtest @@ -11,21 +11,37 @@ command -v cabal || cabal update || die "can't update" -cabal install --dependencies-only --enable-tests || - die "can't install dependencies" -cabal configure --enable-tests || + +if [ -e /etc/arch-release ] +then + # Arch has an unconventional packaging setup + flags=(--disable-library-vanilla --enable-shared --enable-executable-dynamic --ghc-options=-dynamic) +else + flags=() +fi + +cabal install --dependencies-only --enable-tests "${flags[@]}" || + cabal install --dependencies-only "${flags[@]}" || + cabal install --dependencies-only --max-backjumps -1 "${flags[@]}" || + die "can't install dependencies" +cabal configure --enable-tests "${flags[@]}" || die "configure failed" cabal build || die "build failed" cabal test || die "test failed" +cabal haddock || + die "haddock failed" -dist/build/shellcheck/shellcheck - << 'EOF' || die "execution failed" +sc="$(find . -name shellcheck -type f -perm -111)" +[ -x "$sc" ] || die "Can't find executable" + +"$sc" - << 'EOF' || die "execution failed" #!/bin/sh echo "Hello World" EOF -dist/build/shellcheck/shellcheck - << 'EOF' && die "negative execution failed" +"$sc" - << 'EOF' && die "negative execution failed" #!/bin/sh echo $1 EOF diff --git a/test/check_release b/test/check_release new file mode 100755 index 0000000..f3ea9df --- /dev/null +++ b/test/check_release @@ -0,0 +1,84 @@ +#!/usr/bin/env bash +# shellcheck disable=SC2257 + +failed=0 +fail() { + echo "$(tput setaf 1)$*$(tput sgr0)" + failed=1 +} + +if git diff | grep -q "" +then + fail "There are uncommitted changes" +fi + +version=${current#v} +if ! grep "Version:" ShellCheck.cabal | grep -qFw "$version" +then + fail "The cabal file does not match tag version $version" +fi + +if ! grep -qF "## $current" CHANGELOG.md +then + fail "CHANGELOG.md does not contain '## $current'" +fi + +current=$(git tag --points-at) +if [[ -z "$current" ]] +then + fail "No git tag on the current commit" + echo "Create one with: git tag -a v0.0.0" +fi + +if [[ "$current" != v* ]] +then + fail "Bad tag format: expected v0.0.0" +fi + +if [[ "$(git cat-file -t "$current")" != "tag" ]] +then + fail "Current tag is not annotated (required for Snap)." +fi + +if [[ "$(git tag --points-at master)" != "$current" ]] +then + fail "You are not on master" +fi + +if [[ $(git log -1 --pretty=%B) != "Stable version "* ]] +then + fail "Expected git log message to be 'Stable version ...'" +fi + +if [[ $(git log -1 --pretty=%B) != *"CHANGELOG"* ]] +then + fail "Expected git log message to contain CHANGELOG" +fi + +i=1 j=1 +cat << EOF + +Manual Checklist + +$((i++)). Make sure none of the automated checks above failed +$((i++)). Run \`build/build_builder build/*/\` to update all builder images. +$((j++)). \`build/run_builder dist-newstyle/sdist/ShellCheck-*.tar.gz build/*/\` to verify that they work. +$((j++)). \`for f in \$(cat build/*/tag); do docker push "\$f"; done\` to upload them. +$((i++)). Run test/distrotest to ensure that most distros can build OOTB. +$((i++)). Make sure GitHub Build currently passes: https://github.com/koalaman/shellcheck/actions +$((i++)). Make sure SnapCraft build currently works: https://build.snapcraft.io/user/koalaman +$((i++)). Format and read over the manual for bad formatting and outdated info. +$((i++)). Make sure the Hackage package builds locally. + +Release Steps + +$((j++)). \`cabal sdist\` to generate a Hackage package +$((j++)). \`git push --follow-tags\` to push commit +$((j++)). Wait for GitHub Actions to build. +$((j++)). Verify release: + a. Check that the new versions are uploaded: https://github.com/koalaman/shellcheck/tags + b. Check that the docker images have version tags: https://hub.docker.com/u/koalaman +$((j++)). If no disaster, upload to Hackage: http://hackage.haskell.org/upload +$((j++)). Push a new commit that updates CHANGELOG.md +EOF +exit "$failed" diff --git a/test/distrotest b/test/distrotest index 0465c3a..128ee44 100755 --- a/test/distrotest +++ b/test/distrotest @@ -16,10 +16,21 @@ and is still highly experimental. Make sure you're plugged in and have screen/tmux in place, then re-run with $0 --run to continue. + +Also note that dist*/ and .stack-work/ will be deleted. EOF exit 0 } +echo "Deleting 'dist', 'dist-newstyle', and '.stack-work'..." +rm -rf dist dist-newstyle .stack-work + +execs=$(find . -name shellcheck) + +if [ -n "$execs" ] +then + die "Found unexpected executables. Remove and try again: $execs" +fi log=$(mktemp) || die "Can't create temp file" date >> "$log" || die "Can't write to log" @@ -57,19 +68,18 @@ done << EOF debian:stable apt-get update && apt-get install -y cabal-install debian:testing apt-get update && apt-get install -y cabal-install ubuntu:latest apt-get update && apt-get install -y cabal-install -opensuse:latest zypper install -y cabal-install ghc - -# Older Ubuntu versions we want to support -ubuntu:18.04 apt-get update && apt-get install -y cabal-install -ubuntu:17.10 apt-get update && apt-get install -y cabal-install - -# Misc haskell:latest true +opensuse/leap:latest zypper install -y cabal-install ghc +fedora:latest dnf install -y cabal-install ghc-template-haskell-devel findutils libstdc++-static gcc-c++ +archlinux:latest pacman -S -y --noconfirm cabal-install ghc-static base-devel -# Known to currently fail -centos:latest yum install -y epel-release && yum install -y cabal-install -fedora:latest dnf install -y cabal-install -base/archlinux:latest pacman -S -y --noconfirm cabal-install ghc-static base-devel +# Ubuntu LTS +ubuntu:24.04 apt-get update && apt-get install -y cabal-install +ubuntu:22.04 apt-get update && apt-get install -y cabal-install +ubuntu:20.04 apt-get update && apt-get install -y cabal-install + +# Stack on Ubuntu LTS +ubuntu:24.04 set -e; apt-get update && apt-get install -y curl && curl -sSL https://get.haskellstack.org/ | sh -s - -f && cd /mnt && exec test/stacktest EOF exit "$final" diff --git a/test/shellcheck.hs b/test/shellcheck.hs index 6106d9a..d5e056d 100644 --- a/test/shellcheck.hs +++ b/test/shellcheck.hs @@ -2,23 +2,40 @@ module Main where import Control.Monad import System.Exit -import qualified ShellCheck.Checker import qualified ShellCheck.Analytics import qualified ShellCheck.AnalyzerLib -import qualified ShellCheck.Parser +import qualified ShellCheck.ASTLib +import qualified ShellCheck.CFG +import qualified ShellCheck.CFGAnalysis +import qualified ShellCheck.Checker import qualified ShellCheck.Checks.Commands +import qualified ShellCheck.Checks.ControlFlow +import qualified ShellCheck.Checks.Custom import qualified ShellCheck.Checks.ShellSupport +import qualified ShellCheck.Fixer +import qualified ShellCheck.Formatter.Diff +import qualified ShellCheck.Parser main = do putStrLn "Running ShellCheck tests..." - results <- sequence [ - ShellCheck.Checker.runTests, - ShellCheck.Checks.Commands.runTests, - ShellCheck.Checks.ShellSupport.runTests, - ShellCheck.Analytics.runTests, - ShellCheck.AnalyzerLib.runTests, - ShellCheck.Parser.runTests + failures <- filter (not . snd) <$> mapM sequenceA tests + if null failures then exitSuccess else do + putStrLn "Tests failed for the following module(s):" + mapM (putStrLn . ("- ShellCheck." ++) . fst) failures + exitFailure + where + tests = + [ ("Analytics" , ShellCheck.Analytics.runTests) + , ("AnalyzerLib" , ShellCheck.AnalyzerLib.runTests) + , ("ASTLib" , ShellCheck.ASTLib.runTests) + , ("CFG" , ShellCheck.CFG.runTests) + , ("CFGAnalysis" , ShellCheck.CFGAnalysis.runTests) + , ("Checker" , ShellCheck.Checker.runTests) + , ("Checks.Commands" , ShellCheck.Checks.Commands.runTests) + , ("Checks.ControlFlow" , ShellCheck.Checks.ControlFlow.runTests) + , ("Checks.Custom" , ShellCheck.Checks.Custom.runTests) + , ("Checks.ShellSupport", ShellCheck.Checks.ShellSupport.runTests) + , ("Fixer" , ShellCheck.Fixer.runTests) + , ("Formatter.Diff" , ShellCheck.Formatter.Diff.runTests) + , ("Parser" , ShellCheck.Parser.runTests) ] - if and results - then exitSuccess - else exitFailure diff --git a/test/stacktest b/test/stacktest new file mode 100755 index 0000000..b486c31 --- /dev/null +++ b/test/stacktest @@ -0,0 +1,28 @@ +#!/bin/bash +# This script builds ShellCheck through `stack` using +# various resolvers. It's run via distrotest. + +resolvers=( +# nightly-"$(date -d "3 days ago" +"%Y-%m-%d")" +) + +die() { echo "$*" >&2; exit 1; } + +[ -e "ShellCheck.cabal" ] || + die "ShellCheck.cabal not in current dir" +[ -e "stack.yaml" ] || + die "stack.yaml not in current dir" +command -v stack || + die "stack is missing" + +stack setup --allow-different-user || die "Failed to setup with default resolver" +stack build --test || die "Failed to build/test with default resolver" + +# Nice to haves, but not necessary +for resolver in "${resolvers[@]}" +do + stack --resolver="$resolver" setup || die "Failed to setup $resolver. This probably doesn't matter." + stack --resolver="$resolver" build --test || die "Failed build/test with $resolver! This probably doesn't matter." +done + +echo "Success"