Add Default impl for GlobSet to satisfy clippy

Add CodSpeed performance benchmarks

.github/workflows/GnuComment.yml

@@ -0,0 +1,83 @@
+name: GnuComment
+
+# Post the GNU grep testsuite comparison (produced by the GnuTests workflow)
+# as a comment on the pull request.
+
+on:
+  workflow_run:
+    workflows: ["GnuTests"]
+    types:
+      - completed
+
+permissions: {}
+jobs:
+  post-comment:
+    permissions:
+      actions: read  #  to list workflow run artifacts
+      pull-requests: write  #  to comment on the pr
+
+    runs-on: ubuntu-latest
+    if: >
+      github.event.workflow_run.event == 'pull_request'
+    steps:
+      - name: 'Download artifact'
+        uses: actions/github-script@v7
+        with:
+          script: |
+            // List all artifacts from GnuTests
+            var artifacts = await github.rest.actions.listWorkflowRunArtifacts({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              run_id: ${{ github.event.workflow_run.id }},
+            });
+
+            // Download the "comment" artifact, which contains a PR number (NR) and result.txt
+            var matchArtifact = artifacts.data.artifacts.filter((artifact) => {
+              return artifact.name == "comment"
+            })[0];
+
+            if (!matchArtifact) {
+              console.log('No comment artifact found');
+              return;
+            }
+
+            var download = await github.rest.actions.downloadArtifact({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              artifact_id: matchArtifact.id,
+              archive_format: 'zip',
+            });
+            var fs = require('fs');
+            fs.writeFileSync('${{ github.workspace }}/comment.zip', Buffer.from(download.data));
+      - run: unzip comment.zip || echo "Failed to unzip comment artifact"
+
+      - name: 'Comment on PR'
+        uses: actions/github-script@v7
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          script: |
+            var fs = require('fs');
+
+            // Check if files exist
+            if (!fs.existsSync('./NR')) {
+              console.log('No NR file found, skipping comment');
+              return;
+            }
+            if (!fs.existsSync('./result.txt')) {
+              console.log('No result.txt file found, skipping comment');
+              return;
+            }
+
+            var issue_number = Number(fs.readFileSync('./NR'));
+            var content = fs.readFileSync('./result.txt');
+
+            if (content.toString().trim().length > 7) { // 7 because we have backquote + \n
+              await github.rest.issues.createComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                issue_number: issue_number,
+                body: 'GNU grep testsuite comparison:\n```\n' + content + '```'
+              });
+            } else {
+              console.log('Comment content too short, skipping');
+            }

.github/workflows/GnuTests.yml

@@ -0,0 +1,180 @@
+name: GnuTests
+
+# Run the upstream GNU grep testsuite against the Rust grep implementation to
+# track and guard byte-for-byte compatibility. See util/run-gnu-testsuite.sh.
+
+on:
+  pull_request:
+  push:
+    branches:
+      - '*'
+
+# End the current execution if there is a new changeset in the PR.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+env:
+  DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
+  TEST_FULL_SUMMARY_FILE: 'grep-gnu-full-result.json'
+
+jobs:
+  native:
+    name: Run GNU grep testsuite
+    runs-on: ubuntu-24.04
+    steps:
+    - name: Checkout code (grep)
+      uses: actions/checkout@v4
+      with:
+        path: 'grep'
+        persist-credentials: false
+    - uses: dtolnay/rust-toolchain@master
+      with:
+        toolchain: stable
+    - uses: Swatinem/rust-cache@v2
+      with:
+        workspaces: "./grep -> target"
+
+    - name: Fetch GNU grep testsuite
+      shell: bash
+      run: |
+        ## Download and extract the upstream GNU grep release tarball
+        mkdir -p gnu.grep
+        cd gnu.grep
+        bash ../grep/util/fetch-gnu.sh
+
+    - name: Build Rust grep binary
+      shell: bash
+      run: |
+        cd 'grep'
+        cargo build --release
+
+    - name: Run GNU grep testsuite
+      shell: bash
+      run: |
+        cd 'grep'
+        export GNU_GREP_DIR="../gnu.grep"
+        ./util/run-gnu-testsuite.sh --json-output "${{ env.TEST_FULL_SUMMARY_FILE }}" || true
+
+    - name: Upload full json results
+      uses: actions/upload-artifact@v4
+      with:
+        name: grep-gnu-full-result
+        path: grep/${{ env.TEST_FULL_SUMMARY_FILE }}
+        if-no-files-found: warn
+
+  aggregate:
+    needs: [native]
+    permissions:
+      actions: read
+      contents: read
+      pull-requests: read
+    name: Aggregate GNU test results
+    runs-on: ubuntu-24.04
+    steps:
+    - name: Initialize workflow variables
+      id: vars
+      shell: bash
+      run: |
+        ## VARs setup
+        outputs() { step_id="${{ github.action }}"; for var in "$@" ; do echo steps.${step_id}.outputs.${var}="${!var}"; echo "${var}=${!var}" >> $GITHUB_OUTPUT; done; }
+        TEST_SUMMARY_FILE='grep-gnu-result.json'
+        outputs TEST_SUMMARY_FILE
+
+    - name: Checkout code (grep)
+      uses: actions/checkout@v4
+      with:
+        path: 'grep'
+        persist-credentials: false
+
+    - name: Retrieve reference artifacts
+      uses: dawidd6/action-download-artifact@v6
+      continue-on-error: true
+      with:
+        workflow: GnuTests.yml
+        branch: "${{ env.DEFAULT_BRANCH }}"
+        workflow_conclusion: completed
+        path: "reference"
+        if_no_artifact_found: warn
+
+    - name: Download full json results
+      uses: actions/download-artifact@v4
+      with:
+        name: grep-gnu-full-result
+        path: results
+
+    - name: Extract/summarize testing info
+      id: summary
+      shell: bash
+      run: |
+        ## Extract/summarize testing info
+        outputs() { step_id="${{ github.action }}"; for var in "$@" ; do echo steps.${step_id}.outputs.${var}="${!var}"; echo "${var}=${!var}" >> $GITHUB_OUTPUT; done; }
+
+        RESULT_FILE="results/${{ env.TEST_FULL_SUMMARY_FILE }}"
+        if [[ ! -f "$RESULT_FILE" ]]; then
+          echo "::error ::Result file $RESULT_FILE not found"
+          find results -type f || true
+          exit 1
+        fi
+
+        TOTAL=$(jq -r '.summary.total // 0' "$RESULT_FILE")
+        PASS=$(jq -r '.summary.passed // 0' "$RESULT_FILE")
+        FAIL=$(jq -r '.summary.failed // 0' "$RESULT_FILE")
+        SKIP=$(jq -r '.summary.skipped // 0' "$RESULT_FILE")
+
+        output="GNU grep tests summary = TOTAL: $TOTAL / PASS: $PASS / FAIL: $FAIL / SKIP: $SKIP"
+        echo "${output}"
+        if [[ "$FAIL" -gt 0 ]]; then
+          echo "::warning ::${output}"
+        fi
+
+        outputs TOTAL PASS FAIL SKIP
+
+    - name: Compare test failures VS reference
+      shell: bash
+      run: |
+        ## Compare current results against the reference summary from the default branch
+        REF_SUMMARY_FILE='reference/grep-gnu-full-result/${{ env.TEST_FULL_SUMMARY_FILE }}'
+        CURRENT_SUMMARY_FILE="results/${{ env.TEST_FULL_SUMMARY_FILE }}"
+        IGNORE_INTERMITTENT="grep/.github/workflows/ignore-intermittent.txt"
+
+        # Set up comment directory for the GnuComment workflow.
+        COMMENT_DIR="reference/comment"
+        mkdir -p ${COMMENT_DIR}
+        echo ${{ github.event.number }} > ${COMMENT_DIR}/NR
+        COMMENT_LOG="${COMMENT_DIR}/result.txt"
+        : > "${COMMENT_LOG}"
+
+        COMPARISON_RESULT=0
+        if test -f "${REF_SUMMARY_FILE}"; then
+          python3 grep/util/compare_test_results.py \
+            --ignore-file "${IGNORE_INTERMITTENT}" \
+            --output "${COMMENT_LOG}" \
+            "${CURRENT_SUMMARY_FILE}" "${REF_SUMMARY_FILE}" || COMPARISON_RESULT=$?
+        else
+          echo "::warning ::Skipping test comparison; no prior reference summary at '${REF_SUMMARY_FILE}'."
+        fi
+
+        if [ ${COMPARISON_RESULT} -eq 1 ]; then
+          echo "::error ::Found new non-intermittent test failures"
+          UPLOAD_EXIT=1
+        else
+          echo "::notice ::No new test failures detected"
+          UPLOAD_EXIT=0
+        fi
+        echo "UPLOAD_EXIT=${UPLOAD_EXIT}" >> $GITHUB_ENV
+
+    - name: Upload comparison log (for GnuComment workflow)
+      if: success() || failure()
+      uses: actions/upload-artifact@v4
+      with:
+        name: comment
+        path: reference/comment/
+
+    - name: Report test results
+      if: success() || failure()
+      shell: bash
+      run: |
+        echo "::notice ::GNU grep testsuite: TOTAL ${{ steps.summary.outputs.TOTAL }} / PASS ${{ steps.summary.outputs.PASS }} / FAIL ${{ steps.summary.outputs.FAIL }} / SKIP ${{ steps.summary.outputs.SKIP }}"
+        # Fail the job if the comparison found new non-intermittent regressions.
+        exit "${UPLOAD_EXIT:-0}"

.github/workflows/codspeed.yml

@@ -0,0 +1,37 @@
+name: CodSpeed
+
+on:
+  push:
+    branches:
+      - "main"
+  pull_request:
+  # `workflow_dispatch` allows CodSpeed to trigger backtest
+  # performance analysis in order to generate initial data.
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  id-token: write
+
+jobs:
+  codspeed:
+    name: Run benchmarks
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Rust toolchain, cache and cargo-codspeed binary
+        uses: moonrepo/setup-rust@v0
+        with:
+          channel: stable
+          cache-target: release
+          bins: cargo-codspeed
+
+      - name: Build the benchmark target(s)
+        run: cargo codspeed build
+
+      - name: Run the benchmarks
+        uses: CodSpeedHQ/action@v4
+        with:
+          mode: simulation
+          run: cargo codspeed run

.github/workflows/ignore-intermittent.txt

@@ -0,0 +1,7 @@
+# List of intermittent test names to ignore in result comparisons
+# Format: one test name per line, lines starting with # are comments
+#
+# Add test names that are known to be flaky or environment-dependent
+# Example:
+# basic_substitution
+# line_address_test

.pre-commit-config.yaml

@@ -0,0 +1,55 @@
+# See https://pre-commit.com for more information
+# See https://pre-commit.com/hooks.html for more hooks
+exclude: ^tests/fixtures/
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+      - id: check-added-large-files
+      - id: check-executables-have-shebangs
+      - id: check-json
+        exclude: '\.vscode/(cSpell|extensions)\.json' # cSpell.json and extensions.json use comments
+      - id: check-shebang-scripts-are-executable
+        exclude: '.+\.rs' # would be triggered by #![some_attribute]
+      - id: check-symlinks
+      - id: check-toml
+      - id: check-yaml
+        args: [ --allow-multiple-documents ]
+      - id: destroyed-symlinks
+      - id: end-of-file-fixer
+      - id: mixed-line-ending
+        args: [ --fix=lf ]
+      - id: trailing-whitespace
+
+  - repo: local
+    hooks:
+      - id: rust-linting
+        name: Rust linting
+        description: Run cargo fmt on files included in the commit.
+        entry: cargo +stable fmt --
+        pass_filenames: true
+        types: [file, rust]
+        language: system
+      - id: rust-clippy
+        name: Rust clippy
+        description: Run cargo clippy on files included in the commit.
+        entry: cargo +stable clippy --workspace --all-targets --all-features -- -D warnings
+        pass_filenames: false
+        types: [file, rust]
+        language: system
+      - id: cargo-lock-check
+        name: Cargo.lock sync check
+        description: Ensure Cargo.lock and fuzz/Cargo.lock are up-to-date.
+        entry: bash -c 'for dir in . fuzz; do if [ -d "$dir" ]; then ( cd "$dir" && cargo fetch --quiet ); fi; done'
+        pass_filenames: false
+        files: 'Cargo\.(toml|lock)$'
+        language: system
+      - id: cspell
+        name: Code spell checker (cspell)
+        description: Run cspell to check for spelling errors (if available).
+        entry: bash -c 'if command -v cspell >/dev/null 2>&1; then cspell --no-must-find-files -- "$@"; else echo "cspell not found, skipping spell check"; exit 0; fi' --
+        pass_filenames: true
+        language: system
+
+ci:
+  skip: [rust-linting, rust-clippy, cargo-lock-check, cspell]

Cargo.toml

@@ -27,5 +27,10 @@ onig_sys = { version = "*", default-features = false }
 uucore = "0.8.0"
 walkdir = "2.5"
 
+[[bench]]
+name = "grep_bench"
+harness = false
+
 [dev-dependencies]
+criterion = { version = "4.7.0", package = "codspeed-criterion-compat" }
 uutests = "0.8.0"

README.md

@@ -4,6 +4,7 @@
 [![dependency status](https://deps.rs/repo/github/uutils/grep/status.svg)](https://deps.rs/repo/github/uutils/grep)
 
 [![CodeCov](https://codecov.io/gh/uutils/grep/branch/main/graph/badge.svg)](https://codecov.io/gh/uutils/grep)
+[![CodSpeed](https://img.shields.io/endpoint?url=https://codspeed.io/badge.json)](https://codspeed.io/uutils/grep?utm_source=badge)
 
 # Grep, now in Rust
 
@@ -29,10 +30,15 @@ cargo build --release
 cargo test
 ```
 
+## Pre-commit hooks
+
+This project uses [pre-commit](https://pre-commit.com); run `pre-commit install` to enable the git hooks.
+
 ## Known Issues
 
 * Does not take `LANG`, etc., into account for handling file encodings (non-UTF8 matches are treated as binary)
 * No localization support yet
+* Performances need to be improved
 
 ## Contributing
 

benches/grep_bench.rs

@@ -0,0 +1,288 @@
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
+use criterion::{Criterion, black_box, criterion_group, criterion_main};
+use uu_grep::matcher::Matcher;
+use uu_grep::{BinaryMode, ColorConfig, Config, DeviceMode, DirectoryMode, GlobSet, RegexMode};
+
+fn make_config<'a>(
+    patterns: &'a [&'a str],
+    regex_mode: RegexMode,
+    ignore_case: bool,
+    invert_match: bool,
+    word_regexp: bool,
+) -> Config<'a> {
+    Config {
+        directory_mode: DirectoryMode::Read,
+        device_mode: DeviceMode::Default,
+        follow_symlinks: false,
+        include_globs: GlobSet::new(),
+        exclude_globs: GlobSet::new(),
+        exclude_dir_globs: GlobSet::new(),
+        label: "(standard input)",
+        #[cfg(windows)]
+        strip_cr: false,
+        binary_mode: BinaryMode::Binary,
+        max_count: None,
+        before_context: 0,
+        after_context: 0,
+        has_context: false,
+        patterns,
+        regex_mode,
+        ignore_case,
+        invert_match,
+        word_regexp,
+        line_regexp: false,
+        quiet: true,
+        count: false,
+        show_filename: false,
+        files_with_matches: false,
+        files_without_match: false,
+        only_matching: false,
+        byte_offset: false,
+        line_number: false,
+        initial_tab: false,
+        null_separator: false,
+        null_data: false,
+        line_buffered: false,
+        no_messages: true,
+        group_separator: None,
+        use_color: false,
+        color_config: ColorConfig {
+            matched_selected: "",
+            matched_context: "",
+            filename: "",
+            line_number: "",
+            byte_offset: "",
+            separator: "",
+            selected_line: "",
+            context_line: "",
+            reverse_video: false,
+            no_erase: false,
+        },
+    }
+}
+
+fn bench_compile(c: &mut Criterion) {
+    let mut group = c.benchmark_group("compile");
+
+    group.bench_function("fixed_string", |b| {
+        b.iter(|| {
+            let patterns: &[&str] = &["hello world"];
+            let config = make_config(patterns, RegexMode::Fixed, false, false, false);
+            let matcher = Matcher::compile(black_box(&config)).unwrap();
+            let _ = black_box(&matcher);
+        })
+    });
+
+    group.bench_function("basic_regex", |b| {
+        b.iter(|| {
+            let patterns: &[&str] = &[r"[0-9]\{4\}-[0-9]\{2\}-[0-9]\{2\}"];
+            let config = make_config(patterns, RegexMode::Basic, false, false, false);
+            let matcher = Matcher::compile(black_box(&config)).unwrap();
+            let _ = black_box(&matcher);
+        })
+    });
+
+    group.bench_function("extended_regex", |b| {
+        b.iter(|| {
+            let patterns: &[&str] = &[r"[0-9]{4}-[0-9]{2}-[0-9]{2}"];
+            let config = make_config(patterns, RegexMode::Extended, false, false, false);
+            let matcher = Matcher::compile(black_box(&config)).unwrap();
+            let _ = black_box(&matcher);
+        })
+    });
+
+    group.bench_function("perl_regex", |b| {
+        b.iter(|| {
+            let patterns: &[&str] = &[r"\d{4}-\d{2}-\d{2}\s+\d{2}:\d{2}:\d{2}"];
+            let config = make_config(patterns, RegexMode::Perl, false, false, false);
+            let matcher = Matcher::compile(black_box(&config)).unwrap();
+            let _ = black_box(&matcher);
+        })
+    });
+
+    group.bench_function("multiple_patterns", |b| {
+        b.iter(|| {
+            let patterns: &[&str] = &["error", "warning", "critical", "fatal", "panic"];
+            let config = make_config(patterns, RegexMode::Fixed, false, false, false);
+            let matcher = Matcher::compile(black_box(&config)).unwrap();
+            let _ = black_box(&matcher);
+        })
+    });
+
+    group.finish();
+}
+
+fn bench_match(c: &mut Criterion) {
+    let mut group = c.benchmark_group("match");
+
+    // Fixed string match - hit
+    {
+        let patterns: &[&str] = &["ERROR"];
+        let config = make_config(patterns, RegexMode::Fixed, false, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+        let line = b"2024-01-15 10:30:45 ERROR: Connection timeout on server-42";
+
+        group.bench_function("fixed_string_hit", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(line))))
+        });
+    }
+
+    // Fixed string match - miss
+    {
+        let patterns: &[&str] = &["CRITICAL"];
+        let config = make_config(patterns, RegexMode::Fixed, false, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+        let line = b"2024-01-15 10:30:45 INFO: Server started successfully";
+
+        group.bench_function("fixed_string_miss", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(line))))
+        });
+    }
+
+    // Extended regex match
+    {
+        let patterns: &[&str] = &[r"[0-9]{4}-[0-9]{2}-[0-9]{2}"];
+        let config = make_config(patterns, RegexMode::Extended, false, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+        let line = b"2024-01-15 10:30:45 ERROR: Connection timeout";
+
+        group.bench_function("extended_regex_hit", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(line))))
+        });
+    }
+
+    // Case-insensitive match
+    {
+        let patterns: &[&str] = &["error"];
+        let config = make_config(patterns, RegexMode::Fixed, true, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+        let line = b"2024-01-15 10:30:45 ERROR: Connection timeout";
+
+        group.bench_function("case_insensitive_hit", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(line))))
+        });
+    }
+
+    // Inverted match
+    {
+        let patterns: &[&str] = &["ERROR"];
+        let config = make_config(patterns, RegexMode::Fixed, false, true, false);
+        let matcher = Matcher::compile(&config).unwrap();
+        let line = b"2024-01-15 10:30:45 INFO: Server started successfully";
+
+        group.bench_function("inverted_match", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(line))))
+        });
+    }
+
+    // Word boundary match
+    {
+        let patterns: &[&str] = &["error"];
+        let config = make_config(patterns, RegexMode::Fixed, true, false, true);
+        let matcher = Matcher::compile(&config).unwrap();
+        let line = b"2024-01-15 10:30:45 error: Connection timeout";
+
+        group.bench_function("word_boundary_hit", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(line))))
+        });
+    }
+
+    // Multiple patterns
+    {
+        let patterns: &[&str] = &["error", "warning", "critical", "fatal", "panic"];
+        let config = make_config(patterns, RegexMode::Fixed, true, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+        let line = b"2024-01-15 10:30:45 WARNING: High memory usage detected on node-7";
+
+        group.bench_function("multi_pattern_hit", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(line))))
+        });
+    }
+
+    // Long line
+    {
+        let patterns: &[&str] = &["needle"];
+        let config = make_config(patterns, RegexMode::Fixed, false, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+        let mut long_line = "a".repeat(5000);
+        long_line.push_str("needle");
+        long_line.push_str(&"b".repeat(5000));
+        let long_line_bytes = long_line.into_bytes();
+
+        group.bench_function("long_line_hit", |b| {
+            b.iter(|| black_box(matcher.match_line(black_box(&long_line_bytes))))
+        });
+    }
+
+    group.finish();
+}
+
+fn bench_throughput(c: &mut Criterion) {
+    let mut group = c.benchmark_group("throughput");
+
+    // Simulate processing many lines (like searching a log file)
+    let lines: Vec<Vec<u8>> = (0..1000)
+        .map(|i| {
+            if i % 50 == 0 {
+                format!(
+                    "2024-01-15 10:30:{:02} ERROR: Connection timeout on server-{}",
+                    i % 60,
+                    i
+                )
+                .into_bytes()
+            } else {
+                format!(
+                    "2024-01-15 10:30:{:02} INFO: Request processed in {}ms",
+                    i % 60,
+                    i * 3
+                )
+                .into_bytes()
+            }
+        })
+        .collect();
+
+    {
+        let patterns: &[&str] = &["ERROR"];
+        let config = make_config(patterns, RegexMode::Fixed, false, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+
+        group.bench_function("scan_1000_lines_fixed", |b| {
+            b.iter(|| {
+                let mut matches = 0u64;
+                for line in &lines {
+                    if matcher.match_line(black_box(line)).is_some() {
+                        matches += 1;
+                    }
+                }
+                black_box(matches)
+            })
+        });
+    }
+
+    {
+        let patterns: &[&str] = &[r"[0-9]+ *ms"];
+        let config = make_config(patterns, RegexMode::Extended, false, false, false);
+        let matcher = Matcher::compile(&config).unwrap();
+
+        group.bench_function("scan_1000_lines_regex", |b| {
+            b.iter(|| {
+                let mut matches = 0u64;
+                for line in &lines {
+                    if matcher.match_line(black_box(line)).is_some() {
+                        matches += 1;
+                    }
+                }
+                black_box(matches)
+            })
+        });
+    }
+
+    group.finish();
+}
+
+criterion_group!(benches, bench_compile, bench_match, bench_throughput);
+criterion_main!(benches);

src/context_buffer.rs

@@ -1,3 +1,8 @@
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
 pub struct LineView<'a> {
     /// Line content (without the terminator).
     pub line: &'a [u8],

src/lib.rs

@@ -1,6 +1,14 @@
-mod context_buffer;
-mod line_buffer;
-mod matcher;
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
+#[doc(hidden)]
+pub mod context_buffer;
+#[doc(hidden)]
+pub mod line_buffer;
+#[doc(hidden)]
+pub mod matcher;
 mod output;
 mod searcher;
 
@@ -15,7 +23,8 @@ use std::path::Path;
 use uucore::error::{FromIo, UResult, USimpleError};
 
 #[derive(Clone, Copy, PartialEq, Eq)]
-enum RegexMode {
+#[doc(hidden)]
+pub enum RegexMode {
     Fixed,
     Basic,
     Extended,
@@ -23,7 +32,8 @@ enum RegexMode {
 }
 
 #[derive(Clone, Copy, PartialEq, Eq)]
-enum BinaryMode {
+#[doc(hidden)]
+pub enum BinaryMode {
     Binary,
     Text,
     WithoutMatch,
@@ -37,79 +47,84 @@ enum ColorMode {
 }
 
 #[derive(Clone, Copy, PartialEq, Eq)]
-enum DirectoryMode {
+#[doc(hidden)]
+pub enum DirectoryMode {
     Read,
     Skip,
     Recurse,
 }
 
 #[derive(Clone, Copy, PartialEq, Eq)]
-enum DeviceMode {
+#[doc(hidden)]
+pub enum DeviceMode {
     Default,
     Read,
     Skip,
 }
 
-struct ColorConfig<'a> {
-    matched_selected: &'a str,
-    matched_context: &'a str,
-    filename: &'a str,
-    line_number: &'a str,
-    byte_offset: &'a str,
-    separator: &'a str,
-    selected_line: &'a str,
-    context_line: &'a str,
+#[doc(hidden)]
+pub struct ColorConfig<'a> {
+    pub matched_selected: &'a str,
+    pub matched_context: &'a str,
+    pub filename: &'a str,
+    pub line_number: &'a str,
+    pub byte_offset: &'a str,
+    pub separator: &'a str,
+    pub selected_line: &'a str,
+    pub context_line: &'a str,
 
-    reverse_video: bool,
-    no_erase: bool,
+    pub reverse_video: bool,
+    pub no_erase: bool,
 }
 
-struct GlobSet {
+#[doc(hidden)]
+pub struct GlobSet {
     patterns: Vec<glob::Pattern>,
 }
 
-struct Config<'a> {
+#[doc(hidden)]
+pub struct Config<'a> {
     // Searcher
-    directory_mode: DirectoryMode,
-    device_mode: DeviceMode,
-    follow_symlinks: bool,
-    include_globs: GlobSet,
-    exclude_globs: GlobSet,
-    exclude_dir_globs: GlobSet,
-    label: &'a str,
+    pub directory_mode: DirectoryMode,
+    pub device_mode: DeviceMode,
+    pub follow_symlinks: bool,
+    pub include_globs: GlobSet,
+    pub exclude_globs: GlobSet,
+    pub exclude_dir_globs: GlobSet,
+    pub label: &'a str,
     #[cfg(windows)]
-    strip_cr: bool,
-    binary_mode: BinaryMode,
-    max_count: Option<u64>,
-    before_context: usize,
-    after_context: usize,
-    has_context: bool,
+    pub strip_cr: bool,
+    pub binary_mode: BinaryMode,
+    pub max_count: Option<u64>,
+    pub before_context: usize,
+    pub after_context: usize,
+    pub has_context: bool,
 
     // Matcher
-    patterns: &'a [&'a str],
-    regex_mode: RegexMode,
-    ignore_case: bool,
-    invert_match: bool,
-    word_regexp: bool,
-    line_regexp: bool,
+    pub patterns: &'a [&'a str],
+    pub regex_mode: RegexMode,
+    pub ignore_case: bool,
+    pub invert_match: bool,
+    pub word_regexp: bool,
+    pub line_regexp: bool,
 
     // Output
-    quiet: bool,
-    count: bool,
-    show_filename: bool,
-    files_with_matches: bool,
-    files_without_match: bool,
-    only_matching: bool,
-    byte_offset: bool,
-    line_number: bool,
-    initial_tab: bool,
-    null_separator: bool,
-    null_data: bool,
-    line_buffered: bool,
-    no_messages: bool,
-    group_separator: Option<&'a str>,
-    use_color: bool,
-    color_config: ColorConfig<'a>,
+    pub quiet: bool,
+    pub count: bool,
+    pub show_filename: bool,
+    pub files_with_matches: bool,
+    pub files_without_match: bool,
+    pub only_matching: bool,
+    pub byte_offset: bool,
+    pub line_number: bool,
+    pub initial_tab: bool,
+    pub null_separator: bool,
+    pub null_data: bool,
+    pub line_buffered: bool,
+    pub no_messages: bool,
+    pub group_separator: Option<&'a str>,
+    pub use_color: bool,
+    pub color_config: ColorConfig<'a>,
 }
 
 #[uucore::main(no_signals)]
@@ -416,6 +431,10 @@ pub fn uu_app() -> Command {
         .about("Search for PATTERNS in each FILE.")
         .disable_help_flag(true)
         .disable_version_flag(true)
+        // GNU grep accepts repeated options (booleans are idempotent, value
+        // options take the last); make clap replace rather than error. Args
+        // with ArgAction::Append (e.g. -e/-f/--include) still accumulate.
+        .args_override_self(true)
         .after_help(
             "When FILE is '-', read standard input.  If no FILE is given, read standard \
              input, but with -r, recursively search the working directory instead.  With \
@@ -845,8 +864,21 @@ fn expand_num_shorthand(args: impl Iterator<Item = OsString>) -> Vec<OsString> {
     out
 }
 
+impl Default for GlobSet {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
 impl GlobSet {
-    fn with_capacity(capacity: usize) -> Self {
+    /// Create an empty GlobSet.
+    pub fn new() -> Self {
+        Self {
+            patterns: Vec::new(),
+        }
+    }
+
+    pub fn with_capacity(capacity: usize) -> Self {
         Self {
             patterns: Vec::with_capacity(capacity),
         }

src/line_buffer.rs

@@ -1,3 +1,8 @@
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
 use memchr::memchr;
 use std::fs::File;
 use std::io::{self, Read as _};

src/main.rs

@@ -1 +1,6 @@
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
 uucore::bin!(uu_grep);

src/matcher.rs

@@ -1,5 +1,13 @@
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
 use crate::{Config, RegexMode};
-use onig::{EncodedBytes, Regex, RegexOptions, Region, SearchOptions, Syntax, SyntaxBehavior};
+use onig::{
+    EncodedBytes, Regex, RegexOptions, Region, SearchOptions, Syntax, SyntaxBehavior,
+    SyntaxOperator,
+};
 use onig_sys::{OnigEncCtype_ONIGENC_CTYPE_WORD, OnigEncodingUTF8};
 use uucore::error::{UResult, USimpleError};
 
@@ -201,11 +209,23 @@ impl CompiledPattern {
             RegexMode::Fixed => Syntax::asis(),
             RegexMode::Basic => Syntax::grep(),
             RegexMode::Extended => Syntax::gnu_regex(),
-            RegexMode::Perl => Syntax::perl(),
+            RegexMode::Perl => Syntax::perl_ng(),
         };
-        if !matches!(config.regex_mode, RegexMode::Fixed) {
+        if config.regex_mode != RegexMode::Fixed {
+            // GNU grep supports `{,n}` as an alias for `{0,n}`.
             syntax.enable_behavior(SyntaxBehavior::SYNTAX_BEHAVIOR_ALLOW_INTERVAL_LOW_ABBREV);
         }
+        if config.regex_mode == RegexMode::Perl {
+            // GNU grep supports `(?P<name>...)`.
+            // Unfortunately, the onig crate defines the OP2 flag without the
+            // necessary <<32 bit shift, so we need to hotpatch that here.
+            const _: () =
+                assert!(SyntaxOperator::SYNTAX_OPERATOR_QMARK_CAPITAL_P_NAME.bits() == 0x80000000);
+            const FIXED: SyntaxOperator = SyntaxOperator::from_bits_retain(
+                SyntaxOperator::SYNTAX_OPERATOR_QMARK_CAPITAL_P_NAME.bits() << 32,
+            );
+            syntax.enable_operators(FIXED);
+        }
 
         let mut options = RegexOptions::REGEX_OPTION_NONE;
         if config.ignore_case {

src/output.rs

@@ -1,8 +1,14 @@
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
 use crate::Config;
 use crate::context_buffer::LineView;
 use std::ffi::OsStr;
 use std::io::{self, BufWriter, StdoutLock, Write};
 use std::path::Path;
+use uucore::error::strip_errno;
 
 #[cfg(target_pointer_width = "64")]
 const BUF_SIZE: usize = 128 * 1024;
@@ -190,7 +196,12 @@ impl<'a> OutputWriter<'a> {
     /// Write an IO error to stderr.
     pub fn report_io_error(&self, label: &OsStr, err: &io::Error) {
         if !self.config.no_messages && !self.config.quiet {
-            eprintln!("grep: {label}: {err}", label = label.to_string_lossy());
+            // Strip the trailing " (os error XX)" so the message matches GNU grep.
+            eprintln!(
+                "grep: {label}: {err}",
+                label = label.to_string_lossy(),
+                err = strip_errno(err)
+            );
         }
     }
 

src/searcher.rs

@@ -1,3 +1,8 @@
+// This file is part of the uutils grep package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
 use crate::context_buffer::{ContextBuffer, LineView};
 use crate::line_buffer::LineBuffer;
 use crate::matcher::Matcher;

tests/test_grep.rs

@@ -440,6 +440,17 @@ fn files_with_and_without_matches() {
         .fails_with_code(1)
         .stdout_is("hit\nmiss\n");
 
+    // -L with a pattern that DOES match in one file: the matching file is
+    // excluded from the listing, so only the non-matching file is printed.
+    // This exercises the early-return in `session_handle_match` taken when
+    // `files_without_match` is set and a match is found (src/searcher.rs).
+    let (scene, mut c) = ucmd();
+    scene.fixtures.write("hit", "yes\n");
+    scene.fixtures.write("miss", "no\n");
+    c.args(&["-L", "yes", "hit", "miss"])
+        .succeeds()
+        .stdout_is("miss\n");
+
     // -l early-exits after the first match. Verify it doesn't print twice.
     // when the file has many.
     let (scene, mut c) = ucmd();
@@ -615,6 +626,15 @@ fn line_number_and_byte_offset_prefixes() {
         .pipe_in("x\n")
         .succeeds()
         .stdout_contains("1:\tx\n");
+
+    // -T against a real file (not stdin): the line-number field width is
+    // derived from the file size, which only happens on the `File` path in
+    // `process_file` (src/searcher.rs), not the stdin path.
+    let (scene, mut c) = ucmd();
+    scene.fixtures.write("f", "x\n");
+    c.args(&["-T", "-n", "x", "f"])
+        .succeeds()
+        .stdout_contains("1:\tx\n");
 }
 
 #[test]
@@ -883,6 +903,25 @@ fn recursive_no_file_defaults_to_cwd_not_stdin() {
         .stdout_contains("a.txt");
 }
 
+#[test]
+fn recursive_implicit_cwd_strips_dot_prefix() {
+    // `-r` with no path argument searches the implicit ".", so reported paths
+    // come back as "./only.txt" (or ".\\only.txt" on Windows). GNU strips that
+    // leading prefix; `strip_dot_prefix` in src/searcher.rs must do the same.
+    // A single file keeps the output deterministic and lets us assert the exact
+    // line, which `stdout_contains` in `recursive_no_file_defaults_to_cwd_not_stdin`
+    // cannot (it would also pass with a leaked "./" prefix).
+    let (scene, _) = ucmd();
+    scene.fixtures.mkdir_all("flat");
+    scene.fixtures.write("flat/only.txt", "grep me\n");
+
+    let mut c = scene.cmd(env!("CARGO_BIN_EXE_grep"));
+    c.current_dir(scene.fixtures.plus("flat"))
+        .args(&["-r", "grep"])
+        .succeeds()
+        .stdout_is("only.txt:grep me\n");
+}
+
 #[test]
 fn recursive_with_include_exclude() {
     let (scene, _) = ucmd();
@@ -1022,6 +1061,30 @@ fn recursive_skips_fifos_by_default() {
         .stdout_does_not_contain("fifo");
 }
 
+#[cfg(unix)]
+#[test]
+fn device_skip_on_explicit_special_file_arg() {
+    use std::process::Command;
+
+    // A special file (FIFO) named *directly* as an argument, not via recursion.
+    // With `-D skip` it must be dropped without reading (reading would block
+    // forever, so the test returning at all proves it was skipped). This covers
+    // the top-level special-file branch in `process_path` and `is_special_file`
+    // (src/searcher.rs), distinct from the recursive FIFO path.
+    let (scene, _) = ucmd();
+    let fifo_path = scene.fixtures.plus("fifo");
+    let status = Command::new("mkfifo")
+        .arg(&fifo_path)
+        .status()
+        .expect("mkfifo failed");
+    assert!(status.success(), "could not create FIFO");
+
+    let mut c = scene.cmd(env!("CARGO_BIN_EXE_grep"));
+    c.args(&["-D", "skip", "grep", "fifo"])
+        .fails_with_code(1)
+        .no_output();
+}
+
 #[test]
 fn nonexistent_file_is_error() {
     let (_s, mut c) = ucmd();
@@ -1030,6 +1093,23 @@ fn nonexistent_file_is_error() {
         .stderr_contains("does-not-exist");
 }
 
+#[test]
+fn nonexistent_file_error_has_no_os_error_suffix() {
+    // GNU prints "grep: <file>: No such file or directory" with no
+    // " (os error 2)" suffix; strip_errno keeps us byte-compatible. The
+    // underlying OS message text differs on Windows, but in both cases the
+    // trailing " (os error N)" must be absent.
+    #[cfg(not(windows))]
+    let expected = "grep: does-not-exist: No such file or directory\n";
+    #[cfg(windows)]
+    let expected = "grep: does-not-exist: The system cannot find the file specified.\n";
+
+    let (_s, mut c) = ucmd();
+    c.args(&["x", "does-not-exist"])
+        .fails_with_code(2)
+        .stderr_is(expected);
+}
+
 #[test]
 fn dash_argument_means_stdin() {
     let (_s, mut c) = ucmd();
@@ -1138,3 +1218,38 @@ fn help_and_version() {
         .succeeds()
         .stdout_contains(env!("CARGO_PKG_VERSION"));
 }
+
+#[test]
+fn repeated_options_are_accepted() {
+    // GNU grep tolerates options given more than once: boolean flags are
+    // idempotent and value options take the last occurrence. clap would
+    // otherwise error with "cannot be used multiple times".
+
+    // Repeated boolean flags are a no-op (not an error).
+    let (_s, mut c) = ucmd();
+    c.args(&["-n", "-n", "a"])
+        .pipe_in("abc\n")
+        .succeeds()
+        .stdout_only("1:abc\n");
+
+    // Mixed repeated booleans behave like a single occurrence.
+    let (_s, mut c) = ucmd();
+    c.args(&["-i", "-i", "abc"])
+        .pipe_in("ABC\n")
+        .succeeds()
+        .stdout_only("ABC\n");
+
+    // Repeated value options take the last value (here: -m 1 wins).
+    let (_s, mut c) = ucmd();
+    c.args(&["-m", "5", "-m", "1", "x"])
+        .pipe_in("x\nx\nx\n")
+        .succeeds()
+        .stdout_only("x\n");
+
+    // -e (ArgAction::Append) must still accumulate every pattern.
+    let (_s, mut c) = ucmd();
+    c.args(&["-e", "a", "-e", "b"])
+        .pipe_in("a\nb\nc\n")
+        .succeeds()
+        .stdout_only("a\nb\n");
+}

util/compare_test_results.py

@@ -0,0 +1,173 @@
+#!/usr/bin/env python3
+
+"""
+Compare the current GNU test results to the last results gathered from the main branch to
+highlight if a PR is making the results better/worse.
+Don't exit with error code if all failing tests are in the ignore-intermittent.txt list.
+"""
+
+import json
+import sys
+import argparse
+from pathlib import Path
+
+
+def load_ignore_list(ignore_file):
+    """Load list of intermittent test names to ignore from file."""
+    ignore_set = set()
+    if ignore_file and Path(ignore_file).exists():
+        with open(ignore_file, "r") as f:
+            for line in f:
+                line = line.strip()
+                if line and not line.startswith("#"):
+                    ignore_set.add(line)
+    return ignore_set
+
+
+def extract_test_results(json_data):
+    """Extract test results from JSON data."""
+    if not json_data or "summary" not in json_data:
+        return {"total": 0, "passed": 0, "failed": 0, "skipped": 0}, []
+
+    summary = json_data["summary"]
+    tests = json_data.get("tests", [])
+
+    # Extract failed test names
+    failed_tests = []
+    for test in tests:
+        if test.get("status") == "FAIL":
+            failed_tests.append(test.get("name", "unknown"))
+
+    return summary, failed_tests
+
+
+def compare_results(current_file, reference_file, ignore_file=None, output_file=None):
+    """Compare current results with reference results."""
+    # Load ignore list
+    ignore_set = load_ignore_list(ignore_file)
+
+    # Load JSON files
+    try:
+        with open(current_file, "r") as f:
+            current_data = json.load(f)
+        current_summary, current_failed = extract_test_results(current_data)
+    except Exception as e:
+        print(f"Error loading current results: {e}")
+        return 1
+
+    try:
+        with open(reference_file, "r") as f:
+            reference_data = json.load(f)
+        reference_summary, reference_failed = extract_test_results(reference_data)
+    except Exception as e:
+        print(f"Error loading reference results: {e}")
+        return 1
+
+    # Calculate differences
+    pass_diff = int(current_summary.get("passed", 0)) - int(
+        reference_summary.get("passed", 0)
+    )
+    fail_diff = int(current_summary.get("failed", 0)) - int(
+        reference_summary.get("failed", 0)
+    )
+    total_diff = int(current_summary.get("total", 0)) - int(
+        reference_summary.get("total", 0)
+    )
+
+    # Find new failures and improvements
+    current_failed_set = set(current_failed)
+    reference_failed_set = set(reference_failed)
+
+    new_failures = current_failed_set - reference_failed_set
+    improvements = reference_failed_set - current_failed_set
+
+    # Filter out intermittent failures
+    non_intermittent_new_failures = new_failures - ignore_set
+
+    # Check if results are identical (no changes)
+    no_changes = (
+        pass_diff == 0
+        and fail_diff == 0
+        and total_diff == 0
+        and not new_failures
+        and not improvements
+    )
+
+    # If no changes, write empty output to prevent comment posting
+    if no_changes:
+        with open(output_file, "w") as f:
+            f.write("")
+        return 0
+
+    # Prepare output message
+    output_lines = []
+
+    # Show current vs reference numbers for debugging
+    output_lines.append("Test results comparison:")
+    output_lines.append(
+        f"  Current:   TOTAL: {current_summary.get('total', 0)} / PASSED: {current_summary.get('passed', 0)} / FAILED: {current_summary.get('failed', 0)} / SKIPPED: {current_summary.get('skipped', 0)}"
+    )
+    output_lines.append(
+        f"  Reference: TOTAL: {reference_summary.get('total', 0)} / PASSED: {reference_summary.get('passed', 0)} / FAILED: {reference_summary.get('failed', 0)} / SKIPPED: {reference_summary.get('skipped', 0)}"
+    )
+    output_lines.append("")
+
+    # Summary of changes
+    if pass_diff != 0 or fail_diff != 0 or total_diff != 0:
+        output_lines.append("Changes from main branch:")
+        output_lines.append(f"  TOTAL: {total_diff:+d}")
+        output_lines.append(f"  PASSED: {pass_diff:+d}")
+        output_lines.append(f"  FAILED: {fail_diff:+d}")
+        output_lines.append("")
+
+    # New failures
+    if new_failures:
+        output_lines.append(f"New test failures ({len(new_failures)}):")
+        for test in sorted(new_failures):
+            if test in ignore_set:
+                output_lines.append(f"  - {test} (intermittent)")
+            else:
+                output_lines.append(f"  - {test}")
+        output_lines.append("")
+
+    # Improvements
+    if improvements:
+        output_lines.append(f"Test improvements ({len(improvements)}):")
+        for test in sorted(improvements):
+            output_lines.append(f"  + {test}")
+        output_lines.append("")
+
+    # Write output
+    output_text = "\n".join(output_lines)
+    if output_file:
+        with open(output_file, "w") as f:
+            f.write(output_text)
+    else:
+        print(output_text)
+
+    # Return appropriate exit code
+    if non_intermittent_new_failures:
+        print(
+            f"ERROR: Found {len(non_intermittent_new_failures)} new non-intermittent test failures"
+        )
+        return 1
+
+    return 0
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Compare GNU test results")
+    parser.add_argument("current", help="Current test results JSON file")
+    parser.add_argument("reference", help="Reference test results JSON file")
+    parser.add_argument(
+        "--ignore-file", help="File containing intermittent test names to ignore"
+    )
+    parser.add_argument("--output", help="Output file for comparison results")
+
+    args = parser.parse_args()
+
+    return compare_results(args.current, args.reference, args.ignore_file, args.output)
+
+
+if __name__ == "__main__":
+    sys.exit(main())

util/fetch-gnu.sh

@@ -0,0 +1,17 @@
+#!/bin/bash -e
+# This file is part of the uutils grep package.
+#
+# For the full copyright and license information, please view the LICENSE
+# file that was distributed with this source code.
+#
+# Download and extract the upstream GNU grep release tarball into the current
+# directory. Run it from an (empty) directory that will hold the GNU grep tree,
+# e.g.:
+#
+#   mkdir -p ../gnu.grep && (cd ../gnu.grep && bash ../grep/util/fetch-gnu.sh)
+#
+# The extracted tree ships a ready-to-use gnulib test framework under tests/
+# (init.sh + init.cfg + the extensionless test scripts), which
+# util/run-gnu-testsuite.sh drives against the Rust grep binary.
+ver="3.12"
+curl -L "https://ftp.gnu.org/gnu/grep/grep-${ver}.tar.xz" | tar --strip-components=1 -xJf -

util/run-gnu-testsuite.sh

@@ -0,0 +1,359 @@
+#!/bin/bash
+# This file is part of the uutils grep package.
+#
+# For the full copyright and license information, please view the LICENSE
+# file that was distributed with this source code.
+#
+# Run the upstream GNU grep testsuite against the Rust grep implementation.
+#
+# Unlike GNU coreutils, we do *not* build GNU grep here. Instead we reuse the
+# gnulib test framework (tests/init.sh + tests/init.cfg) shipped in the GNU grep
+# release tarball and inject our Rust `grep` binary via PATH, replicating the
+# environment that tests/Makefile.am's TESTS_ENVIRONMENT would normally set up.
+# Each test is classified by its gnulib exit code: 0 = PASS, 77 = SKIP, anything
+# else = FAIL (timeouts and framework failures count as FAIL).
+#
+# Get the GNU grep sources with:
+#   mkdir -p ../gnu.grep && (cd ../gnu.grep && bash ../grep/util/fetch-gnu.sh)
+#
+# Usage: ./util/run-gnu-testsuite.sh [options]
+#
+# Options:
+#   -h, --help                Show this help message
+#   -v, --verbose             Show diagnostics for failing/skipped tests
+#   -q, --quiet               Only print failures and the final summary
+#   --json-output FILE        Write results to FILE as JSON
+#
+# Environment variables:
+#   GNU_GREP_DIR              Path to the extracted GNU grep source tree
+#                             (default: ../gnu.grep)
+#   RUN_EXPENSIVE_TESTS       Set to "yes" to run expensive tests (default: no)
+#   PER_TEST_TIMEOUT          Per-test timeout in seconds (default: 30)
+
+# Don't exit on failure since test failures are expected.
+set -o pipefail
+
+# Configuration
+RUST_GREP_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+GNU_GREP_DIR="${GNU_GREP_DIR:-${RUST_GREP_DIR}/../gnu.grep}"
+GNU_TESTS_DIR=""
+VERBOSE=false
+QUIET=false
+JSON_OUTPUT_FILE=""
+PER_TEST_TIMEOUT="${PER_TEST_TIMEOUT:-30}"
+DETAILED_RESULTS=()
+
+# Statistics
+TOTAL_TESTS=0
+PASSED_TESTS=0
+FAILED_TESTS=0
+SKIPPED_TESTS=0
+
+usage() {
+    echo "Usage: $0 [options]"
+    echo
+    echo "Options:"
+    echo "  -h, --help                Show this help message"
+    echo "  -v, --verbose             Show diagnostics for failing/skipped tests"
+    echo "  -q, --quiet               Only print failures and the final summary"
+    echo "  --json-output FILE        Write results to FILE as JSON"
+    echo
+    echo "Environment variables:"
+    echo "  GNU_GREP_DIR              Path to the extracted GNU grep source tree"
+    echo "                            (default: ../gnu.grep)"
+    echo "  RUN_EXPENSIVE_TESTS       Set to 'yes' to run expensive tests"
+    echo "  PER_TEST_TIMEOUT          Per-test timeout in seconds (default: 30)"
+    echo
+    echo "Setup:"
+    echo "  mkdir -p ../gnu.grep && (cd ../gnu.grep && bash ../grep/util/fetch-gnu.sh)"
+}
+
+log_info()    { [[ "$QUIET" != "true" ]] && echo "[INFO] $1"; return 0; }
+log_success() { [[ "$QUIET" != "true" ]] && echo "[PASS] $1"; return 0; }
+log_skip()    { [[ "$QUIET" != "true" ]] && echo "[SKIP] $1"; return 0; }
+log_warning() { echo "[WARN] $1"; }
+log_error()   { echo "[FAIL] $1"; }
+
+# Generate JSON output (schema shared with ../sed so compare_test_results.py works).
+generate_json_output() {
+    cd "$RUST_GREP_DIR" || return
+
+    local timestamp
+    timestamp=$(date -u +"%Y-%m-%dT%H:%M:%SZ")
+    local rust_version
+    rust_version=$(cargo metadata --no-deps --format-version 1 2>/dev/null | jq -r '.packages[0].version // "unknown"')
+
+    local tests_json="[]"
+    if [[ ${#DETAILED_RESULTS[@]} -gt 0 ]]; then
+        local temp_file
+        temp_file=$(mktemp)
+        printf "%s\n" "${DETAILED_RESULTS[@]}" > "$temp_file"
+        tests_json=$(jq -s '.' < "$temp_file" 2>/dev/null) || tests_json="[]"
+        rm -f "$temp_file"
+    fi
+
+    jq -n \
+        --arg timestamp "$timestamp" \
+        --argjson total "$TOTAL_TESTS" \
+        --argjson passed "$PASSED_TESTS" \
+        --argjson failed "$FAILED_TESTS" \
+        --argjson skipped "$SKIPPED_TESTS" \
+        --argjson duration "$duration" \
+        --arg rust_version "$rust_version" \
+        --arg gnu_testsuite_dir "$GNU_TESTS_DIR" \
+        --argjson tests "$tests_json" \
+        '{
+            timestamp: $timestamp,
+            summary: {
+                total: $total,
+                passed: $passed,
+                failed: $failed,
+                skipped: $skipped,
+                duration_seconds: $duration
+            },
+            environment: {
+                rust_grep_version: $rust_version,
+                gnu_testsuite_dir: $gnu_testsuite_dir
+            },
+            tests: $tests
+        }' > "$JSON_OUTPUT_FILE"
+
+    log_info "JSON results written to: $JSON_OUTPUT_FILE"
+}
+
+# Parse command line arguments
+while [[ $# -gt 0 ]]; do
+    case $1 in
+        -h|--help) usage; exit 0 ;;
+        -v|--verbose) VERBOSE=true; shift ;;
+        -q|--quiet) QUIET=true; shift ;;
+        --json-output) JSON_OUTPUT_FILE="$2"; shift 2 ;;
+        *) echo "Unknown argument: $1"; usage; exit 1 ;;
+    esac
+done
+
+# Validate environment
+if [[ -d "$GNU_GREP_DIR" ]]; then
+    GNU_GREP_DIR="$(cd "$GNU_GREP_DIR" && pwd)"
+    GNU_TESTS_DIR="$GNU_GREP_DIR/tests"
+fi
+
+if [[ ! -f "$GNU_TESTS_DIR/init.sh" ]]; then
+    log_error "GNU grep testsuite not found at: $GNU_GREP_DIR"
+    log_error "Fetch it with:"
+    log_error "  mkdir -p ${RUST_GREP_DIR}/../gnu.grep && (cd ${RUST_GREP_DIR}/../gnu.grep && bash ${RUST_GREP_DIR}/util/fetch-gnu.sh)"
+    exit 1
+fi
+
+if [[ ! -f "$RUST_GREP_DIR/Cargo.toml" ]]; then
+    log_error "Not in a Rust project directory: $RUST_GREP_DIR"
+    exit 1
+fi
+
+# Build the Rust grep implementation
+log_info "Building Rust grep implementation..."
+cd "$RUST_GREP_DIR" || exit 1
+if ! cargo build --release --quiet; then
+    log_error "Failed to build Rust grep implementation"
+    exit 1
+fi
+
+RUST_GREP_BIN="$RUST_GREP_DIR/target/release/grep"
+if [[ ! -x "$RUST_GREP_BIN" ]]; then
+    log_error "Built grep binary not found at: $RUST_GREP_BIN"
+    exit 1
+fi
+log_info "Using Rust grep binary: $RUST_GREP_BIN"
+
+# Create a temporary work tree that mimics a GNU grep build directory.
+TEST_WORK_DIR=$(mktemp -d)
+trap 'rm -rf "$TEST_WORK_DIR"' EXIT
+log_info "Test working directory: $TEST_WORK_DIR"
+
+# A fake $abs_top_builddir whose src/ holds the binaries the tests expect.
+BUILD_DIR="$TEST_WORK_DIR/build"
+BIN_DIR="$BUILD_DIR/src"
+mkdir -p "$BIN_DIR"
+
+# grep, plus the egrep/fgrep wrappers a handful of tests rely on.
+cat > "$BIN_DIR/grep" <<WRAPPER_EOF
+#!/bin/sh
+exec "$RUST_GREP_BIN" "\$@"
+WRAPPER_EOF
+cat > "$BIN_DIR/egrep" <<WRAPPER_EOF
+#!/bin/sh
+exec "$RUST_GREP_BIN" -E "\$@"
+WRAPPER_EOF
+cat > "$BIN_DIR/fgrep" <<WRAPPER_EOF
+#!/bin/sh
+exec "$RUST_GREP_BIN" -F "\$@"
+WRAPPER_EOF
+chmod +x "$BIN_DIR/grep" "$BIN_DIR/egrep" "$BIN_DIR/fgrep"
+
+# Empty config.h: tests that probe it for build-time features just skip.
+: > "$BUILD_DIR/config.h"
+
+# get-mb-cur-max is a tiny standalone helper used by the locale require_ checks.
+if [[ -f "$GNU_TESTS_DIR/get-mb-cur-max.c" ]]; then
+    if cc -I"$BUILD_DIR" -o "$BIN_DIR/get-mb-cur-max" "$GNU_TESTS_DIR/get-mb-cur-max.c" 2>/dev/null; then
+        log_info "Built get-mb-cur-max helper"
+    else
+        log_warning "Could not build get-mb-cur-max; multibyte/locale tests may skip"
+    fi
+fi
+
+# Replicate the PCRE_WORKS probe from tests/Makefile.am's TESTS_ENVIRONMENT.
+PCRE_WORKS=0
+if err=$(echo . | "$BIN_DIR/grep" -Pq . 2>&1); then
+    [[ -z "$err" ]] && PCRE_WORKS=1
+fi
+log_info "PCRE_WORKS=$PCRE_WORKS"
+
+GREP_VERSION=$(basename "$GNU_GREP_DIR" | sed 's/^grep-//')
+[[ "$GREP_VERSION" == "$(basename "$GNU_GREP_DIR")" ]] && GREP_VERSION="unknown"
+HOST_TRIPLET="$(uname -m)-pc-linux-gnu"
+
+# Record a test result (for JSON output)
+record_result() {
+    if [[ -n "$JSON_OUTPUT_FILE" ]]; then
+        DETAILED_RESULTS+=("$(jq -n \
+            --arg name "$1" --arg status "$2" --arg error "$3" \
+            '{name: $name, status: $status, error: $error}')")
+    fi
+}
+
+# Run a single GNU testsuite script with the Rust grep on PATH.
+run_gnu_test() {
+    local test_script="$1"
+    local test_name
+    test_name=$(basename "$test_script")
+
+    TOTAL_TESTS=$((TOTAL_TESTS + 1))
+
+    local test_output_file="$TEST_WORK_DIR/test_output_$$"
+    local test_exit_code=0
+
+    # When not the process-group leader (e.g. in CI), GNU timeout falls back to
+    # "foreground" mode and SIGTERMs the whole group on timeout. Shield the
+    # parent script so a single hung test doesn't take the run down.
+    trap '' TERM
+
+    (
+        cd "$TEST_WORK_DIR" || exit 99
+        # init.cfg refuses to run if these are set.
+        unset GREP_COLOR GREP_COLORS TERM CDPATH
+        export PATH="$BIN_DIR:$PATH"
+        export srcdir="$GNU_TESTS_DIR" abs_srcdir="$GNU_TESTS_DIR"
+        export abs_top_srcdir="$GNU_GREP_DIR" top_srcdir="$GNU_GREP_DIR"
+        export abs_top_builddir="$BUILD_DIR"
+        export CONFIG_HEADER="$BUILD_DIR/config.h"
+        export built_programs="grep egrep fgrep"
+        export AWK=awk PERL=perl SHELL=/bin/sh MAKE=make CC=cc
+        export LC_ALL=C MALLOC_PERTURB_=87
+        export VERSION="$GREP_VERSION" PACKAGE_VERSION="$GREP_VERSION"
+        export host_triplet="$HOST_TRIPLET"
+        export PCRE_WORKS="$PCRE_WORKS"
+        export GREP_TEST_NAME="$test_name"
+        export RUN_EXPENSIVE_TESTS="${RUN_EXPENSIVE_TESTS:-no}"
+
+        # fd 9 is the framework's stderr (init.cfg's stderr_fileno_=9).
+        if [[ "$test_name" == *.pl ]]; then
+            exec timeout --kill-after=5 "$PER_TEST_TIMEOUT" \
+                perl -w -I"$GNU_TESTS_DIR" -MCoreutils -MCuSkip "$test_script" 9>&2
+        else
+            exec timeout --kill-after=5 "$PER_TEST_TIMEOUT" \
+                /bin/sh "$test_script" 9>&2
+        fi
+    ) </dev/null >"$test_output_file" 2>&1
+    test_exit_code=$?
+
+    trap - TERM
+
+    # Strip NUL bytes: some tests (e.g. z-anchor-newline) emit binary output,
+    # which would otherwise trigger a "ignored null byte" warning from $(...).
+    local test_output=""
+    [[ -f "$test_output_file" ]] && test_output=$(tr -d '\0' < "$test_output_file")
+    rm -f "$test_output_file"
+
+    # 124 = GNU timeout, 125 = uutils timeout, >=128 = killed by signal.
+    if [[ $test_exit_code -eq 124 || $test_exit_code -eq 125 || $test_exit_code -ge 128 ]]; then
+        log_error "$test_name (timeout)"
+        FAILED_TESTS=$((FAILED_TESTS + 1))
+        record_result "$test_name" "FAIL" "Test timed out after ${PER_TEST_TIMEOUT}s"
+        return
+    fi
+
+    case $test_exit_code in
+        0)
+            log_success "$test_name"
+            PASSED_TESTS=$((PASSED_TESTS + 1))
+            record_result "$test_name" "PASS" ""
+            ;;
+        77)
+            log_skip "$test_name"
+            SKIPPED_TESTS=$((SKIPPED_TESTS + 1))
+            [[ "$VERBOSE" == "true" ]] && echo "$test_output" | head -3 | sed 's/^/  | /'
+            record_result "$test_name" "SKIP" "$test_output"
+            ;;
+        *)
+            log_error "$test_name (exit $test_exit_code)"
+            FAILED_TESTS=$((FAILED_TESTS + 1))
+            [[ "$VERBOSE" == "true" ]] && echo "$test_output" | head -10 | sed 's/^/  | /'
+            record_result "$test_name" "FAIL" "Exit code: $test_exit_code"
+            ;;
+    esac
+}
+
+# Discover the canonical test list from tests/Makefile.am's TESTS variable.
+collect_tests() {
+    awk '
+        /^TESTS *\+?=/ { collect=1; sub(/^TESTS *\+?=/, "") }
+        collect {
+            line=$0
+            cont=sub(/\\[ \t]*$/, "", line)
+            n=split(line, a, /[ \t]+/)
+            for (i=1; i<=n; i++) if (a[i] != "") print a[i]
+            if (!cont) collect=0
+        }
+    ' "$GNU_TESTS_DIR/Makefile.am"
+}
+
+log_info "Discovering tests from $GNU_TESTS_DIR/Makefile.am"
+mapfile -t TEST_LIST < <(collect_tests | sort -u)
+log_info "Found ${#TEST_LIST[@]} tests"
+
+log_info "Starting test execution..."
+start_time=$(date +%s)
+
+for t in "${TEST_LIST[@]}"; do
+    [[ -z "$t" ]] && continue
+    test_path="$GNU_TESTS_DIR/$t"
+    [[ -f "$test_path" ]] || { log_warning "Listed test not found: $t"; continue; }
+    run_gnu_test "$test_path"
+done
+
+end_time=$(date +%s)
+duration=$((end_time - start_time))
+
+# Print summary
+echo
+echo "========================================="
+echo "GNU grep testsuite results"
+echo "========================================="
+echo "Total tests:   $TOTAL_TESTS"
+echo "Passed:        $PASSED_TESTS"
+echo "Failed:        $FAILED_TESTS"
+echo "Skipped:       $SKIPPED_TESTS"
+echo "Duration:      ${duration}s"
+
+if [[ -n "$JSON_OUTPUT_FILE" ]]; then
+    generate_json_output
+fi
+
+if [[ $((PASSED_TESTS + FAILED_TESTS)) -gt 0 ]]; then
+    pass_rate=$(( (PASSED_TESTS * 100) / (PASSED_TESTS + FAILED_TESTS) ))
+    echo "Pass rate:     ${pass_rate}%"
+fi
+
+# Mirror the script's exit convention to ../sed: nonzero if anything failed.
+[[ $FAILED_TESTS -eq 0 ]] && exit 0 || exit 1