Skip to content

Changelog

All notable changes to go-crap will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

v0.5.1 - UNRELEASED

Added

  • --progress flag — force-enable progress indicators (default: auto-detect terminal)
  • --no-progress flag — force-disable progress indicators
  • Progress bars on stderr showing scan phases: "Discovering modules", "Running coverage tests", "Analyzing complexity", "Processing results"
  • Docker images published to docker.io/padiazg/go-crap and ghcr.io/padiazg/go-crap (multi-arch linux/amd64 + linux/arm64) — run with docker run --rm -v "$PWD:/code" ghcr.io/padiazg/go-crap scan
  • scan now accepts multiple Go package patterns (same syntax as go list / go build): ./... (recursive), ./pkg (single), github.com/x/y (import path). Patterns resolved via go list -json -e, with fallback to path-based discovery for nested modules

Changed

  • No-arg scan default changed from current directory (.) to whole module (./...); accepts multiple positional patterns instead of a single path argument

v0.5.0 - 2026-07-28

Added

  • Partial coverage preserved on test failure — when go test fails in a module, go-crap still parses the coverage profile from any passing tests instead of discarding all coverage data
  • extractFailedTests helper parses --- FAIL: TestName lines from stderr and logs failed test names at Warn level (use --verbose to see them)
  • --coverage-profile flag — supply an existing coverage profile instead of running go test (collaboration with @mem)
  • --timeout flag — configurable timeout for the scan command (collaboration with @matiasinsaurralde)
  • Profile field on ModuleCoverage — stores the path to the coverage profile file
  • Baseline comparison engine — load previous JSON reports for delta analysis
  • New --baseline <path> flag — compare current CRAP against a previous report
  • New --fail-regression flag — exit code 1 when functions regressed vs baseline
  • New --fail-regression-ignore-covered flag — exclude fully covered functions from regression failures (still shows them with ~)
  • New --fail-regression-threshold <float> flag (default 0.01) — minimum delta to trigger
  • New --show-unchanged flag — in baseline mode, also show functions whose CRAP score did not change (requires --baseline)
  • Baseline struct with LoadBaseline() — parses JSON reports into comparison-ready data
  • AnnotateWithBaseline() — sets BaselineCRAP on each CRAP entry for delta tracking
  • ComputeSummaryWithBaseline() — aggregate stats with baseline/compare deltas
  • FindRegressions() — exported regression detection for CLI use
  • ErrRegression sentinel error — distinct from threshold exceedance
  • BaselineCRAP field on CRAPEntry — stores previous report CRAP value (-1 = new function)
  • Delta field on CRAPEntry — CRAP delta from baseline

Changed

  • runTests no longer deletes the temp coverage profile on error; the profile is kept for parseCoverProfile and cleaned up after parsing
  • scanModule no longer returns early on runTests failure; coverage is always parsed and partial data is returned alongside the error
  • Scan preserves the full ModuleCoverage struct from scanModule, retaining Functions alongside the wrapped error
  • JSON schema bumped to 1.1.0 — per-entry baseline_crap/delta fields, summary object with baseline/compare
  • PR comment formatter: header shows Combined/Average CRAP with deltas vs baseline; ASCII badges ([OK], [!!], [ERROR], [NEW]); "New Functions" and "Regressions" sections; summary table
  • Table formatter: Δ column when baseline available; per-function delta with +X ↑ / -X ↓; delta footer lines
  • GitHub formatter: summary ::notice:: annotation when baseline provided
  • CLI: --fail-regression without --baseline returns error
  • JSON output: schema field updated to 1.1.0, new summary object with baseline/compare stats

Fixed

  • Coverage profile deduplication — when go test -coverprofile produces a merged profile from multiple test binaries, the same source blocks appeared multiple times. Entries are now deduplicated by (startLine, endLine) position using OR semantics (a block is covered if any test binary covered it), fixing inaccurate coverage percentages (reported by @corani)
  • Coverage profile temp files now cleaned up after successful scan (collaboration with @matiasinsaurralde)
  • Field alignment for better memory layout
  • Tests added for Spash function

v0.4.1 - 2026-06-18

Added

  • Coverage unavailable warning — propagates go test failure messages through merge → score → formatters
  • CoverageWarning field on MergedEntry and CRAPEntry — set when coverage data is unavailable due to module-level test errors
  • table formatter: coverage column shows N/A ‼ with deduplicated footer warning when coverage unavailable
  • json formatter: coverage is null and coverage_warning field populated when coverage unavailable
  • github formatter: emits ::warning annotation when coverage is unavailable
  • sarif formatter: new result with RuleID: "go-crap/coverage-unavailable" when coverage unavailable
  • pr-comment formatter: "Coverage Unavailable" section listing affected functions
  • Issue templates (bug report + feature request)
  • Dependabot configuration
  • .golangci.yml with full linter configuration
  • internal/scan/entries.go with EntryList test helpers
  • Extensive test coverage across all packages
  • Mutation testing robustness improvements across all internal packages
  • permissions: read-all to GitHub Actions workflow

Changed

  • Refactored cmd/scan.go output pipeline to use config struct
  • Refactored internal/scan/scan.go with Options struct and helper extraction
  • DRY refactors across internal packages
  • Dead code removal
  • Doc comment cleanup across internal packages
  • Bumped golangci-lint config version to v1.65.0
  • Bumped golangci-lint-action from v6 to v9
  • Removed dead code in pkg/slogger and pkg/dummylogger

Fixed

  • crappy[:25] slice bounds panic when filtered entries < 25 items
  • Linting issues across codebase
  • Typo ThresholdExeeded → ThresholdExceeded
  • Typo in version command description ("TestGen" → "go-crap")
  • Logic bugs in merge, score, mutation, report, and scan packages
  • Typos and embarrassment fixes across codebase

v0.4.0 - 2026-06-08

Added

  • New --mutation-report flag — path to a gremlins JSON mutation report to validate coverage reliability
  • New internal/mutation package — parses gremlins mutation reports and annotates CRAP entries
  • New CoverageUntrusted field on CRAPEntry — set to true when lived mutants are found in a function
  • New MutationScore field on CRAPEntry — killed/(killed+lived) ratio for the function
  • New EffectiveCRAP field on CRAPEntry — CRAP score recalculated with 0% coverage when CoverageUntrusted is true
  • Mutation score included in JSON output (mutation_score, coverage_untrusted, effective_crap fields)
  • Coverage warning flag (⚠) in table and pr-comment formatters when coverage is unreliable
  • Coverage-untrusted warnings in github and sarif formatters (SARIF adds a second coverage-untrusted result)
  • New "Unreliable Coverage" section in pr-comment output listing affected functions with mutation scores
  • New --detailed flag — includes mutation failure details (original/replacement code, line, type) in report output
  • New MutationDetail struct on CRAPEntry — stores survived mutant details when mutation report is provided
  • JSON output includes mutation_details array per entry when --detailed is set, with type, mutator_name, file, line, status, original_text, and replacement_text fields
  • SARIF appends survived mutation details (type, line, code diff) to warning messages when --detailed
  • PR Comment adds Survived Mutants column with code snippets when --detailed
  • New OriginalCode and ReplacementCode fields on mutation Mutant struct, parsed from Gremlins JSON report

Changed

  • ThresholdExceeded() now uses EffectiveCRAP instead of CRAP for threshold comparison
  • --fail-above now checks EffectiveCRAP against the threshold

Fixed

  • Functions with lived mutants now show their true risk level via EffectiveCRAP (CRAP at 0% coverage)
  • --top no longer drops functions flagged with lived mutation mutants — CoverageUntrusted entries always survive truncation alongside the top trusted entries
  • --min no longer filters out CoverageUntrusted entries, even when their CRAP score is below the threshold
  • github format now emits ::warning for CoverageUntrusted entries regardless of threshold
  • pr-comment format now always renders the "Unreliable Coverage" section when mutation report is provided, even when no function exceeds the CRAP threshold

v0.3.0 - 2026-06-05

Added

  • New sarif output format — SARIF 2.1.0 compliant JSON for static analysis tools
  • New pr-comment output format — markdown table formatted for pull request comments
  • New --verbose flag — enables debug-level logging via pkg/logger and pkg/slogger
  • New --output / -o flag — write output to a file instead of stdout

v0.2.0 - 2026-06-04

Added

  • New internal/scan package — unified scan pipeline orchestrating coverage, complexity, merge, and score in a single call
  • New pkg/utils package with BuildExcludeRegex and MatchExclude helpers
  • Method receiver tracking — CRAP entries now include receiver info for struct methods
  • EntryList wrapper type with ThresholdExceeded() method for CI integration
  • Skipped field on CRAPEntry for functions excluded by --missing skip policy
  • JSON output now uses EntryList structure with optional receiver field

Changed

  • --exclude flag now uses regex matching instead of glob matching
  • Refactored cmd/scan.go to use new internal/scan package
  • Coverage and complexity packages now accept *regexp.Regexp instead of []string for exclude patterns
  • Merge now constructs full function names from receiver + method name
  • Report formatters now accept *score.EntryList instead of []score.CRAPEntry
  • GitHub formatter now includes function name (and receiver for methods) in warning messages

Fixed

  • Properly match function names with method receivers (e.g. (*MyType).Method)
  • Handle zero-coverage entries correctly in merge step
  • runCoverTool now runs in correct module directory (modDir set as cmd.Dir)
  • Correct coverage path matching in CI environments — buildSuffix now uses 3 path components instead of 2, bridging Go module paths (github.com/.../file.go) and absolute filesystem paths (/home/runner/.../go-crap/file.go)
  • Fix normalizeFuncName failing to strip value-receiver prefixes — methods like Level.String or Logger.Debug were not matched against coverage data, causing false 0% coverage reports for all value-receiver methods
  • Fix merge discarding 0% coverage as "missing" — functions with known 0% coverage (coverage data present but no statements executed) are now correctly distinguished from functions with no coverage data at all

v0.1.0 - 2026-06-02

Added

  • Initial release of go-crap
  • scan command — analyze Go modules and calculate CRAP scores
  • AST-based cyclomatic complexity via internal/complexity (adapted from gocyclo, BSD-3-Clause)
  • Coverage profiling via internal/coverage (adapted from test-finder, MIT)
  • Double-index merge — joins coverage and complexity by (filepath, funcname) without CWD path resolution issues
  • CRAP formula — CC² × (1 - coverage/100)³ + CC
  • Missing coverage policy — pessimistic (default), optimistic, skip
  • Output formatters:
  • table — human-readable with status symbols (✓, ▲, ✗) and coverage bars
  • json — structured output with schema URL
  • github — GitHub Actions workflow annotation format
  • Filtering — --top N, --min score, --exclude glob
  • CI integration — --fail-above exits with code 1, --format github produces workflow annotations
  • GitHub Actions CI/CD workflow (tests + CRAP check)
  • GitHub Pages documentation deployment