Lines 3654-3694markdown
3654 D005 C# vulnerabilities defect: parser written against synthetic
3655 output, never validated against real tool output. Fixed by
3656 matching the canonical `\): error \w+:` regex. Caught by adding
3657 the C# row to the new lint matrix; the row failed because the
3658 parser returned 0 despite exit code != 0 and visible violations
3659 in the output. (`src/languages/csharp.ts`, **D016**)
3663- **Cross-ecosystem matrix layer** (`test/integration/cross-ecosystem.test.ts`).
3664 New `BENCHMARK_LANGUAGES` table at the top of the file is the
3665 single source of truth for which languages participate and where
3666 each fixture's deliberate findings live. Each `describe('matrix —
3667 <report>')` block iterates the table to produce one uniform
3668 assertion per language — adding a new feature is one new
3669 optional field per row + one new `matrix —` describe; adding a
3670 6th language is one row append + one fixture dir + one CI install.
3671 No search-and-replace across describe blocks.
3673- **`matrix — secrets` (Phase 10i.0.1)** — 5 hardcoded fake AWS
3674 access keys (`AKIA1234567890ABCDEF` — patterned digits/letters
3675 that pass gitleaks' `aws-access-token` regex but fail real AWS
3676 validation and GitHub push protection). One per benchmark
3677 ecosystem. Asserts `dxkit vulnerabilities` surfaces a
3678 `SecretFinding` (category=secret, tool=gitleaks,
3679 rule=aws-access-token) for each.
3681- **`matrix — lint` (Phase 10i.0.2)** — 5 deliberate idiomatic
3682 linter violations (Python ruff F401 unused-import, Go gosimple
3683 S1002 bool-comparison, Rust clippy unused_variables, C#
3684 dotnet-format whitespace × 2). Asserts `dxkit quality` reports
3685 the expected linter and ≥1 lint finding. CI workflow now
3686 installs `ruff` (pipx), `golangci-lint` (curl install script),
3687 and `clippy` (rustup component) alongside the existing depVulns
3688 toolchains; `dotnet format` ships in the .NET 8 SDK.
3690- **`matrix — duplications` (Phase 10i.0.3)** — two near-identical
3691 helpers per fixture, sized comfortably above jscpd's
3692 `--min-lines 5 --min-tokens 50` defaults (initial pass had
3693 ~30-token bodies that fell below the threshold; widened on the
3694 way in). Asserts `metrics.duplication.cloneCount > 0`.