description: "Verify closed bug fixes haven't regressed — dynamically discovers and checks" argument-hint: "--issues <N,N,N> --limit <N> --label <label>"

Regression Verification Audit

Confirm that previously-fixed bugs are still fixed. This audit dynamically discovers closed bug issues from GitHub, locates each fix and its guard test, and reports any fix that has gone missing as a Regression of #NNN.

See .claude/commands/_audit-common.md for project layout, severity, dedup, context rules, and the per-finding format. See .claude/commands/_audit-severity.md for the severity scale. This file only adds the regression-specific flow.

Parameters (from $ARGUMENTS)

--issues <N,N,N>: Verify only these issue numbers (e.g., --issues 9,16,1516).
--limit <N>: Max closed issues to verify (default: 50).
--label <label>: Issue label filter (default: bug).

Step 1 — Discover fixed issues

gh issue list --repo matiaszanolli/ByroRedux --state closed --label bug \
  --limit 50 --json number,title,body,closedAt,labels

With --issues, fetch those numbers directly (gh issue view <N> --repo matiaszanolli/ByroRedux --json number,title,body,closedAt,labels) instead.

Default --label bug structurally misses closed issues carrying only the documentation / doc-rot labels (e.g. #1818, a doc-rot fix) — pass --label bug,documentation,doc-rot to include doc-rot regressions in the discovery pass. To scope a regression sweep to one title, filter on the game axis instead: --label game:skyrim (likewise game:fnv, game:fo3, game:fo4, game:fo76, game:oblivion, game:starfield).

For each issue, pull out:

Number + title — the regression handle.
File references — backtick-quoted paths in the body (crates/nif/...).
Acceptance criteria / fix description — what the fix is supposed to do.
Related #NNNN — phased fixes split across several issues (e.g. #1210 → #1255 → #1257) regress as a set; verify the whole chain, not just the head.

Discovery window caveat. The repo has 1600+ closed issues. The default --limit 50 only covers the most-recently-closed bugs, so older high-value fixes get no coverage unless you raise --limit or pass them via --issues. The unconditional Step 4 fragile-area checks are the safety net for fixes that landed as proactive refactors and were never an issue at all — run them every time regardless of which issues Step 1 surfaced.

Fresh verification candidates (recent decompiler-safety + LC wave). Recently-closed, high-churn fixes worth an explicit --issues pass while they're still warm: #1815 (decompiler recursion-depth cap in the boolean-collapse pass), #1816 (translate_pex missing catch_unwind), #1728 (Skyrim-BE/Starfield round-trip test for the .pex reader), #1740 (DA10 .pex byte-equality parity test), #1731 (VWD record-header flag parse + expose), #1718 (ragdoll bone/constraint-drop telemetry on bone-name miss). Note #1651 (BGSM/BGEM GL→Gamebryo blend factors) was itself a WRONG fix — its premise was disproven and reverted by #1823; don't re-verify #1651 as if it still holds. Several of these touch the import→material boundary that Step 4 already pins — cross-check there.

Step 2 — Locate each fix and its guard

For each issue, work the fix → guard-test chain:

Find the fix commit. git log --oneline --grep="#<N>" (commits use Fix #<N>: … and fix/<N>-… branch merges). git show <commit> --stat shows which files moved.
Confirm the fix is present in the live tree. Read the referenced file(s) at the symbol named in the issue/commit (prefer grep -n "fn <name>" over a line number — the post-Session-34/35 module splits invalidate old line refs).
Find the guard test. Tests live as *_tests.rs siblings next to the module they cover (e.g. crates/nif/src/blocks/interpolator_tests.rs), or as #[cfg(test)] mod tests inline. To locate one:
- grep -rn "<N>" crates/ byroredux/ --include='*.rs' — many tests cite the issue number in a name or comment (fn fix_1516_…, // #1516).
- Failing that, grep the fixed symbol or a keyword from the title across *_tests.rs siblings of the fix file.
Run the guard to prove it still passes. Crate packages are named byroredux-<crate> (e.g. cargo test -p byroredux-nif <test_name>, cargo test -p byroredux-renderer, cargo test -p byroredux-core).

Step 3 — Assign a status

PASS — fix code confirmed present and a guard test exists (ideally run green).
PARTIAL — fix code present but no guard test. Flag as a hardening gap.
FAIL — fix code missing or its guard now fails. This is the regression. Report it with Status: Regression of #<N> per the _audit-common finding format.
UNVERIFIABLE — the issue body names no file/symbol and no fix commit is findable. Note it and move on; don't guess.

Step 4 — Unconditional fragile-area checks

These guard fixes/contracts whose breakage is invisible to GitHub-issue discovery — most landed as refactors, not closed bugs — so check them every run regardless of Step 1's window. A FAIL here is still reported as a regression (reference the relevant issue if one exists, else describe the contract).

NIFAL canonical-translation tier (spec: docs/engine/nifal.md; see also /audit-nifal for the dimension-level checklist):

Single material boundary. byroredux/src/material_translate.rs (fn translate_material) must remain the only ImportedMesh → Material site — per-game material classification lives here, never in a shader. Material (crates/core/src/ecs/components/material.rs) metalness / roughness must stay plain resolved f32 fields — no reintroduced Option<f32> and no render-time classifier. The resolve-once contract is the boundary filling overrides + Material::resolve_pbr (which calls classify_pbr_keyword) filling only the unresolved slots.
Typed particle emitters. NiPSysEmitter / NiPSysEmitterCtlr / NiPSysEmitterCtlrData / NiPSysGrowFadeModifier must still parse as typed blocks (crates/nif/src/blocks/particle.rs, dispatched in crates/nif/src/blocks/mod.rs), feed extract_emitter_params / extract_emitter_rate (crates/nif/src/import/walk/mod.rs → ImportedEmitterParams in crates/nif/src/import/types.rs), and be consumed by apply_emitter_params (byroredux/src/systems/particle.rs). A regression to opaque NiPSysBlock shows up as zero-sized emitters or clobbered colors.
Collision shape coverage. BhkMultiSphereShape + BhkConvexListShape must still translate to a CollisionShape in crates/nif/src/import/collision/mod.rs (they were previously dropped to None).

Disney BSDF + GPU struct contracts (recent shader wave):

The Disney/Burley lobe now lives in crates/renderer/shaders/include/pbr.glsl (split out of triangle.frag; the GLSL-PathTracer MIT attribution block stays top-of-triangle.frag, Burley 2012 cite). The per-reservoir resRadiance[] array was retired (#1369 factoring → commit 218b425b, which removed the ReSTIR reservoir G-buffer attachment): WRS is register-local now, recomputing the unshadowed radiance from the light index via shadowableLightRadiance in crates/renderer/shaders/include/lighting.glsl. A regression here is a reintroduced per-thread reservoir array or a re-added G-buffer reservoir attachment — verify the array stays gone, not "intact".
#[repr(C)] GPU structs hold their size pins in crates/renderer/src/vulkan/scene_buffer/gpu_instance_layout_tests.rs: GpuInstance = 128 B (gpu_instance_is_128_bytes_std430_compatible) and GpuCamera = 352 B (gpu_camera_is_352_bytes). Run them: cargo test -p byroredux-renderer gpu_.

Output

Write to: docs/audits/AUDIT_REGRESSION_<TODAY>.md (YYYY-MM-DD).

Per-issue entry

## #<ISSUE>: <Title>
- **Status**: PASS | PARTIAL | FAIL | UNVERIFIABLE
- **Closed**: <date>
- **Fix commit**: <hash> (or "not found")
- **Fix site**: `<path>` (`<symbol>`)
- **Fix present**: Yes / No / Unknown
- **Guard test**: `<test name>` in `<path>` — passes / fails / none
- **Notes**: <concerns>

Summary table

| Issue | Title | Status | Fix Present | Guard |
|-------|-------|--------|-------------|-------|

For any FAIL, surface it as a Regression of #NNN finding (base format in _audit-common.md) and suggest: /audit-publish docs/audits/AUDIT_REGRESSION_<TODAY>.md

Audit de non-régression

Recommandé pour