source-command-audit-regression

name	source-command-audit-regression
description	Verify previously fixed issues have not regressed; check that the fix code is still present and tests exist

source-command-audit-regression

Use this skill when the user asks to run the migrated source command audit-regression.

Command Template

Regression Verification Audit

Verify that ALL previously fixed issues and recent critical fixes have not regressed. For each fix, confirm the code is still present and check whether regression tests exist.

Shared protocol: Read .Codex/commands/_audit-common.md first for project layout, severity framework, methodology, deduplication rules, and GitHub issue template.

Fix Discovery (Dynamic)

Build the full list of fixes to verify by combining BOTH sources below.

Source 1: Git History

Run these commands to discover recent fixes dynamically:

cd /mnt/data/src/matchering
git log --oneline --grep="fix" -30
git log --oneline --grep="Fix" -30

For each fix commit found, read the changed files and understand what was fixed.

Source 2: Closed GitHub Issues

Run this command to find recently closed bug fixes:

gh issue list --state closed --label bug --limit 50 --json number,title,closedAt

For each closed issue, verify the fix is still present in the current code.

Source 3: Seed Registry

These are known critical invariants that must ALWAYS be verified, regardless of what git history shows. If a fix commit is listed, verify that specific commit's changes are still present.

Fix Description	Commit	File(s) to Check	What to Verify
Equal-power crossfade between mastering chunks	0a5df7a3	auralis-web/backend/core/chunked_processor.py	Crossfade uses equal-power (sqrt) curve, not linear. Overlap region is 3s.
Parallel processing for sub-bass control	8bc5b217	auralis/core/simple_mastering.py	Sub-bass processing uses parallel path to prevent excessive loss
Double-windowing removal in EQ	cca59d9c	auralis/dsp/	No double-windowing in VectorizedEQProcessor
Audio loading thread safety	53cef6b4	auralis/analysis/fingerprint/	Audio loading doesn't block on KeyboardInterrupt
Cursor-based pagination in cleanup	bd94fd59	auralis/library/	cleanup_missing_files uses ID-cursor, not offset pagination
SQLAlchemy engine disposal	8adb8d0a	auralis/library/migration_manager.py	Engine is disposed in MigrationManager.close()
Sample count preservation in DSP pipeline	—	auralis/core/hybrid_processor.py, auralis/dsp/stages.py	len(output) == len(input) invariant maintained across all processing stages
Copy-before-modify pattern	—	auralis/core/simple_mastering.py, auralis/dsp/stages.py	audio.copy() called before any in-place operations
Thread-safe player state (RLock)	—	auralis/player/enhanced_audio_player.py	All state mutations protected by RLock
SQLite thread-safe pooling	—	auralis/library/manager.py	pool_pre_ping=True and proper connection pooling configured
Repository pattern (no raw SQL)	—	auralis/library/repositories/	All database access goes through repository classes, no raw SQL
Gapless playback engine	—	auralis/player/gapless_playback_engine.py	No gap or click at track boundaries

Note: This registry should be updated when new critical fixes are made. Add entries when closing important bugs.

Verification Method

For each fix (from all sources):

Step 1: Confirm Fix Code

1. Read the file(s) listed or changed in the commit
2. Look for the specific code described in "What to Verify"
3. Verdict: FIX PRESENT or FIX MISSING (regression)

Step 2: Check for Regression Tests

1. Search for test files related to the fix: grep -r "<keyword>" tests/
2. Look for test files named after the fix behavior
3. Check if the specific invariant is asserted in any test
4. Verdict: TESTS PRESENT (list files) or NO TESTS

Step 3: Assign Status

• PASS: Fix present + tests exist
• PARTIAL: Fix present but no regression tests
• FAIL: Fix missing or broken (REGRESSION)
• N/A: Fix pending or not applicable

Phase 1: Audit

Write your report to: docs/audits/AUDIT_REGRESSION_<TODAY>.md (use today's date).

Per-Fix Format

## <Fix Description>
- **Status**: PASS | PARTIAL | FAIL | N/A
- **Source**: Git commit <hash> | GitHub issue #NNN | Seed registry
- **File checked**: `<path>:<line>`
- **Fix present**: Yes / No
- **Fix description**: <what the fix does, confirmed in code>
- **Tests exist**: Yes / No
- **Test files**: `<path>` (if applicable)
- **Notes**: <concerns, known limitations>

Summary Table

| Fix | Source | Status | Fix Present | Tests | Notes |
|-----|--------|--------|-------------|-------|-------|
| Equal-power crossfade | 0a5df7a3 | PASS | Yes | Yes | — |
| ...                    | ...      | ...  | ... | ... | ... |

Results: X PASS, Y PARTIAL, Z FAIL, W N/A

Phase 2: Report Finalization

1. Save the report to docs/audits/AUDIT_REGRESSION_<TODAY>.md
2. Do NOT create GitHub issues directly
3. Inform the user the report is ready and suggest:

   /audit-publish docs/audits/AUDIT_REGRESSION_<TODAY>.md
   

Note: When publishing, FAILs become HIGH severity issues and PARTIALs become LOW enhancement issues.