docs(audit): refresh the ledger PR column — every cluster now has a PR - #2442
docs(audit): refresh the ledger PR column — every cluster now has a PR#2442noahgift wants to merge 3 commits into
Conversation
The PR column was written before waves 3 and 4 and showed 30 findings covered. Regenerated from the live issue-to-PR map: 166 rows updated, 57 findings now carry a merged fix. All 37 clusters have at least one PR. Several are deliberately partial and say so in their own bodies - "Refs #2374 (partial) - remaining: 4, 9, 10, 12" - rather than claiming a clean sweep, which is why merged PRs outnumber closed issues: an issue closes when its LAST finding is covered. Added a short key for the column, and recorded which fixes have been verified against binaries built from main and SHA-matched via `apr --version`. That matching matters here: the workspace target dir is shared across every worktree, so an "after" binary can silently be another worktree's build - one fix agent hit exactly that and measured a binary carrying neither of its commits. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Explains the #NNNN / checkmark notation, records that all 37 clusters have a PR with several deliberately partial, and lists which fixes were verified on SHA-matched main builds. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
Parking to let the merge queue drain — branch untouched, this will be reopened. Six PRs are in the merge queue and their The queue is the only path by which anything actually merges, so it gets the runners until it is empty. Reopening immediately afterwards. |
Conflict was the whole ledger file: #2430 landed the "How this count is kept honest" methodology section on main while this branch was rewriting the same table's PR column. Rather than hand-merge two rewrites of one table, took main's file wholesale as the base and re-ran the column derivation against the live issue-to-PR map on top of it. Both changes survive, and the column now reflects everything merged to date rather than the snapshot this branch was cut from. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
Parking again to hold the CI host under its limit — branch untouched, will be reopened. Nine audit PRs in flight took the shared runner box to 84% disk and a load average of 112. At that level Reopening as the queue drains. This is pacing, not a problem with the change. |
1 participant
The ledger's
Fixed bycolumn was written before waves 3 and 4 and showed 30 findings covered. Regenerated from the live issue→PR map: 166 rows updated, 57 findings now carry a merged fix.All 37 clusters have at least one PR. Several are deliberately partial and say so in their own bodies —
Refs #2374 (partial) — remaining: 4, 9, 10, 12— rather than claiming a clean sweep. That's why merged PRs outnumber closed issues: an issue closes when its last finding is covered.Also added a short key for the column, and recorded which fixes have been verified against binaries built from
mainand SHA-matched viaapr --version:--offline,validate, SSE streaming,rosetta, the native serve routes plus the remote abort, and MCPapr.serve.That SHA match is not ceremony. The workspace target dir is shared across every worktree, so an "after" binary can silently be another worktree's concurrent build — one fix agent hit exactly that and measured a binary carrying neither of its own commits.
Docs only.
Audit epic: #2373
🤖 Generated with Claude Code