You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Signature index v2 edge cases from the #265 review: migration retry on signature-free ingest, reindex vs concurrent serve ingest #270
Two P2 Codex notes on #265 left for a follow-up (no trust-gate impact; both are fail-safe):
If the startup legacy migration (.sc/signatures -> .sc/signatures.v2) failed transiently and a later incoming pack contains NO signatures, ingest does not retry the migration, so legacy entries stay unread until the next open/ingest that does. Retry migration on every ingest (cheap when already migrated), or on the next Repo::open.
signatures::reindex is not locked across its scan; if a serve-path ingest targets the same destination during a clone's post-clone reindex, concurrent appends can be lost. Take the index lock around the rebuild's write phase (or merge appends made since the scan started).
Activity
fix(repo): reindex re-judges signatures whose target landed after the…
fix(repo): reindex verifies late signatures outside the index lock (ref
test(repo): cover reindex's locked pass on its own (ref #270)
fix(repo): signature index v2 edge cases: migration retry and reindex…