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
fix(sidebar): follow-ups from the finish-scope cache review (#2183) #2201
Follow-up to #2183 (non-blocking notes from the merge pass's review).
Cache tests patch Map.prototype globally (omo MEDIUM, test style): src/lib/seenFinishes.test.ts ~72 spies on .values and src/lib/sidebarActivity.hook.test.ts ~737 stubs .set, so a behaviour-preserving refactor (e.g. for...of) could break them or make them vacuous. Use a test-only counter or an identity check across loadSeen calls.
Departed PCs stay in memory (Codex LOW, src/lib/seenFinishes.ts ~187, src/lib/sidebarActivity.tsx ~138): cycling 2,025 distinct PCs kept 2,026 empty scopes and epoch entries (main capped epochs at 2,000). Bound or evict departed-PC entries while keeping stored acknowledgements.
Closing under the new follow-up bar: a follow-up issue is kept only for a reproduced, user-visible MEDIUM finding. The MEDIUM is test style (Map.prototype stubs in tests), not user-visible, and the departed-PC cache growth is LOW (2,025 synthetic PCs). Below the follow-up bar: LOW/unreproduced; recorded on PR #2183 and in this issue for reference.
Activity
devswha commentedon Oct 28, 2026
Closing under the new follow-up bar: a follow-up issue is kept only for a reproduced, user-visible MEDIUM finding. The MEDIUM is test style (Map.prototype stubs in tests), not user-visible, and the departed-PC cache growth is LOW (2,025 synthetic PCs). Below the follow-up bar: LOW/unreproduced; recorded on PR #2183 and in this issue for reference.