LotusGuild/cinny
3
0
Fork 0
Code Issues 1 Pull Requests 2 Actions Packages Projects Releases Wiki Activity

docs(todo): record composer autocomplete-insert crash fix (477df4ae)
CI / Build & Quality Checks (push) Successful in 10m45s
Details
CI / Trigger Desktop Build (push) Successful in 7s
Details

Browse Source 
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
jared
2026-07-26 16:13:34 -04:00
co-authored by Claude Opus 4.8
parent 477df4ae32
commit 1176bea0ee
1 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
LOTUS_TODO.md
+4
View File
@@ -143,6 +143,10 @@ The web-side nav fix (`0ddf86c6`) makes the native rich-toast path live for the
- [x] **[Low]** Export-history date-range early-break can over-paginate + mislabel "truncated" in E2EE rooms (`oldestRawTs` only advances on decrypted `m.room.message`, so undecryptable old events never move it). `ExportRoomHistory.tsx:104,136`. **FIXED** (`3ff8fb8e`): boundary now advances on every event (getTs is envelope metadata), above the type/decryption filters; guarded `ts > 0` so a bogus 0-ts can't cause the opposite (silent under-pagination). 2-agent reviewed.
- [x] **[Info/doc]** `PolicyListViewer` is a manual room-ID/alias viewer with **no** subscribe/unsubscribe controls and no subscribed-lists listing — `LOTUS_FEATURES.md:1287` describes both. Docs oversell; not a runtime bug. **FIXED** (`8a461610`, doc): LOTUS_FEATURES corrected to describe the read-only room-ID/alias viewer (no subscribe controls).
### ✅ Composer autocomplete-insert crash (reported 2026-07) — FIXED (`477df4ae`)
Picking an autocomplete item (mention/emoji/command) occasionally tripped the composer error boundary ("encountered an error" → forced refresh) even though the element inserted. Root-caused (3 agents, incl. a headless slate simulation) to `moveCursor` deferring its cursor work to `setTimeout`, leaving the caret on the just-inserted inline-void's zero-width edge; slate-react's commit-phase `setBaseAndExtent(voidEdge, 1)` then threw `IndexSizeError` mid-render → boundary. **Fix:** do `Transforms.move` (escape the void) + `insertText(' ')` synchronously in the same commit as the insert, so the caret is a resolvable text point when the selection sync runs. Plus a recoverable boundary ("Reload composer" + `onReset` deselect) so any residual composer crash no longer needs a page refresh. (A first "sync insertText without move" attempt was caught in review — the void guard drops the space + traps the caret; `move` is required.)
### ✅ Unread/read-receipt flakiness (reported 2026-07) — FIXED (pending prod QA)
Room unread dots were inconsistent: reading a message sometimes cleared the dot, sometimes left it stuck, sometimes it resurrected. Root cause (confirmed by tracing + diffing upstream cinny `dev`): **our own "N4" change.** `handleReceipt` recomputed via `getUnreadInfo`, which reads `room.getUnreadNotificationCount()` — server-computed and **stale on the synchronous synthetic receipt echo** (SDK only zeroes it immediately when the last event is your own message) → it PUT the stale non-zero count back → stuck/resurrecting. Compounded by `hasUnread = !!unread` lighting the dot on any present map entry, incl. phantom `{0,0}` PUTs from our `UnreadNotifications` listener. Plus a Mark-as-Unread (MSC2867) flag that never cleared on opening an already-read room (no receipt → no auto-clear).