The Share Room QR was fetched from the third-party api.qrserver.com, leaking
which rooms a user shares (and failing offline / under strict CSP). Now rendered
locally via qrcode.react (QRCodeSVG) — no network request, works offline. Added a
white quiet-zone container so the code scans on any theme; dropped the qrError
fallback (local generation can't fail the same way). Removed api.qrserver.com
from the prod CSP img-src (matrix repo). Build verified (rolldown interop OK).
Verification steps added to LOTUS_TESTING.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Wave-3 bug-hunt fixes (findings in LOTUS_TODO), reviewed + gate-green:
- 🔴 ACL editor [H1–H4]: block saving an empty allow-list (was a one-click
federation brick), warn on self-ban (case-insensitive glob match of
mx.getDomain() vs allow/deny), accept real globs (1.2.3.*, *.evil.*), and
gate Save behind a confirm dialog.
- 🔴 [P1] room context menu no longer acts on the wrong room after a live
reorder (key by roomId, not list index). 🔴 [P2] status writes no longer
force presence to online over Invisible/DND (shared presenceStateFromSetting).
- 🟠 [P3] timed mutes restored on boot; [P4] custom-status auto-clear now fires
(always-mounted StatusExpiryMonitor); [P5] timezone also PUT to the m.tz
profile field so it's visible to others; [H6] RoomInsights single-pass
min/max (was Math.min(...spread) stack overflow); [H7/H8] mod-log labels.
- 🟡 [P6/P7] favorites collapse+filter, [P8] charCount reset, [P9] DM preview
refresh on decrypt; theming [T-P1] lazy decorations, [T-P2] drop the redundant
always-on body animation, [T-P4] live useReducedMotion, [T-P5] decoration key.
- NATIVE-CINNY LAW: notification presets + Powers permissions use folds icons.
DEFERRED: [H5] invite-QR is fetched from api.qrserver.com (third-party leak);
local generation needs a bundled QR lib (not added). tsc/eslint/prettier clean,
build OK, 677 tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Notification profile presets (P5-27) used literal emoji (🎮/💼/🌙) instead of
folds Icons → Gaming=Ball, Work=Monitor, Sleep=BellMute.
- Permissions "Powers" list used ✅/❌ text emoji for has/no-power → folds
Icons.Check / Icons.Cross (colored via the row).
Reviewed the rest of the UI: seasonal-theme picker emoji kept (folds has no
holiday-icon equivalents; a distinctly-Lotus visual feature), soundboard clip
emoji kept (user-chosen clip identity), URL-preview brand glyphs + upstream
device-verification emoji + keyboard key-symbols left as-is.
(Also records the F2 URL-preview decision: keep default-on.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Web fixes from the Wave-2 bug-hunt (findings in LOTUS_TODO):
- F1 (security): wipe the decrypted-plaintext search index on SERVER-FORCED
logout too (token expiry / remote sign-out) — only manual logout did before.
F4: the delete no longer reports success while onblocked (waits, 3s cap).
- M1/M2 (data-loss): useBookmarks + useUserNotes account-data writes are now
serialized at MODULE scope (single queue + latestRef per client, echo-driven),
fixing the cross-instance lost-update clobber (useBookmarks mounts per message
row, so a per-instance queue was insufficient — caught in review).
- M6: room-history export gets a 200-page cap + Cancel + unmount-abort +
correct date-range early-break (raw paginated ts). M4: image compression
skips PNG (was flattening transparency to black), bakes EXIF orientation via
createImageBitmap, .jpg-renames, and falls back to the original on decode
failure instead of dropping the file. M5: MediaGallery lightbox opens the
right item (shared thumb guard). M8: audio speed survives async decrypt.
- Desktop web wiring: D2 badge sums leaf rooms only (space double-count, like
the favicon fix); D3 useTauriDnd re-hydrates from get_tray_dnd on mount; D5
updater has a terminal state.
Reviewed; M7 reverted (past-time clamp is an intentional, tested contract).
tsc/eslint/prettier clean, build OK, 678 tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- C-H1: forceState only on FIRST join; on EC reconnect re-arm the fork handlers
(resendForkState — deafen+quality only) instead of clobbering live mic/video/
deafen back to the join-time snapshot.
- C-H2: AFK auto-mute reads the fork's io.lotus.call_state VAD of the LOCAL
published track instead of getUserMedia on the browser DEFAULT mic (which could
measure silence while the user spoke on another device → auto-mute an active
speaker). Fails safe (never mutes) when call_state is null OR empty.
- C-H3: control observer re-binds after EC re-renders (body subtree:true + 100ms
debounce) with an early-return so unchanged state doesn't re-render.
- C-M3 setQuality join-gated; C-M4 hangup 4s fallback dispose (idempotent);
C-M5 PTT no longer silently un-deafens; C-M6 screenshare-audio mute resets on
stop; C-L4 deafen key works in the iframe; C-L6 setState-after-unmount guards.
Reviewed (C-H2 [] fail-safe + C-H3 re-render guard applied). tsc/eslint/prettier
clean, build OK, 677 tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- T1 (🔴): markThreadAsRead no longer receipts the thread ROOT (a 2nd instance
of the read-marker-corruption regression — opening a thread whose root is old
re-lit the whole room). Extracted to a pure threadReceipt.ts + 5 regression
tests.
- N1 (🔴): favicon/tab-title unread count now sums only leaf rooms (was double-
counting every ancestor-space aggregate in roomToUnread).
- N2 (🔴): notifications/sounds dedupe on the event id, not the unread count —
fixes "read a DM, next message never notifies again".
- T4 (🟠): the thread notification path no longer re-gates on the room count, so
an explicit per-thread "All replies" override in a Mentions-only room fires.
- N3 (🟠): getUnreadInfos skips phantom {0,0} entries (muted-thread-only rooms no
longer light the nav row / pollute unread filters).
- N4 (🟠): the Receipt handler recomputes unread instead of blanket-DELETE, so a
threaded receipt can't wipe a room's valid main-timeline badge.
- T2 (🟠): thread "Jump to Latest" re-anchors the virtual window (was landing on
a stale mid/old event).
Gates: tsc/eslint/prettier clean, build OK, 678 tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
4 parallel deep-audit agents over the Tier-1 high-risk areas. Findings only (no
source changes). Top 🔴: markThreadAsRead corrupts the main read marker via a
thread-root receipt (a SECOND instance of the P6 read-receipt regression, likely
a live cause of "unread won't clear"); favicon/title count double-counts space
aggregates; deliverNotification dedupe cache never cleared on read → missed
notifications/sounds. Plus 🟠 (thread "All" override defeated, phantom
muted-thread dot, receipt-DELETE badge race, thread jump-to-latest, call
forceState-on-reconnect clobber, AFK wrong-mic auto-mute, stale control observer)
and a long 🟡 tail. Recorded in LOTUS_TODO for prioritized fix passes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CI check:eslint failed with 28 errors in two test files: callSounds.test.ts
(lines-between-class-members on mock classes) and lotusDenoiseUtils.test.ts
(prefer-arrow-callback on `function AudioWorkletNode(){}` constructor mocks —
arrows aren't constructable, so auto-fixing would break the test). Both are
stylistic false-positives for test code; relax them in the existing test-file
override next to max-classes-per-file. `npm run check:eslint` now exits 0.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The prior thread-receipt change (8192da5a) broke read receipts globally. Exact
cause: markAsRead used `thread.lastReply() ?? thread.rootEvent`. When a thread's
replies weren't loaded (lastReply() null — common on room open), it sent a
receipt for the thread ROOT. Since roots are "in the main timeline",
threadIdForReceipt() makes that a MAIN receipt at an old event; when the root
isn't in the loaded timeline the SDK's backward-guard falls back to timestamp
and applies it, moving the main read receipt onto an event we don't have, so
getEventReadUpTo() returns null and roomHaveUnread() reports the room unread —
re-broken on every mark-read, amplified by the bulk mark-all-orphan-rooms-read
callers.
Fix: main unthreaded receipt unchanged; the thread loop now sends a threaded
receipt ONLY for a genuine loaded thread reply (thread.lastReply()), never the
root — if replies aren't loaded, skip. New notifications.test.ts locks the
regression (null lastReply → no root receipt) + the main/threaded/no-op cases.
Gates: tsc/eslint/prettier clean, build OK, 672 tests (7 new).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two federated-room bugs surfaced by the desktop build:
1. markAsRead only sent one unthreaded receipt at the main-timeline tail. With
threadSupport enabled, thread replies leave the main timeline, so a reply
newer than that tail was never covered — its per-thread notification count
(which the room dot sums) lingered, so the unread dot never cleared even
after reading. It also early-returned when the main timeline was already
read. Now also send a threaded receipt at each unread thread's latest reply.
2. useAvatarDecoration never cached non-404 failures, so every avatar mount
re-requested io.lotus.avatar_decoration for federated users whose homeserver
403s/502s the field — a refetch storm that spammed the console and hammered
our homeserver's federation. Now cache definitive rejections (400/403/404)
and give up after ~2 transient (429/5xx) attempts per session.
Gates: tsc/eslint/prettier clean, build OK, 665 tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Toggling custom chrome expanded the screen and sent the message feed
auto-scrolling into the past. Two causes:
- DesktopChrome used height:100vh while html/#root use 100dvh; in the Tauri
webview 100vh can exceed the visible height after decorations are stripped,
making the timeline's scroll container taller than the viewport → the virtual
paginator runs away paginating backwards. Switched to 100dvh.
- Toggling live reflowed the whole app while the timeline was mounted. The
setting now persists + reloads so the layout is rebuilt cleanly (description
updated: "reloads to apply").
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- index.tsx: request navigator.storage.persist() for logged-in sessions so the
browser can't evict the IndexedDB rust-crypto store (eviction while the
localStorage session survives resurrects the device with a blank store → the
KE-1 "one time key already exists" upload storm). Guarded, checks persisted()
first, best-effort.
- Docs: remove HANDOFF_ELEMENT_CALL_FORK.md, LOTUS_E2EE_INVESTIGATION.md, and
LOTUS_BUGS.md. Port their live content into the three kept docs — verification
backlog → LOTUS_TESTING; open bugs + E2EE (KE-1..4) + an Element Call fork
operational reference (publish steps + io.lotus action catalog) → LOTUS_TODO.
Fix all dangling references (README, code comments, cross-doc links). Full
history of the removed docs remains in git.
Gates: tsc/eslint/prettier clean, build OK, 665 tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Code-read + upstream-issue triage this session:
- 41.7.0 / crypto-wasm 18.3.1 does NOT fix KE-1 (no OTK/upload change; #5200
still open) — the SDK-pin remediation lever is closed.
- Confirmed root cause = rust-crypto store <-> Synapse OTK divergence; the
leading web trigger is that cinny never requests persistent storage, so the
IndexedDB crypto store is evictable while the localStorage session survives.
- New buildable preventive mitigation: navigator.storage.persist() on login
(+ multi-tab guard, 400-loop recovery prompt). Added as §6 with a secondary
KE-2 to-device-validation hypothesis and capture discriminators.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add HSTS + Permissions-Policy + the standard X-Frame/X-Content/Referrer set to
the contrib nginx (443 block) and caddy examples; fix the caddy SPA try_files
fallback (stray space). Generic (no homeserver-specific CSP). The real prod
config lives in the matrix repo. P6-4 trimmed to headers only — patch-package /
types-drift / build-config skipped (see LOTUS_TODO).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Forward: checkbox multi-select room picker + "Send to N rooms" batch send
(Promise.allSettled). Full success auto-closes; partial failure keeps the dialog
open with a "Forwarded to X/N — failed: …" summary and prunes the selection to
only the failures (retry won't duplicate to already-sent rooms). Content builder
extracted to a unit-tested forwardContent.ts (edit-forwarding, reply-strip,
undecryptable-refused; 4 tests).
Bookmarks: BookmarksPanel resolves each saved message's live event (useRoomEvent)
so previews reflect edits and show a deleted indicator for redactions; the stored
snapshot stays as the fallback while loading, on fetch failure, or after leaving
the room. Stored bookmark shape unchanged.
Gates: tsc/eslint/prettier clean, build OK, 665 tests. Reviewed (dup-resend on
retry + Checkbox readOnly fixed).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CallControl now sends the new io.lotus.set_deafen action (join-gated via
forceState) on every deafen / screenshare-audio-mute toggle + on join, ALONGSIDE
the retained iframe-DOM .muted hack (transitional). Against the current pinned
bundle the action is immediately error-replied + swallowed by .catch — inert, no
timeout. Reordered toggleSound() to commit state before setSound() so the sent
deafen value isn't inverted.
Phase 2 (after the fork is published): bump the pin lotus.1 -> lotus.2 and delete
the DOM hack. Docs: HANDOFF §12.4, LOTUS_TODO P6-2, LOTUS_BUGS.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- `?` shortcut now stopImmediatePropagation so RoomView's type-to-focus handler
doesn't steal focus into the composer behind the dialog (and swallow Escape) —
CONFIRMED review finding.
- Typing live region stays mounted (empty when idle) so the FIRST "X is typing"
is reliably announced (a status region added with its text isn't always read).
- Removed a stray empty `{}` JSX expression in MediaGallery (leftover from an
auto-fix).
Reviewer verified the rest: collapsed-message labels, focus-return
classification (4 dialogs fixed, popouts correctly left), and all aria fixes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Enables ARIA-correctness rules (aria-props/proptypes/role/unsupported-elements,
role-has/supports-aria-props, no-redundant-roles, anchor/heading-has-content)
+ label-has-associated-control as errors — a regression gate for accessible
names + valid ARIA. control-has-associated-label deliberately NOT enabled (the
repo's <Text as="label" htmlFor> component pattern defeats its static analysis);
the real gaps it surfaced were fixed directly. Also disable max-classes-per-file
for test files (mock classes).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Focus returns to the trigger when closing 4 genuine dialogs (room-topic
viewer, reaction viewer, header topic, Search) — 20 inline popouts/menus
correctly left as-is (returning focus to a hover target would be wrong).
- Typing indicator announced via a visually-hidden role="status" region;
the visual text is aria-hidden to avoid double announcement.
- New keyboard-shortcuts help dialog (press ?, ignored while typing),
mounted in ClientNonUIFeatures.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Accessible names for ~15 controls that lacked them: invite/join/create-room/
account-data/image-pack/private-note/power-level inputs (visible <label htmlFor>
where a label exists, else aria-label); the two range sliders (night-light
intensity, noise-gate threshold); the soundboard file input; media <video>
elements; and the Media Gallery (region) + Search (dialog) overlays. Hidden
notification/preview <audio> marked aria-hidden.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Each message is role="article"; collapsed messages (consecutive from one
sender) now carry an aria-label with sender + time — previously a screen
reader heard only the body with no attribution (the biggest a11y gap).
Pure messageAriaLabel() reuses the existing time utils (+3 tests).
- Editing a message announces "Editing message from <sender>" (ariaLabel
threaded MessageEditor → CustomEditor; the main composer is unaffected).
- System emoji get role="img" + aria-label from the shortcode; custom
emoticons always have an accessible name.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fills the gap where LOTUS_BUGS referenced test IDs (P3-8/P4-1/P4-4/P4-8/N97a/
AW-1..4) with no matching procedures in the testing guide.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Dep triage recorded (zero shipped exposure; SDK now 41.7.0 stable; dompurify
removed); Needs Verification rows for the audit-wave fixes (scheduled-cancel,
emoji lazy-load, SW precache, desktop CSP smoke).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
HEAD-checks the copied denoise worklet/wasm/model assets for the selected model
and console.warns a single line listing anything missing — a silent asset skew
between the EC fork's expectations and vite's copied files would otherwise
disable noise suppression with no signal. Fire-and-forget; never blocks call
setup.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- emojibase (~965 KB) is now fully lazy: plugins/emoji.ts loads compact data +
shortcode maps via a memoized dynamic import (rejections reset the memo so a
mid-deploy chunk 404 can retry); reaction labels degrade to the raw glyph
until loaded. Consumers get FRESH array references on load (the module arrays
populate in place — same-ref state updates would skip re-render and leave
emoji search empty; reviewer-caught). Verified out of the eager graph.
- Service worker precaches hashed assets (workbox precacheAndRoute, 82 entries
~10.8 MB incl. the crypto wasm): repeat visits stop re-downloading the app.
index.html is NOT precached — navigations stay network-first so deploys are
picked up immediately; the media-auth fetch handler is untouched.
- ReactPrism: curated 21-language set — chunk 574 KB → 71 KB.
- Timeline inline images get loading="lazy".
- Removed dead dompurify (+types); sanitize-html is the real sanitizer.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- ScheduledMessagesTray: cancel prunes local state ONLY on confirmed server
cancel; failures keep the item + show an inline error (was: a failed cancel
looked cancelled but still sent at the scheduled time).
- Escape semantics: the composer consumes Escape (preventDefault+stopPropagation)
iff autocomplete is open or a reply draft is set; the thread panel and Room's
markAsRead act only on unconsumed Escape, and markAsRead defers entirely while
a thread panel is open (listener order made it fire before the panel closed).
- Room: thread panel / media gallery are mutually exclusive (most-recently-
opened wins); on mobile at most one right panel renders (thread > gallery >
members) instead of stacked fullscreen overlays.
- RemindMeDialog: busy-disabled presets (no more double-click duplicates),
try/catch with inline error, close only on success.
- ThreadTimeline: "Jump to Latest" floating chip when scrolled up (RoomTimeline
idiom).
From the 4-auditor deep-audit wave; reviewer-verified.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The deletions from the git-mv in 992d2b83 were unstaged by a concurrent
worktree operation before commit, so the pushed tree contained BOTH
threadSummary.ts and threadSummaryData.ts (and the Windows case-collision
persisted). This commit removes the stale originals; caseCollision.test.ts
would have failed CI on the incomplete state.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
threadSummary.ts (pure helpers) and ThreadSummary.tsx (chip component) lived in
the same directory differing only by case. On the case-insensitive Windows
release runner, RoomTimeline's extensionless import of ./thread/ThreadSummary
resolved .ts BEFORE .tsx and matched the helper module → rolldown
MISSING_EXPORT "ThreadSummary" — invisible on every Linux/macOS build (and the
cause of the earlier masked pdf.worker failure). Helper module renamed to
threadSummaryData.ts (+ test), 3 importers updated.
Prevention: new caseCollision.test.ts walks src/ and fails on any same-directory
names differing only by case (extensionless compare, so Foo.tsx vs foo.ts is
caught) — verified it fails on the pre-rename tree. Runs in the hard CI gate.
Gates: tsc clean, eslint/prettier clean, build OK, 658/659 tests (1 IDB skip).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Soundboard v2 — a near-parallel of the custom-emoji image-pack system for
in-call audio clips.
- Data model: 3-tier packs mirroring MSC2545 — room/space pack (state event
io.lotus.soundboard, inherited by child rooms via parent-space aggregation),
global refs (io.lotus.soundboard_rooms), and the personal pack
(io.lotus.soundboard account data; the v1 flat-list content is migrated to the
pack shape on read). New plugins/soundboard/ (readers, SoundboardPack, utils) +
hooks/useSoundboardPacks (useRelevantSoundboardPacks = user U global U room,
deduped). Unit-tested (migration + slug).
- Management: reusable SoundboardPackEditor (name + emoji + per-clip volume +
delete + upload + batched save), power-level-gated for room packs like emoji
packs; a Soundboard page wired into Room + Space settings.
- In-call: CallSoundboard rewritten as a Discord-style grid grouped by pack
(emoji + name tiles), sourcing room+parent-space U personal clips; a Manage
toggle embeds the editors; per-clip volume x master volume on playback.
- Spam guard: host gates on a playing key (fork enforces one clip at a time).
- Control bar: Mute-Screenshare moved next to the Screenshare button.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Audit of ForwardMessageDialog, fixes:
- Search input was intrinsic-width (sat in a default Row Box with no grow) —
now a Column Box stretches it full-width, matching every other search input.
- Search field is auto-focused on open (FocusTrap initialFocus; was nothing).
- Edited messages now forward the LATEST edit (m.new_content via
getEditedEvent) instead of the stale pre-edit body.
- Reply fallbacks stripped (trimReplyFromBody + <mx-reply> block) along with
m.relates_to, so forwards stand alone instead of quoting the old room.
- Undecryptable events are refused with an inline error (previously forwarded
m.bad.encrypted junk); send failures now show an error instead of silently
resetting.
- sendEvent uses the typed threadId-aware overload (explicit null) instead of
an untyped (mx as any) call relying on the SDK's legacy arg-sniffing.
- Room list + filter memoized (was re-sorting all rooms every keystroke).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
closeBundle also runs when the build FAILED mid-render (dist/ absent); the
plugin's copyFileSync then threw ENOENT and vite reported THAT instead of the
actual render error — exactly what hid the real failure in the Windows desktop
CI run. Now: warn-and-skip on any error, mkdir the dest dir when copying.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- useRoomsListener now PREPENDS the emitting Room (was appended): the SDK emits
RoomEvent.UnreadNotifications with VARIABLE arity (0/1/2 args), so a trailing
extra arg landed in the wrong positional slot on the most common room-count
sync path — room.isSpaceRoom() threw inside the SDK emit loop and the badge
PUT never ran. Both consumers updated (CONFIRMED HIGH review finding).
- roomToUnread: SpaceChild RESET now passes the thread prefs so muted-thread
subtraction survives space-child state changes.
Reviewer also verified: badge subtraction math exact (no double-subtraction),
encrypted thread replies caught by the timeline guard (m.relates_to is
cleartext), fresh prefs flow to handlers, single-owner wiring load-bearing.
Documented-acceptable: hasCurrentUserParticipated can lag until the server
bundle refreshes after your first reply; dedupe maps grow per-session only.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Default = Participating: thread replies notify only when you've posted in the
thread or are @mentioned; per-thread override to All / Mentions-only / Mute via
a bell menu in the thread panel header. Modes sync across devices in
io.lotus.thread_notifications account data (pruned on write: left rooms, >180d,
cap 200/room). Muted threads: no notifications/sounds, chip badge suppressed
(+BellMute glyph), and their counts are subtracted from the room's sidebar
badge (client-side; clamped ≥0).
Also fixes the thread notification path itself: thread replies are now owned by
exactly ONE handler (room-level ThreadEvent.NewReply via a new useRoomsListener
hook, with per-thread dedupe, panel-aware focus suppression, and per-thread OS
tag coalescing) — the existing RoomEvent.Timeline handlers in the notifier and
the unread binder are explicitly thread-guarded, eliminating the previously
un-gated/double path. Room badges now also refresh live on
RoomEvent.UnreadNotifications (surgical per-room PUT; fixes thread-badge lag).
Pure decision core shouldNotifyThreadReply (13-case matrix) + prune + unread
subtraction: +32 tests (648 total). E2EE caveat documented: mentions-only may
under-notify pre-decryption (same class as the existing path).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two-reviewer audit of the thread stack; confirmed findings fixed:
- ThreadTimeline: wrap encrypted events in EncryptedContent so a live-arriving
E2EE reply re-renders when its key decrypts (decryption emits neither
RoomEvent.Timeline nor ThreadEvent.Update — previously stuck at "Unable to
decrypt").
- ThreadPanel: mark-read deduped on the latest event id (RoomEvent.Timeline
re-emits per backfilled event/edit/reaction; previously up to N receipt POSTs
per panel open) + rejection handled with retry.
- RoomTimeline: ThreadSummary chips now mount only for events carrying thread
data (each chip holds a room-level listener; one per rendered message would
blow the SDK's 100-listener emitter cap) with a single room-level
ThreadEvent.New tick for new-thread liveness.
- useThreadPendingEvents: keep a sent reply visible through the /send-response→
/sync window (was flashing out of the pending strip before landing).
- ThreadTimeline: reseed the window on RoomEvent.TimelineReset (gappy sync left
a detached timeline).
Documented-acceptable (reviewer-noted): thread typing shows as room typing (no
per-thread typing in the spec; Element matches), thread panel + members drawer
can be open together, scheduled-send is thread-unaware but unreachable there.
Gates: tsc clean, eslint 0 errors, build OK, 616/617 tests (1 IDB skip).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Right-side thread drawer (MembersDrawer pattern; mobile fullscreen):
- ThreadPanel: header + close/Escape, ThreadTimeline, its own RoomInput
(threadRootId prop; drafts/replies/uploads isolated per roomId::threadId;
schedule + slash-commands off in threads v1) and threaded mark-as-read.
- ThreadTimeline: lean reimplementation over thread.liveTimeline — copied
useTimelinePagination pattern (/relations back-pagination + decryption),
virtualized, root event emphasized + "N replies" divider, reactions/edits/
redactions, and a pending strip (chronological local echo never enters the
thread timelineSet — rendered from LocalEchoUpdated instead).
- ThreadSummary chips on root messages (server-aggregated bundle or live
Thread; unread badge via getThreadUnreadNotificationCount) keep threads
discoverable now that replies leave the main timeline.
- Reply-in-Thread menu + thread indicators open the panel; deep links to
thread events redirect into it.
- State: roomIdToActiveThreadIdAtomFamily + getThreadDraftKey (+18 tests).
Gates: tsc clean, eslint 0 errors, build OK, 616/617 tests (1 IDB skip).
Awaiting live QA; release note: threaded replies no longer render inline.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
threadSupport:true makes matrix-js-sdk partition m.thread relations into Thread
objects (replies leave the main timeline; roots stay). markAsRead now sends
UNTHREADED receipts so one receipt still clears room + thread notification
counts — without this, badges would stick unread. The thread panel + summary
chips land in the same push.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TODO: P4-7 already-implemented [x]; P4-6 mozilla test enablement verified live;
Audit-3 researched → deferred tracking MSC4427 (banner_url proposal, unmerged);
P3-8 Thread Panel now carries the complete SDK-evidence-backed build plan
(threadSupport side effects, local-echo gap, receipt fix, 4-agent partition) —
ready for its own session. BUGS: N127 removed, Big #5 (backgrounds/seasonal)
done, CDN env-var closed (VITE_DECORATION_CDN exists), test count updated, KE
section points at the new investigation kit.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
LOTUS_E2EE_INVESTIGATION.md: per-KE capture runbook (console signatures, synapse
log greps + SQL against the documented LXC deployment, the KE-1⇒KE-2 causality
decision tree, ranked remediations incl. what a crypto-store reset wipes; SDK
finding: stable 41.6.0 has no OTK fix over our RC pin). Client: capture-only
console ring buffer (cryptoDiagLog, KE-signature-matched, max 200) + a Crypto
Diagnostics card in Developer Tools with a download-report button. ClientRoot
installs the capture hook at module load and mounts useSessionSync (cross-tab
sessions, prior commit).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Session now persists as ONE atomic cinny_session_v1 JSON write (blob-wins read,
transparent migration from the ~10 legacy keys, dual-write kept one release for
rollback). subscribeSessionChanges + useSessionSync reload a tab whose session
was changed/removed by another tab (logout/login/token rotation). OIDC refresher
already routes through setFallbackSession, so rotations stay atomic. Tests 7→22.
Full token-protection redesign remains tracked in LOTUS_BUGS.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Raw-IndexedDB cache (lotus-search-cache: messages keyed [roomId,eventId] +
per-room coverage) merged into local search with in-memory-wins dedupe. OPT-IN
(default off) via a standalone atom — stores decrypted text at rest, so it ships
with a privacy note, a Clear button, and an unconditional wipe on logout
(initMatrix). All IDB errors degrade to cache-miss. +8 tests (1 IDB skip in node).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
jared