Until now a send that failed (offline, homeserver down, a blip) went
straight to "Failed to send": nothing retried it, and a reload dropped it
without trace (chronological pending ordering keeps local echoes in memory
only).
- Outbox (utils/outbox.ts + features/outbox/OutboxFeature): own message
sends (text, stickers, reactions, polls; not call signalling or
redactions) are mirrored to localStorage from their first local echo until
the server confirms them or the user cancels.
- After a reload they come back as local echoes via room.addPendingEvent,
same shape as the SDK's own. Recent ones (< 1 h) are sent again with the
same txnId; older ones come back as "Failed to send" for the user to
retry or cancel. Ones the server already has (transaction id seen in
/sync) are dropped, so no duplicates.
- Retries: network failures (ConnectionError, 408/429/5xx) are re-sent when
the connection returns (sync recovers or the browser goes back online),
and after a blip while online (5 s, backing off, max 10 per message).
Oldest first, in order per room. 4xx / consent / encryption failures are
left to the user.
- UI: a network failure while offline shows a clock, "Queued. Will send
when you're back online" (thread view too), not the red ✕. The ✕ is now
a button: click to retry.
- Logout wipes the outbox with the other plaintext caches (the content is
decrypted, like drafts).
Tested end to end against a local Synapse (Chromium): offline → queued →
sent once on reconnect; homeserver unreachable → queued → sent once; failed
send → reload → sent once and shown once; server accepted but response lost
→ reload → no duplicate; 2 h old entry → failed, not sent, click ✕ → sent;
cancel → gone after reload; encrypted room → restored message goes out as
m.room.encrypted with no plaintext and decrypts; one-off failure retried by
itself in ~5 s; no page errors. Unit tests for the pure parts; Playwright
20 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Lotus permalinks (#130) were built from window.location.origin. In the
desktop app that's the local tauri-plugin-localhost server (hash-routed), so
"Copy Lotus Link" copied e.g. http://localhost:…/#/home/!room…, which works
for nobody else. And the link recogniser only knew that local base, so a real
https://chat.lotusguild.org/home/… link in a message opened the browser
instead of the room.
- useLotusShareBase: in the desktop app, links for other people use config
`webAppUrl` (https only, set by cinny-desktop #23) in web path-routing
form; otherwise the origin, as before. Used by "Copy Lotus Link" on
messages, the space menu and space tabs.
- The recogniser accepts several bases: the origin, plus `webAppUrl` in the
desktop app.
- The web app is unchanged.
Verified with a simulated desktop (Tauri bridge + webAppUrl) against a local
Synapse. Copied links are https://chat.lotusguild.org/home/<room>/<event> and
https://chat.lotusguild.org/<space>. A public link in a message renders as
the room pill and clicking it opens the room in-app, with nothing sent to the
system browser. The web app still copies origin links. Unit tests for the
base selection; Playwright 20 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The Linux desktop app runs on WebKitGTK, which ships without WebRTC (2.52
has no RTCPeerConnection; 2.54 disables it outright pending a libwebrtc
backend around 2.56), so calls can't work there. Until now the call button
just disappeared, the call room said "Your browser does not support WebRTC"
with Join disabled, and an incoming call couldn't be answered.
In the desktop app (isTauri) without WebRTC:
- call rooms: "Calls aren't available in the desktop app on Linux yet: its
web engine has no WebRTC" + an "Open in browser" button;
- incoming-call overlay: the same, with "Answer in browser";
- room header: the call button stays, and opens the room in the browser.
The link is the room in the web app (config.json `webAppUrl`, https only,
new key); the user presses Join there. Deliberately not an auto-join link:
a crafted URL must not be able to join a call and open someone's mic. It
opens through the desktop's new-window handler (web/mail schemes only → the
system browser). Without `webAppUrl` the explanation shows with no button;
browsers without WebRTC keep the old message.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
matrix-widget-api's host transport handled a message from ANY window on the
page as long as it carried the widget's id; its strictOriginCheck only
compares with the host's own origin and is off by default. The call's id is
the fixed 'call-embed', so any other frame (a room widget, a URL-preview
embed) could post fromWidget actions as the call. Reproduced locally: an
opaque-origin frame posting one io.lotus.hotkey keydown for the PTT key
turned a push-to-talk user's mic on ("● Live").
restrictWidgetMessages() swaps each ClientWidgetApi transport's listener
for one that requires ev.source === the widget iframe's window and
ev.origin === the widget's origin. Applied to the call and to room widgets
(so one widget can't impersonate another). Verified: the spoof no longer
opens the mic; PTT/deafen from inside the call, screenshare, speaking
indicator and room widgets (capability prompt, send, live events) unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
markAsRead ran on every room visit (timeline at the bottom and focused) and
sent a threaded receipt for every unread thread, so a reply in a thread you
started or replied in lost its unread badge the moment you glanced at the
room, without opening the thread.
Reads from just viewing the timeline are now "passive":
- threads you follow (started, replied in, or were mentioned in) stay unread
until their panel is opened; other threads are still cleared so they don't
keep the room dot lit forever;
- while a followed thread has an unread reply, the main receipt is scoped to
the main timeline instead of unthreaded, because an unthreaded receipt
also reads every older thread reply (the next main message would clear the
thread anyway). The check also asks whether the latest reply is read, since
the thread's count lags when the reply and a main message share a sync;
- the thread open in the panel is skipped, as the panel sends its own
receipt (was two identical receipts per reply).
Explicit "mark as read" (room menu, Escape, bulk actions) still clears
everything. Unit tests for each rule plus a local-homeserver e2e that checks
the server's per-thread count survives a reply + newer main message.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Thread images opened the room media lightbox, which looks the event up in
the room's detached media timeline. Thread replies never reach that
timeline, so a thread image always showed "1 / 1" with no prev/next, after
paging the room's media up to six times. The thread panel now builds the
viewer's items from its own root + loaded replies, and "Go to message"
scrolls the thread panel instead of the room.
The viewer gains a "Copy image" button: fetches the displayed media (blob
URL for E2EE, authenticated URL otherwise), re-encodes to PNG when needed,
and writes it via ClipboardItem with a promise so Safari keeps the click's
user activation. Hidden where ClipboardItem is missing. No "open in new
tab": an E2EE blob URL is revoked when the viewer closes and authenticated
media 401s in a bare tab.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The reaction button's aria-label used getShortcodeFor(), which returns
undefined until the lazily loaded emoji data arrives. The same button read
"🎉 reaction, 1 person" on first render and "tada reaction, 2 people" after
any later re-render. It now always uses the emoji itself (screen readers
speak it by its proper name, e.g. "party popper").
Custom (mxc) emoji were labelled just "custom emoji"; they now use the
shortcode carried on the reaction event (":lotus_blob: reaction").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
With the mention list open, pressing Enter sent the half-typed text
("hey @bo") instead of inserting the highlighted person — found while
investigating the "composer hit a snag" report; only Tab picked. Enter now
picks the top suggestion for people, rooms and commands, as in Discord,
Slack and Element.
useAutocompleteEnter listens in the window capture phase (so it runs before
the composer's submit handler) and only while the list has suggestions —
with no command match Enter still sends. Emoji suggestions are deliberately
left on Tab only, so "lol :p" + Enter still sends rather than inserting an
emoji. IME composition and Shift/Ctrl+Enter are untouched. Tab now inserts
the member's display name, same as a click.
Verified: "@bo" + Enter inserts the mention and sends nothing; the next
Enter sends "hey @bob hi" with m.mentions; ":smi" + Enter still sends.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Reported on chat.lotusguild.org: picking a name from the @-mention list
showed "The message composer hit a snag." — clicking OK continued with the
draft intact. The composer's known failure mode is a transient render error
right after an autocomplete insert (slate-react's DOM selection sync racing
the model; 477df4ae fixed one such path). It did not reproduce locally in 60
attempts (click / Enter / Tab, fast and slow, display names with spaces and
emoji, Chromium and WebKit), so this makes it harmless and diagnosable:
- ComposerErrorBoundary: the first failure clears the selection and remounts
the composer immediately, with no notice (the draft is intact). A second
failure within 5 s shows the notice, now saying the draft is safe, with
"Reload composer" and "Copy details" (time, error, stack, component stack,
browser). Every failure is logged to the console as "[composer] render
error".
- Used for the room composer and, newly, the thread panel composer (which had
no boundary, so the same error took the whole panel down).
Verified with a temporary injected crash (not committed): one crash recovers
with the draft kept and no notice; two within 5 s show the notice, Copy
details copies the report, Reload composer restores the full draft.
Chromium e2e 19 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
When a multi-line paste looks like code, a chip above the composer asks
"That looks like code. Format it as a code block (js)?" with Format as code /
No thanks. It never converts on its own; typing, No, or 8 s dismiss it, and
Settings → Editor → "Offer to Format Pasted Code" turns it off.
- utils/looksLikeCode.ts (pure, tested): ≥ 3 lines and ≥ 2 of indentation
with depth changes, statement terminators, operators + brackets, keywords,
monospace/<pre> clipboard HTML, or SQL clause lines. Prose guards: quoted
replies, URL lists, long sentence-punctuated lines, plain-word lines,
markdown lists. Language guess only when fairly sure (js/ts/python/sql/
rust/c/php).
- Accept rebuilds the pasted paragraphs as one code block from their plain
text (code lines hold text only) and leaves the caret after it.
- Code blocks carry an optional lang → <code class="language-js"> (whitelisted
identifier only).
Verified in Chromium: pasting a JS function shows the chip; Format as code →
sent formatted_body is <pre><code class="language-js"> with indentation intact;
a three-line prose paste shows no chip; typing after a paste dismisses it.
Unit tests: 14 detector fixtures + 2 output tests; chromium e2e 19 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The SDK gives a thread root its own id as threadRootId, so the
`!isThreadedMessage` guard hid the thread action on exactly the message that
has a thread; the "N replies" chip was the only way in. Roots now get the
action in both the hover bar and the ⋮ menu, labelled "View thread" /
"View Thread (14 replies)" with the thread icon; other messages keep
"Reply in Thread". Same handler, which opens the thread panel.
Verified in Chromium: on a 14-reply root both entries show and open the panel;
an ordinary message still shows "Reply in thread".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
A row in the Threads list only opened the thread panel; there was no way to
get to the root message in the room's timeline. Each row now has a small
"Go to message" button (a sibling of the row button, laid over its corner —
no nested buttons) that navigates the room to the root event. On a phone,
where the list covers the timeline, it also closes the list, and the button
is the bigger touch size.
Verified in Chromium: desktop jumps to the root with the list still open;
Pixel 7 jumps and closes the list (34 px target).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The subtitle was only the room name. It now reads e.g. "Thread Lab · started
by alice, Fri 09:18 PM" (the user's clock/date preferences via the #139
formatter), which is what you want to know when arriving from the Threads
list. Verified in the 360 px desktop panel: fits without clipping.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
A captioned image/video (MSC2530: `filename` set and `body` differs) put the
caption in the header, truncated to one line, so a long caption was
unreadable. The header now shows the file name; the caption is shown in full
under the media (wrapped, scrolls past 25vh). The image alt text uses the
caption, and Download now saves under the real file name instead of the
caption.
Verified in Chromium: header "lake-sunset.png", the 150-character caption
fully visible below the image.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The viewer's arrows were 32 px at the screen edges on a phone, under the
44 px target the rest of the mobile work uses. On a coarse pointer they use
the next IconButton size (48 px measured on a Pixel 7); desktop is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
- sendRoomMessage (composer's fire-and-forget sends: text, location, voice,
files, GIFs) and sendRoomEvent (polls, poll responses/ends, forwards,
reactions, edits) in utils/room.ts carry the one cast each needs
(`keyof TimelineEvents` / `RoomMessageEventContent`, no `any`).
- sendRoomMessage also swallows the rejected promise: a failed send already
shows on the local echo (Failed to send + Retry, or the consent prompt), so
it no longer surfaces as an unhandled error in the console.
- getAccountData narrows to `keyof AccountDataEvents`; ForwardMessageDialog's
guard now narrows `contentToSend` itself (same behaviour).
- `as any` 39 → 23; eslint warnings 46 → 36, ratchet tightened to 36.
Verified in Chromium on a local Synapse: a text message, a quick reaction and
an edit all reach the server with the right content; a consent-blocked send
no longer logs an unhandled MatrixError. 1219 unit tests pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Zoom on a phone was only the +/- buttons; the pinch people try first did
nothing. usePinchZoom tracks a two-finger gesture on the viewer's media area
and scales from the zoom it started at (clamped 1×–5×, snapping back to
exactly 1× when released near it, so one-finger swipe navigation re-arms).
The area gets touch-action: none so the browser doesn't zoom the page
instead. One-finger swipe and pan already ignore multi-touch.
Verified on an emulated Pixel 7 with CDP two-point touch: spread 80→200 px
gives 250 %, closing to 120 px gives 150 %, closing fully returns to 100 %,
and a one-finger swipe afterwards still moves 2/2 → 1/2.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Room widgets were limited to display-only capabilities because the driver
couldn't serve anything else. Now:
- classifyWidgetCapabilities: display caps are still granted silently; reading
or sending events/state in the widget's OWN room is offered to the user;
everything else (other rooms' timelines, to-device, account data, uploads,
user directory, delayed events) stays denied. Writing protected state
(power levels, join rules, encryption, membership, ACLs, widgets, …) is
never offered, and the driver refuses it again at send time.
- WidgetPermissionPrompt names the widget, the host that runs it and who added
it; each request in plain words ("Send messages of type m.text in this
room · as you"); reading is pre-ticked, sending is not; Deny / Escape grant
nothing extra. "Remember my choice" stores it per viewer (localStorage),
tied to the widget URL, so a changed URL asks again.
- GeneralWidgetDriver implements sendEvent / readRoomTimeline / readRoomState
/ readEventRelations, each refusing any room but the widget's own;
RoomWidgetView feeds the room's live (decrypted) events and state updates,
which ClientWidgetApi forwards only if the widget holds the capability.
Verified in Chromium with a cross-origin test widget against a local Synapse:
power-levels and timeline:* requests are never shown; after allowing
send+read the widget's message lands on the server, its power-levels write
is rejected, it receives live messages, and after a reload the remembered
choice skips the prompt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
16 menus (room header, Home/Direct/Space sidebar tabs, create tab, lobby,
room-nav item, history visibility, client root) set
returnFocusOnDeactivate: false, so pressing Escape dropped focus to
<body> and a keyboard user had to start over from the top of the page.
That option was there so focus wouldn't be pulled back to the menu button
when an item opens a dialog; since ce8ed89f dialogs move focus into
themselves (their initial focus runs after the menu's return), so it's no
longer needed. The autocomplete menu (focus must stay in the editor) and
the two menus whose opener disappears are left as they were.
Checked: room header menu → Escape → focus on "More options" (was
<body>); menu → Leave Room → focus inside the Leave Room dialog on
Cancel, same as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Follow-up to ce8ed89f for the modals that have no visible heading to
point at: the three "Seen by" reader lists (read-receipt pill, "is
following", message menu), View source, the text/PDF file viewers, the
room-card join error and the user-profile modal get role="dialog",
aria-modal and an aria-label; their traps move focus in (fallbackFocus
on the dialog) where the trap is local.
Verified: the receipt pill opens a "Seen by" dialog with focus on
Close; Escape closes it and focus returns to the pill.
The remaining unnamed Modal/Dialog uses are startup/loading and error
screens (config, feature check, spec versions, client root, password
reset) and wrappers around components that carry their own role (image
viewer).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Most modals rendered a folds Dialog/Modal with no role and no name, and
their focus traps used `initialFocus: false`, so focus stayed behind the
modal and a screen reader never announced it.
- 32 dialogs with a visible heading: role="dialog", aria-modal,
aria-labelledby → the heading (given an id), tabIndex=-1.
- 4 dialogs that already had a name (Leave Room, room topic viewer,
server ACL, room-nav prompt): role + aria-modal.
- Their focus traps drop `initialFocus: false` for focus-trap's default
(keep an already-focused autoFocus field, else the first tabbable
element) with the dialog itself as fallbackFocus, so a dialog without
a tabbable node can't crash the trap. Traps that live in a parent
(UIA stages, Logout, Forward, Invite) get the semantics only.
- The file drop overlay is deliberately left alone (not a dialog).
Checked at runtime: Join with Address, Delete Message, Report Message,
Leave Room and Logout open as named dialogs with focus inside and close
with Escape (Tab first when a text field has focus — the shared
stopPropagation keeps Escape from discarding typed text, by design).
The axe e2e spec (6 tests) passes; eslint warnings unchanged (46).
17 modals with no heading (image/file viewers, loading screens) remain.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The "N replies · <time>" chip under a thread root was named only by its
visible text, so a screen reader never said it opens the thread, and the
unread dot / muted bell were visual-only. Its accessible name is now the
visible text first (WCAG 2.5.3) plus "view thread", "unread replies" and
"muted" as applicable, e.g. "1 reply · 12:16 PM, view thread".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
- Skip link: following #main-content left focus on <body> (the <main>
target wasn't focusable), so nothing was announced and the URL got a
stray fragment. <main> is now tabIndex=-1 and the link focuses it.
- Keyboard-shortcuts dialog (?): no role, and focus stayed in the
timeline, so it opened silently. Now role="dialog" aria-modal, and focus
moves into it (Escape still returns focus to where you were).
- Reaction viewer (both the reaction context-menu path and "View
Reactions"): same — role="dialog" aria-modal aria-label="Reactions",
focus moves in.
Keyboard-only checks (Playwright): Tab → skip link → Enter focuses
<main>; Tab reaches the room list, Enter opens a room, typing lands in the
composer, Enter sends (verified on the server); focus ring visible. "?"
opens the dialog with focus inside, Escape returns to the same element,
"?" in the composer stays text. Topic viewer, Search dialog and reaction
viewer all return focus to their opener.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
RoomView focuses the composer on any "typing" key pressed while nothing
editable has focus, and Escape counted as typing. Closing the GIF picker
with Esc (its focus trap returns focus to the GIF button) therefore
landed in the composer instead, and so did Esc on any other room control.
Escape, CapsLock, Insert, ContextMenu, PrintScreen and Pause are now
excluded like the other non-typing keys.
Verified at 1300 px and 320 px: Esc closes the picker and focus is back
on "Insert GIF"; the 320 px picker fits with no page overflow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Removed every `as any` in src/, then restored only the ones tsc still
needs: 163 → 36. 112 were in lotus-terminal.css.ts (`'… !important' as
any` etc. — vanilla-extract's style types accept these strings as-is),
so its file-wide eslint-disable goes too. The casts were type-only, so
emitted code and generated CSS are unchanged.
eslint warnings 49 → 46; the ratchet is tightened to match.
tsc clean; 1202 unit tests pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
A "Mark all read" chip appears in the Threads panel header whenever any
thread in the room has unread replies, and sends one threaded receipt per
unread thread (never the root — a root lives in the main timeline and a
receipt there would drag the MAIN read marker backwards, the P6
regression). Honours the private-receipt settings.
Verified: two threads with unread replies → the chip appears, the rows
say "unread" in their labels; after clicking, no unread rows and the chip
is gone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Touch had no way to move between images except the small arrow buttons.
A one-finger horizontal swipe now goes next/previous, and is inert while
the image is zoomed in (where the same gesture pans). The classifier is
pure and unit-tested: ≥ 60 px horizontal, not mostly vertical, under
800 ms, single finger.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
utils/haptics.ts: tick('ptt-on' | 'ptt-off' | 'reaction') → 10/10/8 ms
navigator.vibrate, a no-op without the API (iOS), when the system prefers
reduced motion, or when the new Settings → Calls "Haptic Feedback" switch
(default on, only rendered where the API exists) is off. PTT is observed
once through pttActiveAtom so the keyboard, global-hotkey and on-screen
paths all tick; reactions tick where the reaction event is sent in the
room and thread timelines (quick bar, hover bar, sheet and emoji board
all funnel there).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Verified first: after a paste the focus stayed in the composer and the
caption needed a click. Now, when exactly one image is pasted or dropped
into an empty composer, its upload card's caption input takes focus;
Enter there sends the board (with any composer text) and Escape returns
focus to the composer. Multi-file drops and non-empty composers are
unchanged. The target card is matched by file name + mtime because the
metadata strip re-wraps the File.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
A reply to an image, video or sticker used to quote just the filename.
The quote (timeline) and the composer's reply-draft preview now carry a
36 px thumbnail from the event's own thumbnail, decrypted for E2EE media
via the same hook the gallery uses — never the full-size file. Clicking
still jumps to the original.
useDecryptedMediaUrl and getThumbMxc moved out of MediaGallery into
hooks/ and utils/ so components/message can use them without a cycle.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Metadata was only dropped as a side effect of opt-in compression, so a
phone photo carried its GPS fix, camera model and timestamp into the room
and the media store. utils/stripImageMetadata.ts now removes it at the
container level, without touching pixels: JPEG drops APP1/APP13/COM
(writing back a minimal EXIF holding only Orientation when it isn't 1,
so sideways-stored photos still display upright), PNG drops eXIf and the
text chunks XMP lives in, WebP drops EXIF/XMP and clears the VP8X flags.
Other types pass through.
Applied before encryption on every composer path (attach, paste, drop,
share target) and to user/room avatar picks; GIF upload is excluded.
Setting → General → Privacy "Remove Photo Metadata Before Sending",
default on. The upload card says "Photo metadata removed".
Unit tests on generated fixtures with a GPS IFD (JPEG orientation 6,
JPEG + comment, PNG with eXIf + XMP, WebP with EXIF); verified end to
end: the bytes stored by Synapse decode fine and carry only Orientation.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Event handlers typed as React.MouseEvent, join errors as Error, the
service-worker session setter as strings, the UIA policy map and the
webkitAudioContext probe as narrow shapes. No behaviour change.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
public/manifest.json declares share_target (POST multipart to
/share-target: title/text/url + image/video/audio/pdf/text files). The
service worker answers that POST itself: it stashes the form in a Cache
API bucket and 303s to the in-app /share page, which lists what arrived,
offers a room search, and on pick writes the files into that room's
upload-board atom (encrypting first for E2EE rooms via the composer's
shared filesToUploadItems) and the title/text/url into its draft, then
opens the room — the user still presses Send. The stash is cleared once
placed; reopening /share afterwards says so.
nginx/caddy examples and the prod image config gain a 303 for
/share-target so a POST that reaches the origin before the worker
controls the page lands on /share instead of a 405. iOS has no share
target support and ignores the manifest entry.
Verified headless against the built preview: SW-controlled page → POST
/share-target (two PNGs + title + text) → /share lists both files and the
text → pick the DM → composer shows both files on the upload board and
the text in the draft → /share reports nothing pending.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
e2e/a11y.spec.ts runs @axe-core/playwright (WCAG 2.x A/AA) over the login
page, room timeline + composer, message options menu, thread panel, user
settings and room settings, failing on critical/serious findings other
than colour contrast (reported, not gated: generated avatar colours and
portal false positives). Aria snapshots of the composer, message menu,
thread panel and settings nav catch lost names/roles/live regions.
Burned down what the first run found:
- NavItem: callers' aria-selected is not valid on a div (axe critical);
it now drives data-selected for styling and aria-current="page".
- Composer placeholder at 0.5 opacity was ~2.3:1; now P300.
- Voice-limit and explore custom-limit number inputs had no label.
- Thread panel is an <aside aria-label="Thread">; the settings modal is a
role=dialog; the settings sections are a <nav>; the message action
menu carries data-message-menu + a label.
Also allows WebKit's CI wording for the well-known probe ("Could not
connect … Connection refused") that failed run #2003.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Client-side only: every file is still its own standard m.image/m.video
event, so Element and friends keep seeing N plain images. In Lotus a run
of media from one sender — contiguous, ≤ 60 s apart, no reply/thread/edit
relation, up to 10 — renders once, at its last event, as a 2–4 column
grid of square thumbnails (blurhash placeholder, video play badge,
tap-to-load when media auto-load is off). A member with reactions or a
thread closes its group so those stay visible under the rendered event.
Tapping a tile opens the lightbox on just that group in send order
(←/→, zoom, download, jump). "Show separately" splits a group back into
individual messages for the session; "Show as gallery" undoes it.
Planning is lazy per render pass (utils/mediaGroups.ts, unit-tested):
the first media event met plans its whole run in both directions, so a
virtual window that starts mid-run agrees with one that starts before it.
Verified: 5 files dropped at once in an encrypted room — both sender and
recipient see one 5-tile grid with decrypted thumbnails; desktop + phone;
a reaction on photo 3 yields [1–3]+👍 and [4–5].
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Audit of every rendered time found five families of ad-hoc formatting:
the shared Time component + copies of its today/yesterday branch
(forwarded header, thread summary, read receipts, device tile, moderation
alerts, edit history), locale-default toLocale*String calls that ignored
the user's 12/24 h and date-format settings (scheduled tray, reminders,
schedule preview, notification snooze, bookmarks, threads list, search
cache line, room insights, media gallery), a hard-coded en-US date in the
activity log, and three relative-age variants.
utils/formatTimestamp.ts now holds the rules — today → time; yesterday /
tomorrow → day word + time; last 6 days → weekday + time; older → date +
time in dateFormatString — plus autoDate / time / date / dateTime styles,
formatDayDivider (full weekday), formatShortAge (room list) and
formatRelativeAge (list rows). useTimestampFormatter binds them to the
settings. 11 unit tests with an injected 'now'.
Visible changes are limited to consistency: 12 h times keep the existing
zero-padded hh:mm A; the a11y label and Created-by line use the user's
date format instead of a fixed long month.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Fat-finger guard for phones. On coarse-pointer devices the first tap on a
sticker (emoji board Sticker tab) or a GIF parks it in a small bar with a
thumbnail, name, Send and Cancel; a second tap on the same item or Send
sends it; tapping a different item switches the preview; tapping empty
picker space or Cancel clears it. Mouse clicks, keyboard activation and
screen-reader activation (bare click without touch events) still send in
one step, and the emoji tab is untouched.
useRecentTouch records touches inside the picker at the document level so
it survives the Sticker tab remount, and reports false without a coarse
pointer.
Also fixes the compact composer's GIF picker opening mostly off-screen:
end-aligning the 312px popout to a button near the left edge of the
overflow row pushed it to x≈-95; it now anchors to the row itself.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Many distinct reactions used to wrap into a tall stack (16 reactions =
4 rows at phone width). Now only the first row is shown, ending in a
"+N" chip; tapping it expands inline and a "less" chip collapses again.
Expanded state is remembered per message for the session. Nothing changes
when the reactions already fit on one row.
Overflowing chips stay in the DOM (invisible, aria-hidden, untabbable,
clipped by max-height) so the container keeps its natural width — which
keeps shrink-to-fit bubble layout stable — and each chip stays measurable.
utils/reactionOverflow.ts holds the unit-tested fit calculation; a
ResizeObserver re-fits on width changes. "+N" is forced LTR for RTL UIs.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Hovering or focusing "(edited)" shows a tooltip with only the most recent
edit as a word diff — removed words struck, added words bold — plus a
+N/−N summary. Clicking still opens the full history viewer. On touch, a
long-press on the label shows the same diff as a popout (a plain tap opens
the viewer; the message's own long-press action sheet is not triggered).
utils/wordDiff.ts is a unit-tested LCS over words that ignores whitespace-
only changes and gives up past 400 words. Only plain-text bodies are
diffed; formatted edits fall back to the viewer.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Clicking a reply quote (or arriving via a thread-panel / search / permalink
jump, i.e. an eventId route change) remembers the message that was in the
middle of the viewport and shows a Primary chip next to Jump to Latest —
'Back to where you were' — which scrolls/navigates back to it. Nothing is
rendered normally; the chip expires after 20 s or once used. Verified headless:
reply quote → original 120 messages up is shown + chip; chip → back at the
reply, chip gone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
One small Surface chip next to the header controls, only when a rule exists:
'0/10' (participants / io.lotus.voice_limit) and/or a muted-camera glyph when
the room's io.lotus.room_quality forbids camera and screenshare; the tooltip
and aria-label carry the words ('Voice limit 10 (0 in call) · Audio-only
room'). Desktop only — the mobile header is already tight. Verified headless:
appears with the rules set, disappears when they are cleared.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
On coarse-pointer devices the hover action bar is unreachable and iOS Safari
has no contextmenu on long-press. useLongPress (450 ms, < 10 px movement,
cancelled by scroll/lift) now opens an ActionSheet — slides up from the
bottom, drag handle, swipe-down or backdrop tap to dismiss — with the quick
reactions row + 'Add reaction' on top and the same folds Menu the desktop
popout uses (extracted to a shared menuJSX) beneath. Android's contextmenu
echo of the same press is dropped; the press-induced text/image selection is
cleared; the finger-lift's synthetic click is swallowed so it can't hit a menu
item; the sheet ignores 'outside' clicks for its first 600 ms. Desktop is
unchanged (right-click → anchored popout).
Verified with Playwright touch emulation (Pixel 7, CDP touch events):
long-press → sheet with 👍 / Add Reaction / Reply / Forward / Bookmark /
Remind Me / Reply in Thread / Copy Link / Pin / Delete; swipe-down dismisses;
a scroll gesture does not open it; tapping Reply closes it and arms the
composer. Still wants a real iPhone + Android pass before it's called done.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The ring event (m.rtc.notification) now renders — only once the ring is over
(its lifetime lapsed or the caller hung up) and only if we did not join between
the ring and the hang-up — as a red 'Missed call from bob' line with the ring's
time, in the same style as the call-membership summaries. Derived entirely from
events already in the timeline; nothing is sent; a call we answered shows
nothing extra, and a later separate call cannot retro-actively 'answer' an
earlier missed one.
Verified headless in a DM: ring ignored → line appears once bob hangs up; a
second call that alice answers adds no line and leaves the first one in place.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
'Unable to decrypt message' now carries one sentence per matrix-js-sdk
DecryptionFailureCode (describeDecryptionFailure, unit-tested against every
code so no raw code can leak into the copy) and, where something fixes it,
one button: no key backup → 'Set up key backup'; backup exists but this
session can't open it / key withheld for an unverified session → 'Unlock key
backup' / 'Verify this session' (both open Settings → Devices via a new
settingsRequestAtom that SettingsTab consumes); backup working or unknown
session (rust-crypto re-requests keys itself) → 'Retry', which re-runs
decryptEventIfNeeded. Sender-side problems are plain text. The raw code sits
in the placeholder's tooltip for support.
Verified headless on a fresh session in the encrypted seed room: each event
shows 'Sent before you signed in here, and no key backup exists…' with
tooltip HISTORICAL_MESSAGE_NO_KEY_BACKUP; the button opens Settings → Devices.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Clicking an image in the room or thread timeline now opens the same viewer
the media gallery uses — dark backdrop, sender + date, 1/N counter, ←/→ across
the room's media, +/-/0 and wheel/double-click zoom, download, jump to
message — positioned at the clicked event. RoomMediaLightbox feeds it from the
detached media timeline (#163); when the event isn't in the loaded window it
pages back (bounded, 6 pages) and shows the clicked image alone meanwhile, so
the viewer opens instantly. ImageContent gains onOpenViewer (RenderMessageContent
passes onOpenImageViewer); its built-in viewer remains for stickers, search
results, pins, notifications and avatars.
Verified headless: click → 'Media viewer' dialog focused, counter 1/2, '+' →
120 %, Jump to message present, Esc closes; an older image 10 pages up →
found at 10/45 after paging.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Two causes. (1) Three right-hand columns don't fit under ~1500 px: the member
drawer now yields to an open content panel (thread / gallery / widgets /
threads list) on desktops ≤ 1500 px and returns when it closes, like Element.
(2) The thread composer rendered the full ten-button toolbar inside a 360 px
panel, leaving 24 px for the input ('S…') at every width — RoomInput gains a
compactLayout prop (the phone one-row '+ | input | emoji | send' layout) and
ThreadPanel uses it.
Measured headless: 1400 px thread open → drawer hidden, composers 328/168 px
(was 422 → 'Send a mes…' + 24 px); 1600 px → drawer stays, thread composer 168 px.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Auto-scroll on a live event only ran while the at-bottom flag was true, so a
stale flag (in-place growth like a poll gaining a row briefly pushes the anchor
out of view; or simply having scrolled up) left your own just-sent message
below the fold behind 'Jump to Latest'. An own non-relation, non-state,
non-reaction event now always brings the live end into view: advance the
range + smooth scroll when the window is at the live end, otherwise reset to
the live timeline like the Jump to Latest button. Others' messages keep the
current behaviour.
Verified headless: scrolled up 600 px → Send → message visible, no Jump pill
(before: Jump pill, message off-screen).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
rateLimitedActions now collects non-429 failures (and a 429 that exhausted its
retries) and returns them instead of swallowing them; existing callers ignore
the return. /kick and /ban turn the list into a CommandError whose message
names who and why, using the server's own sentence (MatrixError.data.error),
never the URL-bearing toString(); RoomInput's toast shows it verbatim.
Verified headless as a non-moderator: '/kick @alice' → "Could not kick
@alice:localhost: You cannot kick user @alice:localhost."; '/ban @nobody
@alice' → "Could not ban @nobody:localhost, @alice:localhost: You don't have
permission to ban".
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Gallery lightbox: FocusTrap had initialFocus:false, so focus stayed on the
tile behind the overlay and ←/→/Esc/+/- were dead until the user clicked
inside. Timeline ImageViewer: no keyboard zoom, no role/aria-modal/label,
focus landed on the <img>; now mirrors the lightbox (+ = / - / 0,
double-click toggles 1×↔2×, role=dialog aria-modal labelled by the file
name, focuses on open).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Inside the thread panel each reply rendered a 'Thread ↩ <root author> <root
text…>' quote because the spec's fallback reply relation (is_falling_back
+ m.in_reply_to root) was treated like a real reply — noise on every row
when the root is already pinned at the top. Genuine reply-to-a-reply
quotes are kept. The Threads list and Widgets panel had desktop-only
header buttons and no way to open them on a phone; both are now in the
mobile ⋮ menu next to Members / Media Gallery.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Lotus CIandClaude Opus 5.5
jaredandClaude Opus 5