Two mirror-image bugs from voice/DM-call transitions leaving stale state.
DM call → space channel (stuck "Connecting…"):
The last participant to leave a DM call for a space channel receives a
`dm_call_ended` echo (server empties the DM room on their `voice_join`).
The handlers called `disconnectFn()` unconditionally, tearing down the
space room they had just connected to. Route `dm_call_ended` /
`dm_call_rejected` / terminal `dm_call_undeliverable` through a new
`teardownDmCall()` that only disconnects LiveKit when not in a space
channel (`currentVoiceChannelId` null).
Space channel → DM call (still shown as "in" the voice channel):
1. Entering a DM call never cleared `currentVoiceChannelId`, so
`VoiceChannel` mapped the DM call's live LiveKit participants onto the
old space channel. Add `clearSpaceVoiceForDmCall()`, called in
`connect()` when `isDm`, restoring the invariant that a DM call has no
`currentVoiceChannelId`.
2. `dm_call_accepted` gated the caller's connect on `!isLiveKitConnected`,
so a caller already in a space channel was never connected to the DM
room. Gate on `wasOutgoingCall` only (connect() de-dupes same-room).
Tests: teardownDmCall.test.ts, clearSpaceVoiceForDmCall.test.ts.
Docs: docs/systems/voice.md.
The Pi's Caddyfile carries extra vhost blocks (e.g. other-site.example.com) that
aren't in this repo; add it to the rsync excludes so deploys don't overwrite it.
- Project status: drop the test-instances mention and the self-contradictory
'no tagged public release yet' framing → 'early-stage and under active development'.
- Moderation: 'Bans with reason and audit trail' overstated it — there's no
append-only audit log and unban hard-deletes the record. Reworded to
'Bans with reason and moderator attribution (who, why, and when)' to match the code.
- Features list: 'read states' -> 'unread tracking (badges + unread divider)',
which is what the code actually does (no read receipts).
- Clarify the PWA bullet (add-to-home-screen standalone app) and the
mobile-responsive touch layout.
- Add a Mobile section: working mobile-browser/PWA today, native iOS/Android
app-store apps planned once traction and license funding are secured.
Replace the placeholder Screenshots section with a hero voice/video shot
and a curated 6-shot grid (chat, screen-share controls, spaces, federated
DMs, social discovery, federation admin), linking to a new
docs/screenshots.md gallery of all 14 shots grouped by feature.
Images are optimized to 1600px WebP (~1.2MB total vs ~45MB raw); raw
originals in demo-pictures/ are gitignored.
The flat app-icon.svg's gradient B mark has a bright (#fff) sheen that runs
to the badge perimeter with no dark separation. At favicon sizes (16/32px)
that edge anti-aliases into a white halo that reads as a border around the
icon — visible in Safari browser tabs, and the same defect in the small
Windows .ico / Linux launcher reps that also rendered from the flat SVG.
The committed 3D raster masters (used by every >=128px output already) frame
the mark in a dark surround and stay clean down to 16px. Set RASTER_THRESHOLD
0 so all app-icon sizes route through the raster path; the flat SVG is kept
as a gated source, re-enablable only with a corrected flat mark. Regenerated
favicons + small desktop reps; output remains byte-deterministic. Updated the
generator header/comments, README source matrix, and the dated icon spec.
The server's `playable` flag is computed Chromium-first, but HEVC
web-playability is browser-dependent: WebKit (Safari on macOS/iOS) decodes
HEVC via the OS while Chromium/Firefox/Electron can't. Treating the flag as
global wrongly showed Safari users the download fallback for files they can
play inline.
The client now treats `playable === false` as "needs a capability check": it
pre-renders the fallback only when the current browser also can't decode the
format, gated on a one-time canPlayType probe (BROWSER_SUPPORTS_HEVC). Capable
browsers attempt inline playback; the <video> onError handler remains the
safety net for genuine failures.
macOS screen recordings are HEVC inside a .mov container, which Chromium,
Firefox and stock Electron can't decode. The file uploaded fine and a
server-side ffmpeg poster was generated, but inline <video> playback failed
silently — stuck at 0:00 with no error, since AttachmentRenderer had no error
handling. Root cause: the system had no concept of web-playability.
Server detects, client degrades:
- mediaPlayable.ts: classifyVideoPlayable(mimetype, codec) — tri-state
(false = known-undecodable e.g. HEVC/ProRes, true = web codec in web
container, null = unknown/optimistic). Never widens `false` beyond codecs
that fail everywhere, so ffmpeg-less instances keep prior behaviour.
- probeMediaMeta now captures the video codec_name; the upload finish hook
stores the verdict in the new attachments.playable column (migration 0007).
- Flag propagated through every serializer: space messages, DMs, WS, and
federation relay (outbound + inbound) — federation-compatible.
- VideoAttachment component: playable===false renders a download card (poster
+ "Can't play here — download" + name/duration/size) with no dead-player
flash; otherwise plays inline with an onError fallback to the same card.
Specs updated: uploads.md, database.md, federation.md.
Voice presence (voiceStates/voiceUserStates/spaceVoiceStates) was only ever
delivered in the WS `ready` payload — i.e. at connect/reload. A user joining a
space mid-session got `member_joined` (no voice data) and a bare space object;
`GET /api/spaces/:id` (the channel-sidebar hydrator) carries no voice state
either. So members already sitting in a voice channel stayed invisible in the
new member's sidebar until a full page reload.
Fix at the systemic root: ConnectionManager.addUserSpace — the single chokepoint
every join path funnels through (invite, public join, join-request approval),
and which is NOT used on reconnect (that path uses setUserSpaces) — now pushes a
scoped `space_voice_state` snapshot to the joining user. The snapshot is built by
a new buildSpaceVoiceState(spaceId, userId) helper that is also the single source
of truth feeding buildReadyPayload (refactored to use it), so the connect-time
and join-time paths can never drift.
Robustness:
- Delivered over the same ordered WebSocket as voice_state_update deltas — no
REST snapshot-vs-event-stream race.
- VIEW_CHANNEL-filtered via computePermissions exactly like `ready`: a joiner is
never told who occupies a voice channel they cannot see.
- Client applies it scoped to the space (utils/voiceStateSync.applySpaceVoiceState):
merges occupants/statuses and rebuilds only that space's restriction keys,
never disturbing voice state in other spaces.
- Skipped when the space has no active voice and no restrictions (e.g. space
creation).
Tests: server helper behavior, the join push, and private-channel exclusion;
client scoped-apply. Specs updated (websocket.md, voice.md, spaces.md).
The sidebar's visibleChannels filter is keyed on the channelPermissions
Map. Creating a channel raced two state updates: the optimistic create
(added to channels with no permission entry) and the channel_created WS
event (the only thing that set the permission). When the optimistic add
won the race, the WS handler hit its dedup guard, skipped setChannels,
and set the permission by mutating the Map in place — no new reference,
so visibleChannels never recomputed and the channel stayed hidden until
loadSpace rebuilt the maps (i.e. leaving and returning to the space).
Centralize the logic in a new upsertChannel store action that replaces
channels and channelPermissions with fresh references, used by both the
create path and the channel_created handler. Also return the creator's
computed myPermissions (and isPrivate) from POST so the channel renders
immediately from the response, independent of WS timing.
Adds spaceStore.upsertChannel.test.ts covering the reference-identity
regression and the optimistic-reconcile path.
The scroll container's space-y-[2px] utility set margin-bottom:0 (via
Tailwind's space-y reverse mechanism) on every category wrapper except
the first, with higher specificity (.space-y > :not ~ :not = 0,3,0) than
each category's mb-[19px] (0,1,0). This zeroed the 19px separator for all
but the first category, so the inter-category gap rendered only below the
first category and shifted when categories were reordered.
Remove the redundant container-level space-y; the per-category mb-[19px]
is the intended separator and now applies uniformly.
Pin floating tags to versions already running on both live boxes:
- livekit/livekit-server:latest -> v1.9.11
- caddy:2-alpine -> 2.11.1-alpine
No operational change; tags proven-good in production.
Finding 1: no-arg branch used 'ls ... | while' which, under set -euo
pipefail, exits 1 on an empty backups dir (glob matches nothing). Replace
with shopt nullglob array check that prints a clear message and exit 0.
Finding 2: pass $TS and $SNAP_NAME as positional args to the inner alpine
shell instead of interpolating them into the sh -c string, making the swap
injection-proof. Behavior unchanged: pre-restore copy -> clear WAL/SHM -> install.
Replace ripped Discord audio with self-authored Ogg Vorbis files (avoids
copyright on open-source launch) and fix three cue bugs:
- Anti-pop envelope: playSound now applies a 10ms fade-in (+ tail fade-out
for one-shots) so buffers no longer start at a non-zero amplitude. Kills
the pop on the looping call cues. Mirrors playTestTone.
- Deafen no longer plays mute+deafen at once: toggleDeafen flips isMuted and
isDeafened atomically, so SoundController saw both transitions. New tested
pure helper selectVoiceStateSound() suppresses the side-effect mute cue.
- Viewer-side watch feedback: stream_user_joined/left now play locally on the
watcher's own machine for explicit Watch/Stop actions, via
handleViewerWatchToggle. Direct local playback (not a watchingStreams diff)
keeps auto-teardown silent and works identically on Safari and Electron.
Docs: docs/systems/sounds.md updated (Ogg, dual-audience cue rows, mute/deafen
selection, viewer feedback, anti-pop envelope).
Worktrees are embedded git repos; if any future agent runs `git add -A` they
get committed as a submodule pointer, which causes partial-checkout breakage
on clones. Always-ignore so the trap isn't reachable.
Eliminates the visible push-down/snap-back when the pagination loading
skeleton appears at the top of the message list. Achieved by rendering a
constant-height (~200px) top-of-list slot above messages whenever
hasMore===true; the skeleton's grey-bar contents toggle inside that slot
rather than the slot itself mounting/unmounting. Companion changes:
load-more trigger raised to fire before the slot enters the viewport with
an iOS Safari rubber-band guard, prepend scroll-restore math corrected to
handle non-zero prevScrollTop, useDelayedLoading threshold lowered to
50ms for pagination only.
Brings the parked mobile voice + screenshare polish from wip/mobile-polish
(commit dbb9b2c) into main. Conflicts in MainContent.tsx and mobile-ui.md
expected per the WIP commit message — resolved manually.
Three layered bugs all manifesting as "the group name doesn't show / the
placeholder is a 40-character wall of names":
1. **WS ready payload was missing `name`/`icon`.** The handler serialized
DmChannel rows with only `id, federatedId, ownerId, createdAt, members,
lastMessage`. The optional metadata fields were silently dropped, so
the client store never received `dm.name` until a subsequent
`dm_channel_updated` event fired (i.e. only mid-session renames worked,
never the initial render). `ownerHomeUserId`, `ownerHomeInstance`, and
`metadataUpdatedAt` were also missing — added too because federated
routing depends on `ownerHomeInstance` (`getDmOwnerHomeInstance`).
2. **Header surfaces silently dropped `dm.name`.** `MainContent` (desktop
chat header) and `MobileChatScreen` always rendered the joined member
names, even when `dm.name` was set. Five other surfaces (`DmListItem`,
`MobileDmsScreen`, `MessageList` welcome hero, `MobileGroupDmInfo`,
`GroupDmSettings`) honored it correctly, so a renamed group showed
different titles depending on which surface you looked at.
3. **Message-input placeholder rendered joined names.** Once a group
has 4+ members "Message #Alice, Bob, Charlie, Dave" overflows the
textarea and obscures the call-to-action.
Consolidates the display-name logic behind two utilities in
`dmFormatters.ts`:
- `formatDmHeaderName(dm, currentUser)` — `dm.name` verbatim if set,
else joined names (excluding self); falls back to `'Group'` /
`'Direct Message'`. Used by all 5 header surfaces (was inlined
5 different ways).
- `formatDmInputLabel(dm, currentUser)` — `'#<name>'` if set,
`'the group'` for unnamed groups (collapses the unreadable
joined-names form), `'@<partner>'` for 1-on-1.
`MessageInput` accepts an optional `placeholder` prop that bypasses the
default `Message {#|@}<channelName>` derivation; DM call sites use it
to inject the `formatDmInputLabel`-based form. 1-on-1 DMs keep the
canonical-view lookup so replicated aliases still surface the home
account's displayName; the placeholder reuses the canonical `dmName`
so header + placeholder stay aligned even when raw partner ≠ canonical.
13 new unit tests covering `formatDmHeaderName` (8 cases: named, blank,
joined, federated-username base, empty group, 1-on-1, no-displayName,
no-partner) and `formatDmInputLabel` (4 cases: named, unnamed,
whitespace-only, 1-on-1). 365 → 377 web tests, 1053 server tests,
typecheck clean.
Each AvatarTile rendered at `size × size` with a 2px border under
`box-sizing: border-box`, giving a content area of `(size − 4) × (size − 4)`.
The inner `<Avatar size={size}>` exceeded the padding box, and the
`overflow-hidden + rounded-full` clip — centered on the wrapper — combined
with the avatar contents anchored at the padding-edge top-left to displace
photos and especially the centered initials gradient + letter toward the
lower-right of the visible disc, leaving a sliver of background opposite.
Compounding this, `Avatar`'s `inline-flex` root sat on the line-box text
baseline, so any inherited `line-height ≠ 1` (the DM list inherits the
row's line-height) drifted the avatar a further several px vertically.
Two corrections:
• Size the inner Avatar to `size − 2·TILE_BORDER_WIDTH` (matches the
padding box) — extracted as a constant so the dependency between
`border-2` and the inner size is visible.
• Center geometrically via `flex items-center justify-center` on the
wrapper, bypassing inline-flow placement so the Avatar is anchored
regardless of inherited type metrics.
Verified in the live app (DM sidebar, chat header, welcome header) and
with a 4-tile diamond at sizes 32 and 80: visible offset is now exactly
the border width on every tile, letters/photos sit dead-center.
Adds a regression test that pins both invariants (flex centering classes
present + inner Avatar style.width === tileSize − 4) so a future change
that re-introduces the bug fails fast. 12/12 AvatarStack tests, 365/365
web tests, typecheck clean.
design-system.md spec updated with the AvatarTile geometry contract.
Manual ownership transfers between two federated instances diverged because
`dm_channels.ownerHomeInstance` was stored as a BARE host (`orbit.ddns.net`)
for federated owners — via `transferGroupDmOwnership` copying `users.homeInstance`
verbatim — while `sourceInstance` always arrives as a full URL on the wire.
`processOwnershipTransferEvent` and `processMemberRemoveEvent` then compared the
two with strict equality and rejected legitimate inbound events as
`unauthorized_source`, keeping ownership permanently divergent across peers.
Live DB inspection on the two test instances confirmed both rows (nova + orbit)
had a BARE `owner_home_instance`, matching the bug report exactly.
Three compounding fixes:
1. Receiver authority checks now compare via `normalizeOriginForCompare` so
legacy bare-vs-full rows accept legitimate transfers (and kicks).
2. New `canonicalizeHomeInstance` helper in `federationAuth.ts`; every write
site that persists `ownerHomeInstance` (`transferGroupDmOwnership`, group DM
creation, lazy federation in member-add, `processMemberAddEvent` bootstrap,
`processOwnershipTransferEvent` receiver storage) routes through it. Full URL
is the canonical storage form, matching how `sourceInstance` arrives.
3. `dm_owner_updated` WS event extended with optional `newOwnerHomeUserId` and
`newOwnerHomeInstance` fields. Client `updateDmOwner` writes them when
present and leaves existing values untouched otherwise (legacy-server safe).
Without this, `getOwnerInstanceForDm` returned the previous owner's home
after a successful WS broadcast, routing the next owner-only op to the wrong
instance.
Coverage: new `federation.ownershipTransfer.test.ts` (7 receiver tests including
the headline bare-vs-full regression and the dedup replay guard); new bare-vs-full
case in `federation.kick.test.ts`; two new client-side cases in
`groupDm.ownerRouting.test.ts` covering both the extended-payload write path and
the legacy-server passthrough. Tests: 1053 server + 364 web, all green.
Specs updated: `dm-system.md` historical bugs + frontend handler table + WS
state-change events table; `federation.md` `ownership_transfer` receiver flow;
`websocket.md` event-fields table.
Transferring ownership or kicking a member surfaced "Target user is not a
member of this DM channel" whenever the target was a federated user.
Root cause: the client passed `canonical.id` from `useCanonicalUserView`,
which returns the user's HOME id when the home view is in the userViews
cache. After owner-routing the request to the owner instance, that
instance's `dm_members.userId` (its own local replicated id) never
matched the home id, so `isDmMember` returned false. The same failure
mode applied across any cross-instance scenario where the
channel-serving instance and the owner-serving instance disagree on the
local replicated user id for the same federated user.
Fix: both endpoints now accept federated identification, mirroring the
existing pattern on `POST /api/dm/:id/members`:
- `POST /api/dm/:id/transfer` body: `{ newOwnerId? } | { homeUserId, homeInstance }`.
Federated args win when both are supplied (strictly more specific).
- `DELETE /api/dm/:id/members/:targetUserId` reads optional
`?homeInstance=<origin>` query; when present, the URL segment is
treated as a homeUserId and resolved via `resolveOrCreateReplicatedUser`.
Client `api.dm.kickMember` and `api.dm.transferOwnership` gain an
optional `federated` argument; `DmRosterPanel` and `MobileGroupDmInfo`
pass it whenever the target has `homeUserId` + `homeInstance` populated.
Adds 5 server tests (3 transfer + 2 kick) covering federated targets,
the federated-wins-over-local precedence rule, and federated-non-member
rejection. Updates 2 client routing tests and 2 DmRosterPanel test
assertions for the new signature. Updates `docs/systems/dm-system.md`
and `docs/systems/api.md`.
Server: 965 tests pass (was 960). Web: 362 tests pass (was 360).
The old 2x2 grid for 3+ members produced a cramped, misaligned look
(circles cut off inside their own border, no overlap, just four small
disks in a grid). The 2-member overlap aesthetic — equal-size tiles
that overlap diagonally — was the visual gold standard but only
existed for that one count.
Extend the same "huddle of overlapping faces" aesthetic to all
member counts:
• 3 members → equilateral triangle of 62%-size tiles (top, bl, br)
• 4 members → diamond of 58%-size tiles (top, right, bottom, left)
• 5-10 members → diamond with `+N` overflow occupying the bottom
slot (z-index boosted above neighbors so the digit is never clipped)
Tiles are positioned radially around the box center on a circle of
radius (S − T) / 2 so the farthest edges of each tile graze the
bounding rect — no clipping, no wasted whitespace. Z-index descends
clockwise from the top slot so each tile tucks slightly under its
clockwise neighbor, mirroring the 2-member z-stack.
The 0/1/2-member cases and the icon-override branch are unchanged.
Props interface unchanged — no call-site updates needed.
Tests updated to assert the new layout markers (`triangle`, `diamond`)
and to verify the geometry (top tile in triangle has smallest `top`;
overflow tile in diamond has largest `top`). 11/11 AvatarStack tests
pass; 360/360 web tests pass; typecheck clean.
design-system.md spec updated with the new layout table + geometry
section.
Snapshot of in-progress work parked here so group-DM-polish can land
cleanly on main. Touches MainContent + mobile-ui.md which overlap with
group-DM-polish; rebase onto post-merge main and resolve conflicts on
those two files manually.
Files: MainContent, MessageInput, MobileVoiceFullScreen (+test), StreamTile,
VoiceUser, useLiveKit, useVisualViewportInset, AudioManager, voiceStore,
voice utils, mobile-ui.md, voice.md, mobile-parity handoff doc.
getProfileReferencedFilenames() didn't include dm_channels.icon, so the
storage janitor deleted group DM icons within ~1 hour:
- Owner instance: PATCH /api/dm/:id leaves an attachments row with
messageId=null and dmMessageId=null. After 1h, getUnlinkedAttachments
flags it and cleanupStorage phase 2 deletes the file because the
filename isn't in profileReferenced.
- Receiver instance: downloadProfileAsset writes the icon directly to
uploadDir with no attachments row. cleanupStorage phase 1 treats it
as orphaned and deletes it.
Fix: include dm_channels.icon (non-null, not soft-deleted, not http://)
in the profile-referenced set. Soft-deleted DMs are excluded so their
files still get reaped by cleanupSoftDeletedDmChannels. Absolute URLs
are skipped because they live on a remote instance.
Also mirror the avatar precedent at the PATCH endpoint by deleting the
new icon's standalone attachment row — the file is now protected via
dm_channels.icon, matching users.ts:473.
Refactors the leave-DM destructive core into a shared `removeDmMember`
helper and adds a kick endpoint that reuses it. Both endpoints write the
`member_removed` system message, delete the dm_members + read_states
rows, broadcast `dm_member_removed`, and queue a federation
`member_remove` event with the appropriate `reason` ('leave' | 'kick').
Branching invariants preserved by the helper:
- Ownership transfer fires only on self-leave when the leaver was the
owner. Kicks cannot orphan a group (the owner is still present), so
the transfer block is skipped.
- Soft-delete on last-member-empty fires only on self-leave. Kicks are
guaranteed to leave the owner behind, so the channel can never be
empty after a kick.
Endpoint validation:
- 1-on-1 DM → 400 'Cannot kick from a 1-on-1 DM'
- Caller not the owner → 403 'Only the group owner can remove members'
- Self-target → 400 'Owners cannot kick themselves; use leave instead'
- Target not a member → 404
- Channel missing or soft-deleted → 404
The kicked user is also evicted from the DM voice room (mirroring the
self-leave path) and receives `dm_channel_closed` so the client closes
the channel locally.
config.ts requires JWT_SECRET at module-load time. Since `.env` is gitignored
it doesn't propagate to git worktrees, fresh clones, or CI without secrets —
producing 200+ confusing cascade failures (vitest module-mock errors that
look unrelated to the real cause).
Add a vitest setupFile that sets JWT_SECRET to a fixed test-only value via
`??=`, so a real .env still wins where one exists. Tests are now hermetic:
clone the repo, `pnpm install`, `pnpm --filter server test` → 477/477 pass.
Multi-pass mobile chat polish landed across this session.
- MessageInput is now a floating glass-bubble (`position: absolute`) on both desktop and mobile — last messages scroll *behind* the translucent bubble. MobileChatScreen wraps MessageList + MessageInput in a `relative` parent so absolute positioning resolves. Removed the prior in-flow mobile branch that clipped message-list bottom against an invisible barrier.
- iOS PWA keyboard handling: new `useVisualViewportInset` hook subscribes to `visualViewport.resize/scroll` AND polls `vv.height` for ~600ms after focusin (iOS PWA standalone often fails to dispatch resize for keyboard transitions). MobileShell sizes its container to `vv.height` when keyboard is open — `bottom: 0` on the composer naturally lands flush with the keyboard top, regardless of how reliably resize events fire. Composer uses 6px gap above home indicator (keyboard closed) and 0px gap above keyboard (keyboard open). Added `interactive-widget=resizes-content` viewport meta as the cleaner native equivalent for Chrome/Android.
- MessageList bottom padding is dynamic via `--composer-clearance` CSS variable. MessageInput writes `composerHeight + bottomOffset + 12px` to its parent via ResizeObserver — re-fires on textarea autosize, reply banner, attachment tile growth, parent resize. Last message always has 12px breathing room above the bubble regardless of composer state.
- AttachmentRenderer generic file chip: `max-w-full sm:max-w-[400px]` on outer + `min-w-0` + `flex-shrink-0` on icon + `flex-1` on text + `flex-wrap` on badge row. Long filenames now ellipsize cleanly on narrow viewports instead of pushing the chip off-screen.
- New `useDragToClose` hook: shared bottom-sheet drag-down-to-dismiss gesture. Spread on handle/header only (body scrolling unaffected). 6px deadzone, 100px or 0.5px/ms velocity threshold, 200ms `cubic-bezier(0.22, 1, 0.36, 1)` close-out animation, rAF-staged transform for a stable from-value. `hasInteracted` latch prevents the open keyframe from re-firing mid-close (the bounce-up-then-vanish bug). Wired into InputPopover (emoji/GIF), MobileVoiceJoinSheet, MobileFolderSheet.
Specs: docs/systems/mobile-ui.md (Floating Composer + Drag-to-Close sections), docs/systems/message-list.md (--composer-clearance), docs/systems/design-system.md (glass-bubble row references).
Brings mobile to skeleton parity with desktop. Two mobile-specific renders, plus a fix for the empty-state mascot flashing on every space switch.
- MobileSpacesScreen: shimmer skeleton for the channel list while `useSpaceStore.loadingSpaceId === selectedSpaceId` (uncategorized rows + category header + categorized rows). Mirrors desktop ChannelSidebar's `showChannelSkeleton`. Gated through `useDelayedLoading` so cached/fast loads don't flash the placeholder.
- MobileMembersScreen: shimmer skeleton for the member list (two role-section headers + circular avatars + name bars). Mirrors desktop MemberSidebar's `showMemberSkeleton`.
- Empty-state "No channels yet." mascot was firing on EVERY space switch for ~50–200 ms — `state.channels` only ever holds the most-recently loaded space's channels, so `spaceChannels` filters to `[]` immediately on selection change while `loadSpaceDetail` is still in flight. New `loadedSpaceIds: Set<string>` on useSpaceStore, populated only on `loadSpaceDetail` success and pruned on space removal events. The mascot is now gated on `loadedSpaceIds.has(selectedSpaceId)` so it only fires when the space is *settled* with zero channels.
Reuses the existing `.skeleton` / `.skeleton-bar` / `.skeleton-circle` CSS primitives + `useDelayedLoading` hook (no new shared `<Skeleton />` primitive — sites need row-specific geometry, parameterizing would either over-abstract or duplicate the inline approach).
Specs: docs/systems/mobile-ui.md (Loading Skeletons inventory + per-screen skeleton bullets + settle-gate rationale); docs/systems/spaces.md (Client Load State subsection covering loadingSpaceId vs loadedSpaceIds).
Two compounding root causes produced a "PiP-style grey view" briefly visible after Join Voice on mobile:
1. MobileSpacesScreen.handleVoiceJoin pushed 'voice' onto the mobile stack. MobileShell.screenMap only has 'voice-full' — the renderer returned null for 'voice' while the root screen sat under `visibility: hidden` (stack length > 0). Corrected to 'voice-full' with an explanatory comment.
2. AppLayout mounted the desktop <PictureInPicture /> in the mobile branch too. Its visibility check `(isInServerVoice || isInDmCall) && !voiceFullscreen` is true the moment currentVoiceChannelId is set, so the desktop PiP rendered as the only visible UI on top of the hidden mobile shell. Removed from the mobile branch; kept on desktop. Added a comment to prevent regression.
Mobile owns its voice overlay chrome exclusively via MobileVoiceMiniBar + MobileVoiceFullScreen.
Spec: docs/systems/mobile-ui.md gains a "Voice Join Flow (Mobile)" subsection and a leading paragraph in "Voice Overlay" stating that PiP is desktop-only.
Closes the mobile parity push.
- MobileVoiceJoinSheet: full pre-join camera preview. Dormant-by-default (never auto-fires getUserMedia), explicit user gesture to arm, hard-bound disarm on sheet close (any path). Camera picker popup portaled to document.body so a long device list stays scrollable above the aspect-video preview tile (max-height min(50vh,320px), iOS scroll momentum). iOS Safari-safe: autoPlay playsInline muted + post-await play().
- MobileSettingsScreen: Keybinds + Desktop sections gated on isElectron(); they reuse the existing KeybindsPanel/DesktopPanel which are already mobile-fit. screenMap entries added in MobileShell. (Verify on Electron desktop build at narrow viewport.)
- ToastContainer: real overlap was hiding the voice-fullscreen control bar. Mobile branch now reads isMobile/mobileStack/currentVoiceChannelId and computes the bottom offset across five mobile states (voice-full / pushed+voice / pushed / root+voice / root), all with safe-area-inset-bottom; left-3 right-3 + items-center keeps toasts in the safe-tap zone. Desktop bottom-6 right-6 unchanged.
Specs: docs/systems/mobile-ui.md (Toast Positioning section, screenMap rows, Electron-entry rationale, z-index row); docs/systems/voice.md (Mobile pre-join preview subsection covering lifecycle + camera picker portal).
Adds sticky `lastSelectedSpaceId` to useSpaceStore. Updated on every
setCurrentSpace(non-null) and on loadSpaceDetail success; NOT cleared by
setCurrentSpace(null), so the @me URL effect in AppLayout no longer wipes
the memory. Cleared only when the remembered space is actually removed
(deleteSpace / leaveSpace / removeSpace / removeInstanceSpaces / reset).
Ephemeral — URL drives initial state on reload.
MobileBottomNav.handleTab('spaces') and MobileSpacesScreen's initial
useState now resolve via `currentSpaceId ?? lastSelectedSpaceId`. The
prior Object.keys(lastChannelPerSpace)[0] fallback was wrong twice over:
first-inserted ordering locked the tab to whichever space the user
opened first in their session, and any approach reading currentSpaceId
alone breaks after a /channels/@me detour.
Also fixes a pre-existing render loop in MobileSpacesScreen's
currentSpaceId sync effect — `selectedSpaceId` removed from deps,
functional setState makes the update idempotent.
Spec: docs/systems/mobile-ui.md updates Tab Tap Behavior, MobileScreenHeader,
and LocalStorage Persistence sections.
iOS Safari auto-zooms (and shifts the viewport right) on focus when the
computed font-size is <16px. Single @media (max-width: 767px) block in
globals.css covers all four input tiers, every relevant bare <input>
type, <textarea>, <select>, and contenteditable surfaces. `!important`
is required because Tailwind utilities (text-[15px] on the chat composer,
text-sm on form fields) emit after @layer components and would otherwise
override the input-tier rules.
Replaces ad-hoc per-input `text-base md:text-sm` overrides with a single
global truth. New inputs/contenteditables get the fix for free.
Spec: docs/systems/design-system.md gains an "iOS Auto-Zoom Suppression"
section under Input Tiers.
Jannis Braun