The dot badge (8px circle) was positioned at top:-2px, left:-2px
which placed its CENTER 2px inside the avatar corner rather than
on the corner itself. Changed to top:-4px, left:-4px so the dot
is centered exactly on the top-left corner of the avatar, matching
how the count badge (top:-6px for ~12px height) is positioned.
Also updated the folder compound variant dot position from +2px to 0px
for consistency.
Fixes#293
RoomCallButton was calling startCall(room) without passing preferences,
which caused Element Call to use its own defaults (camera on) instead of
the user's saved settings. Now reads preferences via useCallPreferences()
and passes them to startCall(), matching the behavior of PrescreenControls.
Fixes#285
Overrides browser find-in-page (Ctrl+F) to open Sable's search modal
instead, matching the behavior users expect from chat applications.
The existing Ctrl+K hotkey remains functional.
Fixes#179
WebKit (Safari/iOS) has a rendering quirk where colorful emoji inside <b>
tags can contaminate the paint color of adjacent sibling text nodes. This
caused display names following a user with an emoji (e.g. '🔴 Alice') to
render with the emoji's color instead of the expected text color.
Apply -webkit-text-fill-color: inherit on <b> name tags and
-webkit-text-fill-color: currentColor on the following <Text as="span">
elements. This explicitly resets the WebKit fill color chain at each step,
preventing emoji paint properties from cascading into sibling spans.
Fixes#267
The thread drawer had padding: S600 0 (24px top+bottom) on both the root
message box and the reply messages box, plus S200 padding on the reply count
label. This stacked up to ~84px of empty space between the root message and
the first reply, especially noticeable in newly created threads.
Fix:
- Root message box: change bottom padding S600→S200 (top kept at S400)
- Reply messages box: remove top padding (reply count label already provides
visual separation)
Fixes#294
The old code used IntersectionObserver anchors to determine whether scroll
arrows should be shown. This caused false positives in two scenarios:
1. When a UrlPreviewCard renders null (fetch error), the empty anchor div
was sometimes not detected as intersecting, leaving a stale arrow visible.
2. When loading (Spinner makes card 400px wide), the IO would mark the front
anchor as not visible, showing the arrow; after load failure the arrow
stayed stuck.
Replace with direct scrollWidth/clientWidth measurement via ResizeObserver
(fires on container and content resize) plus a scroll event listener. Arrows
now only appear when the horizontal scroll area actually has overflowed content.
Fixes#220