### Description
- switch Matrix custom HTML sanitization from `sanitize-html` to
`DOMPurify`
- enforce the Matrix v1.18 tag and attribute allowlist exactly,
including scheme and `mxc://` restrictions
- sanitize reply previews before trimming/parsing so untrusted
`formatted_body` never reaches the reply extraction path unsanitized
- move `data-mx-color` and `data-mx-bg-color` translation to render time
and explicitly ignore incoming `style` in the HTML parser
- use `sanitizeText` for plain-text interpolation paths that were
previously running through the HTML sanitizer
- add regression tests for sanitizer allowlist behavior, reply preview
sanitization, parser color translation, and parser-side style stripping
#### Type of change
- [x] Bug fix (non-breaking change which fixes an issue)
- [ ] New feature (non-breaking change which adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to not work as expected)
- [ ] This change requires a documentation update
### Checklist:
- [x] My code follows the style guidelines of this project
- [x] I have performed a self-review of my own code
- [x] I have commented my code, particularly in hard-to-understand areas
- [ ] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
### AI disclosure:
- [ ] Partially AI assisted (clarify which code was AI assisted and
briefly explain what it does).
- [x] Fully AI generated (explain what all the generated code does in
moderate detail).
This PR replaces the Matrix custom HTML sanitizer with a DOMPurify-based
implementation that matches the Matrix v1.18 allowlist instead of the
broader legacy rules. The sanitizer now validates tag-specific
attributes, rejects incoming `style`, limits links to
`https/http/ftp/mailto/magnet`, limits images to `mxc://`, strips
`mx-reply`, and keeps the existing nesting guard. Reply previews now
sanitize `formatted_body` before any reply trimming or parsing. The
React HTML parser now turns `data-mx-color` and `data-mx-bg-color` into
React styles at render time and ignores raw incoming `style` attributes
even if unsanitized HTML somehow reaches the parser. The plain-text HTML
interpolation paths in forwarding, the room composer, and the message
editor now use text escaping instead of the HTML allowlist sanitizer.
The new tests lock down the sanitizer contract, the reply-preview
regression, parser color translation, and parser-side style stripping.
<!-- Please read
https://github.com/SableClient/Sable/blob/dev/CONTRIBUTING.md before
submitting your pull request -->
### Description
Adds a room abbreviations feature: moderators can define term/definition
pairs in room settings. Defined terms are highlighted with a dotted
underline in messages and show a tooltip on hover (tap-to-pin on
mobile). Abbreviations defined in a parent space are inherited by child
rooms and nested spaces, with overrides allowed at each level. The
settings UI shows a flat list of inherited and locally-defined entries.
Fixes #
#### Type of change
- [ ] Bug fix (non-breaking change which fixes an issue)
- [x] New feature (non-breaking change which adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to not work as expected)
- [ ] This change requires a documentation update
### Checklist:
- [x] My code follows the style guidelines of this project
- [x] I have performed a self-review of my own code
- [ ] I have commented my code, particularly in hard-to-understand areas
- [ ] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
### AI disclosure:
- [ ] Partially AI assisted (clarify which code was AI assisted and
briefly explain what it does).
- [X] Fully AI generated (explain what all the generated code does in
moderate detail).
<!-- Write any explanation required here, but do not generate the
explanation using AI!! You must prove you understand what the code in
this PR does. -->
The code adds changes that allow for setting and viewing room
abbreviations, with an on-hover tooltip (or press on mobile -
semi-persistent). It stores these in the room state under a new key, and
recursively fetches abbreviations from the space level. This also adds
regex so that abbreviations correctly work at word boundaries. Also adds
the needed hooks to `room.ts` and `slidingSync.ts`
Previously, abbreviation term highlighting was skipped for any message
that has a formatted_body (virtually all messages from modern clients
using a rich-text editor), because RenderBody returned early from the
customBody path before reaching the abbreviation logic.
The fix adds an optional replaceTextNode callback to
getReactCustomHtmlParser. Since opts closes over itself, renderChildren
inside element handlers recurses through the same replace function, so
text nodes nested inside <p>, <strong>, <em>, list items, etc. are all
processed correctly.
buildAbbrReplaceTextNode is exported from RenderBody.tsx and wired
into both RoomTimeline and ThreadDrawer via useRoomAbbreviationsContext.
### Description
Fixes#350
This adds a dynamic multiline layout to `CustomEditor`. When the message
input grows past a single line (word wrap, explicit newlines, or
multiple blocks), the editor shifts from a single-row layout into a
two-row grid, before/textarea on top, with responsive content spanning
the full width below. It drops back to single-row once the text fits on
one line again.
The old `replacementContent` prop (which swapped out the textarea on
mobile during voice recording) is gone. It's replaced by
`responsiveAfter`, which renders inline next to the send button in
single-line mode and moves down to the footer row in multiline mode.
There's also a `forceMultilineLayout` prop for when you want to push
into multiline immediately, which is what we use when the audio recorder
opens. This means the recorder now takes the same code path on desktop
and mobile instead of having a separate mobile-only branch.
A few other things got fixed along the way:
- Holding the record button on mobile was selecting the "following the
conversation" text underneath it. Fixed with `userSelect: none`.
(related to #350)
- The stop button's `onClick` was being skipped on mobile entirely, so
if `pointerup` got lost mid-drag there was no way to stop recording. Now
`onClick` works as a fallback when recording is already active. (a form
of recovery in case more situations like #350 happen again)
- The waveform bar count is now calculated from the actual container
width via `ResizeObserver` instead of a hardcoded number.
- `cleanupMediaRecorder` wasn't being called in all the
stop/delete/restart/error/unmount paths, which could leave the mic track
open. That's fixed.
- Firefox couldn't seek voice messages because `audio/webm` doesn't
include a seek index. Codec preference now puts `audio/ogg;codecs=opus`
first so Firefox records seekable blobs. Also added a `seekTo` helper
that waits for `loadedmetadata` if the audio element isn't ready yet,
and `audio.preload = 'auto'` + `audio.load()` on mount so Firefox parses
metadata right away.
- Multiline wrap detection is now debounced through
`requestAnimationFrame` and uses a `ResizeObserver` on the scroll
container to cut down on flicker and extra re-renders.
- Added cloudflared tunnels for easier testing on mobile
Also sorry for the cursed line height measurer, there was legit no other
way to detect it that didn't cause other issues.
Good amount of this was created with help of AI (like figuring out
options on how to fix the edge cases), especially the edge case unit
tests, but It had to be fixed by hand in many places and frankly I can't
tell you exactly where was what. I've been working on this for couple
days now.. (losing my mind over it)
### Screenshots:
https://github.com/user-attachments/assets/bcff3766-6c60-46c7-ac9e-a8b943a665d3https://github.com/user-attachments/assets/367cdcc4-666e-4324-8e81-1b415feb0aca
#### Type of change
- [x] Bug fix (non-breaking change which fixes an issue)
- [x] New feature (non-breaking change which adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to not work as expected)
- [ ] This change requires a documentation update
### Checklist:
- [x] My code follows the style guidelines of this project
- [x] I have performed a self-review of my own code
- [x] I have commented my code, particularly in hard-to-understand areas
- [ ] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
### AI disclosure:
- [x] Partially AI assisted (clarify which code was AI assisted and
briefly explain what it does).
- [ ] Fully AI generated (explain what all the generated code does in
moderate detail).