We forgot to tell React that we need the audio renderer to react to changes in the set of MatrixRTC participants; instead we had it referencing rtcSession.memberships non-reactively.
Now, I'm not 100% confident that this is going to fix the "speaking from the void" issues observed in the wild, because I can't reproduce them and, in my testing, the InCallView component always seemed to be rendered redundantly when the MatrixRTC participants change, even though we hadn't explicitly stated that it needs to react. (This makes sense as we haven't memoized the component.) But it's worth a shot.
* Simplify key local storage management.
* Refactor useLivekit to only ever connect to one room.
This change also tries to make the code more explicit so that we only do the things we really need to do and rely less on react updating everything correctly.
It also surfaces, that we are currently implementing useLivekit in a way, so that we can change the encryption system on the fly and recreate the room. I am not sure this is a case we need to support?
* simplify the useLivekit hook even more
This is possible because we concluded that we do not need to be able to hot reload the e2ee system.
* review
* linter
* Update src/room/InCallView.tsx
Co-authored-by: Robin <robin@robin.town>
---------
Co-authored-by: Robin <robin@robin.town>
This gives us the additional insurance of breaking the Safari media acquisition loop at the source by admitting that they can be spurious in practice. Safari, why!?
Because we're now requiring the 'Prevent blocked' check to pass before merging a PR, GitHub Actions now expects it to be associated with the latest Git ref of the PR's branch whenever the branch is updated. Therefore we need to re-run the workflow on the 'synchronize' event.
Because the latest version requires eslint-plugin-unicorn v57, which requires eslint v9, and eslint-plugin-matrix-org is not yet compatible with eslint v9.
