We need to be consistent about whether we import matrix-js-sdk from `src` or
`lib`, otherwise we get two copies of matrix-js-sdk, and everything explodes.
* Fix issues detected by Knip
Including cleaning up some unused code and dependencies, using a React hook that we unintentionally stopped using, and also adding some previously undeclared dependencies.
* Replace remaining React ARIA components with Compound components
* fix button position
* disable scrollbars to resolve overlapping button
---------
Co-authored-by: Timo <toger5@hotmail.de>
* Install Knip
* Clarify an import that was confusing Knip
* Fix issues detected by Knip
Including cleaning up some unused code and dependencies, using a React hook that we unintentionally stopped using, and also adding some previously undeclared dependencies.
* Run dead code analysis in lint script and CI
---------
Co-authored-by: Timo <toger5@hotmail.de>
* Fix coverage reporting
Codecov hasn't been working recently because Vitest doesn't report coverage by default.
* Suppress some noisy log lines
Closes https://github.com/element-hq/element-call/issues/686
* Store test files alongside source files
This way we benefit from not having to maintain the same directory structure twice, and our linters etc. will actually lint test files by default.
* Stop using Vitest globals
Vitest provides globals primarily to make the transition from Jest more smooth. But importing its functions explicitly is considered a better pattern, and we have so few tests right now that it's trivial to migrate them all.
* Remove Storybook directory
We no longer use Storybook.
* Configure Codecov
Add a coverage gate for all new changes and disable its comments.
* upgrade vitest
---------
Co-authored-by: Timo <toger5@hotmail.de>
* Stop sharing state observables when the view model is destroyed
By default, observables running with shareReplay will continue running forever even if there are no subscribers. We need to stop them when the view model is destroyed to avoid memory leaks and other unintuitive behavior.
* Hydrate the call view model in a less hacky way
This ensures that only a single view model is created per call, unlike the previous solution which would create extra view models in strict mode which it was unable to dispose of. The other way was invalid because React gives us no way to reliably dispose of a resource created in the render phase. This is essentially a memory leak fix.
* Add simple global controls to put the call in picture-in-picture mode
Our web and mobile apps (will) all support putting calls into a picture-in-picture mode. However, it'd be nice to have a way of doing this that's more explicit than a breakpoint, because PiP views could in theory get fairly large. Specifically, on mobile, we want a way to do this that can tell you whether the call is ongoing, and that works even without the widget API (because we support SPA calls in the Element X apps…)
To this end, I've created a simple global "controls" API on the window. Right now it only has methods for controlling the picture-in-picture state, but in theory we can expand it to also control mute states, which is current possible via the widget API only.
* Fix footer appearing in large PiP views
* Add a method for whether you can enter picture-in-picture mode
* Have the controls emit booleans directly
The buttons were scrolling with the view instead of always being visible in a fixed location on the tile, and the indicators were not adopting the correct width.
The code path for when all tiles can fit on screen was failing to realize that it could sometimes get by with fewer columns. This resulted in wasted space for 4 person calls at some window sizes.
We've gotten feedback that it's distracting whenever the same video is shown in two places on screen. This fixes the spotlight case by showing only the avatar of anyone who is already visible in the spotlight. It also makes sense to hide the speaking indicators in spotlight layouts, I think, because this information is redundant to the spotlight tile.
This is because our layouts for flat windows are good at adapting to both small width and small height, while our layouts for narrow windows aren't so good at adapting to a small height.
If you were the only one in the call, you could get a broken-looking view in which the local tile is shown in the spotlight, and it's also shown in the PiP. This is redundant.
There were a couple of cases where the lack of margins after the new layout changes just looked odd. Specifically, when the header is hidden (as in embedded mode), there would be no margin at the top of the window. Also the floating tile would run directly up against the sides of the window.
Due to an oversight of mine, 2440037639 actually removed the ability to see the one-on-one layout on mobile. This restores mobile one-on-one calls to working order and also avoids showing the spotlight tile unless there are more than a few participants.
If no one had spoken yet, we were still showing the local user in the spotlight. We should instead eagerly switch to showing an arbitrary remote participant in this case.
* Add DeviceMute widget action `io.element.device_mute`.
This allows to send mute requests ("toWidget") and get the current mute state as a response.
And it will update the client about each change of mute states.
* review + better explanation
* review
* add comments
* use `useCallback`
We've concluded that this behavior is actually more distracting than it is helpful, and we want to try out what it's like to just have the importance ordering and visual cues help you find who's speaking.
We're finding that if we reorder participants based on whether their mic is muted, this just creates a lot of distracting layout shifts. People who speak are automatically promoted into the speaker category, so there's little value in additionally caring about mute state.
The Compound design tokens package is now set up to generate React components for every icon, so we no longer need to use our more error-prone method of importing the SVGs.
Ensure that they don't interfere with say, using spacebar to press a button, and also ensure that they won't do surprising things like scroll the page at the same time.
Follow-up to ea2d98179c
This took a couple of iterations to find something that works without creating update loops, but I think that by automatically informing Grid whenever a layout component is re-rendered, we'll have a much easier time ensuring that our layouts are fully reactive.
We no longer allow individual tiles to be put in full screen, because we're seeing what it's like to just stretch the spotlight tile edge-to-edge and keep the margins minimal.
Includes the mobile UX optimizations and the tweaks we've made to cut down on wasted space, but does not yet include the change to embed the spotlight tile within the grid.
Because we were hiding even the local participant during initial connection, there would be no participants, and therefore nothing to put in the spotlight. The designs don't really tell us what the connecting state should look like, so I've taken the liberty of restoring it to its former glory of showing the local participant immediately.
react-rxjs is the library we've been using to connect our React components to view models and consume observables. However, after spending some time with react-rxjs, I feel that it's a very heavy-handed solution. It requires us to sprinkle <Subscribe /> and <RemoveSubscribe /> components all throughout the code, and makes React go through an extra render cycle whenever we mount a component that binds to a view model. What I really want is a lightweight React hook that just gets the current value out of a plain observable, without any extra setup. Luckily the observable-hooks library with its useObservableEagerState hook seems to do just that—and it's more actively maintained, too!
If not set, legacy call membership state events are sent instead.
Even if set, legacy events are sent in rooms with active legacy calls.
---------
Co-authored-by: Timo <16718859+toger5@users.noreply.github.com>
