* Listen to call decline to stop ringing when declined from other device
* use proper swift naming convention for Id/ID
* fix Force unwrapping
* fix lint
* An approach without the custom publisher.
* review: correct naming convention
* review: revert some invisible char/tab changes
* add ref to the room proxy in the closure
---------
Co-authored-by: Doug <douglase@element.io>
* Stop sending call notifications manually
The Element Call widget can now assume responsibility for sending them when you start a call.
* Bump Element Call embedded package to v0.14.0
* Delegate the sending of call notifications to Element Call
* Update the SDK.
* Use 'is' prefix for boolean
---------
Co-authored-by: Doug <douglase@element.io>
* EC: native switch for audio outputs/inputs and earpiece
also includes proximity sensor enabling when the earpiece is used.
* test disabling and re-enabling audio on audio change
* Revert "test disabling and re-enabling audio on audio change"
This reverts commit 907c791b8df7009214f5c164ecf8fe7d90a8addc.
* pr suggestions
* pr suggestion and small refactor of the widgetAction handler
* Bump the RustSDK to v25.03.11
* Replace oidc login prompt with nil following the changes from https://github.com/matrix-org/matrix-rust-sdk/pull/4761
```
/// * `prompt` - The desired user experience in the web UI. No value means
/// that the user wishes to login into an existing account, and a value of
/// `Create` means that the user wishes to register a new account.
```
* Fix trailing closure warnings
* Update the client proxy after making `getNotificationSettings()` and `cachedAvatarUrl()` async (they used to be blocking on the rust side).
* Move `Room.isEncrypted` to the info publisher and manually update the encryption state when creating the room.
* Bump the SDK again to v25.03.12 - This introduces a new way to configure the tokio runtime that we can use to have extensions use less memory
- introduce a new Target struct that takes care of setting up rust services (tracing and tokio) for our various targets
- cleanup MXLog and friends
* Address PR comments
* Bump the SDK again, switch back to using `.consent` as the OIDC login prompt (which was reintroduced in matrix-org/matrix-rust-sdk/pull/4791)
Element Call now sends a 'close' widget action when the widget is ready to close. Usually this will be sent immediately after the 'hangup' action, but it could be sent later if the widget wants to present an error screen before closing. So by listening to the 'close' action rather than the 'hangup' action, we get to see these error screens.
* Introduce a `MediaEventsTimelineFlowCoordinator`
* Update SDK API and architecture
* Add a feature flag, add translations
* Move the media events timeline presentation under the room flow coordinator state machine
* Rename `TimelineViewState.timelineViewState` of type `TimelineState` to `timelineState`
* Enabled SwiftLint's `trailing_closure` rule and fix the warnings.
* Fix a race condition observing room info updates for calls.
* Fix a bug where call observation wasn't set up if the call comes when the app has been killed.
* Fixes#1227 - Add support for receiving and interacting with incoming session verification requests.
* Fix a couple of random small warnings
* Move static view config to the view state
* Update snapshots
* Use native video call picture in picture!
* Handle isSupported, isPossible and web view errors.
* Use new canEnterPip method.
* Hide a room's Call button when already joined to the call.
* Tweak feature flag description.
* PR comments.
* Remove incorrect message send call and fix typo.
* Add overlay coordinator presentation to the NavigationRootCoordinator.
* Remove GenericCallLinkCoordinator, merging it into CallScreen.
This will allow for picture in picture on call links when available.
- we're currently working around the widget not properly aquiring media streams by manually terminating a call before starting the web view
- it seems that there's a race condition between that and the system opening the app
- as such we need to delay terminating the call until the system gets a chance to handle the redirect
- CallKit decides what action to to take when pressing the hardware lock button based on where the audio is routed to. Using the speaker will make it lock the phone while using the headset will make it end the call.
- Before the introduction of CallKit we had to manually setup the AudioSession to keep the audio live in the background. This is now clashing with CallKit making it bevelie the audio is not routed to the speaker and causing the call be terminated when pressing the lock button.
- not completely sure why this happens but `reportOutgoingCall` interferes with how the WebView gets access to video streams
- the call itself isn't really necessary so removing it is the simplest way forward
Valere Fedronic
Valere Fedronic
Doug
Copilot
Robin
Mauro Romito
Stefan Ceriu
Mauro
Stefan Ceriu
Hugh Nimmo-Smith
manuroe
fkwp