compose/ui/web: fix incorrect interpretation of ontouchend events

[TheLortex]

In ComposeWindow.w3c.kt, the touch handler does not correctly handle the touchend event. Indeed the logic selecting between targetTouches and changedTouches is flawed. For touchend both should be taken into account, also changedTouches contains the un-pressed pointer and targetTouches contains the rest of the remaining pressed pointers.

The fix consists in removing the logic that compares the size of both targetTouches and changedTouches to pick the largest one.
Instead, it checks the pointer event type to determine whether only targetTouches or both targetTouches and changedTouches are needed while correctly accounting which touches are pressed depending on the source.

Fixes: https://youtrack.jetbrains.com/issue/CMP-9030/wasm-canvas-incorrect-handling-of-multi-touch-input

Testing

I have added a failing test in GestureTest.kt which is subsequently fixed by the implementation commit. The test consists in using three touches that are progressively pressed (1, 2, 3) and unpressed (3, 2, 1), logging and comparing the behavior of the pointerInput modifier with the expected events.

Release Notes

Fixes - Web

• Fix incorrect interpretation of ontouchend events.

[eymar]

Would you like to add a demo that you have created in the bug report? https://youtrack.jetbrains.com/issue/CMP-9030/wasm-canvas-incorrect-handling-of-multi-touch-input

It can be added here

https://github.com/JetBrains/compose-multiplatform-core/tree/jb-main/compose/mpp/demo

[eymar]

As I see the final touches list always contains event.targetTouches.

In this case, wdyt about changing this to:

val touches = event.targetTouches.asList().fastMap { it to true }.toMutableList()
if (eventType == PointerEventType.Release) {
    touches.addAll(event.changedTouches.asList().fastMap { it to false })
}

[TheLortex]

yes, that's cleaner

[TheLortex]

Thank you for your review. I have applied your suggestion.

Would you like to add a demo that you have created in the bug report?

I'd like to keep the changeset small if that's fine for you ! I don't think the demo is needed in this case