| Summary: | "Frame End" timeline point is wrong with UI-side compositing | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Simon Fraser (smfr) <simon.fraser> |
| Component: | Web Inspector | Assignee: | Nikos Mouchtaris <nmouchtaris> |
| Status: | RESOLVED FIXED | ||
| Severity: | Normal | CC: | inspector-bugzilla-changes, simon.fraser, webkit-bug-importer |
| Priority: | P2 | Keywords: | InRadar |
| Version: | WebKit Nightly Build | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| See Also: |
https://bugs.webkit.org/show_bug.cgi?id=249750 https://bugs.webkit.org/show_bug.cgi?id=249736 https://bugs.webkit.org/show_bug.cgi?id=249807 https://bugs.webkit.org/show_bug.cgi?id=260360 |
||
Currently on iOS, "frame end" will happen before "composite end" in some cases. Pull request: https://github.com/WebKit/WebKit/pull/13661 Committed 263917@main (f11ea318b8fc): <https://commits.webkit.org/263917@main> Reviewed commits have been landed. Closing PR #13661 and removing active labels. |
`InspectorTimelineAgent::internalStart` uses this mechanism to detect "frame end": m_frameStopObserver = makeUnique<RunLoopObserver>(static_cast<CFIndex>(RunLoopObserver::WellKnownRunLoopOrders::InspectorFrameEnd), [this]() { This doesn't work with UI-side compositing (i.e. iOS, and future macOS), because a "frame" can extend over multiple turns of the CFRunLoop while we wait for backing store flushing (see where RemoteLayerTreeDrawingArea calls didCompleteRenderingUpdateDisplay()).