| Summary: | REGRESSION(r244372): [GTK][WPE] fast/images/icon-decoding.html and others are failing | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | WebKit | Reporter: | Alicia Boya García <aboya> | ||||||
| Component: | WebKitGTK | Assignee: | Miguel Gomez <magomez> | ||||||
| Status: | RESOLVED FIXED | ||||||||
| Severity: | Normal | CC: | aperez, bugs-noreply, clopez, commit-queue, magomez, mcatanzaro, zan | ||||||
| Priority: | P2 | ||||||||
| Version: | WebKit Nightly Build | ||||||||
| Hardware: | Unspecified | ||||||||
| OS: | Unspecified | ||||||||
| See Also: | https://bugs.webkit.org/show_bug.cgi?id=203240 | ||||||||
| Attachments: |
|
||||||||
|
Description
Alicia Boya García
2019-04-24 15:28:53 PDT
That's going to be "ScalableImageDecoder: don't forcefully decode image data when querying frame completeness, duration" again Several tests are failing on the WPE bot as well, and as Michael comments, the culprit seems to be r244372 "ScalableImageDecoder: don't forcefully decode image data when querying frame completeness, duration" fast/images/icon-decoding.html [ Failure ] fast/images/reset-image-animation.html [ Failure ] fast/images/slower-animation-than-decoding-image.html [ Failure ] fast/images/slower-decoding-than-animation-image.html [ Failure ] fast/images/animated-image-loop-count.html [ ImageOnlyFailure ] fast/images/animated-webp.html [ ImageOnlyFailure ] I've been giving a look to this. There are 3 different issues. First: fast/images/icon-decoding.html just needs new baseline Second: https://github.com/WebKit/webkit/blob/1240cfe9a01ee4469fdaaacdf3056c4401b01b24/Source/WebCore/platform/image-decoders/ScalableImageDecoder.cpp#L205 is causing that even non complete frames have a duration of 100ms. We should return 0_s if !frame.isComplete(), and avoid the 11ms change. This fixes these tests: fast/images/reset-image-animation.html fast/images/slower-animation-than-decoding-image.html fast/images/slower-decoding-than-animation-image.html fast/images/animated-image-loop-count.html Then css3/shapes/shape-outside/shape-image/shape-image-025.html imported/w3c/web-platform-tests/css/css-shapes/shape-outside/shape-image/shape-image-025.html are a bit trickier. They are testing that when using shape-outside on an animated image, the shape must be generated from the first frame of the animation. I've been debugging the code and I can see how the shape is being generated for each frame of the animation (both before and after the regression), and I think that these were passing before just by chance. The thing is that just decoding the first frame would be enough for the shape image, but nothing indicates that and the complete animation is decoded. With each decoded frame, RenderBox::imageChanged() ends up being invoked and that causes the shape to be marked as dirty, so the next time we want to use it we'll generate it again from the new frame. I think we should somehow indicate that we don't want to decode the full animation when drawing the first frame (BitmapImage::draw()), but I need to check how the Apple ports deal with this, as it's not failing there. I'm going to split this bug into 2, cause there are 2 issues here, one in ScalableDecoder and another one that seems to be in the multiplatform code handling the shape-outside property. I leave this bug for the ScalableImageDecoder problem and I've created https://bugs.webkit.org/show_bug.cgi?id=203240 to handle the (possible) multiplatform issue. *** Bug 197501 has been marked as a duplicate of this bug. *** Created attachment 381541 [details]
Patch
Comment on attachment 381541 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=381541&action=review There is a small typo in the ChangeLog; please fix it before landing. > Source/WebCore/ChangeLog:10 > + Covered by existent tests. s/existent/existing/ Created attachment 381543 [details]
Patch
Comment on attachment 381543 [details] Patch Clearing flags on attachment: 381543 Committed r251431: <https://trac.webkit.org/changeset/251431> All reviewed patches have been landed. Closing bug. |