Bug 216915
Summary: | Test WebGL draw video to canvas software path on iOS | ||
---|---|---|---|
Product: | WebKit | Reporter: | Kimmo Kinnunen <kkinnunen> |
Component: | WebGL | Assignee: | Nobody <webkit-unassigned> |
Status: | NEW | ||
Severity: | Normal | CC: | dino, kbr, webkit-bug-importer |
Priority: | P2 | Keywords: | InRadar |
Version: | Other | ||
Hardware: | iPhone / iPad | ||
OS: | iOS 13 | ||
See Also: | https://bugs.webkit.org/show_bug.cgi?id=205734 | ||
Bug Depends on: | 218970, 215908, 218637 | ||
Bug Blocks: | 198948, 218971 |
Kimmo Kinnunen
Test WebGL draw video to canvas software path on iOS
In bug 215908 we had a regression where webgl texImage2D with video did not produce output.
The problem was that GPU YUV->RGB codepath had a bug.
However, the software conversion codepath should have taken effect, but in fact just produced transparent texture.
Interestingly normal "MP4" file worked, where as a HLS stream did not. It's unsure if this is a factor.
The problem should have been caught with simulator testing and/or real hardware testing.
Attachments | ||
---|---|---|
Add attachment proposed patch, testcase, etc. |
Radar WebKit Bug Importer
<rdar://problem/69824545>
Kimmo Kinnunen
In bug 218637 we had a regression where texImage2D with video did not produce output.
The problem was that the video decoder produced a frame in a format that was unexpected by yuv->rgb conversion codepath. However, the software conversion codepath should have taken effect, but in fact just produced transparent texture.
Kenneth Russell
On Bug 218971 I just suggested to see also Bug 205734. A couple of the existing layout tests seem to have videos that (maybe accidentally) go down the software decoding path at least on macOS. It would be great to add some internal method to the WebKitTestRunner to force the software decoded path for testing purposes.