Bug 282791
| Summary: | [GStreamer][MSE] media/media-source/media-source-vp8-hiddenframes.html is a permanent failure | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Jean-Yves Avenard [:jya] <jean-yves.avenard> |
| Component: | Media | Assignee: | Philippe Normand <philn> |
| Status: | REOPENED | ||
| Severity: | Normal | CC: | bugs-noreply |
| Priority: | P2 | ||
| Version: | WebKit Nightly Build | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
Jean-Yves Avenard [:jya]
in bug 278833, I removed the automatic skip that incorrectly disabled the test on all platform. The reftest indicates that the GStreamer MSE player produces garbage output for vp8 content that has alternative reference frames.
https://ews-build.s3-us-west-2.amazonaws.com/GTK-WK2-Tests-EWS/ca015363-67461-stress-mode/results.html
diff: https://ews-build.s3-us-west-2.amazonaws.com/GTK-WK2-Tests-EWS/ca015363-67461-stress-mode/media/media-source/media-source-vp8-hiddenframes-diff.png
alternative reference frames are typically packed together in a webm block and have all the same timestamps.
On cocoa platforms, one issue was that when separately added to the SourceBuffer's TrackBuffer, the last frames of the block override the earlier one as the timestamp is used as key.
A workaround we added to our webm demuxer was to combined all MediaSamples in a single MediaSample, and split it up again when decoding.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Philippe Normand
I think this is a duplicate of bug 283457.
*** This bug has been marked as a duplicate of bug 283457 ***
Philippe Normand
Locally this test is passing but not on bots. Re-opening :/