Add WebM VP9/VP9 Alpha tests
A new field was introduced, codec-alpha, this is now produced by vp9parse. Also add both VP8 and VP9 streams to test the playback.
Related to:
Merge request reports
Activity
mentioned in merge request gst-plugins-bad!2199 (merged)
added 1 commit
- f7b8a12d - Add medias to cover VP8/VP9 WebM with alpha channels
added 1 commit
- f7c85fd3 - Add medias to cover VP8/VP9 WebM with alpha channels
added 1 commit
- 36bed6a0 - Add medias to cover VP8/VP9 WebM with alpha channels
added 1 commit
- 9b2cf1d8 - Add medias to cover VP8/VP9 WebM with alpha channels
I guess I'm starting to get it, you can't just
~
in the test list, you have to add them to the blacklist. Now, I didn't want the transcode to be enabled, since it made little sense to transcode to opaque, but then though maybe it should work. VP8 works, it's a video only WebM, but VP9 is failing here. Note sure yet what is happending. First meaningful warning indicates a bug in matroskademux when dealing with audio:0:00:00.166745760 275929 0x1644b60 WARN validate gst-validate-reporter.c:201:gst_validate_report_valist: <matroskademux0:audio_0> 2487 (warning) : buffer: buffer was received before a segment : Received buffer before Segment event
- Resolved by Nicolas Dufresne
Ok, though if I disable vp9alphadecodebin it works, so it's arguably something specific to this setup. I'll investigate. In fact, now I notice that vp9dec0 (the main video stream decoder) fails silently.
0:00:07.613332354 278167 0x7fcc340289e0 DEBUG GST_SCHEDULING gstpad.c:4443:gst_pad_chain_data_unchecked:<vp9dec0:sink> called chainfunction &0x7fcc7e13c2c0 with buffer 0x7fcc300b3c60, returned error
added 1 commit
- 5d47806f - Add medias to cover VP8/VP9 WebM with alpha channels
So,
validate.file.playback.seek_with_stop.opus_vp9-alpha_webm
failed first and got retried, this was real failure, but racy. I made a fix for this one.The remaining,
validate.file.transcode.to_mp3_and_h264_in_mp4.opus_vp9-alpha_webm
is an unrelated bug, I haven't digged yet, but the opus caps are reset to mono somewhere and gst-validate hard fails on that.- Resolved by Nicolas Dufresne
This really looks like a false positive, gst-validate complains because opusdec took care of converting from 2 to 1 channels. I'll keep looking, but perhaps there was two set_caps() or something like this ?
mentioned in commit ndufresne/gst-plugins-base@9df5b172
mentioned in merge request gst-plugins-base!1142 (merged)
mentioned in commit ndufresne/gst-plugins-base@259542a6
mentioned in commit ndufresne/gst-plugins-base@848f9985
mentioned in commit ndufresne/gst-plugins-base@725763f8