User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/68.0.3440.84 Safari/537.36 Steps to reproduce: Playback of the test video below with mixed video sizes (webcam: 640x360, video 960x540, screenshare 1904x1088) with hardware acceleration. Test video: https://cdn.ubivent.com/web/test/firefox/videotest.html Actual results: The video is displayed in the maximum dimensions found. Wenn a smaller segment is played, the excess area is green. Screenshots: https://cdn.ubivent.com/web/test/firefox/firefox_video_hardware_acceleration_largest_segment.jpg https://cdn.ubivent.com/web/test/firefox/firefox_video_hardware_acceleration_small_segment.jpg https://cdn.ubivent.com/web/test/firefox/firefox_video_hardware_acceleration_smallest_segment.jpg Firefox Nightly and Beta have same results. Expected results: The video element should have adjusted its size to the current video segment and there should be no green area around the video. If hardware acceleration is turned off, the issue does not occur. Screenshot without acceleration: https://cdn.ubivent.com/web/test/firefox/firefox_video_smallest_segment.jpg Chrome with hardware acceleration works as expected. User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:61.0) Gecko/20100101 Firefox/61.0 Hardware information: GPU #1 Active: Yes Description: Intel(R) HD Graphics 620 Vendor ID: 0x8086 Device ID: 0x5916 Driver Version: 23.20.16.4973 Driver Date: 2-28-2018 Drivers: igdumdim64 igd10iumd64 igd10iumd64 igd12umd64 igdumdim32 igd10iumd32 igd10iumd32 igd12umd32 Subsys ID: 224517aa RAM: Unknown More information in about:support attachment.

I changed the original test Video, as the audio was broken. The screenshots are from the original video.

i am having the same issue. the provided test video shows this: https://www.seuffert.biz/tmp/ff-webm-fail.jpg GPU #1 Active Yes Description Intel(R) UHD Graphics 620 Vendor ID 0x8086 Device ID 0x5917 Driver Version 24.20.100.6194 Driver Date 6-20-2018 Drivers igdumdim64 igd10iumd64 igd10iumd64 igd12umd64 igdumdim32 igd10iumd32 igd10iumd32 igd12umd32 Subsys ID 225b17aa RAM Unknown Firefox 62.0b15 Windows_NT 10.0 The issue does not occur If i switch to the second GPU (NVIDIA GeForce MX150).

technically, those webm are invalid...

We need something similar to the H264Converter for VP9/VP8, also required for webrtc....

The source of the webm file is a MediaSource stream where the segments are merged together into a webm file. The same issue also appears in the original MediaSource stream. The idea was that a single webm is easier to test than a MediaSource stream.

That doesn't make for a valid webm... If you want to use MediaSource and adding a new webm segment containing a different resolution then you *must* insert an init segment first (https://w3c.github.io/media-source/webm-byte-stream-format.html#webm-init-segments) If getting the situation you describe with MSE, you must not be properly adding the init segment. Which would still be an invalid use of MSE.

another observation: the issue only happens if the video starts with larger dimensions and changes to smaller dimensions.

Bug 1497951 will allow such broken streams to play

So bug 1497951 doesn't solve this particular problem seems to do with the Intel VP8 decoder itself as it returns a frame smaller than what the VP8 bytestream indicate. Would you have the segment that is 960x540 on its own by any chance?

I have included all the segments of the demo video. The [n]-ue.webm are the original files, 0-ue.webm being the init segment. The [n].webm files have the same initial init segment prepended. The issue itself is apparent in segments 5 and 11. Segments 1 to 4 are 960x540 Semgments 6 to 7 are 640x360 Segments 8 to 10 are 960x540 Semgments 12 to 13 are 640x360 https://cdn.ubivent.com/web/test/firefox/segments.zip

And use it to determine if a frame is a keyframe, its size and display size.

We also set the TrackInfoSharedPtr to each decoded sample so that on decoder that can be recycled (android) the frames are displayed with the right size. Depends on D11588

:bryce could you please action this review please?

Pushed by jyavenard@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/76b88f20b082 P1. Implement VP8/VP9 frame header parser. r=TD-Linux https://hg.mozilla.org/integration/autoland/rev/6b87d74e97c9 P2. Use new VPx frame parser to detect content change. r=padenot

https://hg.mozilla.org/mozilla-central/rev/76b88f20b082 https://hg.mozilla.org/mozilla-central/rev/6b87d74e97c9

I have tested it in the Nightly and the sample video mixed video sizes now works. Thank you!