On Nightly, the contents of the first Macbook Pro screen on http://www.apple.com/macbook-pro/ are distorted. They're applying a mask image / clip path to the video, which means that it's going to be rendered on the main thread in Firefox.

[Tracking Requested - why for this release]: regression This is a regression from bug 1273823.

Yea, I confirmed the problem on my mac book pro. Stride calculation seems wrong with the video.

attachment 8805477 [details] [diff] [review] addressed the problem on my mac book pro. Before bug 1273823 fix, data source's stride was handled as aligned to 32.

Tracking 52+ for this regression so we can correct stride circulation.

Pushed by sikeda@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/9b5b04d1a15a Handle width alignment of 32 r=nical

Backed out for failing gl tests test_conformance__textures__misc__texture-npot-video.html and test_2_conformance__textures__video__tex-2d-rgb-rgb-unsigned_byte.html, and dom/media/test/test_bug879717.html: https://hg.mozilla.org/integration/mozilla-inbound/rev/15bfa2b75e1bada9920d9d7dfe0522c845d1d057 Push with failures: https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=9b5b04d1a15a7c238fe40093f5167dcab34e5ef2 Please check the gl and mda logs on OS X. There might be more failing tests than in the commit message.

Track 51+ as regression.

Hmm, we might need to limit optimization path only when width is aligned to 32.

(In reply to Sotaro Ikeda [:sotaro] from comment #12) > Hmm, we might need to limit optimization path only when width is aligned to > 32. Data is aligned to 32, but the data might not be mapped correctly.

(In reply to Sotaro Ikeda [:sotaro] from comment #13) > (In reply to Sotaro Ikeda [:sotaro] from comment #12) > > Hmm, we might need to limit optimization path only when width is aligned to > > 32. > > Data is aligned to 32, but the data might not be mapped correctly. And current libyuv seems to have a problem around crop handling. attachment 8806586 [details] [diff] [review] is going to bypass it for now.

Pushed by sikeda@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/93515fab001d Revert Bug 1273823 if width is not aligned to 32 r=nical

Hi :sotaro, Since this is a regression in 51, do you think the patch is worth uplifting to 51?

Comment on attachment 8806586 [details] [diff] [review] patch - Revert Bug 1273823 if width is not aligned to 32 Approval Request Comment [Feature/regressing bug #]: bug 1273823 [User impact if declined]: Some h264 video is not rendered correctly when BasicCompsitor is used. [Describe test coverage new/current, TreeHerder]: locally tested. [Risks and why]: Low. The patch just partially back-out bug 1273823 fix. [String/UUID change made/needed]: none.

Comment on attachment 8806586 [details] [diff] [review] patch - Revert Bug 1273823 if width is not aligned to 32 Fix a regression related to video rendering. This will affect user experiences. Take it in 51 aurora.

Comment on attachment 8866270 [details] xss.html (marking unrelated xss.html files as obsolete)