With attachment 8380443 [details] [diff] [review] (bug 975928) applied, I also get this test failure: failed | bug516323.indexed.ogv: First range end should be media end - got 4.208333, expected 4.208 It turns out that the index in bug516323.indexed.ogv has a timestamp denominator of 1000; so its timestamps are in milliseconds. Whereas our media stack now has microsecond precision. So the index is reporting a duration of 4.208 seconds. Whereas our Ogg GetBuffered() code doesn't use the index to calculate the buffered ranges, it examines the cached ranges and parses them to extract timestamps. So it gets a more precise duration, and our test fails due to the difference. We can prevent this failure by increasing the precision of bug516323.indexed.ogv's index.
I updated the index by patching OggIndex to write indexes with a timestamp denominator of 100000 instead of 1000. So I'll need to upstream that patch. I'm not even sure how to do that now.
Attachment #8380445 - Flags: review?(giles)
Comment on attachment 8380445 [details] [diff] [review] Patch: Update bug516323.indexed.ogv's index Upstream for OggIndoex seems to be https://git.xiph.org/?p=OggIndex.git You should have commit access. So just push the patches to email@example.com:OggIndex.git If you don't still have access, send me a new key, or attach the patch here and I'll push it upstream.
Attachment #8380445 - Flags: review?(giles) → review+
Pushed patch upstream. Thanks!
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla30
You need to log in before you can comment on or make changes to this bug.