on SnowLeopard: http://build.webkit.org/results/SnowLeopard%20Intel%20Release%20%28Tests%29/r54766%20%285575%29/media/video-display-aspect-ratio-pretty-diff.html on Leopard: http://build.webkit.org/results/Leopard%20Intel%20Release%20%28Tests%29/r54766%20%2810624%29/media/video-display-aspect-ratio-pretty-diff.html on Windows: http://build.webkit.org/results/Windows%20Release%20%28Tests%29/r54766%20%289057%29/media/video-display-aspect-ratio-pretty-diff.html On Leopard the width difference is only 1 pixel, maybe we could have platform-specific results for it. I will skip the test on these platforms for now.
The commit that introduced the new test is http://trac.webkit.org/changeset/54766
Test skipped as of r54768
Hmm, it also fails on Tiger now. When I look at the test movie with QuickTime, it reports 427 as width. Are you sure 426 is right? Also on my leopard box, it reports 427.
I think the movie is the problem here, it has a width 426.666. The width field in the video track header is 01 AA AA AA. Track width and height are 16.16 fixed point, so 0x01AA -> 426 and 0xAAAA/0x00010000 -> 0.6666564941. I don't know what the rules are for rounding a number like this, but I wouldn't be surprised if they are ambiguous. What is this movie supposed to test?
It is meant to test that the display aspect ratio (DAR) of the video is applied correctly during naturalSize() calculation. In the GStreamer player we come with 426 by applying the PAR (4:3) to the video resolution and rescaling against the calculated DAR. The GStreamer qt demuxer correctly retrieves the display resolution but doesn't advertize it downstream in the pipeline. Instead it calculates the PAR using this, as a int/int fraction, hence the loss of precision.
Created attachment 48945 [details] Proposed patch
Patch looks fine, but don't you also want to modify the skipped lists? Most platforms skip the test because of inconsitent results..
Philippe, I won't be able to get back to this until tomorrow my time so please feel free to take it from here if you want.
Created attachment 48982 [details] updated patch
(In reply to comment #9) > Created an attachment (id=48982) [details] > updated patch I unskipped the test only on Leopard. On SnowLeopard and Windows it is still skipped because the test results are much more different from the expected results.
Comment on attachment 48982 [details] updated patch OK.
Landed in r55125. Should we keep the bug opened nevertheless? Test is still skipped on SnowLeopard and Windows.
Comment on attachment 48982 [details] updated patch clearing flags, patch was commited
Closing as the patch was commited. See Bug 35297 from now on ;)