I haven't yet had a chance to look into the differences between the images, but reftests that match background images to background colors that should be identical, and that work on Mac and Linux, fail on Windows.
Attachment #256008 - Attachment mime type: text/plain; charset=windows-1252 → text/html; charset=windows-1252
So it turns out they also fail on the Linux tinderbox, so it's not platform...
Summary: reftests matching background images to background colors fail on Windows → reftests matching background images to background colors fail on some machines
(I'd note that it's possible I made the PNGs wrong, although I'm not sure how to tell. I also plan to look at the differences more closely tomorrow.)
(In reply to comment #3) > Created an attachment (id=256009) [details] > reftest output from Linux testing tinderbox In the first failure in this log, the images come out at #008200 instead of #008000. The reference comes out (correctly) as #008000.
(In reply to comment #1) > Created an attachment (id=256008) [details] > reftest output Same for the first failure in this (Windows) log -- #008200.
Note that I've now worked around this for all but one of the tests by switching to blue (#0000ff). The only test that still shows the problem is check-image.html / check-image-ref.html. (I made a separate check-image-blue.html / check-image-blue-ref.html that's marked as passing for that one case.) So it's no longer blocking getting other tests to pass.
This test isn't failing anymore, so I removed the 'random' marking. But the underlying bug may just be covered up by bug 371867.
My best guess is that this may have had to do with color management. Do the PNG's have a color profile? Also, is this still a problem?
You need to log in before you can comment on or make changes to this bug.