Created attachment 456893 [details] log After enabling reftests for Direct 2D we are hitting this failure in this test suite.
Argh, everything from comment 2 up to here is actually a different bug from this. Note that this bug is filed for a D2D-specific issue, and all the logs since comment 2 are non-D2D (you can tell because of the "-opt-u-reftest" in the test filename, rather than "opt-u-reftest-d2d") Sorry for starting the cascade of spam. This bug here is d2d-specific, and it's characterized by a 48px mismatch around the happy face's nose. The non-D2D bug (the logs since comment 2) have only a 4px mismatch (based on just one of the logs -- haven't looked at all of them), and the mismatching pixels are at various points in the page's "blank space". Filing a new bug for the non-D2D issue...
Filed Bug 579548 for the non-D2D issue here (comment 2 thru 8). For now (until win7+D2D is un-hidden on the tbpl page), any failures in this test on win7 that show up on tbpl should be noted on Bug 579548, not here.
Comments 10 through 15 are all non-d2d and are hence all Bug 579548, not this bug.
Comment 17 and 18 are also Bug 579548. /me updates summary to make this bug less inviting.
Seems like this just caused by rasterization differences with direct2d.
Switching the summary, now that non-D2D is what you won't ever be seeing.
Well, and I guess I could have shown I was paying minimal attention by switching the nature of what's UNEXPECTED, since the problem is that we expect it to fail if(d2d), but it doesn't always fail for debug Win7, only sometimes.
Created attachment 499783 [details] [diff] [review] random-if I don't think we're learning anything by letting the random passes be unexpected, and I'm pretty sure that the annotation pointing to a stale bug about multiple failures which no longer exist isn't helping.
Comment on attachment 499783 [details] [diff] [review] random-if Sounds good to me.