User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:25.0) Gecko/20100101 Firefox/25.0 (Beta/Release) Build ID: 20131025151332 Steps to reproduce: Fuzzy font when scrolling in flyoutpanel

I've been seeing this on scrollable form drop downs as well. Also probably the same as the overlay button rendering bug 936609.

Bug 936500 is also likely the same bug.

regression range: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=770de5942471&tochange=9ba3faa35c96

Some potential suspects below. I don't see much in the way of gfx landings though. Kats had a couple apzc landings but this rendering bug shows up in scrollable chrome vs. apzc content. Matt any chance your landings might have caused this bug or bug 936609? Matt Woodrow — Bug 923434 - Use DrawSurface in ThebesLayerBuffer instead of FillRect. Matt Woodrow — Bug 926023 - Use the value of SetPermitSubpixelAA for DrawTargetCairo.

I backed out "Don't use operator source for D2D" locally and confirmed it was the cause of this particular bug.

Bas, any idea why this might happen? I can't see anything wrong with the changes, it seems like it should be identical.

(In reply to Matt Woodrow (:mattwoodrow) from comment #6) > Bas, any idea why this might happen? I can't see anything wrong with the > changes, it seems like it should be identical. These fonts aren't fuzzy.. they're non-antialiased. The opposite of fuzzy, I'm going to think about this a little more.

(In reply to Bas Schouten (:bas.schouten) from comment #7) > (In reply to Matt Woodrow (:mattwoodrow) from comment #6) > > Bas, any idea why this might happen? I can't see anything wrong with the > > changes, it seems like it should be identical. > > These fonts aren't fuzzy.. they're non-antialiased. The opposite of fuzzy, > I'm going to think about this a little more. Bas this blocks our merge for fx28. Can we maybe back out the offending patch from bug 923434? Or if def it maybe? Alternatively if you think we can come up with a fix that can land on aurora, we can bump this out to relerase28.

Matt, the original bug got uplifted to aurora, which means that it will be beta in a couple of days. Is this regression less of an issue, or do we think we should need to uplift this fix to match?

I think we should uplift for this, it's a fairly bad regression in some cases. The fix is fairly simple too.

Is this fixed on m-c? I suspect it will have been fixed by bug 946475.

(In reply to Matt Woodrow (:mattwoodrow) from comment #12) > Is this fixed on m-c? I suspect it will have been fixed by bug 946475. yes it does.