+++ This bug was initially created as a clone of Bug #1124543 +++ Bug 1124543 does not fix the graphical glitch Steps to reproduce: 1. Extract attachment 8552868 [details] 2. Open index.html Actual results: See attached screenshot See attachment 8552867 [details] Pushlog: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=2a61df4eaa2d&tochange=24ba8274ed60 In local build(win32 non pgo) Last Good: 2a61df4eaa2d First Bad: 823227372483 Regressed by 823227372483 Bas Schouten — Bug 1107297: Only recomposite the damaged rect with D3D11. r=jrmuizel
3 years ago
This works for me in the latest nightly on Windows 7, D2D1.1/D3D11. Alice, do you still see this problem?
For the original bug, we concluded (bug 1124543 comment 4) "...layout incorrectly reporting the damaged rect to the layer manager...", and some of it got fixed. Bas, is there a good way to confirm this is still the case with the remaining issue? For example, if that was the case, would you expect to see the same problem without hardware acceleration?
(In reply to Milan Sreckovic [:milan] from comment #3) > For the original bug, we concluded (bug 1124543 comment 4) "...layout > incorrectly reporting the damaged rect to the layer manager...", and some of > it got fixed. Bas, is there a good way to confirm this is still the case > with the remaining issue? For example, if that was the case, would you > expect to see the same problem without hardware acceleration? Not really actually. Without hardware acceleration I believe GDI will automatically 'clip' to the actual damaged area which could lead to significantly different symptoms.
I am experiencing this bug in Firefox 47.0 on GNU/Linux. I'm not sure whether to report my issues as another bug, or simply comment here. I figure I will post here first since it doesn't hurt. I'm experiencing the exact same effect as what is shown in the attached screenshot from comment #0, but I am also getting these weird graphical glitches with the animation that plays when downloads finish and with the red loading bar at the top of YouTube. Oddly, though, the specific graphical glitch shown in comment #0 doesn't happen in fullscreen.
Not likely to be the same problem, given that the above was D2D specific, but anything is possible. Could you get a regression range using mozregression tool and let us know what you get? In this bug is fine, we can split them off easily enough.
Okay, so I tried getting a regression range using mozregression for a couple of versions back, but I didn't get my specific issue to occur in any of them. However, while the bug described in comment #0 doesn't occur in any of the >= 40 versions, it does occur in all of them when using a non-e10s window. For my more specific issues, I can confirm that the issue seems to have first occurred in Firefox 47.0. I have tested many of the official releases, and none of the earlier releases before 47 have the issue. I am looking at finding a way to reliably reproduce it.