Mozilla/5.0 (Windows NT 6.1; WOW64; rv:39.0) Gecko/20100101 Firefox/39.0 ID:20150318055750 CSet: 41a61514461e 1. New profile, start firefox 2. Navigate to http://imgur.com/gallery/RaGvZ 3. Observe as the images are decoding, there are horizontal black lines across the image, which will disappear periodically and after decoding has completed 4. Click on one of the images to increase its size 5. Observe a white flash the screen as the image reaches its maximum size. The same effect occurs when you click on the image a second time to reduce its size back to what it was. CTRL+F5 will show the issue again if you need to see it. I assume both these problems are related; maybe they should have been filed separately. If I set gfx.vsync.compositor and gfx.vsync.hw-vsync.enabled to false, the problem still persists. If I disable hardware acceleration the problem still persists.

I just tried to reproduce and couldn't see the issue here on OS X. I'll try again on Windows later today.

Only one person on Mozillazine forum has seen it too, most others cannot reproduce. I include my graphics details: Adapter Description NVIDIA GeForce GTX 660 Ti Adapter Drivers nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um Adapter RAM 2048 Device ID 0x1183 Direct2D Enabled true DirectWrite Enabled true (6.2.9200.16571) Driver Date 2-5-2015 Driver Version 9.18.13.4752 GPU #2 Active false GPU Accelerated Windows 1/1 Direct3D 11 (OMTC) Subsys ID 36623842 Vendor ID 0x10de WebGL Renderer Google Inc. -- ANGLE (NVIDIA GeForce GTX 660 Ti Direct3D11 vs_5_0 ps_5_0) windowLayerManagerRemote true AzureCanvasBackend direct2d 1.1 AzureContentBackend direct2d 1.1 AzureFallbackCanvasBackend cairo AzureSkiaAccelerated 0 If it's worth anything I'm on a 120Hz monitor; I have tried changing my refresh rate back to 60Hz but it didn't make a difference. Tested on rev 2e2222a40262 and still same problem.

[Tracking Requested - why for this release]: Pushlog: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=161eb635963d&tochange=d45aaa6d2217 Triggered by: Bug 1124084 image.downscale-during-decode.enabled = false helps

Steve, I can reproduce the white flash on Windows, but not the black lines. Could you take a screenshot of the black lines and attach it to the bug? Seeing exactly what they look like might give a clue about where the problem is.

So the white flash is caused by a bug in the patch for bug 1130802. We were supposed to always prefer decoded surfaces, but we don't check whether the "perfect match" check in LookupBestMatch resulted in a surface that was decoded. This means that if we want to draw at 1000x1000 and we have a partially decoded surface of size 1000x1000 and a fully decoded surface of size 900x900, we'll return the partially decoded 1000x1000 surface, which is clearly wrong.

Here's the patch.

Thanks for the quick review! Setting checkin-needed because I think fixing this is pretty urgent, but inbound is closed right now.

Here's a try job (Linux-only, since this has been tested locally at least to some degree on Windows and OS X already): https://treeherder.mozilla.org/#/jobs?repo=try&revision=878574106c65

The white flash problem is fixed on latest m-i build(include 29a208a74d8f). However, The black line problem is STILL reproducible on the build. https://hg.mozilla.org/integration/mozilla-inbound/rev/e4f2552c20fa Mozilla/5.0 (Windows NT 6.1; WOW64; rv:39.0) Gecko/20100101 Firefox/39.0 ID:20150319010140

I'm also seeing the black lines (Windows, latest nightly, e10s, NVIDIA Quadro 1000M): http://img.ctrlv.in/img/15/03/19/550adec07a871.jpg They disappear once the image has completed loading.

And if it helps - the black lines occur at every point where you could see the progressive loading of the image pause for a second. eg: N more pixels vertically load <black line inserted> then next N pixels <another black line> etc.

Given that Alice has confirmed the white-flash problem is fixed in comment 11, do I need to refile the 'black lines problem' as another bug and adjust the title of this for the white-flash problem? Clearing the needinfo flag as Alice has provided a screenshot in comment 9

(In reply to Steve England [:stevee] from comment #14) > Given that Alice has confirmed the white-flash problem is fixed in comment > 11, do I need to refile the 'black lines problem' as another bug and adjust > the title of this for the white-flash problem? Probably.

The follow up is in bug #1145560.