1122031 - Rendering issue with Master Slider plugin in Firefox 35 on OS X Yosemite

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect)

Description

[olau.apm]

Attachment: Screen Shot 2015-01-15 at 16.54.32.png

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_1) AppleWebKit/600.2.5 (KHTML, like Gecko) Version/8.0.2 Safari/600.2.5

Steps to reproduce:

Firefox 35 on OS X Yosemite v10.10.1
- visit this website which is a demo site for a Wordpress theme using the popular Master Slider plugin http://www.industrialthemes.com/explicit/
- check the master slider area as shown in the attached picture


Actual results:

cells on the left of the master slider disappear after a while. They reappear when using the mouse to scroll the cells up/down


Expected results:

cells should not disappear. This is working fine on latest Chrome and Safari browsers . Problem is only with Firefox.

Comment 1

[:Gijs (he/him)]

Unfortunately, I can't reproduce on 10.9 with Firefox 35. Can you reproduce if you restart Firefox in safe mode (Help > Restart with add-ons disabled) ?

Comment 2

[olau.apm]

I restarted Firefox in safe mode and the problem cannot be seen anymore. The cell sliding is sluggish though, looks like some hardware acceleration is not used in safe mode, and probably this is where the problem comes from. Indeed if I open the debugger and remove some CSS translations at runtime the non-visible content reappears.
Also, while not being in safe mode, I disabled all extensions and plugins and this didn't solve the problem.
So I think the issues is related to some hardware acceleration used for rendering which is enabled in normal mode but disabled in safe mode.
You don't have the problem on 10.9 so maybe it is only on 10.10.

Comment 3

[olau.apm]

I installed Firefox 34.0.5 on the same machine and have no problem -> this is a regression in Firefox 35. As this is a rendering problem, the issue may be met in other areas too (to be confirmed). Also the bug does not appear on Firefox 35 running on Windows 8.1. So far it looks like an issue on OS X Yosemite only.

Comment 4

[:Gijs (he/him)]

Do you have time to use http://mozilla.github.io/mozregression/ to track down exactly where between 34 and 35 this broke? That would be really helpful. Assuming you have a relatively fast connection and the problem is quickly visible on the page you cited, finding the regression range will take maybe 10-15 minutes. You can use this invocation:

mozregression --good 2014-01-01

to be a little bit on the safe side in terms of when this still worked (in case we enabled a feature only on nightly but not on our release builds...)

Comment 5

[:Gijs (he/him)]

Oh, also, could you go to about:support and copy/paste the "graphics" section into this report so we know what gpu/driver/display combination you're seeing this with? Thanks!

Comment 6

[olau.apm]

here is the content of the graphics section of about:support on my machine (new Mac Pro). About the mozregression stuff I'll try to take a look but no time at this moment.

Device ID	0x6798
GPU Accelerated Windows	1/1 OpenGL (OMTC)
Vendor ID	0x1002
WebGL Renderer	ATI Technologies Inc. -- AMD Radeon HD - FirePro D700 OpenGL Engine
windowLayerManagerRemote	true
AzureCanvasBackend	quartz
AzureContentBackend	quartz
AzureFallbackCanvasBackend	none
AzureSkiaAccelerated	0

Comment 7

[:Gijs (he/him)]

(In reply to olau.apm from comment #6)
> here is the content of the graphics section of about:support on my machine
> (new Mac Pro). About the mozregression stuff I'll try to take a look but no
> time at this moment.
> 
> Device ID	0x6798
> GPU Accelerated Windows	1/1 OpenGL (OMTC)
> Vendor ID	0x1002
> WebGL Renderer	ATI Technologies Inc. -- AMD Radeon HD - FirePro D700 OpenGL
> Engine
> windowLayerManagerRemote	true
> AzureCanvasBackend	quartz
> AzureContentBackend	quartz
> AzureFallbackCanvasBackend	none
> AzureSkiaAccelerated	0

Thanks!

I rebooted into yosemite (separate partition) and I can't repro on my machine (early 2013 retina MBP using a 0x166 NVIDIA GT 650M). I expect it's related to the card / driver. :-(

Markus, do you have ideas what could be going on here and/or do you have a machine with a similar GPU? (Also CC'ing Milan because that's what I normally do with graphics issues...)

Comment 8

[olau.apm]

I ran the regression tool:
19:52.73 LOG: MainThread Bisector INFO Last good revision: 0753f7b93ab7
19:52.73 LOG: MainThread Bisector INFO First bad revision: 3be45b58fc47
19:52.73 LOG: MainThread Bisector INFO Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=0753f7b93ab7&tochange=3be45b58fc47

Comment 9

[:Gijs (he/him)]

(In reply to olau.apm from comment #8)
> I ran the regression tool:
> 19:52.73 LOG: MainThread Bisector INFO Last good revision: 0753f7b93ab7
> 19:52.73 LOG: MainThread Bisector INFO First bad revision: 3be45b58fc47
> 19:52.73 LOG: MainThread Bisector INFO Pushlog:
> https://hg.mozilla.org/mozilla-central/
> pushloghtml?fromchange=0753f7b93ab7&tochange=3be45b58fc47

Thank you!

Comment 10

[Markus Stange [:mstange]]

I can't reproduce this, unfortunately, and my GPU is a different one (looks like I have exactly the same machine as you do, Gijs). Not sure what to do with this. The regression range contains some stuff by Matt Woodrow, CC'ing him.

Comment 11

[olau.apm]

Attachment: The bug in action

attaching video showing bug reproduction on nightly build http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2014/09/2014-09-06-03-02-04-mozilla-central/firefox-35.0a1.en-US.mac.dmg

Comment 12

[olau.apm]

I could reproduce the problem on another Mac, this time a MacBook Pro 2011 with the following specs:

Fenêtres avec accélération graphique    1/1 OpenGL (OMTC)
ID du périphérique    0x 116
ID du vendeur    0x8086
Rendu WebGL    ATI Technologies Inc. -- AMD Radeon HD 6490M OpenGL Engine
windowLayerManagerRemote    true
AzureCanvasBackend    quartz
AzureContentBackend    quartz
AzureFallbackCanvasBackend    none
AzureSkiaAccelerated    0

To reproduce the problem, please check the attached video. You need to left-click around the area that scrolls (like on top or left of it) then go into the area and use the mousewheel, then back outside and click left-click around a few times, also try to move the mouse into other areas. After a few attempts the problem should show up.

Comment 13

[olau.apm]

I have narrowed down the set of changes to only a few of them, last time I ran the regression tool it stopped due to timeout, this time I ran it until the end:

271:12.41 LOG: MainThread Bisector INFO Oh noes, no (more) inbound revisions :(
271:12.41 LOG: MainThread Bisector INFO Last good revision: a8f2b6d4fa2c
271:12.41 LOG: MainThread Bisector INFO First bad revision: c4cb0bd2dbf1
271:12.41 LOG: MainThread Bisector INFO Pushlog:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=a8f2b6d4fa2c&tochange=c4cb0bd2dbf1

Comment 14

[:Gijs (he/him)]

(In reply to olau.apm from comment #13)
> I have narrowed down the set of changes to only a few of them, last time I
> ran the regression tool it stopped due to timeout, this time I ran it until
> the end:
> 
> 271:12.41 LOG: MainThread Bisector INFO Oh noes, no (more) inbound revisions
> :(
> 271:12.41 LOG: MainThread Bisector INFO Last good revision: a8f2b6d4fa2c
> 271:12.41 LOG: MainThread Bisector INFO First bad revision: c4cb0bd2dbf1
> 271:12.41 LOG: MainThread Bisector INFO Pushlog:
> https://hg.mozilla.org/integration/mozilla-inbound/
> pushloghtml?fromchange=a8f2b6d4fa2c&tochange=c4cb0bd2dbf1

Looks like another victim of bug 982338. :-(
Matt, do you have time to take a look?

Comment 15

[olau.apm]

this issue seems to be fixed in Firefox 35.0.1

Comment 16

[ovidiu boca[:Ovidiu]]

Based on Comment 15 from the reporter, changing status to RESOLVED WORKS FOR ME. 

I also tested this issue on Mac OS 10.10 with Firefox Nightly 48.0a1 and Firefox release 45 and I can't reproduce this issue, unfortunately I discovered another bug that I will log.