1194358 - OMTC causes multiple tab display refresh problems on Linux

Status: RESOLVED INCOMPLETE

(Core :: Graphics, defect)

Description

[M Winiarski]

Attachment: Troubleshooting info from new profile

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:40.0) Gecko/20100101 Firefox/40.0
Build ID: 20150807095123

Steps to reproduce:

Running on Ubuntu 12.04 LTS ...
# lsb_release -a
Distributor ID:	Ubuntu
Description:	Ubuntu 12.04.5 LTS
Release:	12.04
Codename:	precise
# uname -a
Linux xxxxx-xxxxxx 3.13.0-61-generic #100~precise1-Ubuntu SMP Wed Jul 29 12:06:40 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

Upon recent upgrade from FF39.x to FF40, started encountering tab content refresh/redisplay issues when switching between multiple tabs. When switching to a new tab, many elements of the previously active tab would continue to be displayed, instead of the newly selected tab being fully rendered/refreshed/redisplayed. When in this state, mousing over parts of the page will cause some elements of the new tab to be refreshed. The tab page content can be properly displayed by doing a F5 (refresh) or by minimizing/restoring the FF window.

Also noted as part of this issue is that the scrollbar leave remnants of past position and the active tab highlighting does not correspond correctly with the truly active tab.

This was first noticed with my normal profile. I created a new profile and was able to reproduce the issue. Troubleshooting info from this new profile is attached.

Investigating on the web about what changed in FF40, I saw some discussions about OffMainThreadComposition (OMTC) being changed in Linux builds. Under the new profile, I experimented with about:config by toggling layers.offmainthreadcomposition.enabled from true to false and restarting FF. Running the same trials with multiple tabs gave me no problems with any refresh/redisplay issues -- all tabs were correctly rendered when selected as the active tab. Toggling OMTC back to true caused the issue to return.



Actual results:

Open multiple URLs in multiple tabs. Frequently switch between the tabs. Within a short time, the tab content becomes inconsistent/unsynchronized with what should be displayed in the tab -- remnants from previously selected tabs continue to be displayed. Tab content can be correctly displayed via a F5 refresh or minimize/restore of the FF window.

In about:config, toggling the layers.offmainthreadcomposition.enabled from true (default) to false clears up the issue.



Expected results:

The default setup (using a new profile) should render tab content properly without going into about:config and toggling OMTC enabled (to false).

Comment 1

[M Winiarski]

Attachment: Troubleshooting info after toggling OMTC

Comment 2

[Saulius Gurklys]

Hi,

I can confirm having the same problems, on Ubuntu 12.04 (Linux c8l 3.13.0-62-generic #102~precise1-Ubuntu SMP Wed Aug 12 14:09:54 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux) with Nvidia-352.30.

Trouble shouting info:

Application Basics
------------------

Name: Firefox
Version: 40.0
Build ID: 20150807095123
User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:40.0) Gecko/20100101 Firefox/40.0
Multiprocess Windows: 0/1 (default: false)

Graphics
--------

Adapter Description: NVIDIA Corporation -- GeForce GTX 650/PCIe/SSE2
Asynchronous Pan/Zoom: none
Device ID: GeForce GTX 650/PCIe/SSE2
Driver Version: 4.5.0 NVIDIA 352.30
GPU Accelerated Windows: 0/1 Basic
Supports Hardware H264 Decoding: false
Vendor ID: NVIDIA Corporation
WebGL Renderer: NVIDIA Corporation -- GeForce GTX 650/PCIe/SSE2
windowLayerManagerRemote: false
AzureCanvasBackend: cairo
AzureContentBackend: cairo
AzureFallbackCanvasBackend: none
AzureSkiaAccelerated: 0


Important Locked Preferences
----------------------------

JavaScript
----------

Incremental GC: true

Accessibility
-------------

Activated: false
Prevent Accessibility: 0

Library Versions
----------------

NSPR
Expected minimum version: 4.10.8
Version in use: 4.10.8

NSS
Expected minimum version: 3.19.2 Basic ECC
Version in use: 3.19.2 Basic ECC

NSSSMIME
Expected minimum version: 3.19.2 Basic ECC
Version in use: 3.19.2 Basic ECC

NSSSSL
Expected minimum version: 3.19.2 Basic ECC
Version in use: 3.19.2 Basic ECC

NSSUTIL
Expected minimum version: 3.19.2
Version in use: 3.19.2

Experimental Features
---------------------

Sandbox
-------

Seccomp-BPF (System Call Filtering): true
Seccomp Thread Synchronization: true
User Namespaces: true
Media Plugin Sandboxing: true

Comment 4

[Bruce Williams]

When does a bug go from UNCONFIRMED to NEW? It looks like several people have reproduced this bug.

Comment 5

[u279076]

It seems like all the reports of this issue are with Ubuntu 12.04. Can someone please try this with Ubuntu 14.04 and see if the issue persists? If it doesn't then I'm inclined to think it might be a compatibility issue with some library that hasn't been updated on the 12.04 branch. Doesn't mean it's not a Firefox bug but it'd be an important piece of information to have.

Additionally, it would be useful if someone can downgrade to Firefox 39 to confirm if this is a regression in Firefox 40 or not. I can provide further testing instructions once this is confirmed.

In the meantime I'll see if we can track down hardware to reproduce this locally.

Comment 6

[u279076]

Sotaro, it looks like the closest match is your Thinkpad W530 with a Quadro K1000M (same chipset family as the GTX 650 reported above). Would you be able to test this with Ubuntu 12.04 full updated with the NVidia proprietary drivers installed?

Comment 7

[johna]

I had this exact issue with Firefox 40.0.2 on my Ubuntu 12.04.05 box.  Toggling layers.offmainthreadcomposition.enabled solved the issue.

I did not see this problem with Firefox 39 on that same box.

I am currently running Firefox 40.0.3 on a different Ubuntu 14.04.03 machine and the Firefox display looks stable.  No repainting issues.

Comment 8

[lucas.martinez.bernabeu]

Hi,

I also have this exact issue since I updated from FF 39.x to FF 40.0.3 in Ubuntu 12.04.5 (GTX 660 Ti, Nvidia drivers 331.113).


# firefox -v
Mozilla Firefox 40.0.3

# lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 12.04.5 LTS
Release:	12.04
Codename:	precise

# uname -a
Linux xxxxxxxx 3.2.0-89-generic #127-Ubuntu SMP Tue Jul 28 09:28:15 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

# nvidia-smi
Wed Sep  2 10:38:06 2015       
+------------------------------------------------------+                       
| NVIDIA-SMI 331.113    Driver Version: 331.113        |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 660 Ti  Off  | 0000:01:00.0     N/A |                  N/A |
| 10%   38C  N/A     N/A /  N/A |    395MiB /  2047MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+

Comment 9

[Sotaro Ikeda [:sotaro]]

(In reply to Anthony Hughes, QA Mentor (:ashughes) from comment #6)
> Sotaro, it looks like the closest match is your Thinkpad W530 with a Quadro
> K1000M (same chipset family as the GTX 650 reported above). Would you be
> able to test this with Ubuntu 12.04 full updated with the NVidia proprietary
> drivers installed?

This week, I am on a work week. Next week I could test it.

Comment 10

[Sotaro Ikeda [:sotaro]]

(In reply to Anthony Hughes, QA Mentor (:ashughes) from comment #6)
> Sotaro, it looks like the closest match is your Thinkpad W530 with a Quadro
> K1000M (same chipset family as the GTX 650 reported above). Would you be
> able to test this with Ubuntu 12.04 full updated with the NVidia proprietary
> drivers installed?

I tried the above on VMWare on windows. When I installed default Ubuntu 12.04.5 on WMWare, display show just black after log in to Ubuntu if "Accelerate 3D Graphics(GPU)" is ON. When I disabled "Accelerate 3D Graphics(GPU)", UI was shown normally.

Then I updated graphics driver by the following commands. Then, ui was shown with "Accelerate 3D Graphics(GPU)" is ON. And I did not saw the problem during switching Tabs.

> sudo add-apt-repository ppa:ubuntu-x-swat/x-updates
> sudo apt-get update
> sudo apt-get install nvidia-current

http://askubuntu.com/questions/219236/how-to-update-graphics-card-driver-in-ubuntu-12-04-1

Comment 11

[Sotaro Ikeda [:sotaro]]

(In reply to Sotaro Ikeda [:sotaro] from comment #10)
> 
> I tried the above on VMWare on windows. When I installed default Ubuntu
> 12.04.5 on WMWare, display show just black after log in to Ubuntu if
> "Accelerate 3D Graphics(GPU)" is ON. When I disabled "Accelerate 3D
> Graphics(GPU)", UI was shown normally.

"Accelerate 3D Graphics(GPU)" is VMWare's vm setting.

Comment 12

[u279076]

Thank you Sotaro. Can anyone else experiencing this issue confirm that updating the graphics driver as per comment 10 resolves this issue?

Comment 13

[Fernando Acedo]

Hi,

I got this issue some days ago after update to FF 40.0.3 in Ubuntu 14.04.3

I updated Nvidia driver from 304.125 to 340.76 and the issue still persists.

Troubleshooting info:

Name 	Firefox
Version 	40.0.3
Build ID 	20150826185918
User Agent 	Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:40.0) Gecko/20100101 Firefox/40.0
Multiprocess Windows 	0/1 (default: false)

Adapter Description	NVIDIA Corporation -- GeForce 8400 GS/PCIe/SSE2
Asynchronous Pan/Zoom	none
Device ID	GeForce 8400 GS/PCIe/SSE2
Driver Version	3.3.0 NVIDIA 340.76
GPU Accelerated Windows	0/1 Basic (OMTC)
Supports Hardware H264 Decoding	false
Vendor ID	NVIDIA Corporation
WebGL Renderer	NVIDIA Corporation -- GeForce 8400 GS/PCIe/SSE2
windowLayerManagerRemote	true
AzureCanvasBackend	cairo
AzureContentBackend	cairo
AzureFallbackCanvasBackend	none
AzureSkiaAccelerated	0


By other hand, If I try to run:
firefox -v


I get this error:

(process:5739): GLib-CRITICAL **: g_slice_set_config: assertion 'sys_page_size == 0' failed
Mozilla Firefox 40.0.3

Comment 14

[Mark Lord]

Same issue here, on Mint/MATE 17.3 (based on Ubuntu 14.04).
Same issue as well on previous LTS of Mint/Mate (based on Ubuntu 12.04).
Two different systems, but both have older NVIDIA graphics, using latest 340.xx driver.
Same issue with older NVIDIA drivers as well.

This bug is pushing me to use Chrome, which I really, really detest.

Comment 15

[Mark Lord]

Note also that there are numerous other threads and bug entries for MS-Win and OS/X which report similar sounding repaint issues, all of which points towards a bug somewhere in the portable parts of FF.

Comment 16

[Lee Salzman [:lsalzman]]

(In reply to Mark Lord from comment #14)
> Same issue here, on Mint/MATE 17.3 (based on Ubuntu 14.04).
> Same issue as well on previous LTS of Mint/Mate (based on Ubuntu 12.04).
> Two different systems, but both have older NVIDIA graphics, using latest
> 340.xx driver.
> Same issue with older NVIDIA drivers as well.
> 
> This bug is pushing me to use Chrome, which I really, really detest.

If you try a beta or nightly, does this issue still occur?

Comment 17

[Mark Lord]

Dunno about currently beta or nightly.  I only run released versions.

Comment 18

[Mark Lord]

Speaking of which, it (the bug) just showed up again.  So disabling OMTC doesn't fix it 100%, though it does make it less common here.

Comment 19

[M Winiarski]

Firefox 41.0 became available from the Ubuntu repositories today ... Upgraded from 40.x to 41.0 and the problem still exists. Only known workarounds [for me] are to downgrade to FF39.x or toggle layers.offmainthreadcomposition.enable to 'false'. Problem exhibits in existing profiles and new [clean] profiles.

Comment 21

[Loic]

about:support log from the duped bug report:

Accélération graphique
Description de la carte	NVIDIA Corporation -- GeForce 9300M GS/PCIe/SSE2
Fenêtres avec accélération graphique	0/1 Basic
ID du périphérique	GeForce 9300M GS/PCIe/SSE2
ID du vendeur	NVIDIA Corporation
Prise en charge matérielle pour le décodage H264	false
Rendu WebGL	NVIDIA Corporation -- GeForce 9300M GS/PCIe/SSE2
Version du pilote	3.3.0 NVIDIA 331.113
windowLayerManagerRemote	false
Zoom/Panoramique asynchrones	aucun
AzureCanvasBackend	cairo
AzureContentBackend	cairo
AzureFallbackCanvasBackend	none
AzureSkiaAccelerated	0

Comment 22

[Loic]

Many NVIDIA GPUs involved.

Comment 23

[M Winiarski]

Updated NVIDIA driver became available from the Ubuntu repositories today -- upgraded from 331.113 to 340.93. Using FF 41.0, the issue still exists.

# uname -a
Linux xxxxxx-xxxxxx 3.13.0-65-generic #105~precise1-Ubuntu SMP Tue Sep 22 13:22:42 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

# lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 12.04.5 LTS
Release:	12.04
Codename:	precise

# nvidia-smi
Tue Sep 29 09:32:08 2015       
+------------------------------------------------------+                       
| NVIDIA-SMI 340.93     Driver Version: 340.93         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce 8400M GS    Off  | 0000:05:00.0     N/A |                  N/A |
| N/A   58C    P0    N/A /  N/A |    109MiB /   127MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Compute processes:                                               GPU Memory |
|  GPU       PID  Process name                                     Usage      |
|=============================================================================|
|    0            Not Supported                                               |
+-----------------------------------------------------------------------------+

Comment 24

[ccwatson2]

Same problem here. Noticed after upgrade to FF 40.x Environment details are: 

$ sudo nvidia-smi

Sat Oct  3 14:12:21 2015       
+------------------------------------------------------+                       
| NVIDIA-SMI 340.93     Driver Version: 340.93         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro FX 1600M     Off  | 0000:01:00.0     N/A |                  N/A |
| N/A   59C    P0    N/A /  N/A |    158MiB /   255MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Compute processes:                                               GPU Memory |
|  GPU       PID  Process name                                     Usage      |
|=============================================================================|
|    0            Not Supported                                               |
+-----------------------------------------------------------------------------+
$ firefox -v

Mozilla Firefox 41.0

$ lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 12.04.5 LTS
Release:	12.04
Codename:	precise

$ dkms status | grep nvidia

nvidia-340, 340.93, 3.13.0-65-generic, x86_64: installed

Comment 25

[M Winiarski]

Additional data point -- Short story: I believe that interactions with compiz are causing the issues.

Longer story ...

I am the original submitter, and posted updates in comments #19 and #23. One thing I noticed after updating the NVIDIA drivers (comment #23) was that my general desktop environment seemed more sluggish and I had other repainting issues outside of Firefox.

I typically run a gnome-session-fallback [Gnome Classic] desktop, but the problems were also showing up under Unity desktop. Out of curiosity, I tried logging in via the Gnome Classic (No effects) and Ubuntu 2D session settings. I found that under these sessions, /usr/bin/compiz was not running ... AND, my desktop environment felt more responsive ... AND, Firefox 41.0.1 running with OMTC enabled did not exhibit the issues.

I experimented with existing profiles and new profiles, under Unity and Gnome Classic, with and without /usr/bin/compiz running. The issue only occurs when compiz is running.

This may point to a compiz issue, but I don't think it necessarily rules out Firefox interactions with compiz since I did not see the problem with FF39 and earlier.

Comment 26

[u279076]

(In reply to M Winiarski from comment #25)
> This may point to a compiz issue, but I don't think it necessarily rules out
> Firefox interactions with compiz since I did not see the problem with FF39
> and earlier.

Could you try to confirm this by testing Firefox 39 and 40 again? If you can indeed only reproduce this in 40 then I'll need you to run mozregression to find the regression range in the Firefox 39.0a1 nightlies.

You'll find the releases here:
https://archive.mozilla.org/pub/firefox/releases/

You'll find mozregression here:
http://mozilla.github.io/mozregression/install.html

Thanks

Comment 27

[M Winiarski]

(In reply to Anthony Hughes, QA Mentor (:ashughes) from comment #26)
> (In reply to M Winiarski from comment #25)
> > This may point to a compiz issue, but I don't think it necessarily rules out
> > Firefox interactions with compiz since I did not see the problem with FF39
> > and earlier.
> 
> Could you try to confirm this by testing Firefox 39 and 40 again? If you can
> indeed only reproduce this in 40 then I'll need you to run mozregression to
> find the regression range in the Firefox 39.0a1 nightlies.
> 
> You'll find the releases here:
> https://archive.mozilla.org/pub/firefox/releases/
> 
> You'll find mozregression here:
> http://mozilla.github.io/mozregression/install.html
> 
> Thanks

It is unlikely that I would be able to run the tests you request anytime in the near-term. Sorry that I cannot help at this time. Hopefully, someone else can step in and test regressions.

Comment 28

[Loic]

The tests can be run in 15 min, Mozregression just downloads the builds and launches them for you.

Comment 29

[R. Moeller]

I just wanted to point you to a bug report I filed under Ubuntu Launchpad which is possibly related:

https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/1503310

The problem looks similar, but it also affects other programs, not only firefox. Please see my description of the behavior of e.g. multiple tabs in the Gnome shell.

I guessed that it may have something to do with the nvidia driver, but comment 25 by M. Winiarsky pointing to compiz was interesting since I observed the same effect: The problem disappears when compiz is not running (when logging in with "Gnome Classic (no effects)"). But after all, the nvidia driver may be behind this, since the "Gnome Classic (no effects)" doesn't use window compositing.

Can somebody who has the problem with firefox reproduce the error in other applications, e.g. in the Gnome shell?

Comment 30

[gino.marchal]

Not sure if this provides any additional information:

Just wanted to let you guys know that I have the same problem on my computer at work. Thunderbird also seems to be affected now, although this has rather recently started happening. Firefox was affected first.
Config: Firefox 41.0.2 at the moment, Ubuntu 12.04, Ubuntu NVidia drivers 340.93.
I work occasionally from home and I don't have this repaint problem when logging using a VNC connection.

Comment 31

[Gary Murtagh]

Same issue here.

Ubuntu 12.04.5
Lenovo Thinkpad W530
Nvidia Quadro K1000M
nvidia driver 340.93

FF 41.0.2

First experienced in FF 40, and workaround was to set layers.offmainthreadcomposition.enabled to false
But it has reappeared in FF 41 and same workaround no longer works.

Comment 32

[Ryan VanderMeulen [:RyanVM]]

If either of you would be willing, having a regression range would help a lot for tracking this down. We have a tool that automates the process of downloading different builds so that you end up just saying good/bad to each build it launches. On Ubuntu, you can probably just |pip install -U mozregression| to install it. Then a command like |mozregression --good-release 39 --bad-release 40| would hopefully be enough to get you started.

Comment 33

[Fernando Acedo]

I downgraded the nvidia drivers and after last Firefox update I got the system run stable
 
Firefox 41.0.2
layers.offmainthreadcomposition.enabled false

Nvidia Driver Version: 304.128

Ubuntu 14.04.3 LTS

Anyway, Compiz is raising many errors at boot and when using workspaces. So clearly the problem is located in Compiz but has an high impact in Firefox while other programs are running well.

Comment 34

[Ryan VanderMeulen [:RyanVM]]

(In reply to Gary Murtagh from comment #31)
> Ubuntu 12.04.5
> Lenovo Thinkpad W530
> Nvidia Quadro K1000M
> nvidia driver 340.93

This seems like a config we should be able to replicate?

Comment 35

[u279076]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #34)
> (In reply to Gary Murtagh from comment #31)
> > Ubuntu 12.04.5
> > Lenovo Thinkpad W530
> > Nvidia Quadro K1000M
> > nvidia driver 340.93
> 
> This seems like a config we should be able to replicate?

Apparently, Sotaro has access to a machine like this in Toronto (source: https://wiki.mozilla.org/QA/Platform/Graphics/Inventory#NVIDIA_Chipsets).

Comment 36

[Sotaro Ikeda [:sotaro]]

I moved to Tokyo since last month. And I returned W530 to Jonathan Lin, before I came back to Japan.

milan, can you ask him if he still keeps W530?

Comment 37

[tero gusto]

Ubuntu 14.04.3 LTS
NVIDIA Corporation GF119 [GeForce GT 610] (rev a1)
FF 42.0

Setting layers.offmainthreadcomposition.enabled to false in about:config fixed it for me.

Comment 38

[marek.petlicki]

layers.offmainthreadcomposition.enabled = false did not fix it for me, I'm afraid

Firefox 42.0
Ubuntu 15.10
Dell OptiPlex 960
NVIDIA Corporation GT218GL [Quadro FX 380 LP] (rev a2)
NVIDIA binary drivers v 340.96

Comment 39

[[e-mail]]

Problem for me too. I've not yet taken steps to try to fix this yet.
(Sorry if there is some "me too" button in bugzilla, but I couldn't find it.)

Firefox 42.0
Ubuntu 12.04

Comment 40

[Kerry]

Same problem here.  layers.offmainthreadcomposition.enabled = false didn't fix it for me either.

Firefox 42.0
Ubuntu 15.10
+------------------------------------------------------+                       
| NVIDIA-SMI 352.63     Driver Version: 352.63         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 770     Off  | 0000:01:00.0     N/A |                  N/A |
| 50%   43C    P8    N/A /  N/A |   1129MiB /  2047MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 770     Off  | 0000:02:00.0     N/A |                  N/A |
| 46%   40C    P8    N/A /  N/A |   1076MiB /  2047MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+

Comment 41

[Michael Rolnik]

Same problem here.

FireFox 42.0
Ubuntu 12.04
+------------------------------------------------------+                       
| NVIDIA-SMI 352.63     Driver Version: 352.63         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro K2100M       Off  | 0000:01:00.0      On |                  N/A |
| N/A   48C    P0    N/A /  N/A |    423MiB /  2046MiB |      1%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0      1834    G   /usr/bin/Xorg                                  267MiB |
|    0      3243    G   compiz                                         143MiB |
|    0      4003    G   /usr/lib/firefox/firefox                         1MiB |
+-----------------------------------------------------------------------------+

Comment 42

[Lee Salzman [:lsalzman]]

To anyone currently experiencing this bug, does setting the "gfx.xrender.enabled" preference to false help at all?

Comment 43

[tero gusto]

Ubuntu 14.04.3 LTS
NVIDIA Corporation GF119 [GeForce GT 610] (rev a1)
FF 42.0

# these settings solve the problem:
layers.offmainthreadcomposition.enabled;false
gfx.xrender.enabled;true

# these settings ALSO solve the problem:
layers.offmainthreadcomposition.enabled;true
gfx.xrender.enabled; false

Comment 44

[Kerry]

(In reply to Lee Salzman [:eihrul] from comment #42)
> To anyone currently experiencing this bug, does setting the
> "gfx.xrender.enabled" preference to false help at all?

Hi Lee, your suggestion seems to have fixed it for me.  I've been using FF now for about 15 minutes without the problem returning.

Comment 45

[lucas.martinez.bernabeu]

Following Lee Salzman's suggestion I've tried playing with gfx.xrender.enabled preference (as well as with layers.offmainthreadcomposition.enabled), to no avail.

These settings did NOT solve my problem:

a)
layers.offmainthreadcomposition.enabled;false
gfx.xrender.enabled;true

b)
layers.offmainthreadcomposition.enabled;true
gfx.xrender.enabled; false

c)
layers.offmainthreadcomposition.enabled;false
gfx.xrender.enabled; false

d) (DEFAULT)
layers.offmainthreadcomposition.enabled;true
gfx.xrender.enabled; true


The problem affects also the refreshing of other applications: gedit, Unity, Nautilus...
As in this bug pointed out by R. Moeller: https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/1503310


My current settings:

$ firefox -v
Mozilla Firefox 42.0

$ lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 12.04.5 LTS
Release:	12.04
Codename:	precise

$uname -a
Linux lucas-i5-u12 3.2.0-95-generic #135-Ubuntu SMP Tue Nov 10 13:33:29 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

$ nvidia-smi 
Fri Dec 11 10:31:06 2015       
+------------------------------------------------------+                       
| NVIDIA-SMI 340.96     Driver Version: 340.96         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 660 Ti  Off  | 0000:01:00.0     N/A |                  N/A |
| 10%   24C    P8    N/A /  N/A |    344MiB /  2047MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+

Comment 46

[tero gusto]

Ubuntu 14.04.3 LTS
NVIDIA Corporation GF119 [GeForce GT 610] (rev a1)
FF 44.0

# The lag has returned, with sluggish response to key presses, but with both set to false it seems to work better:
layers.offmainthreadcomposition.enabled;false
gfx.xrender.enabled; false

Comment 47

[tero gusto]

Actually, the lag is still there with the above settings both set to false. The refresh problem is still gone, though.

Comment 48

[tero gusto]

UPDATE: Key presses are only ignored on some web sites, so it's not a general problem.

Comment 49

[Paul Pasca[Away. Please needinfo? Paul Oiegas [:pauloiegasSV]]]

Hi reporter,

I haven't managed to reproduce this issue on the latest release(44.0.2) nor latest Nightly(47.0a1). Switching between tabs didn't cause any redraws/repaints. The browser behaved as expected. For testing I've used Ubuntu 14.04 and I had the latest Nvidia driver installed for my GeForce 210. Looks like this issue got fixed along the way. Can you please confirm this?

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:44.0) Gecko/20100101 Firefox/44.0
Build ID: 20160210153822

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:47.0) Gecko/20100101 Firefox/47.0
Build ID: 20160302030209
Also, can you please try to reproduce this on the latest release(44.0.2) and latest Nightly(47.0a1) and provide the results? When doing this, please try to reproduce with a new clean Firefox profile, maybe even in safe mode, as some of this issues may be caused by third party installed add-ons or custom settings(https://support.mozilla.org/en-US/kb/troubleshoot-and-diagnose-firefox-problems).

Thanks,
Paul.

Comment 50

[M Winiarski]

Problem still exists under FF 44.0.2. Becoming more convinced it is a compiz or nvidia issue.

I re-did the testing that I outlined in Comment #25 -- new profiles, existing profiles, Gnome Classic with an without compiz running, Unity desktop, etc.

$ firefox --version
Mozilla Firefox 44.0.2

$ uname -a
Linux XXXXXX-XXXXXX 3.13.0-79-generic #123~precise1-Ubuntu SMP Fri Feb 19 14:48:42 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

$ lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 12.04.5 LTS
Release:	12.04
Codename:	precise

$ nvidia-smi
Thu Mar  3 11:07:43 2016       
+------------------------------------------------------+                       
| NVIDIA-SMI 340.96     Driver Version: 340.96         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce 8400M GS    Off  | 0000:05:00.0     N/A |                  N/A |
| N/A   60C   P10    N/A /  N/A |     89MiB /   127MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Compute processes:                                               GPU Memory |
|  GPU       PID  Process name                                     Usage      |
|=============================================================================|
|    0            Not Supported                                               |
+-----------------------------------------------------------------------------+


I am becoming more convinced that the problem is related to:

https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/1503310

as mentioned in Comment #29.


I have been running Gnome Classic [No Effects] since October 2015 (which avoids starting compiz) and I have not been having any repaint issues. That has been the only workaround that has worked for me.

Comment 51

[Ben N]

I have this problem since a while now, and it's driving me crazy these days. I'm also sure it's a problem with compiz but previous versions of firefox aren't impacted so there has to be something to fix in firefox. So I took the time to run mozregression. Here is the output:

 9:20.69 INFO: Narrowed nightly regression window from [2015-06-25, 2015-06-27] (2 days) to [2015-06-26, 2015-06-27] (1 days) (~0 steps left)
 9:20.69 INFO: Got as far as we can go bisecting nightlies...
 9:20.69 INFO: Last good revision: d45440221297 (2015-06-26)
 9:20.69 INFO: First bad revision: d4c4ce7f060c (2015-06-27)
 9:20.69 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=d45440221297&tochange=d4c4ce7f060c

 9:20.69 INFO: Switching bisection method to taskcluster

[...]

10:37.17 INFO: Narrowed inbound regression window from [d4544022, d4c4ce7f] (3 revisions) to [d4544022, 56e207db] (2 revisions) (~1 steps left)
10:37.17 INFO: Oh noes, no (more) inbound revisions :(
10:37.17 INFO: Last good revision: d45440221297a2f2d7bcf9a9b91578e0affb43d8
10:37.17 INFO: First bad revision: 56e207dbb3bda446e30c8a9fcf16050f35a58fb9
10:37.17 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=d45440221297a2f2d7bcf9a9b91578e0affb43d8&tochange=56e207dbb3bda446e30c8a9fcf16050f35a58fb9

10:38.70 INFO: ************* Switching to mozilla-inbound

[...]

15:23.09 INFO: Narrowed inbound regression window from [d44bc296, 5ddf0a25] (3 revisions) to [cbe9a2ae, 5ddf0a25] (2 revisions) (~1 steps left)
15:23.09 INFO: Oh noes, no (more) inbound revisions :(
15:23.09 INFO: Last good revision: cbe9a2aea9541146ae0a11051ed32d31eaf4e428
15:23.09 INFO: First bad revision: 5ddf0a252b086c7e8e468243c6251af5c0b1da57
15:23.09 INFO: Pushlog:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=cbe9a2aea9541146ae0a11051ed32d31eaf4e428&tochange=5ddf0a252b086c7e8e468243c6251af5c0b1da57

15:24.01 INFO: Looks like the following bug has the changes which introduced the regression:
https://bugzilla.mozilla.org/show_bug.cgi?id=1175556

Comment 52

[Loic]

M Winiarski, could you confirm the regression range of Ben N?
FAQ to use mozregression is here: http://mozilla.github.io/mozregression/

Comment 53

[M Winiarski]

Running mozregression shows that the problem occurs in a much earlier build than Ben's. Once mozregression narrowed it down to a day, it was unable to further narrow down to specific revisions -- it gave WARNING and CRITICAL logs. I've included the tail end of the logging, with some of the WARNING/CRITICAL messages.

Note: mozregression was run in a Gnome Classic environment with compiz running, the same environment I was using when I first encountered the issue.

------------

44:29.84 INFO: Narrowed nightly regression window from [2015-01-12, 2015-01-14] (2 days) to [2015-01-13, 2015-01-14] (1 days) (~0 steps left)
44:29.84 INFO: Got as far as we can go bisecting nightlies...
44:29.84 INFO: Last good revision: 3d846527576f (2015-01-13)
44:29.84 INFO: First bad revision: 63006936ab99 (2015-01-14)
44:29.84 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=3d846527576f&tochange=63006936ab99

44:29.84 INFO: Switching bisection method to taskcluster
44:29.84 INFO: Getting mozilla-central builds between 3d846527576f and 63006936ab99
44:31.67 WARNING: Skipping build 3d846527576f: Unable to find build info using the taskcluster route 'buildbot.revisions.3d846527576fa7fab0ca601f5110b0ed38e145aa.mozilla-central.linux64'
44:31.82 WARNING: Skipping build 63006936ab99: Unable to find build info using the taskcluster route 'buildbot.revisions.63006936ab99b4ea5db20d1d7d67ce6c7f16ebf5.mozilla-central.linux64'
44:32.72 WARNING: Skipping build 5e3790525a92: Unable to find build info using the taskcluster route 'buildbot.revisions.5e3790525a927f94d19c637d1bb6a55a8f5cf3dd.mozilla-central.linux64'

[...]

44:38.23 WARNING: Skipping build b3f84cf78dc2: Unable to find build info using the taskcluster route 'buildbot.revisions.b3f84cf78dc20519db0797f9bd613331647d61e6.mozilla-central.linux64'
44:38.25 CRITICAL: First build 3d846527576f is missing, but mozregression can't find a build before - so it is excluded, but it could contain the regression!
44:39.11 WARNING: Skipping build 21edbaa22144: Unable to find build info using the taskcluster route 'buildbot.revisions.21edbaa221443e1f8211ca4d470c63cdca5d42d5.mozilla-central.linux64'

[...]

44:44.37 WARNING: Skipping build c9162436444e: Unable to find build info using the taskcluster route 'buildbot.revisions.c9162436444e380ba0df573084bcf8947daf0058.mozilla-central.linux64'
44:44.39 CRITICAL: Last build 63006936ab99 is missing, but mozregression can't find a build after - so it is excluded, but it could contain the regression!
44:45.11 WARNING: Skipping build 67257a3edeb5: Unable to find build info using the taskcluster route 'buildbot.revisions.67257a3edeb5f8590ac998f418f4abb67c42efca.mozilla-central.linux64'

[...]

44:46.00 WARNING: Skipping build e978b8bc5c45: Unable to find build info using the taskcluster route 'buildbot.revisions.e978b8bc5c4573ff69d516a7f43fc880b0f15465.mozilla-central.linux64'
44:46.01 INFO: There are no build artifacts on inbound for these changesets (they are probably too old).

Comment 54

[Ben N]

What M Winiarski found was surprising so I checked a random previous version (I picked 2015-02-02). At first I had no problem, but after 5 min of random Internet browsing I got the repaint bug. I guess I missed it because I could usually get the repaint problem within the first minute of browsing. Consider M Winiarski's mozregression tests more relevant than mine so.

Comment 55

[Ryan VanderMeulen [:RyanVM]]

The regression window in comment 53 is interesting. I see a possible suspect in that range. Mike, can you confirm that Firefox 36 (the final release) reproduces the bug, but not Firefox 35?

Comment 56

[Paul Silaghi, QA [:pauly]]

I've noticed some very bad rendering issue on the tab bar and the page contents while switching tabs on 2 Ubuntu machines with NVIDIA GPU, FX 46b11.

Comment 57

[u279076]

Paul, are you able to reproduce this and if so can you please provide a regression window?

Comment 58

[Paul Silaghi, QA [:pauly]]

I couldn't reproduce on 47.0.1, 50.0a1 (2016-07-01), Ubuntu 14.04 x64, Nvidia Geforce GT 620.

Comment 59

[u279076]

(In reply to Paul Silaghi, QA [:pauly] from comment #58)
> I couldn't reproduce on 47.0.1, 50.0a1 (2016-07-01), Ubuntu 14.04 x64,
> Nvidia Geforce GT 620.

Thanks Paul, I am closing this as incomplete for now. Please reopen this bug report if the issue returns and please provide a regression window when you do.