Closed Bug 1108499 Opened 10 years ago Closed 9 years ago

On switching tab tabs content are mixed

Categories

(Core :: Graphics, defect)

Product:
Core
Component:
Graphics
Version:
35 Branch
Platform:
x86_64
Windows 8.1
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED DUPLICATE of bug 1067470

People

(Reporter: elennaro, Unassigned)

Details

Attachments

(3 files)

firefox_changing_tab_bug.PNG
218.13 KB, image/png
Details
Another Screenshot of the bug
2.53 MB, image/png
Details
the-log.txt
11.75 KB, text/plain
Details 
User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:35.0) Gecko/20100101 Firefox/35.0
Build ID: 20141201162954

Steps to reproduce:

It happens pretty often.

Steps to reproduce:
1. Browse some tab.
2. Ctrl+Click to open some link in a new tab.
3. By mouse click on a newly opened tab switch to this tab.

Browser: 34.0-35.0(beta)
Platform: Windows 8.1 Single Language 64bit
Dell Inspiron 15R 5531 (intel Core i7-3517U, AMD Radeon 8730M, *Gb Ram)
(No other bugs related to graphics in any applications including games)


Actual results:

Contents of those pages are graphically mixed. See attachment.
Component: Untriaged → Tabbed Browser
Component: Tabbed Browser → Graphics
Product: Firefox → Core
Can you provide the graphics section from about:support? (this also has driver and hardware acceleration info not in your initial report)
Flags: needinfo?(elennaro)
Graphics
Adapter Description	Intel(R) HD Graphics 4000
Adapter Drivers	igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32
Adapter RAM	Unknown
Device ID	0x0166
DirectWrite Enabled	false (6.3.9600.17111)
Driver Date	3-11-2014
Driver Version	10.18.10.3496
GPU #2 Active	false
GPU Accelerated Windows	0/1 Basic (OMTC)
Subsys ID	05b81028
Vendor ID	0x8086
WebGL Renderer	Google Inc. -- ANGLE (Intel(R) HD Graphics 4000 Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote	true
AzureCanvasBackend	skia
AzureContentBackend	cairo
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0
Flags: needinfo?(elennaro)
I had something similar happen on the Mac in 35 (when it was nightly, and captured in bug 1077636), so I'm not sure it's platform dependent.
Does this happen on all sites, or is there a particular example to reproduce?
See Also: → 1077636
It happens on all sites, just not all the time, once in 3-5 tries. And actually some of my windows friends are also experiencing problem with this bug.
I can share session with team-viewer or something if you want...
I also can make some logs if you teach me how.
And one more thing, I have feeling like when I switch tab and it happens it's like no redraw happens on the page, 'cause after I move mouse some parts of new page are redrawn over the old one.
Logs won't hurt. It is a bit involved if you're running regular install:

1. In about:config, add New integer preference gfx.logging.level, set it to 3
2. Get Console Redictor as described here: https://developer.mozilla.org/en-US/docs/Mozilla/Command_Line_Options?redirectlocale=en-US&redirectslug=Command_Line_Options#-console and find Console Redirector.exe in the zip
3. From command prompt, run "C:\whatever-location\Console Redirector.exe" C:\Program Files (x86)\Mozilla Firefox\firefox.exe"

You should get some logging into the the command window - could you attach whatever you collect when you run into this problem?
This is another one screenshot. I mentioned that this bug happens more often when switching from pinned to common tab.
Attached file The log (obsolete) — 
Here goes the log.
I can't see anything unusual here, however you will get more I hope.
Attached file the-log.txt — 
Some privacy in the log
Oops. Sorry I wanted to delete obsoleted attachment and this comment. But I don't know how :(
See Also: → 1067470
See Also: → 1077644
(In reply to Alex Art from comment #10)
> Oops. Sorry I wanted to delete obsoleted attachment and this comment. But I
> don't know how :(

I think I did "remove" them as much as possible.
Thanks.
Can I do anything else, to help you to find the reason or the way to reproduce this bug?
Yes, I have a guess as to what could be the problem. I'll locate two nightly builds for you to try to test me theory. I'll post it tomorrow.
Flags: needinfo?(bgirard)
I cannot reproduce this bug on any of my machines with hardware acceleration on, either debug builds or optimized builds.
I see this fairly often switching between a pinned twitch tab [1] and a gmail tab [2]. Only some elements draw, with more showing up as you hover over them. The missing stuff seems to be drawn eventually, but may be as much as several seconds delayed. My GPU is an Nvidia GeForce GTX 580.

[1] http://www.twitch.tv/directory/following/live
[2] https://mail.google.com/mail/u/0/#inbox
Oh, and I'm on Windows 7 x64, running 64-bit Nightly, with about:support showing 1/1 Direct3D 11 (OMTC) and direct2d 1.1.
Critically, I believe both Alex Art and I have e10s disabled. I haven't seen anyone mention this with e10s enabled.
Some more logs from yesterday's session (I'm not sure if they are about this error but you better decide):

WARNING: content window passed to PrivateBrowsingUtils.isWindowPrivate. Use isContentWindowPrivate instead (but only for frame scripts).
pbu_isWindowPrivate@resource://gre/modules/PrivateBrowsingUtils.jsm:25:14
NM__onRequestHeader@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/toolkit/webconsole/network-monitor.js:653:34
NM_observeActivity@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/toolkit/webconsole/network-monitor.js:532:7
makeInfallible/<@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/DevToolsUtils.js:83:14
TypeError: Argument 1 of Window.getComputedStyle is not an object.: CssLogic.getComputedStyle@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/styleinspector/css-logic.js:809:10
exports.NodeActor<.computedStyle@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/server/actors/inspector.js:319:12
exports.NodeActor<.isDisplayed@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/server/actors/inspector.js:326:9
WalkerActor<._onReflows@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/server/actors/inspector.js:1254:11
EventEmitter_emit@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/event-emitter.js:137:11
LayoutChangesObserver.prototype<._startEventLoop@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/devtools/server/actors/layout.js:277:7
notify@resource://gre/modules/commonjs/toolkit/loader.js -> resource://gre/modules/commonjs/sdk/timers.js:40:9
Yes Emanuel Hoogeveen I don't think I have Electrolysis enabled. Don't even have an idea how to enable it.
If someone that reproduces the problem can run this tool this would be the most useful thing for us to know. This should take about 15 mins + a bit more time if the reproduction takes time:
http://mozilla.github.io/mozregression/

I'm doing a build that will specifically try my theory. It will be ready in about an hour:
https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=38982cccce8f
Flags: needinfo?(bgirard)
See Also: → 1110696
Moving to NEW since we have at least two users that reproduce it.
Status: UNCONFIRMED → NEW
Ever confirmed: true
Unfortunately I haven't seen this at all today, so I'm not confident of my ability to bisect this. Also I wanted to mention bug 1119066, which looks similar but with tiling (perhaps easier to reproduce with several tiles?).
Benoit Girard. Sorry for a stupid question, but please can you explain me how to run http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/b56girard@gmail.com-38982cccce8f/try-win32/firefox-37.0a1.en-US.win32.zip this version of Firefox?
Extract all files from that archive to your location of choice, close your current instance of Firefox and open firefox.exe from the archive.
Thanks  Emanuel Hoogeveen, I did so. I see all tabs titles are underscored, so e10n is on. I will browse the Interenet using this bug till tomorrow, then I will confirm if there is no bug, or disappoint you with bad news :)
Ok. I've tested this release. And it fixes the bug. Instead of not repainting site from tab 2, or repainting it partially over site from tab 1, it now shows nice and funny (but a little bit annoying) spinning preloader on a white background, then the page content correctly.
Dupe of bug 1067470?
Alright this means we have a bug with the compositor invalidation code.

Bas can we restore that to a good state if we can't fix it on time?
Flags: needinfo?(bas)
[Tracking Requested - why for this release]:
Severe visual regression for affected users.
tracking-firefox35: --- → ?
tracking-firefox36: --- → ?
tracking-firefox37: --- → ?
tracking-firefox38: --- → ?
Definitely tracking this for 36 and later, but as for 35 I don't see this as something we'd want to take the risk on for a dot release and that it should get more time than that to bake with Beta users.  Is there likely to be a lot of users hitting this issue?  Unless we know that there's significant impact *and* landing a backout to a known good state is possible here, I'd lean towards a wontfix for 35.  With more information we could consider this a potential ride along in case of a dot release.
status-firefox35: --- → affected
status-firefox36: --- → affected
status-firefox37: --- → affected
status-firefox38: --- → affected
tracking-firefox36: ? → +
tracking-firefox37: ? → +
tracking-firefox38: ? → +
Oh, I never noticed that Alex filed this for Firefox *35*. I've only noticed this since around the time that bug 1107297 landed, on the Nightly for 37. So either my problem has a different cause, or the cause of the problem was uplifted to beta.
(In reply to Emanuel Hoogeveen [:ehoogeveen] from comment #34)
> Oh, I never noticed that Alex filed this for Firefox *35*.

Because I was confusing this bug for bug 1116557 (too much bugmail..). I've duped that one to this for now, since the issue sounds the same.
Looking on screenshot and description it's definitely a duplicate of bug #1067470.
Status: NEW → RESOLVED
Closed: 9 years ago
status-firefox35: affected → ---
status-firefox36: affected → ---
status-firefox37: affected → ---
status-firefox38: affected → ---
tracking-firefox35: ? → ---
tracking-firefox36: + → ---
tracking-firefox37: + → ---
tracking-firefox38: + → ---
Resolution: --- → DUPLICATE
See Also: 1077636, 1067470, 1077644 →
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: