Closed
Bug 798274
Opened 12 years ago
Closed 11 years ago
crash in gfxContext::PushGroupAndCopyBackground with Direct2D 1.1 (d3d11.dll 6.2)
Categories
(Core :: Graphics: Layers, defect)
Tracking
()
RESOLVED
DUPLICATE
of bug 805406
People
(Reporter: scoobidiver, Assigned: bas.schouten)
References
Details
(Keywords: crash, regression, topcrash-win)
Crash Data
It's currently #3 top crasher in today's build. The regression range is:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=635fcc11d2b1&tochange=4cb8f88213f5
It's likely a regression from bug 797314.
Signature gfxContext::PushGroupAndCopyBackground(gfxASurface::gfxContentType) More Reports Search
UUID 2fffef46-f0f6-4aa6-a9ce-a05e02121005
Date Processed 2012-10-05 08:05:11
Uptime 11462
Last Crash 3.9 hours before submission
Install Age 9.1 hours since version was first installed.
Install Time 2012-10-04 23:00:51
Product Firefox
Version 18.0a1
Build ID 20121004030525
Release Channel nightly
OS Windows NT
OS Version 6.1.7601 Service Pack 1
Build Architecture x86
Build Architecture Info GenuineIntel family 6 model 23 stepping 7
Crash Reason EXCEPTION_ACCESS_VIOLATION_READ
Crash Address 0x0
App Notes
AdapterVendorID: 0x1002, AdapterDeviceID: 0x689c, AdapterSubsysID: 034a1043, AdapterDriverVersion: 8.961.0.0
D2D? D2D+ DWrite? DWrite+ D3D10 Layers? D3D10 Layers+
EMCheckCompatibility True
Adapter Vendor ID 0x1002
Adapter Device ID 0x689c
Total Virtual Memory 4294836224
Available Virtual Memory 3533877248
System Memory Use Percentage 55
Available Page File 9163833344
Available Physical Memory 3827302400
Frame Module Signature Source
0 xul.dll gfxContext::PushGroupAndCopyBackground gfx/thebes/gfxContext.cpp:1533
1 xul.dll mozilla::layers::BasicLayerManager::PushGroupForLayer gfx/layers/basic/BasicLayerManager.cpp:81
2 xul.dll mozilla::layers::BasicThebesLayer::PaintThebes gfx/layers/basic/BasicThebesLayer.cpp:131
3 mozglue.dll je_free memory/mozjemalloc/jemalloc.c:6575
4 xul.dll xul.dll@0xb547f
5 xul.dll mozilla::layers::BasicLayerManager::PaintLayer gfx/layers/basic/BasicLayerManager.cpp:932
6 xul.dll mozilla::layers::BasicLayerManager::PaintSelfOrChildren gfx/layers/basic/BasicLayerManager.cpp:828
7 xul.dll mozilla::layers::BasicLayerManager::PaintLayer gfx/layers/basic/BasicLayerManager.cpp:932
8 xul.dll mozilla::layers::BasicLayerManager::EndTransactionInternal gfx/layers/basic/BasicLayerManager.cpp:578
9 gkmedias.dll mozilla::gfx::DrawTargetCairo::PushClipRect gfx/2d/DrawTargetCairo.cpp:695
10 xul.dll mozilla::layers::BasicLayerManager::EndTransaction gfx/layers/basic/BasicLayerManager.cpp:503
11 xul.dll mozilla::PaintInactiveLayer layout/base/FrameLayerBuilder.cpp:2019
12 xul.dll mozilla::FrameLayerBuilder::DrawThebesLayer layout/base/FrameLayerBuilder.cpp:3224
13 xul.dll mozilla::layers::BasicThebesLayer::PaintThebes gfx/layers/basic/BasicThebesLayer.cpp:139
More reports at:
https://crash-stats.mozilla.com/report/list?signature=gfxContext%3A%3APushGroupAndCopyBackground%28gfxASurface%3A%3AgfxContentType%29
Assignee | ||
Comment 1•12 years ago
|
||
Hrm, this would happen if creating the new DrawTarget would fail, I wonder why that would happen though. It could happen in OOM but that doesn't seem likely at such high frequency. Is anyone able to get STR for this?
Assignee | ||
Updated•12 years ago
|
Assignee: nobody → bas.schouten
Status: NEW → ASSIGNED
Assignee | ||
Comment 2•12 years ago
|
||
Also, it would be interesting to know if is this still happening after bug 797797 landed.
Reporter | ||
Comment 3•12 years ago
|
||
(In reply to Bas Schouten (:bas) from comment #2)
> Also, it would be interesting to know if is this still happening after bug
> 797797 landed.
Yes, it still happens in 18.0a1/20121006 and above but at a lower volume.
Reporter | ||
Comment 4•12 years ago
|
||
Still #4 top browser crasher in today's build.
Comment 5•12 years ago
|
||
Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:19.0) Gecko/19.0 Firefox/19.0 ID:20121010030605 crashed like bp-54a97688-9174-4e9f-abda-6c0072121010 when previewing a tab in the taskbar that was yet to be loaded after a session restore.
Might be related to bug 793175 or to the proxy detection wonkyness I described at http://forums.mozillazine.org/viewtopic.php?p=12361691#p12361691
Comment 6•12 years ago
|
||
Bas - if this is difficult to reproduce/investigate, perhaps we should consider temporarily backing bug 797314 out to double verify it's the cause.
Assignee | ||
Comment 7•12 years ago
|
||
(In reply to Alex Keybl [:akeybl] from comment #6)
> Bas - if this is difficult to reproduce/investigate, perhaps we should
> consider temporarily backing bug 797314 out to double verify it's the cause.
That patch should only have made things better by reducing memory usage/leakage problems. I'd like to look at this a little more. I've found a bug related to the gradient cache that would cause crashes for taskbar previews but they shouldn't have this signature. Let's get a fix for that in at least. I still suspect the people crashing on this out in the open are OOM-ing, but it's hard to say without information from someone seeing specifically this crash.
I'll try to reproduce locally more actively.
Comment 8•12 years ago
|
||
Bas,
Here it crashes rather consistently on Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:19.0) Gecko/19.0 Firefox/19.0 ID:20121010030605
bp-6f01085b-89e7-46c0-9c67-aba8a2121011
Please let me know if you need me to collect anything from the problematic system.
Flags: needinfo?(bas.schouten)
Assignee | ||
Comment 9•12 years ago
|
||
(In reply to alex_mayorga from comment #8)
> Bas,
>
> Here it crashes rather consistently on Mozilla/5.0 (Windows NT 6.1; Win64;
> x64; rv:19.0) Gecko/19.0 Firefox/19.0 ID:20121010030605
>
> bp-6f01085b-89e7-46c0-9c67-aba8a2121011
>
> Please let me know if you need me to collect anything from the problematic
> system.
Memory usage when the crash occurs would be a useful first info. The crash we're seeing in PushClipsToDT is probably the same cause as this one except with ::PushGroup.
Flags: needinfo?(bas.schouten)
Comment 10•12 years ago
|
||
Bas,
Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:19.0) Gecko/19.0 Firefox/19.0 ID:20121011030552 crashed again like bp-73a44b92-bfd1-4416-bbe7-f0be42121011
The Physical Memory usage was ~61% and firefox.exe was using ~550,000 K at the time of the crash.
This laptop has 4 GB of total RAM.
Please see also https://bugzilla.mozilla.org/show_bug.cgi?id=793175#c25 for some screen captures of visual glitches that occur on or around these crashes.
Assignee | ||
Comment 11•12 years ago
|
||
(In reply to alex_mayorga from comment #10)
> Bas,
>
> Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:19.0) Gecko/19.0 Firefox/19.0
> ID:20121011030552 crashed again like bp-73a44b92-bfd1-4416-bbe7-f0be42121011
>
> The Physical Memory usage was ~61% and firefox.exe was using ~550,000 K at
> the time of the crash.
>
> This laptop has 4 GB of total RAM.
>
> Please see also https://bugzilla.mozilla.org/show_bug.cgi?id=793175#c25 for
> some screen captures of visual glitches that occur on or around these
> crashes.
Hrm, so OOM seems unlikely, does that data include the total Working Set (select the Column under the 'view' element in Task Manager) or just the Private Working Set?
Comment 12•12 years ago
|
||
Bas,
~550,000 K was the "Memory (Private Working Set)".
For comparison, right now firefox.exe is using ~740,000 K for "Working Set (Memory)" and ~670,000 K for "Memory (Private Working Set)"
Copying the bug comment I referenced before for clarity:
"In case it is of interest, these are some screen captures of visual oddities that happen on or around these crashes:
http://imm.io/HqKp The URL bar disappears from the preview and the window's border overflows to the second monitor, similar or related to bug 724940
http://imm.io/HqKM The whole preview is blank and the window's border overflows too, similar or related to bug 724940
http://imm.io/HnnJ Nightly button and tabs become transparent, perhaps bug 798224"
Assignee | ||
Comment 13•12 years ago
|
||
Hrm, so no OOM. This is starting to become a real mystery :s. All those values look pretty healthy. We need to find out what error is reported by Azure on DrawTarget creation, are you by any chance in a position to run debug builds?
Comment 14•12 years ago
|
||
Bas,
I can install software in the problematic system, is there a guide somewhere on how "to run debug builds"?
Flags: needinfo?(bas.schouten)
Comment 15•12 years ago
|
||
Bas,
FWIW all the crashes point to http://hg.mozilla.org/mozilla-central/annotate/ec10630b1a54/gfx/thebes/gfxContext.cpp#l1534
But that's a bit above my current skill set =S
Assignee | ||
Comment 16•12 years ago
|
||
(In reply to alex_mayorga from comment #15)
> Bas,
>
> FWIW all the crashes point to
> http://hg.mozilla.org/mozilla-central/annotate/ec10630b1a54/gfx/thebes/
> gfxContext.cpp#l1534
>
> But that's a bit above my current skill set =S
Yeah, the question is -why- mDT is NULL there. If you were running a build with logging it should tell you why. I'm not sure how to run a build with logging without doing your own build. Perhaps someone reading this can enlighten us?
Flags: needinfo?(bas.schouten)
Comment 17•12 years ago
|
||
Maybe if I check that box for "Need additional information from anyone" someone would notice it. Let's see...
Flags: needinfo?
Comment 18•12 years ago
|
||
alex: The needinfo flags doesn't help, nobody will do a search for bugs with that flag and the next comment (this one) will remove it anyway
bas: Is that logging to a file or is a debugger required ?
If he needs only a debug build he could use for example this build:
ftp://ftp.mozilla.org/pub/firefox/tinderbox-builds/mozilla-central-win32-debug/1350117447/firefox-19.0a1.en-US.win32.zip
Flags: needinfo?
Comment 19•12 years ago
|
||
STR:
1. Open Gmail in new tab.
2. Use Windows Taskbar preview and hover Gmail tab.
3. Nightly crash: https://crash-stats.mozilla.com/report/index/bp-abfe1191-53c0-4162-b216-16a052121015
BTW> Only when HWA on.
Reporter | ||
Updated•12 years ago
|
Keywords: reproducible
Reporter | ||
Comment 21•12 years ago
|
||
Crashes stopped spiking after 19.0a1/20121023 and 18.0a2/20121030. The working windows for the spike are:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=48502b61a63e&tochange=93cc1ee94291
http://hg.mozilla.org/releases/mozilla-aurora/pushloghtml?fromchange=97c6c326435e&tochange=2b2d7457e130
It might have been fixed by bug 803949 or bug 758531 as stated by Bas in comment 9.
semtex2, can you still reproduce it?
Keywords: topcrash
Comment 22•12 years ago
|
||
(In reply to Scoobidiver from comment #21)
> Crashes stopped spiking after 19.0a1/20121023 and 18.0a2/20121030. The
> working windows for the spike are:
> http://hg.mozilla.org/mozilla-central/
> pushloghtml?fromchange=48502b61a63e&tochange=93cc1ee94291
> http://hg.mozilla.org/releases/mozilla-aurora/
> pushloghtml?fromchange=97c6c326435e&tochange=2b2d7457e130
> It might have been fixed by bug 803949 or bug 758531 as stated by Bas in
> comment 9.
>
> semtex2, can you still reproduce it?
No, seems to be fixed for me as well.
Reporter | ||
Updated•12 years ago
|
Keywords: reproducible
Comment 23•12 years ago
|
||
The signature seems to be back since the 2013-02-07 build, with multiple crashes in every build since then.
Comment 24•12 years ago
|
||
I get this crash when I install/uninstall Catalyst while Firefox is running.
Windows 8 x64, AMD Radeon HD 7700 Series
Comment 25•12 years ago
|
||
In Firefox 19.0, this crash shows a significant skew towards nvidia graphics hardware and shows up much less frequently with Intel graphics.
Assignee | ||
Comment 26•12 years ago
|
||
(In reply to Benjamin Smedberg [:bsmedberg] from comment #25)
> In Firefox 19.0, this crash shows a significant skew towards nvidia graphics
> hardware and shows up much less frequently with Intel graphics.
I wonder if this is simply occuring on a driver reset/crash. We should really somehow have an automated test for that situation.
Comment 27•12 years ago
|
||
Yes, I think that is very likely. I think I've seen this crash personally when windows update updates my nvidia driver, for example, which causes a graphics reset.
Comment 28•12 years ago
|
||
This has been rising yesterday, and I heard NVidia shipped a driver update, so that could match.
Reporter | ||
Comment 29•12 years ago
|
||
It's #16 browser crasher in 21.0, #22 in 22.0b2 and 23.0a2, and #18 in 24.0a1.
(In reply to Bas Schouten (:bas.schouten) from comment #9)
> Memory usage when the crash occurs would be a useful first info. The crash
> we're seeing in PushClipsToDT is probably the same cause as this one except
> with ::PushGroup.
I think so based on correlations:
99% (440/443) vs. 13% (13551/107853) d3d11.dll
0% (1/443) vs. 0% (7/107853) 6.2.8400.0
4% (16/443) vs. 0% (223/107853) 6.2.9200.16384
17% (75/443) vs. 1% (1582/107853) 6.2.9200.16420
0% (1/443) vs. 0% (46/107853) 6.2.9200.16440
78% (347/443) vs. 11% (11643/107853) 6.2.9200.16492
100% (443/443) vs. 18% (19104/107853) d2d1.dll
0% (2/443) vs. 1% (1085/107853) 6.1.7600.16385
0% (1/443) vs. 1% (1495/107853) 6.1.7601.17514
0% (1/443) vs. 0% (5/107853) 6.2.8400.0
4% (16/443) vs. 0% (187/107853) 6.2.9200.16384
17% (75/443) vs. 1% (1428/107853) 6.2.9200.16420
0% (1/443) vs. 0% (46/107853) 6.2.9200.16440
78% (347/443) vs. 10% (10692/107853) 6.2.9200.16492
Keywords: topcrash
Summary: crash in gfxContext::PushGroupAndCopyBackground → crash in gfxContext::PushGroupAndCopyBackground with Direct2D 1.1 (d3d11.dll 6.2)
Comment 30•11 years ago
|
||
bas, is this related or a dupe of bug 805406
is #19 on 23.0.1 and #56 on 24beta topcrash lists
Flags: needinfo?(bas)
Keywords: topcrash
Assignee | ||
Comment 31•11 years ago
|
||
I believe the cause is exactly the same, so is the underlying bug.
Flags: needinfo?(bas)
Comment 33•11 years ago
|
||
I believe this has morphed into gfxContext::PushGroupAndCopyBackground(gfxContentType) on Fx27
https://crash-stats.mozilla.com/report/index/d59e6096-fc98-4e1f-9df6-bfa872131007 with
100% (16/16) vs. 53% (635/1209) d3d11.dll
It's #9 with the above signature on Nightly (Fx27).
Crash Signature: [@ gfxContext::PushGroupAndCopyBackground(gfxASurface::gfxContentType)] → [@ gfxContext::PushGroupAndCopyBackground(gfxASurface::gfxContentType)]
[@ gfxContext::PushGroupAndCopyBackground(gfxContentType)]
Comment 34•11 years ago
|
||
Current Rank
> Firefox 29: #3 @ 2.01%
> Firefox 28: #8 @ 1.29%
> Firefox 27: #10 @ 0.68%
> Firefox 26: #21 @ 0.52%
Keywords: topcrash-win
Comment 35•11 years ago
|
||
(In reply to Bas Schouten (:bas.schouten) from comment #31)
> I believe the cause is exactly the same, so is the underlying bug.
Between the two bugs (crash signatures) they currently account for 5% of all crashes on 27betas.
Comment 36•11 years ago
|
||
This signature jumped from 0 to ~100 crashes per million ADI from 26 to 27 on beta, it's probably closely related to bug 805406 where we look into this increase already.
Comment 37•11 years ago
|
||
Ah, forgot that the signature morphed as described in comment #33, so 26->27 only made it go up from ~30 to ~100 crashes per million ADI.
Comment 38•11 years ago
|
||
Just crashed like this on Mozilla/5.0 (Windows NT 6.1; WOW64; rv:30.0) Gecko/20100101 Firefox/30.0 ID:20140225030201 CSet: e3daaa4c73dd
Report ID Date Submitted
bp-cfd8dce2-5d24-482c-9dde-e963c2140226 25/02/2014 07:50 p.m.
I was indeed fiddling with the nvidia drivers on the system at the time of the crash.
Comment 39•11 years ago
|
||
Nominating for tracking since this is #14 on Firefox 28.
status-firefox28:
--- → affected
tracking-firefox28:
--- → ?
Comment 40•11 years ago
|
||
If this is related to bug 805406 is it also fixed on 30? We're too late for any kind of speculative work on FF28 so wontfixing based on that. We might even want to dupe this.
status-firefox29:
--- → affected
status-firefox30:
--- → ?
tracking-firefox29:
--- → +
tracking-firefox30:
--- → +
Updated•11 years ago
|
Comment 41•11 years ago
|
||
Given comment #31, I'm going ahead and just duping this to 805406, any action will be tracked there anyhow.
(In reply to Lukas Blakk [:lsblakk] from comment #40)
> If this is related to bug 805406 is it also fixed on 30?
It's also not fixed, even though the patches were supposed to also help this. :-/
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → DUPLICATE
Updated•11 years ago
|
You need to log in
before you can comment on or make changes to this bug.
Description
•