582368 - Crash during the load of a WebGL demo [@ ig4icd32.dll@0x45e6cf ]

Status: RESOLVED WORKSFORME

(Core :: Graphics: CanvasWebGL, defect)

Description

[Scoobidiver (away)]

User-Agent:       Mozilla/5.0 (Windows; Windows NT 6.1; WOW64; rv:2.0b2) Gecko/20100720 Firefox/4.0b2
Build Identifier: Mozilla/5.0 (Windows; Windows NT 6.1; WOW64; rv:2.0b2) Gecko/20100720 Firefox/4.0b2

During the load of a WebGL demo, FF 4.0b2 crashes.

http://crash-stats.mozilla.com/report/index/a5c75140-74f2-44c8-b83c-4e13f2100727

In FF 4.0b1, WebGL could not be enabled because of bug 570474

Graphic card : Intel GMA 4500M
Graphic driver version : 8.15.10.2141 (up-to-date for Intel)
Video driver model : WDDM 1.1
Video shared memory : 1631 MB
DirectX version : DirectX 11

Reproducible: Always

Comment 1

[Benoit Jacob [:bjacob] (mostly away)]

Hm, here on linux x86-64 and a more recent build, it's not crashing. Perhaps it's been fixed?

In any case it's not looking like a webgl crash: the 'raw dump' in your crash report doesn't have any occurence of webgl... also indicated by your report that the crash occurs during the loading.

Could you try with beta3 when it comes out and report if it's still crashing? We'll then try to reassign this to the right people.

Comment 2

[Benoit Jacob [:bjacob] (mostly away)]

...or you could even try today with a nightly build.

Comment 3

[Scoobidiver (away)]

This bug report was not about any WebGL demo but only about the one in the ref URL.
It crashes a few seconds after "loading texture pluto ..."

I tested this URL in minefield b3pre 20100727 and it still crashes the same way.

Comment 4

[Benoit Jacob [:bjacob] (mostly away)]

(In reply to comment #3)
> This bug report was not about any WebGL demo but only about the one in the ref
> URL.

Yes, that's how I understood it :-)

ok, i'll try to see to whom I should reassign this report...

Comment 5

[Benoit Jacob [:bjacob] (mostly away)]

ok, since the stack frame mentions a lot of xul.dll, let's reassign this to XUL people...

Comment 6

[Boris Zbarsky [:bzbarsky]]

xul.dll where?  The only dll I see in the linked crash report is ig4icd32.dll (which is the Intel graphics driver).  In any case, "xul.dll" is just "layout and everything".

Comment 7

[Boris Zbarsky [:bzbarsky]]

Oh, I see where you saw xul.dll.  In the message pump thread for out-of-process plugin stuff.  I _really_ doubt it has anything to do with a crash inside the Intel graphics driver, whereas for webgl I wouldn't be surprised at all.

Comment 8

[Benoit Jacob [:bjacob] (mostly away)]

oohhh... ok. I guess I mis-understood how to read the "raw dump".

So, this is crashing in the intel graphics driver, and we don't have any backtrace leading to this location, right?

What can possibly explain that we don't get any back-trace in the crashing thread?

Comment 9

[Boris Zbarsky [:bzbarsky]]

> So, this is crashing in the intel graphics driver, and we don't have any
> backtrace leading to this location, right?

Yep, as far as I can see.

> What can possibly explain that we don't get any back-trace in the crashing
> thread?

Ted, any ideas?  Other than us not having any symbols for the driver, of course...

Comment 10

[timeless]

I've cc'd people from intel.com who have touched bugzilla since june. My bugmail indicates that they have the following roles (sorted from most likely to be helpful):
 andy is a windows plugin guy (not sure what kind of windows plugin intel has)
 chris is a linux (web)gl guy
 sergey is an optimization guy (cpu, not gpu)
 roger is a meego guy

I'm not certain about the interest areas of the others.

Guys: could one of you please find someone involved in this library and encourage them to comment in this bug? Bugzilla works best when we work together as a community to reach out and address bugs. Similarly, if you ever need help, please don't hesitate to contact me.

Comment 11

[Scoobidiver (away)]

Additional useful data :
  Bios version : Phoenix Technologies LTD V2.10 06/05/2010 (up-to-date for Acer)
  Video Bios version : 1718.0(up-to-date for Acer, seems to be not up-to-date for the last Intel graphic driver)

With the generic Intel driver 8.15.10.2141 (not recommended by Acer), it seems a video bios upgrade must be done among a choice of 4 files (no information about that in the Readme files) :
+ vbios_g4x.zip
+ vbios_gm4x.zip
+ vbios_HD.zip
+ vbios_HDm.zip

To the Intel guys, should I upgrade my vbios ? Which one I choose ? Is it possible to recommend to Acer to upgrade their support center with the latest Intel graphic driver and the latest vbios

Comment 12

[(not currently active) Ted Mielczarek]

FWIW, I downloaded the raw dump and loaded it in WinDBG, and the stack I get there isn't any better, so AFAICT there's not a Breakpad bug. If we had symbols for ig4icd32.dll we might do better.

Comment 13

[Benoit Jacob [:bjacob] (mostly away)]

Thinking about it, the only reasonable explanation that I can imagine for why we don't have any backtrace, is that the stack is corrupted. Could there be a stack-corrupting bug (e.g. infinite recursion) at that location in the intel driver?

Comment 14

[(not currently active) Ted Mielczarek]

Well, alternately, that frame could have a plausible-looking frame pointer in %ebp that the stack walker follows, only to wind up out in random memory. When the stack walker doesn't have symbols for a frame its options are limited, but it tries the frame pointer first (which may not be the smartest decision). Although given that WinDBG can't do any better, you may be right.

Comment 15

[Roger WANG]

(In reply to comment #10)
> I've cc'd people from intel.com who have touched bugzilla since june. My
> bugmail indicates that they have the following roles (sorted from most likely
> to be helpful):
>  andy is a windows plugin guy (not sure what kind of windows plugin intel has)
>  chris is a linux (web)gl guy
>  sergey is an optimization guy (cpu, not gpu)
>  roger is a meego guy
> 
> I'm not certain about the interest areas of the others.
> 
> Guys: could one of you please find someone involved in this library and
> encourage them to comment in this bug? Bugzilla works best when we work
> together as a community to reach out and address bugs. Similarly, if you ever
> need help, please don't hesitate to contact me.

yeah. I've forwarded this bug to one guy I know, who is close to the win gfx team. Hopefully it will reach the right people and give some response here.

Comment 16

[andrzej.trznadel]

Could you please verify the demo with the latest driver from Intel download page (named 15.17.9.64.2182 or 8.15.10.2182)?
We've tested it internally and it seems to work OK now, but of course there might be some OS or HW platforms related issues.

Comment 17

[timeless]

http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=19224&lang=eng

Comment 18

[Scoobidiver (away)]

No more problem with Graphic driver : 8.15.10.2182

Comment 19

[timeless]

thanks scoobidiver, and thanks andrzej