603175 - Crash [@ libOSMesa.so.6.5.3@0x122bb5 ]

Status: RESOLVED FIXED

(Core :: Graphics: CanvasWebGL, defect)

Description

[Alice0775 White]

Build Identifier: Mozilla/5.0 (X11; Linux i686; rv:2.0b8pre) Gecko/20101010 Firefox/4.0b8pre ID:20101010025822

Crash with crash report
bp-ff03249a-273b-4064-b9d0-8ef262101010
bp-e18f4161-2fe4-4479-aa5c-b223f2101010

Reproducible: Always

Steps to Reproduce:
1. Start Minefield with new profile
    webgl.force_osmesa;true
    webgl.osmesalib;/usr/lib/libOSMesa.so (libOSMesa16.so.6.5.3)

2. Open URL ( http://learningwebgl.com/lessons/lesson01/index.html )


Actual Results:
 the following alert box pops up.
  Error: 2001: Invalid external declaration.
  Could not initialise shaders

 And finaly, Browser crashes.

Expected Results:
 not crash

Regression window:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=a0861f461354&tochange=d7b58beb717c

Comment 1

[Alice0775 White]

Browser crashes when open page http://www.iquilezles.org/apps/shadertoy/
And same regression range.

Comment 2

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

Regarding the first stack trace:
This looks like a ANGLE bug. Does the crash persist when you set webgl.shader_validator = false ?
If you can run this in a debugger (using a debug build) can you print the value of the ts pointer at this line?

I don't know what to think about the second stack trace. Plain OSMesa bug?

Vlad: if OSMesa is unmaintained, perhaps we need to stop supporting it? Perhaps there is a way that we can keep supporting it while making it clear enough that it's at the user's own risk?

Comment 3

[Alice0775 White]

(In reply to comment #2)
> Regarding the first stack trace:
> This looks like a ANGLE bug. Does the crash persist when you set
> webgl.shader_validator = false ?
If I set webgl.shader_validator to false, 
the Browser does NOT crash. and it seems no problem.

>If you can run this in a debugger (using a debug build) can you print the value
>of the ts pointer at this line?
It difficult for me.

Comment 4

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

(In reply to comment #3)
> (In reply to comment #2)
> > Regarding the first stack trace:
> > This looks like a ANGLE bug. Does the crash persist when you set
> > webgl.shader_validator = false ?
> If I set webgl.shader_validator to false, 
> the Browser does NOT crash. and it seems no problem.

Great! Does that solve both crashes that you reported in comment 0, or only the first one? (i.e. does the second one still occur?)

> 
> >If you can run this in a debugger (using a debug build) can you print the value
> >of the ts pointer at this line?
> It difficult for me.

Sure, no problem, I'll try to reproduce.

Comment 5

[Alice0775 White]

(In reply to comment #4)
> (In reply to comment #3)
> > (In reply to comment #2)
> > > Regarding the first stack trace:
> > > This looks like a ANGLE bug. Does the crash persist when you set
> > > webgl.shader_validator = false ?
> > If I set webgl.shader_validator to false, 
> > the Browser does NOT crash. and it seems no problem.
> 
> Great! Does that solve both crashes that you reported in comment 0, or only the
> first one? (i.e. does the second one still occur?)
> 
Both pages.

Comment 6

[Alice0775 White]

The crash bug fixed by landing of 73a03305165d    Vladimir Vukicevic —
b=603235, fix string usage; r=bas, a=crash