Brad King wrote:
> Brian Paul wrote:
>> Do those tests use the shading language?  If so, could I see the shaders?
>>
>> Actually, pointers to the location of these programs in your CVS 
>> repository would be helpful.
> 
> The actual test source is here:
> 
> http://www.vtk.org/cgi-bin/viewcvs.cgi/Rendering/Testing/Cxx/TestOpacity.cxx?rev=1.2&view=auto
>  
> 
> 
> I think the source you want is here:
> 
> http://www.vtk.org/cgi-bin/viewcvs.cgi/Rendering/vtkOpenGLRenderer.cxx?rev=1.67&view=auto
>  
> 
> 
> The variable "vtkOpenGLRenderer_PeelingFS" contains a shader program to 
> do "depth peeling".  There is also other code in the source file that 
> checks for all the extensions needed to implement the shader.

OK, I'll look into that.  Probably just a shading language bug.


>>> The "OTHER_FAULT" ones are due to other mesa assertion failures.  The
>>> SEGFAULT ones seem to have some kind of memory corruption caused by
>>> mesa.  These failures started with changes from Feb 2 to Feb 4 in mesa.
>>
>> Hmmm, I don't see anything between Feb 2-4 that looks too likely.  
>> Could you send a stack trace from one of those failures?
> 
> The reason I think it is mesa changes during those dates is because I 
> can use a Mesa from Feb 2 and nothing crashes, but when I update to Mesa 
> from Feb 4 with no other changes things start crashing.  I tried Feb 3 
> Mesa versions but they did not compile.  However, I do not know much 
> about git so these updates were all done with git-diff piped to patch so 
> I do not know for sure I had the proper versions.  Is there a better way 
> to update to a given date?

Well, there's git-bisect which should narrow it down to a particular 
change.  But I think I've fixed a problem in the osmesa driver that was 
causing the renderbuffer assertion.


> For the OTHER_FAULT failures (mesa assertions) I can give stack traces. 
>  For the SEGFAULT ones the stack traces all point to VTK code.  My 
> hypothesis is that mesa is writing over some VTK memory and corrupting 
> it.  Let's focus on the OTHER_FAULT ones for now.  Here is the 
> OSCone-image backtrace:
> 
> #5  0xb4611fb9 in abort () from /lib/tls/i686/cmov/libc.so.6
> #6  0xb4609fbf in __assert_fail () from /lib/tls/i686/cmov/libc.so.6
> #7  0xb62103e4 in _mesa_reference_renderbuffer (ptr=0x82261c8, 
> rb=0x8226460)
>     at main/renderbuffer.c:2172
> #8  0xb6210569 in _mesa_add_renderbuffer (fb=0x82260a0, bufferName=0,
>     rb=0x8226460) at main/renderbuffer.c:2109
> #9  0xb61670bc in OSMesaMakeCurrent (osmesa=0x81eeca8, buffer=0xb3d6c008,
>     type=5121, width=300, height=300) at drivers/osmesa/osmesa.c:1389
> #10 0xb66decef in vtkXOpenGLRenderWindow::MakeCurrent (this=0x81d3d50)
>     at /home/kitware/Dashboards/My 
> Tests/VTK/Rendering/vtkXOpenGLRenderWindow.cxx:1255
> 
> The assertion failure that called abort was:
> 
> main/renderbuffer.c:2172: _mesa_reference_renderbuffer: Assertion 
> `rb->Magic == 0xaabbccdd' failed.

Let me know how things look with the above-mentioned change.

-Brian

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Mesa3d-dev mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

Reply via email to