The following is my attempts to locate the bug. Sorry, it is a bit long, but I hope to be informative.

I launched wine on Poser.exe. I also have difficulties with blender, but not the same kind.
I am using Mesa-6.5.2 with the latest Ubuntu Feisty, I have not found any packet with 6.5.3 and my attempts to compile the original mesa failed -- there are too many patches for debian (why??). However, I feel that the bug does not rely on my specific distribution.


Firstly, the original event :


pierre@ios:~/wine/Poser$ wine Poser.exe
main/renderbuffer.c:2041: _mesa_add_renderbuffer:  l'assertion « bufferName == BUFFER_DEPTH || bufferName == BUFFER_STENCIL || fb->Attachment[bufferName].Renderbuffer == ((void *)0) » a echoue.
wine: Assertion failed at address 0xffffe410 (thread 0009), starting debugger...
Unhandled exception: assertion failed in 32-bit code (0xffffe410).
Register dump:
 CS:0073 SS:007b DS:007b ES:007b FS:0033 GS:003b
 EIP:ffffe410 ESP:0033f550 EBP:0033f568 EFLAGS:00000206(   - 00      - -IP1)
 EAX:00000000 EBX:00005790 ECX:00005790 EDX:00000006
 ESI:0033f608 EDI:b7dd8ff4
Stack dump:
0x0033f550:  0033f568 00000006 00005790 b7cc5df0
0x0033f560:  b7dd8ff4 b7c97a00 0033f694 b7cc7641
0x0033f570:  00000006 0033f608 00000000 00000130
0x0033f580:  7c40db08 00000000 00000000 b7d00dfd
0x0033f590:  0033f5cc 7d492ea8 7d492f0c 0033f6a4
0x0033f5a0:  b7dd8ff4 000000c6 000000c7 0033f678
Backtrace:
=>1 0xffffe410 (0x0033f568)
  2 0xb7cc7641 abort+0x101() in libc.so.6 (0x0033f694)
  3 0xb7cbf43b __assert_fail+0xfb() in libc.so.6 (0x0033f6d8)
  4 0x7de37bac _mesa_add_renderbuffer+0x15c() in unichrome_dri.so (0x0033f6f8)
  5 0x7ddd04c3 in unichrome_dri.so (+0x584c3) (0x0033f728)
  6 0x7ddd0dea viaMakeCurrent+0x1ba() in unichrome_dri.so (0x0033f758)
  7 0x7ddccfa9 in unichrome_dri.so (+0x54fa9) (0x0033f798)
  8 0x7ed1e0ce glXMakeContextCurrent+0xbe() in libgl.so.1 (0x0033f808)
  9 0x7ed1e3b3 glXMakeCurrent+0x23() in libgl.so.1 (0x0033f828)
  10 0x7dffc327 X11DRV_wglMakeCurrent+0xd7() in winex11 (0x0033f888)
  11 0x7e9527b3 wglMakeCurrent+0x63() in gdi32 (0x0033f8b8)
  12 0x008e6819 in poser (+0x4e6819) (0x03ccc2b0)
  13 0x00000000 (0x01501584)
  14 0x00785c40 in poser (+0x385c40) (0x00785bb0)



Not very clear, so I put some fprintf(stderr,... in src/mesa/drivers/dri/unichrome/via_context.c, especially in viaMakeCurrent, before and after the 2 calls to calculate_buffer_parameters, and  :


>>>	  fprintf(stderr,"viaMakeCurrent 1: calculate_buffer_parameters ?\n");
858:	  if (!calculate_buffer_parameters(vmesa, drawBuffer, driDrawPriv)) {
859:	     return GL_FALSE;
860:	  }
>>>	  fprintf(stderr,"viaMakeCurrent 1: calculate_buffer_parameters ok -> true\n");

There are two calls to calculate_buffer_parameters, but only the first one fails.
and 


220:   if (!vmesa->front.Base.InternalFormat) {
221:      /* do one-time init for the renderbuffers */
222:      viaInitRenderbuffer(&vmesa->front, GL_RGBA, dPriv);
223:      viaSetSpanFunctions(&vmesa->front, &fb->Visual);
>>>	  fprintf(stderr,"calculate_buffer_parameters 1 => add_renderbuffer BUFFER_FRONT_LEFT ?\n");
224:      _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &vmesa->front.Base);
>>>	  fprintf(stderr,"calculate_buffer_parameters 1 => add_renderbuffer BUFFER_FRONT_LEFT ok\n");
225:
226:      if (fb->Visual.doubleBufferMode) {

and so on for the 3 other buffers.

This is the dump (having removed the repetitions) :

pierre@ios:~/wine/Poser 5$ wine Poser.exe
viaMakeCurrent 1: calculate_buffer_parameters ?
calculate_buffer_parameters 1 => add_renderbuffer BUFFER_FRONT_LEFT ?
calculate_buffer_parameters 1 ok
calculate_buffer_parameters 2 => add_renderbuffer BUFFER_BACK_LEFT ?
calculate_buffer_parameters 2 ok
calculate_buffer_parameters 3 => add_renderbuffer BUFFER_DEPTH ?
calculate_buffer_parameters 3 ok
calculate_buffer_parameters 4 => add_renderbuffer BUFFER_STENCIL ?
calculate_buffer_parameters 4 ok
viaMakeCurrent 1: calculate_buffer_parameters ok -> true
...
viaMakeCurrent 1: calculate_buffer_parameters ?
calculate_buffer_parameters 1 => add_renderbuffer BUFFER_FRONT_LEFT ?
main/renderbuffer.c:2045: _mesa_add_renderbuffer:  l'assertion « bufferName == BUFFER_DEPTH || bufferName == BUFFER_STENCIL || fb->Attachment[bufferName].Renderbuffer == ((void *)0) » a echoue.
wine: Assertion failed at address 0xffffe410 (thread 0009), starting debugger...



Clearly, the failure appears when vmesa->front.Base.InternalFormat == 0 and when bufferName == BUFFER_FRONT_LEFT. This means that there was an attachment with a renderbuffer in the current framebuffer and we are trying to add another. Just to remember the data structures :

struct gl_renderbuffer_attachment
{
   GLenum Type;  /* GL_NONE or GL_TEXTURE or GL_RENDERBUFFER_EXT */
   ...
   /* IF Type == GL_RENDERBUFFER_EXT: */
   struct gl_renderbuffer *Renderbuffer;
   ...
};

struct gl_framebuffer
{
   ...
   struct gl_renderbuffer_attachment Attachment[BUFFER_COUNT];
   ...
};



Next, I added many prints of every parameters of type viaContext, gl_renderbuffer and gl_framebuffer in viaCreateContext(), viaMakeCurrent(), calculate_buffer_parameters() and _mesa_add_renderbuffer().
This is the result, very long. The actual hex addresses have been replaced by symbols (CONTEXT3...)




... (skipping the beginning)
...


viaCreateContext() => vmesa=CONTEXT3
	via_context vmesa=CONTEXT3
		->depthBits=24
		&->front.Base=RENDERBUF3
	gl_renderbuffer *rb=RENDERBUF3
		->InternalFormat=0
		->Name=0
		->Width=0,Height=0
		->viaScreen->width=1024, height=768

viaMakeCurrent(vmesa=CONTEXT3,drawBuffer=FRAMEBUF3)

calculate_buffer_parameters(vmesa=CONTEXT3,fb=FRAMEBUF3): &vmesa->front.Base=RENDERBUF3 vmesa->front.Base.InternalFormat=0
	via_context vmesa=CONTEXT3
		->depthBits=24
		&->front.Base=RENDERBUF3
	gl_renderbuffer *rb=RENDERBUF3
		->InternalFormat=0
		->Name=0
		->Width=0,Height=0
		->viaScreen->width=1024, height=768
	gl_framebuffer *fb=FRAMEBUF3
		->Name=0
		->Width=640,Height=480
		->Attachment[FRONT]=0
IF==0 => calls to _mesa_add_renderbuffer()
_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=0)
	=>_mesa_new_renderbuffer : RENDERBUF3 attached to fb...[0]
_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=1)
	=>_mesa_new_renderbuffer : 7db0bcb8 attached to fb...[1]
_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=8)
	=>_mesa_new_renderbuffer : 7db0bd64 attached to fb...[8]
_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=9)
	=>_mesa_new_renderbuffer : 7db0be10 attached to fb...[9]



viaCreateContext() => vmesa=CONTEXT4
	via_context vmesa=CONTEXT4
		->depthBits=24
		&->front.Base=RENDERBUF4
	gl_renderbuffer *rb=RENDERBUF4
		->InternalFormat=0
		->Name=0
		->Width=0,Height=0
		->viaScreen->width=1024, height=768

viaMakeCurrent(vmesa=CONTEXT4,drawBuffer=FRAMEBUF3)

calculate_buffer_parameters(vmesa=CONTEXT4,fb=FRAMEBUF3): vmesa->front.Base=RENDERBUF4 vmesa->front.Base.InternalFormat=0
	via_context vmesa=CONTEXT4
		->depthBits=24
		&->front.Base=RENDERBUF4
	gl_renderbuffer *rb=RENDERBUF4
		->InternalFormat=0
		->Name=0
		->Width=0,Height=0
		->viaScreen->width=1024, height=768
		->driDrawable->w=640, h=480
	gl_framebuffer *fb=FRAMEBUF3
		->Name=0
		->Width=640,Height=480
		->Attachment[FRONT]=RENDERBUF3
	gl_renderbuffer *rb=RENDERBUF3
		->InternalFormat=6408
		->Name=0
		->Width=0,Height=0
IF==0 => calls to _mesa_add_renderbuffer()
_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=0):

THE ASSERT WOULD FAIL HERE (with my workaround, it does not fail, returns and continues)

_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=1): no new renderbuffer
_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=8)
	=>_mesa_new_renderbuffer : 7d56a484 attached to fb
_mesa_add_renderbuffer(struct gl_framebuffer *fb=FRAMEBUF3,bufferName=9)
	=>_mesa_new_renderbuffer : 7d56a530 attached to fb
calculate_buffer_parameters(vmesa=CONTEXT4,fb=FRAMEBUF3)=>
gl_framebuffer *fb=FRAMEBUF3
	->Name=0
	->Width=640,Height=480
	->Attachment[FRONT]=RENDERBUF3 = gl_renderbuffer *rb=RENDERBUF3
	->InternalFormat=6408
	->Name=0
	->Width=0,Height=0
gl_renderbuffer *rb=RENDERBUF4
	->InternalFormat=6408
	->Name=0
	->Width=0,Height=0


Finally, I quote these lines in via_context.c

   /* Normally, the renderbuffer would be added to the framebuffer just once
    * when the framebuffer was created.  The VIA driver is a bit funny
    * though in that the front/back/depth renderbuffers are in the per-context
    * state!
    * That should be fixed someday.

Not so funny to my opinion...  ;
-)
But with the help of everyone, we will fix this.