[Mesa3d-dev] [Bug 5791] New: segfault when building vertex program with sse codegen

[<bug...@fr...>]

Please do not reply to this email: if you want to comment on the bug, go to    
       
the URL shown below and enter yourcomments there.     
   
https://bugs.freedesktop.org/show_bug.cgi?id=5791          
     
           Summary: segfault when building vertex program with sse codegen
           Product: Mesa
           Version: CVS
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Mesa core
        AssignedTo: mes...@li...
        ReportedBy: rsc...@gm...


When trying to run quake4 with software vertex programs, I get a segfault,
here's the backtrace:
#0  0x00656e69 in ?? ()
#1  0x3cabee97 in build_vertex_program (cp=0xaf83e7c0)
    at tnl/t_vb_arbprogram_sse.c:1139
#2  0x3cabefbe in _tnl_sse_codegen_vertex_program (p=0x1152d850)
    at tnl/t_vb_arbprogram_sse.c:1201
#3  0x3cabc55f in compile_vertex_program (program=0x9efff38, try_codegen=115 's')
    at tnl/t_vb_arbprogram.c:1130
#4  0x3cabcbf0 in validate_vertex_program (ctx=0x9ac3140, stage=0x47)
    at tnl/t_vb_arbprogram.c:1420
#5  0x3cab5da1 in _tnl_run_pipeline (ctx=0x9ac3140) at tnl/t_pipeline.c:144
#6  0x3ca28b23 in r200WrapRunPipeline (ctx=0x9ac3140) at r200_state.c:2465
#7  0x3cb5743f in _tnl_draw_range_elements (ctx=0x9ac3140, mode=4, max_index=643,
    index_count=71, indices=0x47) at tnl/t_array_api.c:105
#8  0x3cb579ba in _tnl_DrawElements (mode=4, count=2964, type=5125, indices=0x282)
    at tnl/t_array_api.c:373

The corresponding DISASSEM output was this:
0x44cfe80a: DP4 OUT4, IN0, STATE_VAR:1
0x44cfe86c: DP4 RES, IN0, STATE_VAR:2
0x44cfe8bd: MSK OUT4.y, RES
signal caught: Segmentation fault

This happens with q4 demo and arb_vertex_program set to true (and texture_units
set to 6 on a r200). It only happens if linux-dri-x86 target is used. The weird
thing is, I get a pretty similar crash running the arbvptorus demo but only if I
set MESA_EXPERIMENTAL. But in this case, MESA_EXPERIMENTAL is NOT set, and gdb
tells me try_codegen is 114. So either there is an initialization issue or there
was some memory corruption. And btw it looks like MESA_NO_SSE is not honored for
this stuff (though since you're supposed to enable it manually at this stage
that's probably not a problem).          
     
     
--           
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email         
     
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.