Bug 104197

(In reply to bugzilla from comment #0)
> After af2c320190, running either League of Legends or the Blizzard Launcher
> (through wine) causes those programs to crash almost immediately upon
> startup.
> 

Can you get me an apitrace reproducing the issue?

Thanks.

(In reply to Francisco Jerez from comment #1)
> Can you get me an apitrace reproducing the issue?

I've tried but I haven't been able to so far. 

Running
LD_PRELOAD=/usr/lib32/apitrace/wrappers/glxtrace.so wine LeagueClient.exe
gives me an empty file for output. Trying with the 64 bit wrapper gives no output file at all.

apitrace works fine on glxgears.

I'm not very familiar with this so I may be doing something wrong.

(In reply to bugzilla from comment #2)
> (In reply to Francisco Jerez from comment #1)
> > Can you get me an apitrace reproducing the issue?
> 
> I've tried but I haven't been able to so far. 
> 

A shader dump (e.g. using the MESA_SHADER_CAPTURE_PATH environment variable) may be sufficient for me to reproduce the issue.

I saw this when using Valgrind for another program/problem:
------------------------------------------------------
==21860== Conditional jump or move depends on uninitialised value(s)
==21860==    at 0x6DDCB08: fs_visitor::bank_conflict_cycles(fs_inst const*) const (brw_fs_bank_conflicts.cpp:906)
==21860==    by 0x6E5981F: fs_instruction_scheduler::issue_time(backend_instruction*) (brw_schedule_instructions.cpp:1546)
==21860==    by 0x6E5F57B: instruction_scheduler::schedule_instructions(bblock_t*) (brw_schedule_instructions.cpp:1602)
==21860==    by 0x6E5F7B4: instruction_scheduler::run(cfg_t*) (brw_schedule_instructions.cpp:1704)
==21860==    by 0x6E5F9B0: fs_visitor::schedule_instructions(instruction_scheduler_mode) (brw_schedule_instructions.cpp:1730)
==21860==    by 0x6DFB8B5: fs_visitor::allocate_registers(unsigned int, bool) (brw_fs.cpp:6071)
==21860==    by 0x6DFE1BC: fs_visitor::run_fs(bool, bool) (brw_fs.cpp:6454)
==21860==    by 0x6DFF1CA: brw_compile_fs (brw_fs.cpp:6789)
==21860==    by 0x6EB170C: blorp_compile_fs (blorp.c:194)
==21860==    by 0x6EC179C: brw_blorp_get_blit_kernel (blorp_blit.c:1331)
==21860==    by 0x6EC179C: try_blorp_blit (blorp_blit.c:1859)
==21860==    by 0x6EC179C: do_blorp_blit (blorp_blit.c:2000)
==21860==    by 0x6EC4C12: blorp_copy (blorp_blit.c:2525)
==21860==    by 0x6B81E1E: copy_buffer_to_image (anv_blorp.c:391)
------------------------------------------------------

Can you give a try to this [1] patch I attached to bug #104199 to make sure this is not a duplicate?

[1] https://bugs.freedesktop.org/attachment.cgi?id=136094

I tried with git master as of just now (dc07bb5fd1), which looks like it should already include the patch from bug 104199. It does not fix my crash.

MESA_SHADER_CAPTURE_PATH gives only a single output file, with the following contents:

[require]
GL_ARB_vertex_program

[vertex program]
!!ARBvp1.0
PARAM C[66] = { program.env[0..65] };
ADDRESS A0;PARAM zero = {0.0, 0.0, 0.0, 0.0};
ARL A0.x, zero.x;
MOV result.position, C[A0.x + 65];
END

I can't reproduce the crash with that shader.  Can you get me your config.log, and a copy of your Mesa binary libraries? (including i965_dri.so)

Can you confirm whether the patch [1] I just attached to bug 104226 fixes the crash?  This is most likely the same issue.

[1] https://bugs.freedesktop.org/attachment.cgi?id=136240

Should be fixed in master assuming that this is the same problem described in bug #104226.  Please reopen if things still don't work for you.