[Bug 56368] New: hlsl_types_are_equal assert when vertex shaders uses SV_VertexID as lookup index

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

Bug ID: 56368
           Summary: hlsl_types_are_equal assert when vertex shaders uses
                    SV_VertexID as lookup index
           Product: vkd3d
           Version: 1.10
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: hlsl
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 76109
  --> https://bugs.winehq.org/attachment.cgi?id=76109
test.c

Assertion failed: hlsl_types_are_equal(arg1->data_type, arg2->data_type), file
../wine/libs/vkd3d/libs/vkd3d-shader/hlsl.c, line 1320

This assert happens when D3DCompileFromFile or D3DCompile is called on hlsl
vertex shader (vs_5_0 profile) source like this:

struct Input
    {
      float4 c0 : COL0;
      float4 c1 : COL1;
      float4 c2 : COL2;
      float4 c3 : COL3;
      uint   id : SV_VertexID;
    };

float4 vs_main(Input input) : SV_Position
    {
      float4 col[] = { input.c0, input.c1, input.c2, input.c3 };
      return col[input.id];
    }

See attached test.c file that reproduces this assert. On Windows it prints out
"ok = 1" and shader is successfully compiler. It can be compiled also with
fxc.exe on commandline.

[Bug 56368] hlsl_types_are_equal assert when vertex shaders uses SV_VertexID as lookup index

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

--- Comment #1 from Martins  ---
Created attachment 76110
  --> https://bugs.winehq.org/attachment.cgi?id=76110
shader.hlsl

[Bug 56368] hlsl_types_are_equal assert when vertex shaders uses SV_VertexID as lookup index

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #2 from Nikolay Sivov  ---
Confirming. That's something about variable index access:

#0  __pthread_kill_implementation (threadid=<optimized out>,
signo=signo@entry=6, no_tid=no_tid@entry=0) at ./nptl/pthread_kill.c:44
#1  0x00007ffff7d361cf in __pthread_kill_internal (signo=6, threadid=<optimized
out>) at ./nptl/pthread_kill.c:78
#2  0x00007ffff7ce8472 in __GI_raise (sig=sig@entry=6) at
../sysdeps/posix/raise.c:26
#3  0x00007ffff7cd24b2 in __GI_abort () at ./stdlib/abort.c:79
#4  0x00007ffff7cd23d5 in __assert_fail_base (fmt=0x7ffff7e46dc8 "%s%s%s:%u:
%s%sAssertion `%s' failed.\n%n", 
    assertion=assertion@entry=0x7ffff7f866d0
"hlsl_types_are_equal(arg1->data_type, arg2->data_type)",
file=file@entry=0x7ffff7f7c766 "libs/vkd3d-shader/hlsl.c",
line=line@entry=1345, 
    function=function@entry=0x7ffff7f93870 <__PRETTY_FUNCTION__.44>
"hlsl_new_binary_expr") at ./assert/assert.c:92
#5  0x00007ffff7ce13a2 in __assert_fail
(assertion=assertion@entry=0x7ffff7f866d0
"hlsl_types_are_equal(arg1->data_type, arg2->data_type)", 
    file=file@entry=0x7ffff7f7c766 "libs/vkd3d-shader/hlsl.c",
line=line@entry=1345, function=function@entry=0x7ffff7f93870
<__PRETTY_FUNCTION__.44> "hlsl_new_binary_expr")
    at ./assert/assert.c:101
#6  0x00007ffff7f19578 in hlsl_new_binary_expr (ctx=ctx@entry=0x7fffffffd6b0,
op=op@entry=HLSL_OP2_MUL, arg1=0x555555583000, arg2=arg2@entry=0x555555572e00)
at libs/vkd3d-shader/hlsl.c:1345
#7  0x00007ffff7f21e7a in new_offset_from_path_index (loc=<optimized out>,
offset_component=<synthetic pointer>, regset=<optimized out>, idx=<optimized
out>, base_offset=0x555555582f70, 
    type=0x55555556aad0, block=0x7fffffffd490, ctx=<optimized out>) at
libs/vkd3d-shader/hlsl_codegen.c:63
#8  new_offset_instr_from_deref (loc=0x555555571710,
offset_component=<synthetic pointer>, deref=0x555555571738,
block=0x7fffffffd480, ctx=0x7fffffffd6b0)
    at libs/vkd3d-shader/hlsl_codegen.c:132
#9  replace_deref_path_with_offset (ctx=0x7fffffffd6b0, deref=0x555555571738,
instr=0x5555555716e0) at libs/vkd3d-shader/hlsl_codegen.c:171
#10 0x00007ffff7f25492 in hlsl_transform_ir (ctx=ctx@entry=0x7fffffffd6b0,
func=func@entry=0x7ffff7f220a0 <transform_instr_derefs>,
block=block@entry=0x555555569ca0, 
    context=context@entry=0x7ffff7f21c30 <replace_deref_path_with_offset>) at
libs/vkd3d-shader/hlsl_codegen.c:631
#11 0x00007ffff7f282a6 in transform_derefs (block=0x555555569ca0,
func=0x7ffff7f21c30 <replace_deref_path_with_offset>, ctx=0x7fffffffd6b0) at
libs/vkd3d-shader/hlsl_codegen.c:703
#12 hlsl_emit_bytecode (ctx=ctx@entry=0x7fffffffd6b0,
entry_func=0x555555569c50, target_type=VKD3D_SHADER_TARGET_DXBC_TPF,
out=out@entry=0x7fffffffde10)
    at libs/vkd3d-shader/hlsl_codegen.c:5121
#13 0x00007ffff7f1e9db in hlsl_compile_shader (hlsl=hlsl@entry=0x7fffffffdc60,
compile_info=0x7fffffffdd30, out=out@entry=0x7fffffffde10, 
    message_context=message_context@entry=0x7fffffffdc70) at
libs/vkd3d-shader/hlsl.c:3745
#14 0x00007ffff7f5c937 in compile_hlsl (message_context=0x7fffffffdc70,
out=0x7fffffffde10, compile_info=<optimized out>) at
libs/vkd3d-shader/vkd3d_shader_main.c:1656
#15 vkd3d_shader_compile (compile_info=<optimized out>, out=0x7fffffffde10,
messages=0x7fffffffdcf8) at libs/vkd3d-shader/vkd3d_shader_main.c:1725
#16 0x0000555555557040 in main (argc=<optimized out>, argv=<optimized out>) at
programs/vkd3d-compiler/main.c:860

[Bug 56368] hlsl_types_are_equal assert when vertex shaders uses SV_VertexID as lookup index

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

--- Comment #3 from Nikolay Sivov  ---
This fixes it for me https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/909.

[Bug 56368] hlsl_types_are_equal assert when vertex shaders uses SV_VertexID as lookup index

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

--- Comment #4 from Martins  ---
Thanks! I can confirm that change works for me too. Now wine runs the
application with actual shader without problems.

[Bug 56368] hlsl compilatio fails when uint input parameter is used to index arrays

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|hlsl_types_are_equal assert |hlsl compilatio fails when
                   |when vertex shaders uses    |uint input parameter is
                   |SV_VertexID as lookup index |used to index arrays

[Bug 56368] hlsl compilation fails when uint input parameter is used to index arrays

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|hlsl compilatio fails when  |hlsl compilation fails when
                   |uint input parameter is     |uint input parameter is
                   |used to index arrays        |used to index arrays

[Bug 56368] hlsl compilation fails when uint input parameter is used to index arrays

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED
      Fixed by SHA1|                            |d1c2ae3f0eaf0f3b64b0a5f4c50
                   |                            |ad402e822a4eb

--- Comment #5 from Nikolay Sivov  ---
This is now fixed,
https://gitlab.winehq.org/wine/vkd3d/-/commit/d1c2ae3f0eaf0f3b64b0a5f4c50ad4....

[Bug 56368] hlsl compilation fails when uint input parameter is used to index arrays

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56368

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #6 from Nikolay Sivov  ---
Closing bugs fixed in 1.14.