[Mesa-dev] [PATCH] mesa: Use GL_RED for DEPTH_TEXTURE_MODE for #version 300 es shaders.

[Kenneth Graunke]

On 01/21/2013 01:38 PM, Eric Anholt wrote:
> Kenneth Graunke <kenneth at whitecape.org> writes:
>> ---
>>   src/mesa/drivers/dri/i965/brw_state.h             | 3 ++-
>>   src/mesa/drivers/dri/i965/brw_wm.c                | 2 +-
>>   src/mesa/drivers/dri/i965/brw_wm_surface_state.c  | 8 ++++++--
>>   src/mesa/drivers/dri/i965/gen7_wm_surface_state.c | 4 ++--
>>   4 files changed, 11 insertions(+), 6 deletions(-)
>>
>> I believe the BRW_NEW_VERTEX_PROGRAM/BRW_NEW_FRAGMENT_PROGRAM dirty bits
>> on the brw_texture_surfaces atom should cover my new usage of shProg.
>> Eric, can you confirm?  I always get confused by _NEW_PROGRAM vs.
>> CACHE_NEW_WM/VS_PROG vs. BRW_NEW_VERTEX/FRAGMENT_PROGRAM.  Thanks!
>
> CACHE_NEW_*_PROG covers prog_data, BRW_NEW_*_PROGRAM covers the
> brw_*_program and gl_shader_program structs, so that sounds good.  But I
> see an issue anyway...
>
>> diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
>> index 5e99592..b609b09 100644
>> --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
>> +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
>> @@ -685,8 +685,10 @@ brw_get_surface_num_multisamples(unsigned num_samples)
>>    * swizzling.
>>    */
>>   int
>> -brw_get_texture_swizzle(const struct gl_texture_object *t)
>> +brw_get_texture_swizzle(struct gl_context *ctx,
>> +                        const struct gl_texture_object *t)
>>   {
>> +   const struct gl_shader_program *shProg = ctx->Shader._CurrentFragmentProgram;
>
> You're looking at the FS even if we're trying to set up textures for the
> VS.  I think the caller should pass in the particular program, and then
> it'll be more obvious to the caller which state flag is right.

Eric mentioned this on IRC:

<anholt> Kayden: oh, I suppose my comment about looking at the wrong 
program in your depth change doesn't really matter since Version will be 
the same in both.

which is true - otherwise the VS/FS link would have failed.

The trouble is, in at least one of the callers 
(gen7_update_texture_surface), I don't know whether we're setting up 
textures for the VS or for the FS.  So I don't know which I should 
consult.  I could plumb that through, but considering it doesn't matter, 
I'm tempted not to.

Eric, is that okay with you, or would you still like me to pass a 
gl_shader_program rather than gl_context?