[Mesa-dev] [PATCH] util/u_pstipple.c: copy immediates during transformation

[Nicolai Hähnle]

On 18.01.2016 17:54, Jose Fonseca wrote:
> On 18/01/16 20:28, Nicolai Hähnle wrote:
>> From: Nicolai Hähnle <nicolai.haehnle at amd.com>
>>
>> Apparently, nobody has combined stippling with a fragment shader
>> containing immediates in almost five years...
>>
>> Fixes a bug in Kodi with radeonsi reported by Christian König.
>
> Good find.
>
> I don't know how to interpret this: if the fact it wasn't caught for so
> long corroborates the fact that nobody mixes legacy and new OpenGL, or
> if the fact it was found proves people do mix it!

The slightly insane synthesis of those two positions: they do mix it, 
but not enough! ;)

Cheers,
Nicolai

> One thing is sure, we don't really have much test coverage of it, as
> most of our piglit tests tend to aim to the minimum GL a feature was
> introduced, hence rarely cover interactions between older and newer
> features.
>
>> Cc: "11.0 11.1" <mesa-stable at lists.freedesktop.org>
>> Tested-by: Christian König <christian.koenig at amd.com>
>> Reviewed-by: Marek Olšák <marek.olsak at amd.com>
>> ---
>>   src/gallium/auxiliary/util/u_pstipple.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/src/gallium/auxiliary/util/u_pstipple.c
>> b/src/gallium/auxiliary/util/u_pstipple.c
>> index 08dec13..3428172 100644
>> --- a/src/gallium/auxiliary/util/u_pstipple.c
>> +++ b/src/gallium/auxiliary/util/u_pstipple.c
>> @@ -230,6 +230,7 @@ pstip_transform_immed(struct
>> tgsi_transform_context *ctx,
>>      struct pstip_transform_context *pctx =
>>         (struct pstip_transform_context *) ctx;
>>      pctx->numImmed++;
>> +   ctx->emit_immediate(ctx, immed);
>>   }
>>
>
> Ironically, this was fixed in
> src/gallium/auxiliary/draw/draw_pipe_pstipple.c by Brian in
> 7f16246acef4089570abca76a59580691ec6cf68
>
>
> I think we should compare and see if there's been anything else that
> diverged in this files that should be crossported.
>
> Or alternatively, see if there's some way we could have the code once
> (e.g, have draw call u_pstipple.c?  Or force all drivers that use draw
> to also use u_pstipple?)
>
>
> BTW, maybe it would be better to move the emit_immediate() call one line
> up, so it lines up with draw_pipe_pstipple.c.
>
> Otherwise,
>
> Reviewed-by: Jose Fonseca <jfonseca at vmware.com>
>