[Mesa-dev] [PATCH 2/2] mesa: fix GetProgramiv/GetActiveAttrib regression

[Ilia Mirkin]

On Mon, Oct 5, 2015 at 3:33 AM, Tapani Pälli <tapani.palli at intel.com> wrote:
>
>
> On 10/05/2015 10:21 AM, Ilia Mirkin wrote:
>>
>> On Mon, Oct 5, 2015 at 3:14 AM, Tapani Pälli <tapani.palli at intel.com>
>> wrote:
>>>
>>> Commit 4639cea2921669527eb43dcb49724c05afb27e8e added packed varyings
>>> as part of program resource list. We need to take this to account with
>>> all functions dealing with active input attributes.
>>
>>
>> Don't you have this same problem with output attributes as well? Is
>> the situation here that the packed varying pass adds its own varyings
>> on top of the usual ones, and so things were getting (essentially)
>> double-counted?
>
>
> As far as I see the problem is only with active attribute variables,
> glGetActiveAttrib used to take in to account all resources with input type
> (as it was just calling _mesa_program_resource_find_index). Now it
> explicitly checks that varyings are not counted as active attribute
> variables.
>
> For querying outputs via program resource queries I don't see any
> regressions on the current tests so I'm not convinced there is such problem.
>
> For fragment shader outputs there is a problem but I consider it separate
> from this. Mesa lowers fragment output array as series of variables like
> 'gl_out_FragData0, gl_out_FragData1 ..' while user expects there to be
> 'array[0], array[1] ..' and query for array[0] does not match
> gl_out_FragData0. This one I was planning to tackle next.

Oh, well fragment shader outputs don't get packed, so that's not an
issue. There's no way to query the inter-stage stuff? Then what does
this have to do with packing in the first place? Not like vertex
attributes can get packed either...