[Piglit] [PATCH] dsa: avoid stencil8 in gettextureimage-formats

[Ilia Mirkin]

On Wed, Sep 9, 2015 at 2:22 PM, Eric Anholt <eric at anholt.net> wrote:
> Ilia Mirkin <imirkin at alum.mit.edu> writes:
>
>> On Tue, Sep 8, 2015 at 4:55 PM, Eric Anholt <eric at anholt.net> wrote:
>>> Ilia Mirkin <imirkin at alum.mit.edu> writes:
>>>
>>>> On Fri, May 22, 2015 at 8:37 PM, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
>>>>> This causes an assert failure in mesa and it doesn't seem like this test
>>>>> is designed for depth/stencil formats.
>>>>>
>>>>> Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
>>>>> ---
>>>>>
>>>>> Ian, I know you're rewriting half these tests... feel free to fold this
>>>>> into your series if you like.
>>>>>
>>>>>  tests/spec/arb_direct_state_access/gettextureimage-formats.c | 3 ++-
>>>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/tests/spec/arb_direct_state_access/gettextureimage-formats.c b/tests/spec/arb_direct_state_access/gettextureimage-formats.c
>>>>> index e2d637c..ce25205 100644
>>>>> --- a/tests/spec/arb_direct_state_access/gettextureimage-formats.c
>>>>> +++ b/tests/spec/arb_direct_state_access/gettextureimage-formats.c
>>>>> @@ -462,7 +462,8 @@ supported_format_set(const struct test_desc *set)
>>>>>                  set->format == ext_packed_depth_stencil ||
>>>>>                  set->format == arb_texture_rg_int ||
>>>>>                  set->format == arb_depth_texture ||
>>>>> -                set->format == arb_depth_buffer_float) {
>>>>> +                set->format == arb_depth_buffer_float ||
>>>>> +                set->format == arb_texture_stencil8) {
>>>>>                 /*
>>>>>                  * texture_integer requires a fragment shader, different
>>>>>                  * glTexImage calls.  Depth/stencil formats not implemented.
>>>>> --
>>>>> 2.3.6
>>>>>
>>>>
>>>> I recently got a conflict while rebasing this which reminded me that I
>>>> still hadn't pushed this... any objections?
>>>
>>> If we've got a test triggering a Mesa assertion, and this is the only
>>> test doing so, making the test just not trigger the assertion seems bad.
>>
>> But the test isn't supposed to test stencil formats in the first place
>> (see comment in the context).
>
> Sure, but generally we don't avoid Mesa assertions in a piglit test,
> even if the test shouldn't be hitting the path, until we make a separate
> test that *does* hit that path.

OK. TBH I don't even know if there's still a crash... I sent this
patch out 4 months ago. Perhaps in another 4 months someone will
provide such a test and this patch can be pushed. In the meanwhile the
test will fail for drivers that implement ARB_texture_stencil8.

  -ilia