[Piglit] [PATCH] primitive-restart: test primitive-restart with VBO

[Brian Paul]

On 05/11/2012 09:24 AM, Olivier Galibert wrote:
> On Fri, May 11, 2012 at 09:06:21AM -0600, Brian Paul wrote:
>> On 05/10/2012 11:34 AM, Jordan Justen wrote:
>>> On Thu, May 10, 2012 at 9:26 AM, Jordan Justen<jljusten at gmail.com>   wrote:
>>>> On Thu, May 10, 2012 at 8:58 AM, Jordan Justen<jljusten at gmail.com>   wrote:
>>>>> On Thu, May 10, 2012 at 7:10 AM, Brian Paul<brianp at vmware.com>   wrote:
>>>>>> I'd like to have a separate test to check the case of one VBO containing
>>>>>> both vertex data and index data.  Would you be interested in writing such a
>>>>>> test?
>>>>>
>>>>> Sure. Something like tests/spec/arb_vertex_buffer_object/combined-vbo.c?
>>>>
>>>> Hmm, it looks like a combined VBO is used in:
>>>> tests/spec/arb_vertex_buffer_object/mixed-immediate-and-vbo.c
>>>>
>>>> Yet llvmpipe passes arb_vertex_buffer_object-mixed-immediate-and-vbo.
>>>
>>> I can't get llvmpipe to fail in
>>> arb_vertex_buffer_object-mixed-immediate-and-vbo even after tweaking
>>> the code. So, perhaps the restart is important here.
>>
>> Sounds like it.
>>
>>
>>> I'd like to leave the combined VBO for index/vertex data in the
>>> primitive-restart test. Do you agree?
>>
>> I'd rather have separate VBOs to keep the test focused on one thing
>> (primitive restart).  I really don't have any time now to
>> investigate/fix the single-vbo + restart bug (it might happen with all
>> gallium drivers, btw) so the test would just keep failing for the
>> foreseeable future.  I'd rather not leave it in that state.
>
> Doesn't that mean there should be both tests?

Yes.


On 05/11/2012 10:16 AM, Jordan Justen wrote:
 > I could see creating a separate test for VBOs (say
 > primitive-restart-vbo) that has the single VBO case.
 >
 > Or do you prefer that I add VBO support to primitive-restart (with 2
 > VBOs), and add a separate primitive-restart-single-vbo test?

If it were me, I'd just write a new, short test that just hits one of 
the failing cases of a single VBO + prim restart.

BTW, there's also the possible case of vertex data in a VBO but the 
array indexes in user memory.  Or, how about display lists?  The point 
is I'd rather not add a ton of new permutations to the primitive 
restart test.

Like I said, I don't have time to investigate the new failure case and 
I'd rather not see a long-time passing test changed to a long-term 
failing test.

-Brian