[Piglit] [PATCH] gtf: Update EGLConfig ID

[Ian Romanick]

On 10/09/2012 09:01 AM, Kenneth Graunke wrote:
> On 10/09/2012 07:57 AM, Chad Versace wrote:
>> On 10/08/2012 03:40 PM, Kenneth Graunke wrote:
>>> On 10/08/2012 03:21 PM, Chad Versace wrote:
>>>> The mesa commit 7dc0be8 changed the EGLConfig ID's for the Intel
>>>> driver,
>>>> and piglit's GTF integration hardcodes the EGLConfig ID.  (Apparently,
>>>> this implies that piglit's GTF integration only works the Intel
>>>> driver...
>>>> hmm).
>>>
>>> Most people don't have a copy...
>>>
>>>> The old -id=10 and new -id=7 should be the same underlying DRIConfig.
>>>> Anyway, the two configs are close enough that the GTF results don't
>>>> change.
>>>>
>>>> Signed-off-by: Chad Versace <chad.versace at linux.intel.com>
>>>> ---
>>>>    tests/gtf.tests | 2 +-
>>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/tests/gtf.tests b/tests/gtf.tests
>>>> index cade81a..572c729 100644
>>>> --- a/tests/gtf.tests
>>>> +++ b/tests/gtf.tests
>>>> @@ -45,7 +45,7 @@ class GTFTest(ExecTest):
>>>>        pass_re = re.compile(r'Regression PASSED all 1 tests')
>>>>
>>>>        def __init__(self, testpath):
>>>> -        ExecTest.__init__(self, [path.join(testBinDir, 'GTF'),
>>>> '-noimagefileio', '-width=113', '-height=47', '-seed=2', '-id=10',
>>>> '-run=' +
>>>> testpath])
>>>> +        ExecTest.__init__(self, [path.join(testBinDir, 'GTF'),
>>>> '-noimagefileio', '-width=113', '-height=47', '-seed=2', '-id=7',
>>>> '-run=' +
>>>> testpath])
>>>>
>>>>        def interpretResult(self, out, results):
>>>>        if self.pass_re.search(out) is not None:
>>>
>>> NAK.  We used to use 7, but it doesn't have alpha bits, and we
>>> definitely missed
>>> bugs because of that.
>>>
>>> Try 0x47 (71), 0x48 (72), 0x59 (89), or 0x5a (90).  I'm not sure what
>>> the
>>> difference is, since eglinfo just prints ms/ns/b, which isn't ringing
>>> any bells...
>>
>> I NAK it too. This patch is rubbish. Piglit shouldn't hardcode the
>> EGLConfig id
>> at all, because that prevents bisects across commits that alter the id's.
>>
>> Instead, Piglit should appeal to eglinfo to determine the id.
>
> Hmm.  Yeah, that seems reasonable.  We might want to actually include a
> small utility to do that in Piglit itself, rather than relying on the
> Mesa demos repo.

That's my vote.  We can make a tiny utility that prints the ID of a 
single EGL config that will suffice for GTF.