[Piglit] [PATCH v3 2/2] arb_transform_feedback_overflow_query: add error tests.

[Dylan Baker]

Quoting Ilia Mirkin (2016-12-15 11:33:59)
> On Thu, Dec 15, 2016 at 2:27 PM, Rafael Antognolli
> <rafael.antognolli at intel.com> wrote:
> > On Thu, Dec 15, 2016 at 10:58:10AM -0800, Dylan Baker wrote:
> >> Quoting Rafael Antognolli (2016-12-13 14:51:14)
> >> > Add tests to verify that GL_ERROR_VALUE is returned if an invalid
> >> > index is used to begin, end or get a current query.
> >> >
> >> > v2:
> >> >     - Add test to all.py (Ilia Mirkin)
> >> >     - Add check for ARB_transform_feedback_overflow_query extension
> >> > v3:
> >> >     - Skip test if some extensions are not present (Ilia Mirkin)
> >> >
> >> > Signed-off-by: Rafael Antognolli <rafael.antognolli at intel.com>
> >> > ---
> >> >  tests/all.py                                       |   2 +
> >> >  .../CMakeLists.gl.txt                              |   1 +
> >> >  .../arb_transform_feedback_overflow_query/errors.c | 279 +++++++++++++++++++++
> >> >  3 files changed, 282 insertions(+)
> >> >  create mode 100644 tests/spec/arb_transform_feedback_overflow_query/errors.c
> >> >
> >> > diff --git a/tests/all.py b/tests/all.py
> >> > index 888f25f..dbfe225 100644
> >> > --- a/tests/all.py
> >> > +++ b/tests/all.py
> >> > @@ -3680,6 +3680,8 @@ with profile.test_list.group_manager(
> >> >          grouptools.join('spec', 'arb_transform_feedback_overflow_query')) as g:
> >> >      g(['arb_transform_feedback_overflow_query-basic'],
> >> >        'arb_transform_feedback_overflow_query-basic', run_concurrent=False)
> >> > +    g(['arb_transform_feedback_overflow_query-errors'],
> >> > +      'arb_transform_feedback_overflow_query-errors', run_concurrent=False)
> >>
> >> See my comment on the previous patch here too.
> >
> > Hmm... makes sense, I just saw how it is used in other places. I can fix
> > that.
> >
> >> Also, I just noticed that you're setting run_concurrent to False, is there a
> >> particular reason that these tests need to be run serially? Generally only tests
> >> that do front buffer rendering or have CPU/GPU synchronization requirements need
> >> to be set to run_concurrent.
> >
> > No front buffer rendering, but I'm not sure about CPU/GPU
> > synchronization. I mean, when we try to get the result from a query on
> > the CPU, it does wait for the GPU, but I'm not sure that's what you are
> > talking about.
> >
> > I was mainly copying & pasting from
> > arb_transform_feedback3-query_with_invalid_index, which does much less
> > stuff that the overflow_query does, and it still uses
> > run_concurrent=False. It just does some queries with invalid
> > parameters. Is that also not needed?
> 
> Probably not. My guess is that 99% of the existing
> run_concurrent=False settings aren't needed. But in 1% of the cases it
> *is* needed, and it's a pain to read all the tests and verify.
> 
>   -ilia

That's why I ask. Easier to not mark them as serialized when they're committed
than to go back later and sort that out.

Dylan
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: signature
URL: <https://lists.freedesktop.org/archives/piglit/attachments/20161215/ab60a8fc/attachment.sig>