Re: [Piglit] [PATCH] framework: Add generic deqp options to piglit.conf

[Kenneth Graunke]

On Thursday, March 10, 2016 5:44:40 PM PST Dylan Baker wrote:
> Sometimes there are extra arguments to deqp that a developer wants for
> all of the test suites, this adds that option by putting a generic
> [deqp] section and add an extra_args. This new extra args gets joined
> with any suite specific extra_args
> 
> Signed-off-by: Dylan Baker 
> ---
>  framework/test/deqp.py | 28 +---
>  piglit.conf.example|  4 
>  tests/cts.py   |  6 +-
>  tests/deqp_gles2.py|  7 ++-
>  tests/deqp_gles3.py|  6 +-
>  tests/deqp_gles31.py   |  6 +-
>  6 files changed, 42 insertions(+), 15 deletions(-)

I haven't tested this, but it looks good to me and sounds nice.
That way, I can specify --deqp-visibility=hidden once.

Reviewed-by: Kenneth Graunke 


signature.asc
Description: This is a digitally signed message part.
___
Piglit mailing list
Piglit@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/piglit

Re: [Piglit] [PATCH 1/1] cl: Split setKernelArg into subtests

[Dylan Baker]

Quoting Serge Martin (2016-03-10 02:38:20)
> On Wednesday 09 March 2016 12:09:39 Dylan Baker wrote:
> > Quoting Serge Martin (2016-03-05 07:18:29)
> > 
> > > On Monday 29 February 2016 10:45:47 Jan Vesely wrote:
> > > > Signed-off-by: Jan Vesely 
> > > > ---
> > > > applies on top of Serge's changes.
> > > 
> > > In get-mem-object-info.c we only make subtest reporting for positive case,
> > > error path been reported globally.
> > > 
> > > I think reporting error as subtest is a little to verbose. May be we could
> > > make a subtest report for the whole error case?
> > 
> > This breaks a basic assumption of piglit's report modules: that results
> > don't randomly appear and disappear. If they do then you'll have odd
> > failure/pass rates when comparing two results.
> 
> Don't get me wrong about that. When I was talking about "the whole error case"
> I was talking about the input validation checks that could be coalesced in 
> one 
> subtest to report error or failure. Like the v2 Jan posted in fact.
> 
> But, anyway, I'm ok if this "let's join all input validation in one subtest 
> to 
> make it less verbose" idea of mine is not what people are expecting.

Okay, my mistake.

> 
> > 
> > IE: tests should always report the same number of subtests, in both the
> > pass and fail case, the test suite will automatically make the global
> > state the "worst" state of the subtests.
> > 
> > Dylan
> 


signature.asc
Description: signature
___
Piglit mailing list
Piglit@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/piglit

Re: [Piglit] [PATCH 1/1] cl: Split setKernelArg into subtests

[Serge Martin]

On Wednesday 09 March 2016 12:09:39 Dylan Baker wrote:
> Quoting Serge Martin (2016-03-05 07:18:29)
> 
> > On Monday 29 February 2016 10:45:47 Jan Vesely wrote:
> > > Signed-off-by: Jan Vesely 
> > > ---
> > > applies on top of Serge's changes.
> > 
> > In get-mem-object-info.c we only make subtest reporting for positive case,
> > error path been reported globally.
> > 
> > I think reporting error as subtest is a little to verbose. May be we could
> > make a subtest report for the whole error case?
> 
> This breaks a basic assumption of piglit's report modules: that results
> don't randomly appear and disappear. If they do then you'll have odd
> failure/pass rates when comparing two results.

Don't get me wrong about that. When I was talking about "the whole error case"
I was talking about the input validation checks that could be coalesced in one 
subtest to report error or failure. Like the v2 Jan posted in fact.

But, anyway, I'm ok if this "let's join all input validation in one subtest to 
make it less verbose" idea of mine is not what people are expecting.

> 
> IE: tests should always report the same number of subtests, in both the
> pass and fail case, the test suite will automatically make the global
> state the "worst" state of the subtests.
> 
> Dylan

___
Piglit mailing list
Piglit@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/piglit