[Piglit] [PATCH v2 1/6] texturing: make sure s3tc tests require GL 1.1

Emil Velikov emil.l.velikov at gmail.com
Fri Oct 26 16:36:39 UTC 2018


On Fri, 26 Oct 2018 at 16:39, Erik Faye-Lund
<erik.faye-lund at collabora.com> wrote:
>
> On Fri, 2018-10-26 at 16:21 +0100, Emil Velikov wrote:
> > On Fri, 26 Oct 2018 at 14:03, Erik Faye-Lund
> > <erik.faye-lund at collabora.com> wrote:
> > >
> > > The EXT_texture_compression_s3tc spec lists that it requires
> > > OpenGL 1.1, not 1.0 like some of these list.
> > >
> > > In reality, this probably doesn't make a huge difference, as
> > > OpenGL 1.0 hardware/drivers are pretty much extinct, but let's
> > > just get it right to avoid confusion.
> > >
> > > Signed-off-by: Erik Faye-Lund <erik.faye-lund at collabora.com>
> > > ---
> > >  tests/texturing/s3tc-errors.c      | 2 +-
> > >  tests/texturing/s3tc-teximage.c    | 2 +-
> > >  tests/texturing/s3tc-texsubimage.c | 2 +-
> > >  3 files changed, 3 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/tests/texturing/s3tc-errors.c b/tests/texturing/s3tc-
> > > errors.c
> > > index 81fee3e21..ccedf8a22 100644
> > > --- a/tests/texturing/s3tc-errors.c
> > > +++ b/tests/texturing/s3tc-errors.c
> > > @@ -36,7 +36,7 @@
> > >
> > >  PIGLIT_GL_TEST_CONFIG_BEGIN
> > >
> > > -    config.supports_gl_compat_version = 10;
> > > +    config.supports_gl_compat_version = 11;
> > >
> >
> > Ages ago Ilia reported on IRC that some tests (was it 10 or 11, don't
> > recall) failing on the classic swrast and/or nouveau_vieux.
> > That waffle issue should be fixed, although if he can give this patch
> > a try (with platform glx) that'll be appreciated ...
> >
> > Gut feeling suggests that the fix may not have been enough.
>
> I don't think I understand the issue... Some of these tests use 10 and
> some use 11 here, and 11 is the correct value (according to the spec).
> I'm just making it consistent...
>
> This is just a filter under what contexts these tests will run, no? I
> haven't seen an OpenGL 1.0-only implementation, well, ever I think?
> Isn't this just theoretical anyway? I don't understand how this could
> have any practical effect...
>
It's been ~3 years so, my memory and amount of details is flaky.

A while ago Ilia and Brian Paul reported that some tests annotated as
10 (or was it 11)  were failing on nouveau_vieux and classic swrast
respectively.
That was because piglit was passing the explicit version to waffle,
which in turn was using the GLX_ARB_create_context since it's the only
way to request a specific version.

That should be addressed with the following waffle commit [1],
although I'm concerned about the 10 check in the patch.
Sadly I did not document why in the commit log (nor was it asked
during review), perhaps it has something to do with the massive
comment in waffle's glx_context.c [2].

That said, if your patch causes problems I could double-check and fix waffle.
Since I've annoyed Ilia enough times, I wanted to spare another
instance - hence the suggestion.

HTH
Emil

[1] https://github.com/waffle-gl/waffle/commit/e7f0314c7c87e031148d7dee0e6f28d490a97a2a
[2] https://github.com/waffle-gl/waffle/blob/e7f0314c7c87e031148d7dee0e6f28d490a97a2a/src/waffle/glx/glx_context.c#L86


More information about the Piglit mailing list