Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

These example files are from:

 * https://gitlab.com/pages/sphinx.git

 68a1b1da8c74c464c47b5b77e3276c85e96c60a2

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>

This tests unrolling of a loop with a single exit point where the
exact trip count is unknown, but the max iteration count can be
guessed using the size of an array indexed via the induction
variable.

This tests unrolling of some loops with a single exit point but where the
exact trip count is unknown, only the max iteration count is known.

metadata.json contains 'force_glsl' but not 'glsl'.  I guess it must
have gotten renamed at some point.  Fixes the following traceback when
attempting to call 'piglit resume':

Traceback (most recent call last):
  File "/home/kayden/Projects/piglit/piglit", line 178, in <module>
    main()
  File "/home/kayden/Projects/piglit/piglit", line 174, in main
    sys.exit(runner(args))
  File "/home/kayden/Projects/piglit/framework/exceptions.py", line 52, in _inner
    func(*args, **kwargs)
  File "/home/kayden/Projects/piglit/framework/programs/run.py", line 439, in resume
    options.OPTIONS.force_glsl = results.options['glsl']

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

This probes a single pixel at the center of each heart drawn, ensuring
it's the correct color.  Otherwise, failing to draw both halves would
cause the test to pass.

v2: - don't put config inside #ifdef opengl
    - de-duplicate comment

This allows each test to be selected at runtime.

This has config.supports_gl_compat_version = 10, then calls
piglit_require_GL_version(20), so just set the value in the config to 20
and remove the other.

Currently if the first subtest failed then the second subtest (AoA)
would always report fail, this is bad. Instead we want to report fail,
but leave status (which is set to warn in the implementation doesn't
meet certain assumptions by the test), so that the second test can
report accurately.

I can't really test that this works, so just enumerating will have to be
enough.

This doesn't add support for running specific subtests, but it does
enumerate the ones that exist.

Which have never been run, though the array has been present since the
test was added.

This is ground work for using proper subtest handling, it's a big enough
change it makes sense to me to split it into a separate commit.

It's just cosmetic.

It's not immediately obvious whether this test is suitable for
piglit_run_selected_subtests, so just enumerate the subtests it has so
that the framework can catch crashes.

This allows the framework to know what tests will be run, and doesn't
actually make the implementation much more complicated.

This coverts the test to use the piglit_run_selected_subtests function,
which provides a nicer mechanism for running selected subtests, and
enumerates all subtests ahead of time.

Ross Burton authored Nov 30, 2018 and Eric Engestrom committed Dec 06, 2018

If built with -Werror=format-security then Piglit fails to build:

/tests/spec/arb_texture_view/rendering-layers-image.c:150:8:
error: format not a string literal and no format arguments [-Werror=format-security]
         (desc)); \
         ^~~~~~

In this case test->uniform_type is being turned into a string using snprintf()
and then passed to piglit_report_subtest_result() which takes a format string,
but GCC can't verify the format.

As _subtest_report() takes a format string, we can just remove the snprintf()
and let it construct the label.

Also as X is used once and doesn't make the code clearer, just inline it.

Signed-off-by: Ross Burton <ross.burton@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

It was using the same color for the base color, drawing color and
expected color.

As we are here, we also remove the debug names, as the test should
work without names.

Reviewed-by: Józef Kucia <joseph.kucia@gmail.com>

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

Instead of using the fact this one driver currently doesn't support this
one format, let's just use a format that will never be supported by
anyone.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

Ross Burton authored Nov 30, 2018 and Eric Engestrom committed Dec 04, 2018

On ARM, musl does not define FE_* when the architecture does not have VFP (which
is the right interpretation).

As these tests depend on calling fesetround(), skip the test if FE_UPWARD isn't
available.

Signed-off-by: Ross Burton <ross.burton@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

This option only applies to the sample_rgb tests.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

Travis-CI has or will shortly make in early December 2018 a number of beneficial
changes to their Linux continuous integration testing infrastructure [0][1].
An accompanying benefit of this change is that Ubuntu Xenial (16.04 LTS) is now
supported.

The benefits for piglit are primarily:
* Testing against a more modern, supported Ubuntu Xenial (16.04 LTS). [2]
* Removal of a corner-case for Python 3.7, making testing more consistent.
* Modest speed improvements from Travis-CI's move to Linux infrastructure combined
  into one (virtualized), from two previously (virtualized and container-based).

[0] https://blog.travis-ci.com/2018-10-04-combining-linux-infrastructures
[1] https://blog.travis-ci.com/2018-11-19-required-linux-infrastructure-migration
[2] https://docs.travis-ci.com/user/reference/xenial/



Signed-off-by: Rhys Kidd <rhyskidd@gmail.com>
Reviewed-by: Andres Gomez <agomez@igalia.com>

Jan Vesely authored Nov 21, 2018 and Jan Vesely committed Dec 02, 2018

This prevents race condition when a failed kernel launch would no longer
be in the command queue by the time we call clFinish.

Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
Tested-by: Maciej S. Szmigiero <mail@maciej.szmigiero.name>

Jan Vesely authored Nov 21, 2018 and Jan Vesely committed Dec 02, 2018

Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
Tested-by: Maciej S. Szmigiero <mail@maciej.szmigiero.name>

Andrii Simiklit authored Oct 31, 2018 and Tapani Pälli committed Nov 29, 2018

I guess we should not expect a valid location for:
"patch out vec4 tcs_patch;"
because this output variable is declareted
without "layout (location=X)" and according to spec:
   "Not all active variables are assigned valid locations; the
    following variables will have an effective location of -1:

      * uniforms declared as atomic counters;

      * members of a uniform block;

      * built-in inputs, outputs, and uniforms (starting with "gl_"); and

      * inputs or outputs not declared with a "location" layout qualifier,
        except for vertex shader inputs and fragment shader outputs."

Also I fixed some conflicting comments.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108612


Signed-off-by: Andrii Simiklit <andrii.simiklit@globallogic.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>