Though this doesn't follow the spec, we need a bigger toleration
to pass tests on GLES2.0 platforms like Mali400 that use fp16 in
their fragment shader.

Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>

Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>

Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>

Drivers that lower logic ops might need to emit different code when
the render target is multisampled. This is the case for v3d for example.
This patch makes it so that we run the tests for both single-sampled
and multi-sampled targets when the implementation supports
ARB_texture_storage_multisample.

Tested with i965 and v3d.

v2:
 - Indentation fixes (Alejandro)

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

Krzysztof Raszkowski authored Aug 13, 2019 and Kenneth Graunke committed Aug 13, 2019

This test compare pixels color that lie outside the window
connected to the current GL context which regards to spec
is undefined.

Reviewed-by: Alok Hota <alok.hota@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

Signed-off-by: Mark Janes <mark.a.janes@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

Also add a GL_EXT_gpu_shader4 test in case someone enables that
extension on older Intel GPUs that can't do OpenGL 3.0.

See also: mesa/mesa!1597



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

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

The EXT_dsa spec says:
        Should the selector-free versions of various OpenGL 3.0 and
        EXT_framebuffer_object framebuffer object commands not be allowed
        in display lists [...]?

        RESOLVED:  Yes

So stop testing display list support for these.

Fixes: deec4392 ("ext_direct_state_access: add test for Framebuffer functions")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

The shader has size1x32, but the images are 4-component. In such a case,
the results are undefined. Make it a single component.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

From the spec:

    The format of the image unit must be in the "1x32" equivalence class
    in Table X.2 in Section 3.9.X of the OpenGL specification, otherwise
    the atomic operation is invalid.

So we skip these.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

This was useful for some manual testing, may as well have it ready in
the future.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

This should be built only on gles3.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reported-by: Krzysztof Raszkowski <krzysztof.raszkowski@intel.com>
Fixes: ef761094 ("tests: Reproducer for 110796")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

The expected projection wasn't being applied.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

The ARB_shader_image_load_store test suite is quite complete so these tests are
limited to the main differences of behavior between the 2 spec:
  - "size" layout qualifiers (instead of "format")
  - BindImageTextureEXT specific errors
  - atomicIncWrap / atomicDecWrap glsl functions

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

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

James Xiong authored Jul 25, 2019 and Kenneth Graunke committed Jul 26, 2019

Signed-off-by: Xiong, James <james.xiong@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

I've found on older debian platforms that we get EACCES instead of
ENOENT when trying to call wglinfo, which killed the run with an
exception.

Acked-by: Eric Engestrom <eric@engestrom.ch>

This method is not available on any OpenGL ES, so for compatibility it
would be better to avoid it.

Initially I thought about adding another wrapper at
shader_runner_gles_workarounds, but that seemed an overkill for just
one usage of such method. Additionally NamedBufferSubdata requires
OpenGL 4.5 or ARB_direct_state_access, so replacing such call would be
useful also when testing old versions of OpenGL.

v2: fix using wrong parameter for glBindBufferBase (Eric)

Reviewed-by: Eric Anholt <eric@anholt.net>

Abel Briggs authored Jul 15, 2019 and Abel Briggs committed Jul 18, 2019

Reviewed-by: Eric Anholt <eric@anholt.net>

Matt Arsenault authored May 23, 2019 and Jan Vesely committed Jul 17, 2019

Acked-by: Jan Vesely <jan.vesely@rutgers.edu>

This tests that a restored SSO doesn't fail validation once
restored. This was happening in Mesa because we weren't storing
the state of the program parameter PROGRAM_SEPARABLE.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

Since VkRunner commit 6b1d2f51efad6, there are now alternative names
for the commands for setting push constants and UBO data. These names
are a bit less confusing so this patch converts all of the VkRunner
test scripts to use them.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

Jan Vesely authored Jun 28, 2019 and Jan Vesely committed Jul 04, 2019

Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

Adds Test class destructor, that releases memory allocated for test
pattern and manifest program.
Also frees memory allocated for test object itself.

Signed-off-by: Yevhenii Kolesnikov <yevhenii.kolesnikov@globallogic.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

Signed-off-by: Marek Olšák <marek.olsak@amd.com>

These tests preinit the ssbo contents but expect the driver to be doing
std430 packing by default, just specify std430 packing.

Edits done by:

grep -lr ' buffer ' tests/spec/intel_shader_atomic_float_minmax/execution |\
    while read f; do
        sed --in-place -e 's/layout(binding = 0) buffer/layout(binding = 0, std430) buffer/' $f
    done

Reviewed-by: Dave Airlie <airlied@redhat.com>

These tests preinit the ssbo contents but expect the driver to
be doing std430 packing by default, just specify std430 packing.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

Jan Vesely authored Jun 05, 2019 and Jan Vesely committed Jun 26, 2019

Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

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

This was the smallest test case that I could make that would reproduce
the bug, generate actual instructions (instead of constant folding to
nothing), and be small enough to run on any GPU supported by Mesa that
has GLES 2.0 (i.e., i915).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110953


Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

The "variant 1" subtest uses glBlendFunc(GL_DST_ALPHA, ...) so we must initialize
the dst texture or we might get inconsistent results.

This fixes random failures on radeonsi hardware.

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>

Free memory allocated for random image.

Signed-off-by: Yevhenii Kolesnikov <yevhenii.kolesnikov@globallogic.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

PIGLIT_SPIRV_VALIDATE enables the feature, PIGLIT_SPIRV_VAL_BINARY let
the user specify which "spirv-val" to use.  For now, validate against
opengl4.5 target environment.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

In the future we should consider decide this based on the shader_test
requirements.  For now allow us to generate SPIR-V binary in the
expected version -- useful for adding validation later.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

Memory, allocated in glXGetFBConfigs, has to be freed manually.

Signed-off-by: Yevhenii Kolesnikov <yevhenii.kolesnikov@globallogic.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

Sergii Romantsov authored May 24, 2019 and Lionel Landwerlin committed Jun 10, 2019

Test uses asprintf which allocates memory, but not freed it

Fixes: f62e74f6(ARB_texture_gather: add initial execution test)

Signed-off-by: Sergii Romantsov <sergii.romantsov@globallogic.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>

Free memory, allocated by asprintf.

Signed-off-by: Yevhenii Kolesnikov <yevhenii.kolesnikov@globallogic.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

test_copy checks if the outcome of glMapBuffer is the same that some
expected data. But we found that for drivers failing on returnign the
proper data, it is possible/likely that glMapBuffer would just return
NULL. So it would be better to check for NULL to avoid a crash, as
technically the driver is not crashing, but the test (even if the
reason glMapBuffer still needs to be investigated).

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>