John Richardson [Wed, 12 Jul 2017 20:01:21 +0000 (20:01 +0000)]
Add sRGB write decode negative case
am:
6644e7bfef -s ours
Change-Id: I3ca1620efe31a4a7fce3267f6c788d1e5f5ea064
Michael Hadley [Wed, 12 Jul 2017 20:01:18 +0000 (20:01 +0000)]
Add sRGB write decode tests
am:
50d8f31f52 -s ours
Change-Id: Ia9a7bc1a1a23792a3a6d7a22588e74559a5f2b98
John Richardson [Thu, 30 Mar 2017 14:11:43 +0000 (15:11 +0100)]
Add sRGB write decode negative case
Test to check drivers that don't expose EXT_sRGB_write control return
GL_INVALID_ENUM for related gl calls.
New Tests:
- dEQP-GLES31.functional.fbo.srgb_write_control.framebuffer_srgb_unsupported_enum
Change-Id: I605f70894ead6e5dd317a340d061dcec9a6180c8
Michael Hadley [Wed, 8 Mar 2017 13:28:09 +0000 (13:28 +0000)]
Add sRGB write decode tests
New Tests:
- dEQP-GLES31.functional.fbo.srgb_write_control.*
Change-Id: Ieaaee4c7dc49cc530d8fc42a060e40c706d86e5c
Juan A. Suarez Romero [Fri, 30 Jun 2017 22:50:09 +0000 (22:50 +0000)]
x11_egl: use quotes to include glesv1_cm directories
am:
f5e7ab6b79
Change-Id: I839b0667bf29e5695205ff00c3b591dba5df8450
Juan A. Suarez Romero [Fri, 30 Jun 2017 12:22:35 +0000 (14:22 +0200)]
x11_egl: use quotes to include glesv1_cm directories
Modifying the CMAKE_EXE_LINKER_FLAGS requires to set the new options
between quotes.
Otherwise, cmake will join the elements with semicolons, generating
incorrect linker flags.
Test: cmake . -DDEQP_TARGET=x11_egl && make
Change-Id: I4d95dc37b6d91ab787b9c2e5568cbe2f08dd472b
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Pyry Haulos [Wed, 14 Jun 2017 01:53:10 +0000 (01:53 +0000)]
Merge "Remove EGL gles1.rgb565_window from mustpass"
am:
7bff163e7f
Change-Id: I642a906fce854d128ddbf13ca997df01fbece237
Treehugger Robot [Wed, 14 Jun 2017 01:43:04 +0000 (01:43 +0000)]
Merge "Remove EGL gles1.rgb565_window from mustpass"
Pyry Haulos [Wed, 14 Jun 2017 01:22:40 +0000 (01:22 +0000)]
Touch watchdog before freeing objs in max_concurrent
am:
815781b777
Change-Id: Ibe8bbcbbfddca74524f9422b60e14bd7db124f47
Pyry Haulos [Tue, 13 Jun 2017 21:08:33 +0000 (14:08 -0700)]
Remove EGL gles1.rgb565_window from mustpass
Bug:
30041221
Change-Id: Idc8e3d1d78c568fa8d4209c61eaec52ca5696356
(cherry picked from commit
22eb25c58f0a1aa1de401a38778471c8b6700ffd)
Pyry Haulos [Wed, 7 Jun 2017 22:58:05 +0000 (15:58 -0700)]
Touch watchdog before freeing objs in max_concurrent
Affects:
dEQP-VK.api.object_management.max_concurrent.*
Components: Vulkan
Google bug:
37515046
Change-Id: I57eb68764a55b5dceeb3e2676b8b1b298c0ff67e
(cherry picked from commit
252fb76fec02af6e2281da3bc9eea4e0d7972f12)
james.glanville [Thu, 18 May 2017 10:24:06 +0000 (11:24 +0100)]
Apply sRGB conversion to reference in sRGB framebuffer fetch test
Framebuffer fetch returns linear values when the framebuffer format
is sRGB and GL_FRAMEBUFFER_SRGB is enabled (default). This conversion
was not applied when generating reference textures for framebuffer
fetch tests. glReadPixels() does not apply any conversion, so the
final reference value must also be converted back to sRGB.
Affects: dEQP-GLES31.functional.shaders.framebuffer_fetch.
framebuffer_format.srgb8_alpha8
Components: AOSP
Pyry Haulos [Mon, 12 Jun 2017 17:06:12 +0000 (17:06 +0000)]
Merge "Re-remove tessellation explicit_output_array_size_1 from must-pass list" into deqp-dev
Alexander Galazin [Fri, 9 Jun 2017 23:36:06 +0000 (23:36 +0000)]
Skip OOB SSBO fragment tests for ES3.1 GPUs
am:
f5fc3c8409
Change-Id: I611887e4299f25c04add54c3a6b5554c1841f20b
Alexander Galazin [Fri, 9 Jun 2017 14:26:32 +0000 (16:26 +0200)]
Skip OOB SSBO fragment tests for ES3.1 GPUs
GPUs supporting GLES 3.1 may expose 0 SSBOs in fragment shaders.
This change extends fix made in
cf65c56f.
Google bug:
36891454
Components: AOSP
Affects:
dEQP-EGL.functional.robustness.reset_context.shaders.out_of_bounds*.shader_storage_block.fragment
Change-Id: I1d821198d3dba3d3ad1e6b25758d0d434bf08a30
Treehugger Robot [Fri, 9 Jun 2017 21:29:04 +0000 (21:29 +0000)]
Merge "Add support for metamorphic testing" into deqp-dev
Mark Adams [Fri, 9 Jun 2017 17:52:18 +0000 (13:52 -0400)]
Re-remove tessellation explicit_output_array_size_1 from must-pass list
Affected test:
dEQP-GLES31.functional.tessellation.user_defined_io.negative.es32.per_vertex_incorrect_control_explicit_output_array_size_1
Alexander Galazin [Tue, 6 Jun 2017 21:07:54 +0000 (21:07 +0000)]
Ignore A channel in EGL RGB5_A1 clear color tests
am:
9da1b45a8e
Change-Id: I72c8355a27472f559467127051f3108eba9ee581
Alexander Galazin [Tue, 6 Jun 2017 11:22:58 +0000 (13:22 +0200)]
Ignore A channel in EGL RGB5_A1 clear color tests
Dithering can siginifcantly impact alpha channel and the framework
can't fully account it.
Components: AOSP
Affects:
dEQP-EGL.functional.color_clears.*.rgba5551*
Bug:
62383988
Change-Id: Ic8fa4da1a840da56f74af3e076d0728d9abbdadc
Chih-Hung Hsieh [Tue, 6 Jun 2017 02:29:36 +0000 (02:29 +0000)]
Add OWNERS in external/deqp
am:
ea5d0ac339
Change-Id: I22107e5022e4536ba8b5ecc0bb0d7b8d225ef428
Chih-Hung Hsieh [Mon, 5 Jun 2017 17:48:12 +0000 (10:48 -0700)]
Add OWNERS in external/deqp
* Owners are selected from top CL approvals or owners.
They will be suggested to review/approve future CLs.
* OWNERS files are recognized by the new find-owners plugin, see .md files in
https://gerrit.googlesource.com/plugins/find-owners/+/master/src/main/resources/Documentation/
Test: build/make/tools/checkowners.py -c -v OWNERS
Change-Id: I0d256181d2ca0c0ec55cf5479ff34aa68f3190a9
Pyry Haulos [Mon, 5 Jun 2017 17:39:43 +0000 (10:39 -0700)]
Merge aosp/upstream-vulkan-cts-1.0-dev into aosp/deqp-dev
Change-Id: Id9c893e5db0817a29f768cca931adc260d350497
Pyry Haulos [Mon, 5 Jun 2017 17:37:17 +0000 (10:37 -0700)]
Merge vk-gl-cts/aosp-deqp-dev into vk-gl-cts/master
Change-Id: Ib189ff26f3bb7e049f5e0c4345e523cc39e91d4f
Andres Gomez [Sat, 3 Jun 2017 00:30:32 +0000 (02:30 +0200)]
Build fix for x11_egl target builds
The combination of recent refactor on release branch and merge to
master resulted in a compilation error for the XInitThreads symbol
when using the x11_egl built target.
Components: Framework
VK-GL-CTS Issue: 467
Change-Id: Iba0803e8b73462bd66ab7803a6c90bfd40d9164c
Alexander Galazin [Mon, 15 May 2017 09:50:09 +0000 (11:50 +0200)]
Add ES 3.2.4.x mustpass
Components: OpenGL
Change-Id: Ifd8270acae469d80d9243a58f1e17b57bbecf081
Alexander Galazin [Fri, 7 Apr 2017 08:01:45 +0000 (10:01 +0200)]
Remove GTF tests from GLES mustpass
These tests are no longer needed as the same coverage is provided
in dEQP, KHR modules.
Components: OpenGL
Affects: GLES master mustpass
VK-GL-CTS issue: 50
Change-Id: Ib52227b76e40888c979eeadc58d0e7ea6500eebc
Pyry Haulos [Tue, 30 May 2017 01:56:36 +0000 (18:56 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: I0a9fdded2cbfd25f1e2a1c73da97b2d8b8431962
Pyry Haulos [Tue, 30 May 2017 01:56:32 +0000 (18:56 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.1 into vk-gl-cts/vulkan-cts-1.0.2
Change-Id: Ib63e665cb64989f8c5f0c53c9d229e5f17172d62
Boris Zanin [Tue, 16 May 2017 12:32:06 +0000 (14:32 +0200)]
Allow pop-free clipping for lines
Following test checks has been relaxed:
* dEQP-VK.rasterization.primitives.lines_wide
* dEQP-VK.rasterization.primitives.line_strip_wide
* dEQP-VK.rasterization.primitives_multisample_2_bit.lines_wide
* dEQP-VK.rasterization.primitives_multisample_4_bit.lines_wide
* dEQP-VK.rasterization.primitives_multisample_8_bit.lines_wide
Components: Vulkan
VK-GL-CTS issue: 421
Change-Id: I21c04015d36ae55c20ef1c1dda8e2d1fee471221
Yanjun Zhang [Thu, 11 May 2017 22:51:37 +0000 (15:51 -0700)]
Check 'drawIndirectFirstInstance' feature to run indirect draw tests
Component: Vulkan
VK-GL-CTS issue: 413
Change-Id: I13a86721ec76ac50af55e0730fd50e3d7ce159d8
Jason Ekstrand [Fri, 19 May 2017 04:36:15 +0000 (21:36 -0700)]
dEQP-VK: Fix an incorrect layout transition
The tests upload data to the resolve destination using CopyImageToBuffer
prior to doing a resolve to a chunk out of the middle of the image. This
lets the test validate the resolve to a subarea works correctly.
However, they were also transitioning the destination from UNDEFINED to
TRANSFER_DST_OPTIMAL between uploading the data and doing the resolve.
This transition destroys the contents of the image making verifying that
contents incorrect. The correct transition is from TRANSFER_DST_OPTIMAL
to TRANSFER_DST_OPTIMAL.
This affects the following tests:
- dEQP-VK.api.copy_and_blit.resolve_image.partial*
- dEQP-VK.api.copy_and_blit.resolve_image.with_regions*
Change-Id: I9336195d7977d7dd2ffb057525134846b10e8010
Oivind Boge [Tue, 16 May 2017 14:04:01 +0000 (16:04 +0200)]
Fixed invalid usage of VkBufferImageCopy struct
The offset member of VkBufferImageCopy was violating the
valid usage. The offset is now calculated using lowest
common denominator between the pixelSize and 4.
Affects:
dEQP-VK.api.image_clearing.*
Components: Vulkan
VK-GL-CTS issue: 422
Change-Id: I1ef07e06f790a9ed403941b1d94c7840485cec3e
Slawomir Cygan [Mon, 29 May 2017 20:01:49 +0000 (22:01 +0200)]
Build fix for X11-less Linux builds
The combination of recent refactor on release branch and
merge to master resulted in XinitThreads being called from
code that may be compiled even if X11 API is not available.
Components: Framework
VK-GL-CTS Issue: 450
Change-Id: Ia7d1d23eabd1b222d5039868c1fb762e1d42f01f
Alexander Galazin [Mon, 29 May 2017 08:32:47 +0000 (10:32 +0200)]
Update KC CTS ref
Components: OpenGL
VK-GL-CTS issue: 440
Affects:
GL4*-CTS.gtf42.GL3Tests.texture_storage.texture_storage_texture_levels
Change-Id: Id0b338bb74d3b1923744556527b45f3e8db8db38
Jeannot Breton [Thu, 25 May 2017 19:10:31 +0000 (14:10 -0500)]
Use 1.0 for shadow texture expected result
Change the texture clamp test to use 1.0 as the expected value
for shadow texture.
It's now required to pass <format> to generateExpectedResult
Note that the path using compute shader is doing an imageLoad on
depth texture so is not using a depth compare
Affects:
KHR-GL45.sparse_texture_clamp_tests.*
Components: OpenGL
VK-GL-CTS issue: 444
Change-Id: I94c79abe1b9b932471d1d8abdb8819ee3d7cc947
Jeannot Breton [Thu, 25 May 2017 00:40:47 +0000 (19:40 -0500)]
Skip uncommitted depth/stencil test validation
The test in UncommittedDepthStencil expects the stencil buffer to be
unmodified because the shader is doing an uncommitted sparse fetch
This is incorrect, the depth/stencil should be unmodified only if
the depth/stencil buffer are uncommitted and it's not the case,
they are fully-committed non-sparse depth/stencil buffer
This change skip uncommitted depth/stencil test validation
Affects:
KHR-GL45.sparse_texture2_tests.*
Components: OpenGL
VK-GL-CTS issue: 439
Change-Id: I909937081f7058d0fbc57469626ab3d34d43486a
Jeannot Breton [Thu, 25 May 2017 20:48:20 +0000 (15:48 -0500)]
Fix sparse cubemap array tests issues
Add cubeMapArrayRefZDef, a new shader token used as <compare>
parameters on some shadow cube map array texture instructions
Add <CUBE_REFZ_DEF> to some FunctionToken argument list definitions.
It will be replaced by the value of cubeMapArrayRefZDef.
Copy the depth reference value to s.cubeMapArrayRefZDef when we
generate the shader tokens
Change the way cube map array texture coordinate are computed in
the shader to support array with more than 6 layers
Specify that textureGradClampARB on shadow cubemap array is not
supported
Generate cube map array texture coordinate attributes as a vec4
with different coordinate for each layer
Affects:
KHR-GL45.sparse_texture2_tests.*
KHR-GL45.sparse_texture_clamp_tests.*
Components: OpenGL
VK-GL-CTS issue: 445
Change-Id: Ia9fb2240396ab4a34dd697d0b96985da35dfe802
Jeannot Breton [Wed, 24 May 2017 21:21:13 +0000 (16:21 -0500)]
Skip height validation on TEXTURE_1D_ARRAY
In verifyTexStorageInvalidValueErrors the test validate that a
GL_INVALID_VALUE error is returned by the GL when the height
parameters of the texStorage call is invalid.
The height is ignored by the texStorage call for 1d array, so the
error is never generated.
We should skip this validation for 1d array
Affects:
KHR-GL45.sparse_texture_tests.*
Components: OpenGL
VK-GL-CTS issue: 437
Change-Id: Ibe93db467deb560f5394b4b658dfa3f32a953709
Jeannot Breton [Thu, 25 May 2017 00:10:18 +0000 (19:10 -0500)]
Fix texPageCommitment alignment test
In verifyInvalidValueErrors, the test wants to validate that
texPageCommitment returns a GL_INVALID_VALUE error when the page
depth offset is not a multiple of pageSizeZ.
However if pageSizeZ equals 1, any z offset will be a multiple of
pageSizeZ and the test should not return an error.
Logic is added to make sure there's really an alignment error
Affects:
KHR-GL45.sparse_texture_tests.*
Components: OpenGL
VK-GL-CTS issue: 438
Change-Id: If7a00dee071c1dafabbda608f7cdbb14d1eb56b3
Jeannot Breton [Thu, 25 May 2017 18:08:39 +0000 (13:08 -0500)]
textureClamp texcoord with shadow 1d uses vec3
This change modifies the current shader generation for textureClamp
subtests.
s.coordType and s.sizeDef will now use vec3 for sampler1DArrayShadow
and sampler1DShadow
Affects:
KHR-GL45.sparse_texture_clamp_tests.*
Components: OpenGL
VK-GL-CTS issue: 443
Change-Id: I025061981877a62b6eed6bcc6c783f6fdc68c7d3
Jeannot Breton [Thu, 25 May 2017 14:34:52 +0000 (09:34 -0500)]
Initialize FunctionToken for every iteration
The change initializes the variable inside the loop so it gets
re-initialized at every iteration.
This help avoid issues where changes accumulates at every iteration
Affects:
KHR-GL45.sparse_texture_clamp_tests.*
Components: OpenGL
VK-GL-CTS issue: 442
Change-Id: I10eac411a03b3eff5139026ac004dc2df0c025f0
Jeannot Breton [Thu, 25 May 2017 14:02:56 +0000 (09:02 -0500)]
Set uncommited region access expected data to 0
In the GL_TEXTURE_CUBE_MAP path, the expected data is initialized
to 0xff. This seems incorrect.
The test will only fecth the uncommitted region of the texture which
should return 0.
This change fixes this issue and initializes the data to 0
Affects:
KHR-GL45.sparse_texture2_tests.*
KHR-GL45.sparse_texture_clamp_tests.*
Components: OpenGL
VK-GL-CTS issue: 441
Change-Id: Ic00cdfde24248d1640fe458c8f0a547d7d174e3d
Jeannot Breton [Wed, 24 May 2017 20:27:14 +0000 (15:27 -0500)]
Return an error if the pagesize index is invalid
In verifyTexStorageVirtualPageSizeIndexError, the test call
texStorage with a VIRTUAL_PAGE_SIZE_INDEX texture parameter set
to NUM_VIRTUAL_PAGE_SIZES + 1.
However the verifyError after the texStorage expect GL_NO_ERROR.
It should be GL_INVALID_OPERATION
Affects:
KHR-GL45.sparse_texture_tests.*
Components: OpenGL
VK-GL-CTS issue: 436
Change-Id: I43d02dcbee4bbe2c81feb4a17e9fde7386375656
Jeannot Breton [Mon, 22 May 2017 20:26:07 +0000 (15:26 -0500)]
Fix pack alignment issue on odd window width
The test allocates a buffer of width*height elements and expects the
readPixels to only write to this memory.
This is not correct if the window width is not a multiple of 4
Adding a gl.pixelStorei(GL_PACK_ALIGNMENT, 1) before the readPixels
fixes the issue
Affects:
KHR-GL45.indirect_parameters_tests.MultiDrawArraysIndirectCount
Components: OpenGL
VK-GL-CTS issue: 433
Change-Id: I7f8b92af7b5a27a6770e7cde23e51df30508a3c0
James Helferty [Wed, 24 May 2017 15:54:42 +0000 (11:54 -0400)]
Remove OUT_OF_MEMORY coverage of buffer api
The OpenGL spec leaves much of the behaviour around OUT_OF_MEMORY
errors undefined, which makes it difficult to test for them. Regardless
of whether the error is generated or not, the GL context is by
definition left in an undefined state, which could invalidate any
subsequent tests' results.
This change removes OUT_OF_MEMORY test coverage.
VK-GL-CTS issue 435
Component: OpenGL
Affects:
KHR-GL45.direct_state_access.buffers_errors
Change-Id: I6436d4d72a0a7b7cd08336895e400d85bea8da15
Filip Demczewski [Wed, 4 Jan 2017 13:30:28 +0000 (14:30 +0100)]
Added CTS_ARB_shader_ballot specification proposal
affects:
GL45-CTS.shader_ballot_tests.*
components: OpenGL
VK-GL-CTS issue: 143
Change-Id: Ida320ac444665f09edd86d6a69aae9af7515a237
Alexander Galazin [Wed, 24 May 2017 15:15:11 +0000 (17:15 +0200)]
Merge vk-gl-cts/opengl-es-cts-3.2.3 into vk-gl-cts/master
Change-Id: I95690b5479602db484d2f89bdc711a683afe96b4
Graeme Leese [Thu, 11 May 2017 15:17:02 +0000 (16:17 +0100)]
Fix uniform location use in robust_buffer_access
Common code will try to set the texture at uniform location 0, so change
the ES shader to use location 0.
Affects: KHR-GLES32.robust.robust_buffer_access_behavior.texel_fetch
Components: OpenGL
VK-GL-CTS Issue: 416
Change-Id: I6e4e291b33072b16e1ab73eb8592afd00687adc7
Alexander Galazin [Wed, 17 May 2017 09:39:12 +0000 (11:39 +0200)]
Remove dEQP-GLES31.*.vertex_array.*_invalid_map
The tests expects GL_INVALID_OPERATION to be generated if
a non-zero buffer object name is bound to an enabled array
or the element array and the buffer object's data store
is currently mapped.
Such behavior is not required by the specification.
Components: OpenGL
Affects:
dEQP-GLES31.functional.debug.negative_coverage.get_error.vertex_array.draw_elements_base_vertex_invalid_map
dEQP-GLES31.functional.debug.negative_coverage.get_error.vertex_array.draw_elements_instanced_base_vertex_invalid_map
dEQP-GLES31.functional.debug.negative_coverage.get_error.vertex_array.draw_range_elements_base_vertex_invalid_map
VK-GL-CTS issue: 429
Change-Id: Ib7e9c95dde9b86f227b3d014210a2fe732315f31
Pyry Haulos [Tue, 16 May 2017 14:25:35 +0000 (07:25 -0700)]
Remove unused function from es31fSRGBDecodeTests.cpp
Unused function causes compiler warnings to appear with recent GCC
versions.
Change-Id: I941c579994acbc9146df86acf5b9da9ee85341f5
(cherry picked from commit
1ee59ff98641dc0be7511e03e37b5843ff7eed82)
Alexander Galazin [Thu, 11 May 2017 12:27:20 +0000 (14:27 +0200)]
Fix texture/sampler mapping in sRGB tests
sRGB decode tests assume that an API side texture with index i
is connected with a shader sampler variable uTexture with index i.
This relation was however incorrectly established, e.g.
the tests were mapping texture0 to a sampler with location 0, while
such sampler was not necessarily uTexture0.
This change fixes the mapping and makes the relation more obvious.
Components: AOSP
Affects:
dEQP-GLES31.functional.srgb_texture_decode.skip_decode.*.toggled
dEQP-GLES31.functional.srgb_texture_decode.skip_decode.*.conversion_gpu
Google bug:
38257749
VK-GL-CTS issue: 428
Change-Id: I16508f6033b31583b1c5973b70a610dd59e78e8f
(cherry picked from commit
c8efb45eadba4a910ab5770b2207582cf1bb7ca0)
Alexander Galazin [Wed, 10 May 2017 08:36:53 +0000 (10:36 +0200)]
Adjust out of surface verification for wide lines
Wide lines rasterization tests may draw lines out of
the allocated surfaces. Such situation causes verification
failures.
This change adjust the verification algorithm to ignore
missing pixels for out of surface rendering.
Components: Framework
Affects:
dEQP-GLES2.functional.rasterization.primitives.lines_wide
dEQP-GLES2.functional.rasterization.primitives.line_strip_wide
dEQP-GLES2.functional.rasterization.primitives.line_loop_wide
VK-GL-CTS issue: 427
Change-Id: Ia70f33fb6daad0d3774ae9edbd438bd14296f1d3
(cherry picked from commit
c065847203ba5c1ec0bba76fe06e0b7a17cbc6af)
Alexander Galazin [Mon, 8 May 2017 12:59:38 +0000 (14:59 +0200)]
Ensure vertex shader presence if a geometry shader exists
Certain tests verify that INVALID_OPERATION is generated
if a geometry shader is active and mode is incompatible with
the input primitive type of the geometry shader in the currently
installed program object. However these tests have no vertex shader and
this is the real reason why INVALID_OPERATION is generated.
This change ensures that the vertex shader is present.
Components: AOSP
VK-GL-CTS issue: 426
Affects:
dEQP-GLES31.functional.debug.negative_coverage.get_error.vertex_array.draw_elements_base_vertex_primitive_mode_mismatch
dEQP-GLES31.functional.debug.negative_coverage.get_error.vertex_array.draw_elements_instanced_base_vertex_primitive_mode_mismatch
dEQP-GLES31.functional.debug.negative_coverage.get_error.vertex_array.draw_range_elements_base_vertex_primitive_mode_mismatch
Change-Id: I36bcee8892e4f641feadb0b444a53ed35273b727
(cherry picked from commit
f5f087e9bd571b5aee3b810957af7051cd82d2c0)
Alexander Galazin [Tue, 9 May 2017 08:44:12 +0000 (10:44 +0200)]
Fix DrawRangeElementsBaseVertex tests with negative basevertex
DrawRangeElementsBaseVertex tests with negative basevertex call
glDrawRangeElementsBaseVertex with start = 0 and basevertex = -1.
According to the spec: If no element array buffer is bound, the vertex ID of
the ith element transferred is indices[i] + basevertex...
If the vertex ID is larger than the maximum value representable by type,
it should behave as if the calculation were upconverted to 32-bit unsigned
integers (with wrapping on overflow conditions).
Behavior of DrawElementsOneInstance is undefined if the vertex ID is negative
for any element, and should be handled as described in section 6.4.
This change ensures that indexMin + baseVertex will be zero or positive value.
Components: AOSP
VK-GL-CTS issue: 425
Affects:
dEQP-GLES31.functional.draw_base_vertex.draw_range_elements_base_vertex.base_vertex.index_neg_byte
dEQP-GLES31.functional.draw_base_vertex.draw_range_elements_base_vertex.base_vertex.index_neg_short
dEQP-GLES31.functional.draw_base_vertex.draw_range_elements_base_vertex.base_vertex.index_neg_int
Change-Id: I07a9e6bc60d4340784d400b5087bb8fff01f0e57
(cherry picked from commit
e490e2f2158aa9d3aec7d8e385ca3995fefc9ffd)
Alexander Galazin [Mon, 8 May 2017 12:31:15 +0000 (14:31 +0200)]
Fix negative CompressedTexImage2D tests
Certain tests expect GL_INVALID_VAULE for
glCompressedTexImage2D(GL_TEXTURE_2D, -1, 0x8b90, 0, 0, 0, 0, 0).
0x8b90 is a palette format PALETTE4_RGB8 and
the OES_compressed_paletted_texture extension allows
negative values for the level parameter:
Accepted by the <level> parameter of CompressedTexImage2D
Zero and negative values. |level| + 1 determines the number of
mip levels defined for the paletted texture.
This change fixes the tests in presence of
the OES_compressed_paletted_texture extension.
Components: AOSP
Google bug:
38109626
VK-GL-CTS issue: 424
Affects:
dEQP-GLES2.functional.negative_api.texture.compressedteximage2d_neg_level_tex2d
dEQP-GLES2.functional.negative_api.texture.compressedteximage2d_neg_level_cube
Change-Id: Ia37e3fecaf3547abfb209803378114043c63a243
(cherry picked from commit
9de005f0355e389f1782543f34e0fb3c65a0a700)
Graeme Leese [Tue, 9 May 2017 14:21:53 +0000 (15:21 +0100)]
Correct the test conditions for sRGB_write_control
The extension is not part of ES3.2, so the not-supported version of the
tests should be run on ES3.2 implementations without the
GL_EXT_sRGB_write_control extension.
VK-GL-CTS Issue: 389
Components: AOSP
Affects:
dEQP-GLES31.functional.fbo.srgb_write_control.framebuffer_srgb_enabled
dEQP-GLES31.functional.fbo.srgb_write_control.framebuffer_srgb_unsupported_enum
Change-Id: I42a531aa7d1ad38f7a85cde1a5da659f780ec5d5
(cherry picked from commit
2d987e3b1e3c564ddc197124b1c7689aed060b8e)
Alexander Galazin [Wed, 24 May 2017 06:59:51 +0000 (08:59 +0200)]
Update KC CTS ref
Components: OpenGL
VK-GL-CTS issue: 430
Affects:
GTF-GL45.gtf30.GL3Tests.framebuffer_srgb.framebuffer_srgb_draw
Change-Id: I01b5d5bb9ae37b24a44427c1a7cd735f98e60675
David Sodman [Thu, 18 May 2017 16:30:17 +0000 (09:30 -0700)]
Remove dEQP-GLES31.*vertex_array.*_invalid_map tests
Tests made assumptions about GL error returned from glDrawElements*
functions when passed a mapped buffer object, but, that error is not
supported by the specification.
Affects: dEQP-GLES31.*vertex_array.*
Components: AOSP
Bug:
37942569
Change-Id: Ic066dfbe717ff1e7741bec8ee870b09b99362dc7
Pyry Haulos [Mon, 22 May 2017 19:34:20 +0000 (12:34 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: Ie841d0a435d1d587450b7c2d57307b37b3295bbb
Slawomir Cygan [Wed, 11 Jan 2017 16:17:11 +0000 (17:17 +0100)]
Refactor X11 and wayland platforms to run alongside each other
This adds a capability for running Vulkan WSI ane EGL tests on both
X11 and wayland compositors from one binary built in 'default' target.
The change replaces X11 & wayland tcu::Platform-s with one common
'linux' platform, that is parametrized by DEQP_SUPPORT_
(X11|GLX|XCB|WAYLAND) at compile time.
This new platform supports creating WSI displays and EGL contexts
for each of compositors enabled at build-time. To choose EGL compositor
use --deqp-egl-display-type=x11 or Wayland.
No functional change is made to 'wayland' target: it is still possible
to build Wayland-only binary using that target.
VK-GL-CTS issue: 83
Components: Framework
Affects: dEQP-VK.wsi.* and all tests using EGL.
Change-Id: I1d8725d1329e1071ac545f765c374d2d7d5a29df
Iago Toral Quiroga [Thu, 11 May 2017 09:11:48 +0000 (11:11 +0200)]
Fix compute shader in robustness tests to produce an output
The compute shader only writes to shared variables, which means
that it doesn't produce any useful output. Shader compiler
backends can see this and mark all uniform variables declared
by the shader as "inactive" and remove them, which would defeat
the purpose of the 'getnuniform' test and lead to a failure.
Fix this by making the compute shader write to a shader storage
buffer object instead, so that uniforms remain active for being
involved in generating a shader output.
Component: OpenGL
VK-GL-CTS issue: 415
Affects:
KHR-GL45.robustness.getnuniform
Change-Id: I67b3fd8eb3a63012b6c46713e17c86d51b42d82e
James Helferty [Mon, 10 Apr 2017 18:20:30 +0000 (14:20 -0400)]
Tie TCS XFB to Desktop GL + NV_gpu_shader5
This change ties the ability to read PATCHES data directly from a
tessellation control shader via transform feedback to the presence of
NV_gpu_shader5 and a desktop context. Any other use of transform
feedback from a tessellation control shader is considered an error.
As part of this change, we also revert a change that used the wrong
notation for consuming PATCHES data with transform feedback.
(Reverts part of
1ab254234c4f6c45036ba68f67980858e0ef5001)
Affects:
KHR-GLES31.core.tessellation_shader.single.xfb_captures_data_from_correct_stage
KHR-GLES32.core.tessellation_shader.single.xfb_captures_data_from_correct_stage
KHR-GL44.tessellation_shader.single.xfb_captures_data_from_correct_stage
KHR-GL45.tessellation_shader.single.xfb_captures_data_from_correct_stage
Component: OpenGL
VK-GL-CTS issue: 365
Change-Id: Ie5619f9146448c54906401d2805de676000b34e8
Jeannot Breton [Wed, 10 May 2017 19:25:02 +0000 (14:25 -0500)]
Remove copy_image special path for rgb9e5/rg11b10
Copy_image is similar to a memcpy. When analyzing copy_image result,
the test have a special path for rgb9e5 and rg11b10.
Instead of using a memcmp, it converts each component to float and
then compare values.
This could be problematic, when data from other formats is
reinterpreted as rgb9e5/rg11b10, the result can be INF or NAN.
Doing a float equal comparison on 2 NAN values will return false,
so we can have 2 identical 32-bits values that are considered
different by the test because they are reinterpreted as float.
This change remove the special path for rgb9e5 and rg11b10.
I don't think it is required since copy_image should be like a
memcpy, a memcmp should be enough to verify the result
Affects:
KHR-GL45.copy_image.functional
Components: OpenGL
VK-GL-CTS issue: 404
Change-Id: Ifc7f7b4c8d62692e9364510ebc4e4162468e9568
Slawomir Cygan [Wed, 10 May 2017 17:09:07 +0000 (19:09 +0200)]
KHR-GL44.khr_debug: create texture before calling glObjectLabel
The test does not call glBindTexture after glGenTextures, so
texture object is not created.
This causes later glObjectLabel GL error checks to fail due to wrong
GL error codes.
Affects: KHR-GL44.khr_debug.*
Components: OpenGL
VK-GL-CTS Issue: 410
Change-Id: I9036264b2c549c58a5a80fb1ec89e1fdaae4b6b7
Pyry Haulos [Thu, 18 May 2017 23:02:57 +0000 (16:02 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: I305e5144dee58d10e2284e1824d9d4683a9c67d8
Pyry Haulos [Thu, 18 May 2017 22:42:53 +0000 (15:42 -0700)]
Fix Vulkan null driver
Vulkan platform was changed to use vkGetInstanceProcAddr(NULL, pName)
for loading all platform-level functions, except vkGetInstanceProcAddr
itself, in commit
bbd3dce44fafaa4b1952b63564022f19b3a65ca3. Null
driver was not adjusted for that change and that resulted calling into
null function pointer early in the initialization code.
Additionally, null driver now advertises HOST_COHERENT memory.
Change-Id: I65c8b37722950de08e337d7df95aa46a0e826c7b
Gabryelle Chamberlin [Wed, 17 May 2017 16:59:34 +0000 (12:59 -0400)]
Added VS2017 Support to Generator
james.glanville [Wed, 17 May 2017 12:45:37 +0000 (13:45 +0100)]
Reset GL_FRAMEBUFFER_SRGB state for ES tests
GL_EXT_sRGB_write_control adds new state "GL_FRAMEBUFFER_SRGB" which
is enabled by default. This was not reset between tests, meaning state
could leak between tests.
Affects: dEQP-GLES31.functional.fbo.srgb_write_control.
framebuffer_srgb_enabled
Components: AOSP
Graeme Leese [Thu, 11 May 2017 16:03:23 +0000 (17:03 +0100)]
Fix extension check in ES3.2 BoundingBox test
The test is still using the suffixed name to access gl_BoundingBox[EXT],
which means it still needs to enable the extension.
Affects: dEQP-GLES31.functional.android_extension_pack.shaders.es32.
extension_directive.ext_primitive_bounding_box
Components: AOSP
VK-GL-CTS issue: 423
Change-Id: I652e59b6ca29b1d237028eac57fc860731378c35
(cherry picked from commit
abef7613781dfc4b07b8a37cd17ab5fe3dfe0c01)
Pyry Haulos [Tue, 16 May 2017 18:03:58 +0000 (11:03 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: Icf7defc7279b1b0cf9f6b13498de3d1456239e9c
Pyry Haulos [Wed, 10 May 2017 16:54:23 +0000 (09:54 -0700)]
Update vulkan.h.in to 1.0.49 and strip unused exts
This change updates vulkan.h.in to match header version 49.
Vendor-specific extensions and KHX extensions that are not covered by CTS
are removed.
gen_framework.py now auto-detects function class from its first argument,
except for vkGet*ProcAddr, which require special handling.
Components: Vulkan
Change-Id: If6e1d959091a9064585bf29ef9983d03649ce513
Hugues Evrard [Wed, 19 Apr 2017 10:03:12 +0000 (11:03 +0100)]
Add support for metamorphic testing
Add a new test case which compare the renderings of two different yet
semantically equivalent fragment shaders. A couple of shadersets are
also added as associated data, each contains one recipient fragment
shader which is the reference, and a few variants which are semantically
equivalent to the recipient. Each variant included here has proven to
trigger some issue on an Android device, see comment in the variant
source code.
Besides comparing the renderings of both recipient and variant shaders,
the test case also checks that each shader leads to a deterministic
rendering, by performing rendering twice and doing a pixel-per-pixel
comparison on the resulting image. When comparing the recipient and
variant renderings, a slight difference is tolerated to cope with
variations due to floating point arithmetics. Moreover, a sanity check
function may be used in order to verify that the recipient image is
relevant, and thus avoid a e.g. full black image rendering to pass the
test.
Change-Id: Id6cbbe7459864c6fb3acf72faa393ee0f4d8315d
Pyry Haulos [Tue, 16 May 2017 16:12:09 +0000 (09:12 -0700)]
Merge vk-gl-cts/aosp-deqp-dev into vk-gl-cts/master
Change-Id: Iee00ded195e3178e685470b19df25a9b352ddfb5
Pyry Haulos [Tue, 16 May 2017 16:03:30 +0000 (16:03 +0000)]
Remove unused function from es31fSRGBDecodeTests.cpp
am:
1ee59ff986
Change-Id: I0bf132f2d1a09e66768bddc5818906aa0e066967
Pyry Haulos [Tue, 16 May 2017 14:25:35 +0000 (07:25 -0700)]
Remove unused function from es31fSRGBDecodeTests.cpp
Unused function causes compiler warnings to appear with recent GCC
versions.
Change-Id: I941c579994acbc9146df86acf5b9da9ee85341f5
Alexander Galazin [Mon, 15 May 2017 20:17:42 +0000 (20:17 +0000)]
Fix texture/sampler mapping in sRGB tests
am:
c8efb45ead
Change-Id: I23849cbcae67248ac5ca55f443f226d6b1e4389e
Alexander Galazin [Thu, 11 May 2017 12:27:20 +0000 (14:27 +0200)]
Fix texture/sampler mapping in sRGB tests
sRGB decode tests assume that an API side texture with index i
is connected with a shader sampler variable uTexture with index i.
This relation was however incorrectly established, e.g.
the tests were mapping texture0 to a sampler with location 0, while
such sampler was not necessarily uTexture0.
This change fixes the mapping and makes the relation more obvious.
Components: AOSP
Affects:
dEQP-GLES31.functional.srgb_texture_decode.skip_decode.*.toggled
dEQP-GLES31.functional.srgb_texture_decode.skip_decode.*.conversion_gpu
Google bug:
38257749
Change-Id: I16508f6033b31583b1c5973b70a610dd59e78e8f
Filip Demczewski [Mon, 16 Jan 2017 12:27:50 +0000 (13:27 +0100)]
Added CTS_ARB_shader_ballot tests
Added CTS_ARB_shader_ballot cases:
ShaderBallotAvailabilityTestCase
ShaderBallotBitmasksTestCase
ShaderBallotFunctionBallotTestCase
ShaderBallotFunctionReadTestCase
affects:
GL45-CTS.shader_ballot_tests.*
components: OpenGL
VK-GL-CTS issue: 143
Change-Id: I28c724a84adab7e40edb542e70bc019cde0a0156
Boris Zanin [Mon, 15 May 2017 08:00:50 +0000 (10:00 +0200)]
Fix VS2013 build by avoiding using non-standard C extension
Avoid using C++14 0b prefix.
Affects:
* GL45-CTS.shader_atomic_counter_ops_tests.*
Components: OpenGL
VK-GL-CTS issue: 175
Change-Id: Ia500db76e1112933ad3471e9ac29423ccda81d4f
Filip Demczewski [Thu, 16 Mar 2017 15:17:13 +0000 (16:17 +0100)]
Added CTS_ARB_shader_group_vote tests
Added CTS_ARB_shader_group_vote cases:
ShaderGroupVoteAvailabilityTestCase
ShaderGroupVoteAllInvocationsTestCase
ShaderGroupVoteAnyInvocationTestCase
ShaderGroupVoteAllInvocationsEqualTestCase
affects:
GL45-CTS.shader_group_vote_tests.*
components: OpenGL
VK-GL-CTS issue: 211
Change-Id: I479b09277d8386c41b007075465f8ea383fb7f0a
Pyry Haulos [Fri, 12 May 2017 20:46:38 +0000 (13:46 -0700)]
resolve merge conflicts of
633ab6f3 to deqp-dev
Change-Id: I879838e31c3fc6bbceb102c9b2720a2372360c1a
Pyry Haulos [Fri, 12 May 2017 20:30:28 +0000 (13:30 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: Icd63416e0133bca1c1580e3c36b950caa7b1bc23
Pyry Haulos [Fri, 12 May 2017 20:26:36 +0000 (13:26 -0700)]
Import fixes from Android Vulkan CTS 1.0.2 branch
* Android copy of mustpass is updated to exclude waived tests.
* Add missing file to Android.mk
* Add -fwrapv and -Wno-delete-non-virtual-dtor to Android.mk
Change-Id: I9cdb0b0490dbad4134ad085216c3969cda40747a
Pyry Haulos [Fri, 12 May 2017 18:03:48 +0000 (11:03 -0700)]
Remove tcuWin32GLES3Platform.*
GLES3 wrapper is no longer supported so these are clearly broken.
Components: Framework
Change-Id: I656b70aecc5f0380d99b471a516d271f5b820836
Pyry Haulos [Fri, 12 May 2017 17:53:03 +0000 (10:53 -0700)]
Merge vk-gl-cts/aosp-deqp-dev into vk-gl-cts/master
Change-Id: I2d372c38240ffa856509c479b4f24b03ca95649c
Pyry Haulos [Fri, 12 May 2017 17:31:32 +0000 (10:31 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: I94c771f530e058c9ac30185e2708fd0ac6fccf36
Pyry Haulos [Fri, 12 May 2017 17:31:29 +0000 (10:31 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.1 into vk-gl-cts/vulkan-cts-1.0.2
Change-Id: I11828b81ea4c9e67dc879dd98a9cbf91341de2c9
Pyry Haulos [Fri, 12 May 2017 17:31:26 +0000 (10:31 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.0 into vk-gl-cts/vulkan-cts-1.0.1
Change-Id: I6245604cf5ce44a2c4d9c064efb65b058d8f24eb
Iago Toral Quiroga [Fri, 5 May 2017 06:30:15 +0000 (08:30 +0200)]
Add more alternatives to reflect() precision tests
Intel/Mesa started to generate these in some scenarios recently.
Components: Vulkan
Vk-GL-CTS Issue: 392
Affects:
dEQP-VK.glsl.builtin.precision.reflect.*
Change-Id: Ic943cf14d531e392c03adf18605e6330267be4a1
Yanjun Zhang [Mon, 24 Apr 2017 08:59:26 +0000 (01:59 -0700)]
Check sampler_mirror_clamp_to_edge extension in tex filter tests
Components: Vulkan
VK-GL-CTS issue: 384
Affects:
dEQP-VK.texture.filtering.*.combinations.*_mirror_clamp_to_edge_*
Change-Id: Ie829b9247cd374e6c34ac70885e9e4421d284d1e
Pyry Haulos [Mon, 8 May 2017 16:15:25 +0000 (09:15 -0700)]
Do not report warning if OOM testing is incomplete
object_management.alloc_callback_fail tests are trying to exhaustively
test all OOM paths in vkCreate* functions. However, to keep test run-
time reasonable, they have a fixed maximum allocation count and if
that is exceeded report a warning since OOM testing was not
exhaustive.
Since an implementation that requires more allocations than the limit
is valid, this commit changes the tests to report clean Pass instead
of CompatibilityWarning. Note about incomplete OOM testing is included
in the test log contents instead.
Affects:
dEQP-VK.api.object_management.alloc_callback_fail.*
VK-GL-CTS issue: 399
Components: Vulkan
Change-Id: I6890321546ae42f2646b8653007172d4b8affb8a
Pyry Haulos [Fri, 12 May 2017 17:15:26 +0000 (10:15 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.0 into vk-gl-cts/vulkan-cts-1.0.1
Graeme Leese [Thu, 11 May 2017 16:03:23 +0000 (17:03 +0100)]
Fix extension check in ES3.2 BoundingBox test
The test is still using the suffixed name to access gl_BoundingBox[EXT],
which means it still needs to enable the extension.
Affects: dEQP-GLES31.functional.android_extension_pack.shaders.es32.
extension_directive.ext_primitive_bounding_box
Components: AOSP
Change-Id: I652e59b6ca29b1d237028eac57fc860731378c35
Pyry Haulos [Fri, 12 May 2017 16:38:51 +0000 (09:38 -0700)]
Update Vulkan CTS to version 1.0.2.3
New tests:
* Add tests for VK_KHR_shared_presentable_image
* Add tests for VK_KHR_get_surface_capabilities2
Fixes:
* makeFramebufferWithoutAttachments creates invalid framebuffer
* Fix typo in dE-IT.build_info.de_endianness_consistent name
* Fix - uniform buffer incorrectly dereferenced
* Use vkGetInstanceProcAddr(NULL) to load platform funcs
* Imgtec Waiver for texture_gather*cube*depth32f*
* Add VK_KHR_incremental_present to the list of allowed device
extensions
* Use -std=c++03 with GCC and clang
* Fix GCC 6.3 warnings in vulkan-cts-1.0.2
* Improve check_build_sanity.py
* Relax image verification in anisotropic filtering tests
* Respect maxColorAttachments in image tests
* Fix SPIR-V generation in dEQP-VK.spirv_assembly
* Fix - depth/stencil images mandatory only for VK_IMAGE_TYPE_2D
* Add missing barrier in ssbo layout tests
Bug:
38260702
Change-Id: I1c00d4b63c339ba69990b64a5bb225c661957e4c
Pyry Haulos [Fri, 12 May 2017 16:20:30 +0000 (09:20 -0700)]
Merge vk-gl-cts/master into aosp/deqp-dev
Pyry Haulos [Fri, 12 May 2017 16:16:56 +0000 (09:16 -0700)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: Ifb187a8bff7ec809e78fb53da5c84807e1569843
James Helferty [Mon, 8 May 2017 21:31:59 +0000 (17:31 -0400)]
Check for INVALID_OPERATION from TextureStorage
When the effective target differs from the glTextureStorage* function
used, GL should generate an GL_INVALID_OPERATION error instead of
GL_INVALID_ENUM.
Affects:
KHR-GL45.direct_state_access.textures_storage_errors
Components: OpenGL
Issue: 407
Change-Id: I8cbfaeee7ce6bb7519cf2df9a255bfcfb3c84165
Iago Toral Quiroga [Tue, 2 May 2017 11:42:05 +0000 (13:42 +0200)]
Ignore performance messages in KHR_debug tests
The KHR-GL45.khr_debug.receiveing_messages checks that only expected
messages are recorded so it should ignore spurious performance
messages that can be injected by drivers and mess up the expected
message counts.
Components: OpenGL
VK-GL-CTS issue: 394
Affects:
KHR-GL45.khr_debug.receiveing_messages
Change-Id: Iee24cb5d6de3d2c6311dc96da312540512dc2173
Jeannot Breton [Tue, 2 May 2017 19:41:55 +0000 (14:41 -0500)]
Fix texture_size_promotion snorm/RGB10_A2UI issue
Read color clamping was not considered in the snorm expected value,
so the change clamp the expected value
The alpha component in RGB10_A2UI didn't have enough bits to hold
an expected value of 4. So we change the expected alpha value to 1
Affects:
KHR-GL45.texture_size_promotion.functional
Components: OpenGL
VK-GL-CTS issue: 395
Change-Id: I920eb187e558faf9557f886be0269ea0c927918b
Jeannot Breton [Tue, 2 May 2017 20:44:43 +0000 (15:44 -0500)]
Adjust cull_distance position on pixel center
The primitive position is on the pixel edge, it should be on the
pixel center to be rendered at the expected location
Affects:
KHR-GL45.cull_distance.functional
Components: OpenGL
VK-GL-CTS issue: 396
Change-Id: I6f604c559894a4d9b1220a65935eb59b8b8e53e6