1. 29 Nov, 2011 5 commits
  2. 22 Nov, 2011 4 commits
  3. 21 Nov, 2011 1 commit
    • Dave Airlie's avatar
      arb_blend_func_extended: add three api tests · 075b8d99
      Dave Airlie authored
      
      
      The first two are heavily reused version of the gl-3.0 API tests from Ian, just
      adapted to test the two APIs for GL_ARB_blend_func_extended,
      
      the third is a simple test that glBlendFunc can accept the new tokens, and
      the changed behaviour of the GL_SRC_ALPHA_SATURATE token.
      
      tested on fglrx which fails two of the tests because it returns the wrong
      error.
      Signed-off-by: Dave Airlie's avatarDave Airlie <airlied@linux.ie>
      075b8d99
  4. 18 Nov, 2011 4 commits
  5. 17 Nov, 2011 3 commits
  6. 16 Nov, 2011 2 commits
  7. 15 Nov, 2011 2 commits
  8. 14 Nov, 2011 4 commits
  9. 12 Nov, 2011 1 commit
    • Paul Berry's avatar
      Fix redundant static-write-varying-01.frag tests. · 9493b56b
      Paul Berry authored
      
      
      When the static-write-varying-01.frag tests were introduced, they were
      accidentally made the same as the corresponding
      static-write-varying-02.frag.
      
      The original intent (from talking to Chad) was for one set of tests
      verify that varyings can't be written to via direct assignment, and
      the other set of tests to verify that varyings can't be written to via
      passing to a function as an "out" parameter.
      
      This patch updates static-write-varying-01.frag to reflect the
      original intent.
      Reviewed-by: default avatarChad Versace <chad.versace@linux.intel.com>
      9493b56b
  10. 11 Nov, 2011 2 commits
  11. 10 Nov, 2011 1 commit
    • Chad Versace's avatar
      arb_framebuffer_object: Add tests to check GL_DEPTH_STENCIL_ATTACHMENT · 34a56532
      Chad Versace authored
      
      
      Add the following tests:
      	specs/ARB_framebuffer_object/same-attachment-glFramebufferRenderbuffer-GL_DEPTH_STENCIL_ATTACHMENT
      	specs/ARB_framebuffer_object/same-attachment-glFramebufferTexture2D-GL_DEPTH_STENCIL_ATTACHMENT
      
      The tests check that if an object is attached to the
      GL_DEPTH_STENCIL_ATTACHMENT point, then it is also attached to the
      GL_STENCIL_ATTACHMENT and GL_DEPTH_ATTACHMENT points.
      
      v2:
        - Make tests run concurrently.
        - Move test body from piglit_display() to piglit_init().
      Signed-off-by: default avatarChad Versace <chad.versace@linux.intel.com>
      34a56532
  12. 09 Nov, 2011 6 commits
    • Chad Versace's avatar
      glx: Test behavior of glXQueryDrawable() · acb99ccd
      Chad Versace authored
      
      
      Add the following tests:
      
      * glx-query-drawable-GLXBadDrawable
            Calls glXQueryDrawable(draw=0) and expects that the GLXBadDrawable
            error is generated.
      
      * glx-query-drawalbe-GLX_WIDTH
      * glx-query-drawalbe-GLX_HEIGHT
            Calls glXQueryDrawable() with the given attribute.
      
      Tested on Fedora 15 with Xserver 1.10.14. With the proprietary Nvidia
      driver, all tests pass. With Mesa 7.11, the GLXBadDrawable test passes but
      the other two fail.
      
      All tests fail on Fedora 16.
      All tests pass, however, with the proprietary Nvidia driver on Fedora 15.
      Signed-off-by: chadversary's avatarChad Versace <chad@chad-versace.us>
      acb99ccd
    • Chad Versace's avatar
      egl: Test behavior of eglQuerySurface() · 39b276b6
      Chad Versace authored
      
      
      Add the following tests:
      
      * egl-query-surface --bad-surface
            Call eglQuerySurface(surf=0) and expect that error EGL_BAD_SURFACE
            is generated.
      
      * egl-query-surface --bad-attr
            Call eglQuerySurface(attr=EGL_DONT_CARE) and expect that error
            EGL_BAD_ATTRIBUTE is generated.
      
      * egl-query-surface --attr=EGL_WIDTH
      * egl-query-surface --attr=EGL_HEIGHT
            Call eglQueryDrawable() with the given attribute.
      
      All tests pass on Fedora 16.
      They also pass with the proprietary Nvidia driver on Fedora 15.
      Signed-off-by: chadversary's avatarChad Versace <chad@chad-versace.us>
      39b276b6
    • Chad Versace's avatar
      util: Add piglit_expect_egl_error() · 2b0a1583
      Chad Versace authored
      
      
      This function checks for unexpected EGL errors.  If eglGetError() returns
      an unexpected error, then this it prints a diagnostic and terminates the
      test.
      Signed-off-by: chadversary's avatarChad Versace <chad@chad-versace.us>
      2b0a1583
    • Chad Versace's avatar
      egl: Add fields egl_test::window_width,window_height · c0403c04
      Chad Versace authored
      
      
      The default window dimension remains unchanged, 300x300.
      Signed-off-by: chadversary's avatarChad Versace <chad@chad-versace.us>
      c0403c04
    • Chad Versace's avatar
      egl: Change each test to initialize egl_test with common function · 34568a6c
      Chad Versace authored
      
      
      Define egl_test_init(struct egl_test*), which initializes the test to
      default values. Modify each test to use that function.
      
      This is in preparation for allowing an EGL test to specify its window
      dimensions.
      Signed-off-by: chadversary's avatarChad Versace <chad@chad-versace.us>
      34568a6c
    • Chad Versace's avatar
      egl: Add all EGL tests to all_egl.tests · 4762e905
      Chad Versace authored
      
      
      The EGL tests were built by CMake, but since they were not enumerated in
      any Python file one had to run them manually. This meant the EGL tests
      were never ran.
      
      We can't add the EGL tests to all.tests because Windows doesn't support
      EGL. Ideally, we could add the EGL tests to all.tests and the test runner
      would skip the tests if libEGL was not present. However, the test runner
      is not too bright. This patch lists the tests in a separate file as
      a workaround until we find a better solution.
      Signed-off-by: default avatarChad Versace <chad.versace@linux.intel.com>
      4762e905
  13. 08 Nov, 2011 5 commits
    • Ian Romanick's avatar
      Don't set array members that are not used · c4eb4a59
      Ian Romanick authored
      
      
      Optimizing GLSL linkers may reduce the size of the uniform array if
      tail elements are not accessed.  Shader runner will fail the test if
      one of the set uniforms doesn't have a location.
      
      Previously these tests would fail on AMD's closed-source driver for
      this reason.  Now they pass.
      Signed-off-by: default avatarIan Romanick <ian.d.romanick@intel.com>
      Reviewed-by: Paul Berry's avatarPaul Berry <stereotype441@gmail.com>
      c4eb4a59
    • Ian Romanick's avatar
      glsl-1.10: Don't set array members that are not used · 4906ab5b
      Ian Romanick authored
      
      
      Optimizing GLSL linkers may reduce the size of the uniform array if
      tail elements are not accessed.  Shader runner will fail the test if
      one of the set uniforms doesn't have a location.
      
      Previously these tests would fail on AMD's closed-source driver for
      this reason.  Now they pass.
      Signed-off-by: default avatarIan Romanick <ian.d.romanick@intel.com>
      Reviewed-by: Paul Berry's avatarPaul Berry <stereotype441@gmail.com>
      4906ab5b
    • Ian Romanick's avatar
      arb_shader_objects: Test glGetUniformLocation on an array of struct of array · 89e9f949
      Ian Romanick authored
      
      
      This test passes on NVIDIA's closed-source drivera and on AMD's
      closed-source driver.  Currently Mesa fails to get the location of
      array elements embedded in a structure.
      
      v2: Reduce log spam when the test passes.  Initialize pass to true so
      that the test can pass.  Duh.
      Signed-off-by: default avatarIan Romanick <ian.d.romanick@intel.com>
      Reviewed-by: Emma Anholt's avatarEric Anholt <eric@anholt.net>
      89e9f949
    • Paul Berry's avatar
      Transform feedback: test order of recorded vertices. · a6e0199f
      Paul Berry authored
      
      
      This patch adds a test to confirm that the order of vertices output by
      transform feedback matches the order of vertices supplied to the GL
      pipeline, even if some vertices take significantly longer to process
      than others.
      Reviewed-by: default avatarIan Romanick <ian.d.romanick@intel.com>
      a6e0199f
    • Paul Berry's avatar
      Modify EXT_transform_feedback tests to test GL 3.0 core. · 9016314b
      Paul Berry authored
      
      
      Nearly all of the functionality of the EXT_transform_feedback
      extension is available in core OpenGL as of version 3.0 (only one
      function was not promoted to OpenGL 3.0, the BindBufferOffsetEXT
      function).  Therefore, it's possible for an implementation to support
      transform feedback via core OpenGL 3.0 features, without advertising
      support for the EXT_transform_feedback extension.  One such
      implementation, surprisingly, is the nVidia proprietary Linux driver I
      use as my reference platform.
      
      To allow the transform feedback tests to run on these implementations,
      I've created a function piglit_require_transform_feedback(), and the
      following function pointers:
      
      - piglit_BeginTransformFeedback
      - piglit_BindBufferBase
      - piglit_BindBufferRange
      - piglit_EndTransformFeedback
      - piglit_GetBooleani_v
      - piglit_GetIntegeri_v
      - piglit_GetTransformFeedbackVarying
      - piglit_TransformFeedbackVaryings
      
      piglit_require_transform_feedback() checks whether transform feedback
      is provided via EXT_transform_feedback or via core GL 3.0, and sets up
      the function pointers appropriately.  If transform feedback isn't
      provided through either mechanism, then it causes the test to be
      skipped.
      
      This is similar to what we do with piglit_require_GLSL() (see
      tests/util/piglit-shader-gl.c).
      
      I've updated all of the EXT_transform_feedback tests to use these
      function pointers.  This makes it possible to test transform feedback
      on an implementation that supports GL 3.0 but not
      EXT_transform_feedback.
      Acked-by: Marek Olšák's avatarMarek Olšák <maraeo@gmail.com>
      Reviewed-by: default avatarIan Romanick <ian.d.romanick@intel.com>
      9016314b