1. 04 Sep, 2012 1 commit
  2. 31 Aug, 2012 1 commit
  3. 24 Jul, 2012 1 commit
  4. 20 Jul, 2012 1 commit
  5. 08 Jul, 2012 1 commit
  6. 28 Jun, 2012 2 commits
  7. 22 Jun, 2012 1 commit
    • Paul Berry's avatar
      Test that dFdx/dFdy work properly on fbos · d7908a28
      Paul Berry authored
      This test exposes a bug in Mesa on i965 hardware (as of commit
      f2f05e5): the GLSL function dFdy() produces incorrect results when
      rendering to an FBO.  This is a consequence of the fact that FBOs
      place the origin at the upper left, whereas windowsystem framebuffers
      place the origin at the lower left.
      
      Verified using nVidia's proprietary binary driver for Linux.
      Reviewed-by: default avatarChad Versace <chad.versace@linux.intel.com>
      d7908a28
  8. 20 Jun, 2012 1 commit
  9. 07 Jun, 2012 8 commits
    • Paul Berry's avatar
      fbo: Make an alternate function for initializing fbo-formats.h. · 275d8644
      Paul Berry authored
      This patch extracts code from fbo_formats_init() into a separate
      function, fbo_formats_init_test_set(), which skips command line
      parsing and simply initializes fbo-formats.h to use a particular test
      set.  This will allow future tests to make use of fbo-formats.h even
      if they have to use different command-line parsing than what is
      implemented in fbo_formats_init().
      
      In addition, fbo_formats_init() has been changed to a non-static
      function, so that if an fbo test directly calls
      fbo_formats_init_test_set() instead, we won't get a spurious compile
      warning about an unused function.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      275d8644
    • Paul Berry's avatar
      fbo: Split out a function for selecting an fbo test set by name. · d7ccdb0e
      Paul Berry authored
      This patch extracts code from fbo_formats_init() into a separate
      function, fbo_lookup_test_set(), whose only purpose is to find an FBO
      test set given its name.  This will make it easier for future tests to
      make use of fbo-formats.h even if they have to use different
      command-line parsing than what is implemented in fbo_formats_init().
      
      This is also one behavioural change: previously, if multiple
      command-line arguments were specified to fbo_formats_init(), it would
      ignore all but the last.  This was clearly an unintended and
      unnecessary behaviour, so I've replaced it with an error message.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      d7ccdb0e
    • Paul Berry's avatar
      fbo: Eliminate unnecessary variable test_set. · 3ad41247
      Paul Berry authored
      The only purpose of the test_set variable was to point to
      test_sets[test_index], but it was not consistently updated.  This
      patch replaces it with explicit references to test_sets[test_index] to
      reduce confusion.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      3ad41247
    • Paul Berry's avatar
      fbo: Remove test_desc::base. · e7b04a26
      Paul Berry authored
      The field test_desc::base is no longer used.  This patch removes it
      and simplifies the type signatures of the functions that previously
      used it.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      e7b04a26
    • Paul Berry's avatar
      fbo: Use format_desc::base_internal_format rather than test_desc::base. · 2b776b9a
      Paul Berry authored
      The field test_desc::base has two disadvantages: (1) it is only
      populated for depth and depth/stencil formats (for all other formats
      it is zero).  (2) it misclassifies GL_DEPTH32F_STENCIL8 as having a
      base format of GL_DEPTH_COMPONENT, when the correct base format is
      GL_DEPTH_STENCIL.
      
      This patch replaces all references to test_desc::base with
      format_desc::base_internal_format, which doesn't have either of these
      disadvantages.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      2b776b9a
    • Paul Berry's avatar
      fbo: Rearrange if tests for GL_DEPTH32F_STENCIL8 behaviour. · 873ba64f
      Paul Berry authored
      The tests fbo-clear-formats and fbo-generatemipmap-formats need
      special behaviour for GL_DEPTH32F_STENCIL8.  This patch rearranges the
      if statements that implement that special behaviour so that they no
      longer rely on the fact that GL_DEPTH32F_STENCIL8 is (incorrectly)
      categorized as having a base internal format of GL_DEPTH_COMPONENT
      rather than GL_DEPTH_STENCIL.  This will allow the incorrect
      categorization to be removed in a future patch.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      873ba64f
    • Paul Berry's avatar
      fbo: Add base internal format info to fbo-formats.h · 582130b5
      Paul Berry authored
      This patch augments the format_desc structure (used by fbo-formats.h
      to describe a format to be tested) with a base_internal_format field,
      which indicates the base internal format corresponding to each
      particular internal format (e.g. the base internal format of
      GL_R3_G3_B2 is GL_RGB).
      
      The field test_desc::base (which carries the same information, but
      only for depth/stencil formats, and is sometimes incorrect) will be
      removed in a later patch.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      582130b5
    • Paul Berry's avatar
      fbo: Make fbo-formats.h able to be included from a .cpp file. · f9091bbd
      Paul Berry authored
      This required changing format_desc::name to a const char *, renaming a
      variable called "new", and fixing a few signed/unsigned comparison
      warnings.
      Reviewed-by: Anuj Phogat's avatarAnuj Phogat <anuj.phogat@gmail.com>
      f9091bbd
  10. 24 May, 2012 2 commits
  11. 23 May, 2012 1 commit
  12. 20 May, 2012 1 commit
  13. 15 May, 2012 1 commit
  14. 01 May, 2012 1 commit
  15. 23 Mar, 2012 1 commit
  16. 13 Mar, 2012 2 commits
  17. 12 Mar, 2012 1 commit
  18. 08 Mar, 2012 2 commits
  19. 03 Mar, 2012 1 commit
    • Dylan Noblesmith's avatar
      don't write generated header to the source directory · 2cb512bd
      Dylan Noblesmith authored
      It was impossible to have the source directory read-only.
      
      Also add the include_directories() directive for tests/util
      in just one place, under tests/, so that all subdirectories
      inherit it. A bunch of CMakeLists.txt files duplicate it,
      so delete those redundant include flags:
      
      sed -i -e "/^\t\${piglit_SOURCE_DIR}\/tests\/util/ d" \
       `grep piglit_SOURCE_DIR -rl tests/ | grep "CMakeLists\.gl"`
      2cb512bd
  20. 23 Feb, 2012 1 commit
  21. 15 Feb, 2012 1 commit
    • Jian Zhao's avatar
      Add test case of render FBO with point coordinate. · dad51a14
      Jian Zhao authored
      Add test case of rendering with point coordinate, it should work with
      direct rendering, rendering with FBO.
      
      As the es spec 2.0.25 says:
      "the gl-PointCoord fragment shader input defines a per-fragment coordinate
      space (s; t) where s varies from 0 to 1 across the point horizontally
      left-to-right, and t ranges from 0 to 1 across the point vertically
      top-to-bottom."
      
      v2: To simplify the case, set the point size in the program instead of the
      shader. (per Yuanhan's suggestion)
      
      v3: Put the case in the fbo directory, move the FBO setup code into
      piglit_init and will not draw it to the screen when run with '-auto'
      option. (per Ian's suggestion)
      
      Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=44613Signed-off-by: default avatarJian Zhao <jian.j.zhao@intel.com>
      Reviewed-by: Yuanhan Liu's avatarYuanhan Liu <yuanhan.liu@linux.intel.com>
      dad51a14
  22. 27 Jan, 2012 2 commits
  23. 20 Jan, 2012 1 commit
  24. 18 Jan, 2012 1 commit
  25. 05 Jan, 2012 1 commit
    • Brian Paul's avatar
      fbo-blit-d24s8: don't try to blit betweendifferent Z/stencil depths · 68b0f40d
      Brian Paul authored
      The docs for glBlitFramebuffer say GL_INVALID_OPERATION is generated
      if the source/dest depth/stencil buffer formats do no match.  The spec
      is a bit vague on what's meant by formats (do just the component depths
      need to match, or must the hw formats match?) but clearly if the depths
      don't match, the formats can't match.
      
      This fixes a bogus error that's generated when testing swrast which
      usually uses a 16-bit depth buffer for the window and a 24-bit Z
      renderbuffer for the FBO.
      
      v2: just report PIGLIT_SKIP if the bit depths don't match (per Eric)
      Reviewed-by: Eric Anholt's avatarEric Anholt <eric@anholt.net>
      68b0f40d
  26. 02 Jan, 2012 1 commit
  27. 30 Dec, 2011 1 commit
  28. 07 Dec, 2011 1 commit