mesa: c1d35aece0afc2822d6d9f6c22664c04e6fcbba3 (master)

$ ./bin/depth_texture_mode_and_swizzle -auto
Probe at (10,10)
  Expected: 0.500000 0.500000 0.500000 0.500000
  Observed: 0.501961 0.501961 0.501961 1.000000
Probe at (30,10)
  Expected: 1.000000 0.500000 0.500000 0.500000
  Observed: 1.000000 0.501961 0.501961 1.000000
Probe at (130,10)
  Expected: 0.000000 0.000000 0.000000 0.500000
  Observed: 0.000000 0.000000 0.000000 1.000000
Probe at (150,10)
  Expected: 1.000000 0.000000 0.500000 0.000000
  Observed: 1.000000 0.000000 0.501961 1.000000
PIGLIT: {'result': 'fail' }


570ed2be7d776211e1ca2a7a4c44ee6a1d141714 is the first bad commit
commit 570ed2be7d776211e1ca2a7a4c44ee6a1d141714
Author: Carl Worth <cworth@cworth.org>
Date:   Mon Jan 21 12:16:27 2013 -0800

    ReadPixels: Force ALPHA to 1 while rebasing RGBA values for GL_RGB format
    
    When performing a ReadPixels operation, we may be reading from a buffer that
    stores alpha values, but that is actually representing a buffer with no alpha
    channel. In this case, while rebasing the values, touch up all alpha values
    read to 1.0.
    
    This commit fixes the following piglit (sub) tests:
    
    	ARB_texture_float/fbo-colormask-formats
    		GL_RBG16F_ARB
    	EXT_texture_snorm/fbo-colormask-formats
    		GL_RGB16_SNORM
    		GL_RGB8_SNORM
    		GL_RGB_SNORM
    
    It likely improves the results of other tests as well, but a PASS remains
    elusive due to additional bugs.
    
    Reviewed-by: Brian Paul <brianp@vmware.com>
    Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

:040000 040000 144369a7d3779929bad84beca8f3a5b2ccf90640 c25eb37e73f6f6e5435230fe8a799b1b62ed347b M	src
bisect run success