Permalink
Commits on Jun 7, 2017
  1. Fix some typos in comments and documentation (#58)

    stweil committed with Flameeyes Jun 7, 2017
    Most of them were found by codespell.
    
    Signed-off-by: Stefan Weil <sw@weilnetz.de>
Commits on Jan 10, 2017
  1. Merge pull request #28 from ccaccb/master

    Flameeyes committed Jan 10, 2017
    better manpage info for the blackfilter
Commits on Jan 3, 2016
  1. tests: update golden files to reflect the actual output filetype.

    Flameeyes committed Jan 3, 2016
    Recompressing the source png files caused the output test files to be
    emitted in a different format, so fix that.
    
    Proper grayscale tests and golden files should actually be added, but
    separately.
  2. tests: compress source images with zopflipng (both release and git ma…

    Flameeyes committed Jan 3, 2016
    …ster.)
    
    This significantly reduces the size of the tests input corpus, unfortunately the output is the main culprit of the distributable size.
Commits on Dec 2, 2014
  1. Merge pull request #34 from ccaccb/fix_segfault

    Flameeyes committed Dec 2, 2014
    fix a possible printf(NULL) when -debug is used
  2. fix a possible printf(NULL) when -debug is used

    Hannes Franke committed Dec 2, 2014
Commits on Dec 1, 2014
  1. Destroy (most) of the libav data structures created during file loadi…

    Flameeyes committed Dec 1, 2014
    …ng and
    
    saving.
    
    This ensures that we close the context and streams that are created from
    libav; this might still not be complete, but there seems to be different
    APIs between libav and ffmpeg so I'm changing one piece at a time.
    
    Works towards closing Issue #26.
Commits on Nov 29, 2014
  1. Merge pull request #31 from ccaccb/gitignore

    Flameeyes committed Nov 29, 2014
    add compare-image binary to .gitignore
  2. add compare-image binary

    Hannes Franke committed Nov 29, 2014
Commits on Nov 21, 2014
  1. Merge pull request #29 from ccaccb/fix_segfault

    Flameeyes committed Nov 21, 2014
    Fix segfault when inputting three page-arguments
  2. removed redundant 'return -1'

    Hannes Franke committed Nov 21, 2014
  3. fix typo in comment

    Hannes Franke committed Nov 21, 2014
  4. fix segfault and add some debug output

    Hannes Franke committed Nov 21, 2014
Commits on Nov 20, 2014
  1. add handling of short argument -bn as said in manpage

    Hannes Franke committed Nov 20, 2014
  2. better manpage info for the blackfilter

    Hannes Franke committed Nov 20, 2014
Commits on Nov 11, 2014
  1. Remove explicit rule for tests/.dirstamp.

    Flameeyes committed Nov 11, 2014
    Now that we build an object within the tests/ directory, the rule is already emitted by automake.
  2. Use float for the degrees-to-radians conversion, too.

    Flameeyes committed Nov 11, 2014
    It seems like for at leasts some instruction set this is not fully expanded at build time.
  3. Consistently reduce to single precision floating point the rotation c…

    Flameeyes committed Nov 11, 2014
    …ode.
    
    This causes some slight variation in the output depending on the available
    instruction set but still produces a perfectly valid output.
  4. Use float rather than double for rotation detection.

    Flameeyes committed Nov 11, 2014
    The code should be slightly faster this way and the new test system allows us to make this change as the approximation causes some minimal changes.
  5. Rewrite tests to use an image comparison binary instead of md5sum.

    Flameeyes committed Nov 11, 2014
    This allows to check for at most 1% image change with no metadata change
    instead of a perfect match, that fails when using fused-multiply-add
    instructions.
    
    See, for context on why this is needed:
    https://blog.flameeyes.eu/2014/10/the-subtlety-of-modern-cpus-or-the-search-for-the-phantom-bug
Commits on Oct 26, 2014
  1. file: Set the time_base to the AVStream

    lu-zero committed Oct 26, 2014
    From Libav11 only setting the codec time_base triggers a warning.
  2. file: Do not recycle the AVStream codec context

    lu-zero committed Oct 26, 2014
    This the correct pattern since Libav 0.6.
Commits on Aug 19, 2014
  1. Remove INSTALL file.

    Flameeyes committed Aug 19, 2014
    It was not really packaged, and it was the standard automake one, there's no reason to keep it around.
Commits on Aug 18, 2014
Commits on Aug 17, 2014
  1. For PAL8 files, load them in an RGB24 frame instead.

    Flameeyes committed Aug 17, 2014
    This allows loading tiff files with libav without waiting for libav 11 but
    it will create colour files by default, even though the image would be
    grayscale only.
    
    libav 11 supports grayscale tiff files as grayscale rather than palettized,
    but until then, this workaround should work.
  2. Simplify stretching calls.

    Flameeyes committed Aug 17, 2014
    Apply the zoomfactor on the requested sizes, and always call stretch().
    
    If the stretch() would make the image the same size, simply avoid
    processing altogether.
  3. Remove unused function.

    Flameeyes committed Aug 17, 2014