[babl] Created tag BABL_0_1_22

[Øyvind Kolås <ok src gnome org>]

The unsigned tag 'BABL_0_1_22' was created.

Tagger: Øyvind Kolås <pippin gimp org>
Date: Tue Nov 29 22:12:42 2016 +0100

    release 0.1.22

Changes since the last tag 'BABL_0_1_20':

Alexandre Prokoudine (1):
      Post-release version bump to 0.1.21

Edward E (3):
      babl-cache: Add a #define for FALLBACK_CACHE_PATH
      babl-verify: fix win32 compile, with s/setenv/putenv/g
      babl, babl-cache: enable cache write, fix cache path processing

Jehan (1):
      babl-cache: just some styling fix.

SBP (1):
      Add extension with HCY color model, and pixel formats

Simon Budig (1):
      extensions: align rounding of R'G'B'A u8 -> cairo-ARGB32 with reference conversion

Øyvind Kolås (57):
      babl: cache state between babl runs
      bail on db-cache restoration if format lookup fails
      babl_init_db: formats being with first char of line
      babl-fish-path: do not allow parallell construction
      fish cache: sort cached fish paths by descending usage
      build: add generation of git-version.h
      babl.db: store git rev/version in header
      babl.db: if git hash has changed, drop cache
      remove deprecated loss struct member
      babl.db: do not use strtod - it is locale dependent
      move fish cache to own .c file
      babl-cache: store tolerance as part of cache versioning
      babl-cache: add a 1% probability of cached conversions being dropped on load
      reduce default max babl path to 3
      babl-cache: remove unneeded switch
      store cached fishes in ~/.cache/babl
      babl-cache: attempt to make things barely compile on win32
      fast-float: tweak lut generation
      add single precision versions of linear/gamma conversions
      extensions/cairo: replace hard-coded LUT with runtime LUT
      change default BABL_TOLERANCE to 0.000006
      extensions/cairo: remove some arithmetic
      cairo: slight improvements to speed and accuracy
      cairo: pad with 255 instead of copy of color for 24bit format
      fast-float: use higher accuracy LUTs and rework some alpha handling
      extensions/gggl-lies: bump up accuracy of a conversion
      fast-float: implement linear interpolation in LUT
      tools: add babl-verify start of an development aid when improving fast paths
      babl-fish-path: add some specific conversion targeted debug info
      fast-float: make float to cairo conversion take clipping into account
      extensions/cairo: remove ; now unused; LUT
      extension/fast-float: gain some speed by ignoring a couple high mantissa bits
      remove babl_process_cost
      fast-float: improve memory access order
      cairo: reorder memory reads to be linear
      remove bablfishpath
      fast-float: indentation
      gegl-fixups: indentation
      remove babl-fish-stats
      fast-float: fix regression introduced when reindenting
      make pow implementations inlineable - gaining some fixed amount of blitting performance
      fast-float: add RGBA float to R'G'B'A u8
      two-table: add RGBA float to R'G'B'A u8
      babl-verify: set a high tolerance, avoid censoring some underperforming fast paths
      babl-cache: avoid %'ing by 0
      fast-float: tweak precision so that float->8bit tests pass without issue
      babl-cache: use getenv("TEMP") for folder location on win32, and fix mk_ancestry
      fish_cache_path(): simplify through use stack instead of heap
      babl-cache: expect newline trouble on win32
      babl-verify: make it compile on win32, by making setenv use putenv
      mk_ancestry: propagate return value
      babl-cache: remove win32 bail at start of loading code
      babl-cache: rearrange per fish data
      babl-cache: incorporate BABL_PATH_LENGTH in header
      babl-cache: use clearer arithmetic for dropping 1% of cached fishes
      rename define for default BABL_TOLERANCE to match envvar
      configure.ac: 0.1.22 release