[cairo] [cairo-commit] 12 commits - src/cairo-analysis-surface.c src/cairo-directfb-surface.c src/cairo-gstate.c src/cairo.h src/cairoint.h src/cairo-jpeg-info.c src/cairo-jpeg-info-private.h src/cairo-meta-surface.c src/cairo-meta-surface-private.h src/cairo-paginated-surface.c src/cairo-pdf-surface.c src/cairo-pdf-surface-private.h src/cairo-ps-surface.c src/cairo-ps-surface-private.h src/cairo-quartz-surface.c src/cairo-surface.c src/cairo-surface-fallback.c src/cairo-surface-private.h src/cairo-svg-surface.c src/cairo-type1-subset.c src/cairo-type3-glyph-surface.c src/cairo-win32-printing-surface.c src/cairo-win32-surface.c src/cairo-xlib-surface.c src/Makefile.sources src/test-meta-surface.c src/test-paginated-surface.c test/cairo-test.c test/README

[Adrian Johnson]

Behdad Esfahbod wrote:
> Adrian Johnson wrote:
>>  src/cairo-jpeg-info-private.h      |   54 +++++++++++
>>  src/cairo-jpeg-info.c              |  142 +++++++++++++++++++++++++++++
> 
> These two are short enough to live in cairoint.h and cairo-misc.c.

There will be additional code required for each of the remaining four 
image types I will add. Maybe they could all go in a 
cairo-image-format.c or cairo-image-info.c.

 >>[snip]
> 
> This is the wrong approach.  The getter returns NULL for any types it doesn't
> understand.  The idea is to keep anything the user sets, and let the backend
> use as appropriate.
> 
> The way it should be done instead is using a central interned string store
> (cairo_user_data_key_from_string() implemented using a hash table), plus the
> currently in place user-data facilities.

I probably should have explained in the commit message that it is still 
a work in progress and when I add the next image format I would convert 
it to use the user data facilities. Fortunately Chris Wilson has already 
beaten me to it and extended the implementation to as you describe.