1 - It probably makes sense to move the more strange X region API
2 into pixman as well, but guarded with PIXMAN_XORG_COMPATIBILITY
4 - Go through things marked FIXME
6 - Reinstate the FbBits typedef? At the moment we don't
7 even have the FbBits type; we just use uint32_t everywhere.
9 Keith says in bug 2335:
11 The 64-bit code in fb (pixman) is probably broken; it hasn't been
12 used in quite some time as PCI (and AGP) is 32-bits wide, so
13 doing things 64-bits at a time is a net loss. To quickly fix
14 this, I suggest just using 32-bit datatypes by setting
15 IC_SHIFT to 5 for all machines.
17 - Consider whether calling regions region16 is really such a great idea
19 - Run cairo test suite; fix bugs
20 - one bug in source-scale-clip
22 - Remove the warning suppression in the ACCESS_MEM macro and fix the
23 warnings that are real
25 - Add calls to prepare and finish access where necessary.
26 grep for ACCESS_MEM, and make sure they are correctly wrapped in
29 - Make source clipping optional.
30 - done: source clipping happens through an indirection.
31 still needs to make the indirection settable.
33 - Add non-mmx solid fill
37 - Make sure the endian-ness macros are defined correctly.
39 - The rectangles in a region probably shouldn't be returned const as
40 the X server will be changing them.
42 - Right now we _always_ have a clip region, which is empty by default.
43 Why does this work at all? It probably doesn't. The server
44 distinguishes two cases, one where nothing is clipped (CT_NONE), and
45 one where there is a clip region (CT_REGION).
47 - Default clip region should be the full image
49 - Test if pseudo color still works. It does, but it also shows that
50 copying a pixman_indexed_t on every composite operation is not
51 going to fly. So, for now set_indexed() does not copy the
54 Also just the malloc() to allocate a pixman image shows up pretty
59 - Make all the setters not copy their arguments
61 - Possibly combined with going back to the stack allocated
62 approach that we already use for regions.
64 - Keep a cached pixman_image_t around for every picture. It would
65 have to be kept uptodate every time something changes about the
68 - Break the X server ABI and simply have the relevant parameter
69 stored in the pixman image. This would have the additional benefits
72 - We can get rid of the annoying repeat field which is duplicated
75 - We can use pixman_color_t and pixman_gradient_stop_t
76 etc. instead of the types that are defined in