unit test for the problem. Bug found by Bastian Märkisch <bmaerkisch@web.de>.

outside the canvas.

sprintf_s uses that function and VS2010 does not strip it off even if it
is unused. Also do not use 64-bit integer division in the timer code because
that will create unwanted references to ___moddi3 and ___divdi3.

provide the deprecated snprintf(). The mingw32 runtime does not provide the
MS-specific sprintf_s(). Mingw-w64 copes with both. So we switch to sprintf_s
but also provide it as a weak symbol so that mingw32 does not complain.

works with the VS2010 runtime, too. Also reduce the stack size requirements
to avoid depending on __chkstk_ms().

printed cells instead of always returning 0. This is handy if we want to
advance a cursor after each printf() call.

caca_wherey(), in order to match the old <conio.h> naming scheme.

speedups when the calling application knows the dirty rectangle covered by
a complex operation.

display driver to clean up its shit. Fixed X11 driver accordingly.

lines, or on changed chars if we have a mask

different from what was already there.

a more natural (x,y,w,h) 4-tuple to handle.

inner implementation still only handles one dirty rectangle, but this way
we can prepare supporting applictions for the future.

lesson to you: unit tests are good.

that anything outside the dirty rectangle is guaranteed to be unchanged, but
we currently mark far too many cells as dirty. This must be optimised.

are almost there.

split is rendered moot by the plugin system: when enabled, binaries do
not link directly with libX11 or libGL. I hope this is a step towards
more consisteny and clarity.

far better for error checking.

* Use size_t and ssize_t where appropriate.

completely breaks compatibility with previous versions of libcaca on
    64-bit systems.

<cucul.h> will provide the C99 types, even if libcaca has been installed.
  * Rename what's left of "common.h" to "stubs.h".
  * Remove all references to erroneous <inttypes.h> from source files.

* Renamed cucul_*_canvas_frame() to cucul_*_frame() because we bloody well
    know that a frame belongs to a canvas.

* More debugging information in the import/export and font functions.

* Updated TODO.

as bold or underline. No higher level support for these yet, but at
    least everything is consistent.
  * Created cucul_set_attr_ansi() and cucul_set_attr_argb() which are bound
    to replace cucul_set_color() and cucul_set_truecolor().

documentation errors or imprecisions.

fullwidth characters.
  * Added a case to test/fullwidth.c for cucul_blit and fullwidth chars.