[cairo] Supporting the cairo directfb backend

[Chris Wilson]

Hi all,
  over the last couple of days I've attacked cairo's directfb backend
and it's almost entirely passing the test suite. The original motivation
was to simply fix an apparent regression in 1.6:
https://bugs.freedesktop.org/show_bug.cgi?id=15652, but since I was
unable to reproduce it directly I decided to fix as many bugs as I could
identify from the test suite. Having approached the directfb backend
with no prior knowledge of libdirectfb, I may have been overzealous and
too conservative with some of may changes, choosing to trigger a
fallback where perhaps there may have been a better solution. So I am
asking anyone and everyone with a bit of libdirectfb experience to look
over the current backend and see if we can close the last few FAILs.
(And then we start addressing performance issues.)

Looking forward, the directfb backend is the most widely used
unsupported backend. So the question is what issues remain in the
backend and how do we go forward? For example, a critical missing
element is a baseline test environment that Carl and Behdad can run
under distcheck. (Obviously the same/similar environment will be used by
developers as they make infrastructure changes and need to check the
backends.) Immediately, we hit an issue where the directfb backend
chooses different code paths depending upon the available HW
acceleration. (As the fallback paths are tested by other backends in the
test suite, simply forcing the HW accelerated paths makes sense whilst
testing - so, perhaps, system=x11 would suffice?) So I'm open to
suggestions, but I'm optimistically looking for some one to step forward
and develop the backend to supported status during the 1.9 cycle (which
promises to be quite an exciting ride!) and maintain it hence forth.
-- 
Chris