[cairo] Re: Re: 1.1.6 to 1.2.0 performance

[Vladimir Vukicevic]

Sorry, I was wrong earlier -- flush() is called from the generic
surface finalize() handler (not from font code), which is I believe
where this is being hit.  My performance measurement were all being
done after surface creation and before finalization, so I never saw
the effects of the extra XSync.  I'll put together a patch to remove
it outright shortly, sorry 'bout that!

    - Vlad