FrameSvg and masks

[Marco Martin]

Hi all,
speaking with fredrikh about speed with window masks some things have emerged:
returning a qbitmap is a bit wrong, since it will have to be converted to a 
qregion each time, so..
should the framesvg mask function return a qregion? this would of course break 
the api, so quite of a problem
would be source compatible but not binary compatible can still be done before 
4.2? (talk now or be silent forever?) worth it?

also, now the shaped windows like dialogs and tooltips uses clip masks, using 
only input masks when composite is on would be faster, like krunner does since 
today as i can see..

buuuut, for now it's a bit a nooo, because kwin will attempt to use its own 
shadow and krunner looks really nasty now :/

as for this second problem, so it should be postponed for 4.3 i think (and 
still make  krunner set the clip mask always)
and then either find another way for disabling shadows in kwin, or have kwin 
to draw better shadows with shaped windows (and this thing totally doesnt 
depend from us:))

this things i wrote does make sense at all?
what should be done?

Cheers,
Marco Martin