Hi Axel, > I've removed that check now. Although I think it should also return > false if cloning the accelerant failed - but probably not before > having > tried. If cloning the accelerant failed that would be perfectly acceptable I'd think, because CanControlFrameBuffer sits in BWindowScreen: which is unusable then. So, the app_server drawing fault remains. After that, I wouldn't be surprised if BWindowScreen is (nearly) OK.. > BTW I might have fixed the extra processor usage for overlay bitmaps > - > turns out we invalidated the view (causing its Draw() message to be > called and the background to be cleared again) for every overlay > bitmap > update. Can you check? You bet you! :) I just checked background CPU usage of dano and Haiku, these are more or less the same (low), although Haiku is a bit higher. This in any case wasn't the reason either. Another thing, maybe you know by heart: some apps exist that use the PCI busmanager from userspace. Haiku fails these apps (BeRoMeter tested). Intended behaviour? I know on R5 and Dano you get a warning when compiling saying that get_nth_xxx isn't declared, but it's just a warning an the app runs after that OK. Seems like on Haiku this doesn't work however. OK, I'll now do a checkout and rebuild of Haiku on the laptop (I was still working with a bit older (and tweaked for these tests) version, 30 april or so). Will be a while before I check back in I guess :) thanks for your prompt responses!!! At this rate I'll probably be able to run anything on Haiku very soon.. ;-) Bye! Rudolf.