On 2006-01-28 at 19:09:07 [+0100], Stephan Assmus <superstippi@xxxxxx> wrote: > On 2006-01-28 at 17:25:19 [+0100], Ingo Weinhold wrote: > > On 2006-01-28 at 15:40:47 [+0100], Oliver Tappe wrote: > > [...] > > > As the cross compiler isn't tested, I'd be very grateful if someone else > > > could check it out and tell me if my change indeed fixes things... > > > > It does. Thanks! > > > > I tested WonderBrush under qemu and it runs. Although it complains about > > "much too large bitmap bounds" at startup and enters the debugger when > > quitting, due to an unlocked looper. > > Both things happen on a "true" :-) Haiku build too. The Bitmap bounds thing > is a WonderBrush bug that didn't cause any problems on R5 so I fixed Haiku > to handle it too. I have not investigated the other problem yet. Feel free > to do that though! Did that. We were basically not handling not(-yet)-run windows (e.g. offscreen bitmap windows) correctly in the BApplication's window quit loop. The loop must not try to quit those. Now things get a bit further. We do reliably trigger the debugger call in BView::SetViewCursor(). The CanvasView destructor which is indirectly called by the window destructor calls this method (indirectly). Apparently R5 does something differently -- either not detaching the view hierarchy before deleting it, or it doesn't have such a debugger call in SetViewCursor(). CU, Ingo