I traced the problem to wrong coordinates in a XShapeCombineShape... code refactoring missed one obscure corner case... sigh...
All better now... On Nov 27, 2008, at 23:06, Jeremy Huddleston wrote: > > > On Nov 27, 2008, at 21:57, George Peter Staplin wrote: >> >> That sounds odd. I ran into a bug like that in Whim with skype. It >> turned out to be that the offsets in Whim weren't taking into >> account the titlebar size in some cases, so the sent event didn't >> match where the application's window really was. Borders can also >> cause that in some cases. > > It's not actually the time that determines if the eye mask is in the > right place, it's the initial placement of the window... it's as if > its placing the mask at a location in the window using root-relative > coordinates when it should be using window-relative coordinates... > > I am stumped... gonna look at all the x-protocall calls in quartz-wm/ > x- > window.m and do some debug spew on them and compare to the same spew > in the old working version... hopefully that'll reveal the problem... > > I tried looking in xeyes to see how it treats the two layers > differently, and it looks like it treats them exactly the same... =/ > >> Another bug I ran into that is somewhat like that had to do with >> windows jumping when they were initially resized. That was caused >> by resize increments/hints. I learned that you have to apply the >> hints, then resize the window, and update the state carefully. Thus >> the code looks something like this: > > Yeah, it's not happening on resize though... > >> On Nov 27, 2008, at 13:51, Jeremy Huddleston wrote: >> >>> Someone discovered an odd bug that our wm exhibits with xeyes. The >>> "black-lines" are drawn at the right spot, but the ovals that select >>> which region to view move down as quartz-wm starts and stops. >>> ConfigureNotify is always returning the upper left of the "inner" >>> window, and we are doing the correct reparenting offset... anyone >>> seen something like this before of have some guesses for me to >>> persue? >>> >>> I put together an animation of this happening here: >>> >>> http://people.freedesktop.org/~jeremyhu/xeyesbug/xeyes.mov >>> >>> If you don't like .mov, you can see the individual frames: >>> http://people.freedesktop.org/~jeremyhu/xeyesbug >>> >>> Thanks, >>> Jeremy >>> >>> _______________________________________________ >>> xorg mailing list >>> [email protected] >>> http://lists.freedesktop.org/mailman/listinfo/xorg >> > _______________________________________________ > xorg mailing list > [email protected] > http://lists.freedesktop.org/mailman/listinfo/xorg _______________________________________________ xorg mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/xorg
