On Sat, 2016-09-24 at 05:09 +0300, Keith Packard wrote: > > Adam Jackson <a...@nwnk.net> writes: > > > You want to make this non-static, so in the xinerama case you can do > > this translation once at the top level. > > > ... > > > You're not correcting stuff->coordMode in place, which means under > > xinerama you'll mutate geometry once per screen, which ain't right. > > > These seem in conflict to me -- if we have xinerama do the transform > once and smash stuff->coordMode, it seems like we're done, right?
Not really in conflict, just both true. If PanoramiXPolyPoint does the transform and smashes coordMode, then ProcPolyPoint wouldn't need to _also_ do it when xinerama is active. But it does need to do the fixup for the (normal) case where xinerama isn't active. (Please don't key off !noPanoramiXExtension, I'd like to be able to meaningfully test xinerama with just one ScreenRec.) Strictly you could fix up just once in ProcPolyPoint if you also fixed up coordMode, but doing the fixup in xin would get us closer to being able to do thread-per-GPU for xinerama, which might be nice. - ajax _______________________________________________ xorg-devel@lists.x.org: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: https://lists.x.org/mailman/listinfo/xorg-devel