One of the nasty bits about getting backing store's Always mode working again is the asymmetry between map and unmap, namely that sometimes pWin->valdata is a struct and sometimes a magic out-of-band value indicating unmap. This would be fairly easy to clean up except for the overlay visual code in mi, which has to maintain a shadow clip list so it can repaint transparency keys. Unfortunately the only extant user of mioverlay is the nvidia driver on workstation Quadros, so this is a little hard to test. (fglrx also has overlay visual support, but they don't appear to be using mioverlay to accomplish it.)
Also nobody has materially touched that code in something like 12 years, and it's not what you would call "commented", or "obvious". This (compile-tested only) series attempts to eliminate mioverlay's internal use of UnmapValData. It probably makes x11perf's valtree benchmarks marginally faster since we're no longer passing around dead variables, and it should be functionally equivalent. - ajax _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
