I've rewritten the event discarding code more or less from scratch, and hopefully in a less dangerous way. With my testing application fvwm stays barely usable when the application generates 20.000.000 ...NAME... PropertyNotify events (machine with four cores). The X server blocks one cpu at 100%, fvwm uses 50% to 60% on another core, and the application uses about 12% on a third. This looks pretty good to me - as long as fvwm is twice as fast as the X server it cannot be the bottleneck.
Do you notice any change in performance when starting maple? Is fvwm more responsive or less? (The first time maple freaks out responsiveness may be not very good; I think that is caused by the X server or Xlib allocating lots of memory for the events; on a second go fvwm stays usable for me.) The code ist still hardly tested and experimental and may cause any number of bugs. Ciao Dominik ^_^ ^_^ -- Dominik Vogt
