On Sat, May 07, 2005 at 08:42:39PM +0200, Andre Poenitz wrote: > On Sat, May 07, 2005 at 04:49:53PM +0200, Helge Hafting wrote: > > On Fri, May 06, 2005 at 08:19:06PM +0200, Andre Poenitz wrote:
> > > I understand (now) that update() might be called recursively (again). > > > However, this does not serve any purpose but rather should be considered > > > a bug. > > > > I agree. A user typing faster than the UI doesn't want to see every drawing > > operation happen, with lyx lagging more and more behind. Just skipping > > any "stale" drawing operation and do the last one only should be better. > > > > Coarse screen updates when the cpu can't keep up is much better than > > detailed > > but lagging updates. > > And our real problem is that our updates do not serve only the purpose > of showing something on screen, but also to put the coordcache in some > sensible state. This can't be skipped if we expect e.g. <down> to > work even while typing fast. > I see. The coord cache must be updated, but I hope the actual drawing can be skipped ï when lyx knows that it is behind. That improves performance in many cases, such as an overloaded machine or X over a slow network. Helge Hafting