Lars Gullik Bjønnes wrote: > But what is the delay caused by. Not everyone sees it. > Is it the sheer number of fonts, or is it that some fonts are missing?
I think it is the sheer number of fonts. It is at least that people with lots of fonts see the delay, while others with less fonts (like Jean-Marc) don't see it. I have 17383 fonts according to xfontsel (and I have a pretty vanilla SuSE 10 installation) and also have a remarkable delay, not as long as Helge, but he has more fonts. > | I also thought about invoking initMath() at the end of the startup, > | but in the > | background (asynchronously), which would probably be the most elegant > | way. I don't know if this is already possible, though, and how. > | Furthermore, we could of course try to speed up getSymbolFont :-) > | > | > | What do you think? > > For 1.4 only easy fixes that we understand why works. I couldn't agree more. The patch I proposed basically goes back to 1.3 behavior, as far as this is possible. I think I understand it well enough, but more testing should be done nonetheless, especially from people that actually use math. I still think the background initialization is a good idea, but it can certainly also wait. Jürgen
