Re: Emacs 22.2.90 pretest

[Chong Yidong]

Eli Zaretskii <address@hidden> writes:

> I notice that in this pretest, moving point with C-f and C-b or
> inserting a single character is very sluggish: e.g., if I continuously
> press C-f, Emacs cannot keep up(!), although this is a 3.2 GHz
> machine.  It almost feels like working on a remote machine.
>
> This recent change seems to be a likely suspect:
>
>    2008-07-28  Chong Yidong  <address@hidden>
>
>          * textmodes/flyspell.el (flyspell-word, flyspell-large-region)
>          (flyspell-region): Call ispell-maybe-find-aspell-dictionaries.
>
> It seems that its effect is to call ispell-maybe-find-aspell-dictionaries
> on every editing command, which is silly, IMO.  Even if we do need to do
> that on every command (and I'd like to hear a reason why), the call
> should only be made if Ispell is actually Aspell, which in my case it
> isn't.  On top of that, ispell-maybe-find-aspell-dictionaries
> obviously was not designed to be called frequently: it calls
> ispell-check-version, which is expensive and is supposed to be run
> once in an Ispell session (it makes unnecessary destructive changes to
> the Ispell buffer, invokes another Ispell process, etc.).
>
> Why was this change made?

This was bug#232:

http://emacsbugs.donarmstrong.com/cgi-bin/bugreport.cgi?bug=232

Does undoing this change make the problem go away?  If so, it's probably
better to revert it.  (I attempted to backport a change from the trunk
to the branch, but there were other changes to the trunk that apparently
makes the dictionary initialization less expensive.)