Re: texinfo-4.13.93 pretest available

[Patrice Dumas]

On Sat, Jan 05, 2013 at 10:11:30AM +0200, Eli Zaretskii wrote:
> 
> The only problem left is that the Windows batch files have a '+'
> character at the beginning of every line, probably some snafu while
> applying the patch I sent to introduce those files.  Please remove
> those '+' characters, as they will prevent the batch files from
> working.

Done, thanks for the report.

> > If this version goes ok, I'll make one more pretest with wider
> > announcements before the real release.
> 
> Should we do anything about the speed of the new 'makeinfo'?
> According to my measurements, it is about 18 times slower than the C
> implementation from 4.13a.  (I timed it on a Core i7 machine with a
> warm cache producing the Emacs Lisp manual: the 4.13.93 implementation
> took 27 seconds, while the C implementation from 4.13a took just 1.5
> sec.)  I'm afraid that such slowdown will trigger user complains.

That will happen for sure.  But there is not much we can do in the short
term.  Even in the more longer term, I still think that the priority is
still some features.  (Mostly adding macros, include files and other
similar stuff to the tree; refining the HTML customization API;
translation to LaTeX; handling the lilypond manual).  A first pass at a 
profiler didn't show anything obvious.  After makeinfo is stabilized, it
would certainly be relevant to try to improve the speed, maybe
rewritting the time limiting parts in C.  At least this could be
possible to rewrite and optimie the parser core, but I have no idea what
may be done for the converters.

-- 
Pat