RE: Slow compile on Windows XP - font caching problem?

[Trevor Daniels]

Han-Wen, you wrote 17 March 2008 15:18
>
> 2008/3/13, Trevor Daniels <address@hidden>:
> >
> >  OK - here are the results:
> >
> >  2.11.34-1
> >   Compiles in 15 secs
> >   Uses .fontconfig directory
> >
> >  2.11.35-1
> >   Fails with libguile2-17.dll not found
> >     Copied this in, then
> >       Fails with error about ice-9/boot-9.scm 
> not in load
> >  path
> >
> >  2.11.35-2
> >   Compiles in 15 secs, but
> >     Lots of error messages like
> >       Pango CRITICAL
> >       FT_Get_Glyph_Name() error invalid argument
> >       Glyph has no name
> >       Skipping Glyph U+FFFFFFFF ... 
> Century-Schl-Roma.otf
> >   Uses .fontconfig directory
> >
> >  2.11.36-1, 37-1, 41-1, 42-1
> >   All compile slowly - 65 secs
> >   All use .lilypond-fonts directory
> 
> 
> Dang, this happened after a big refactoring of 
> our build system.
> Is there anything interesting (eg. by way of diff 
> -r) of /program
> files/lilypond/usr/etc/ directory between .34 and .36 ?
> 

Not really.  A couple of minor changes in 
fonts/fonts.conf but nothing that seems
significant.

However, I noticed that usr/bin/libfontconfig-1.dll
changed between 34-1 and 35-1, and again between
35-2 and 36.1, looking at the module size.  

Moving usr/bin/libfontconfig-1.dll from 34-1 into
42-1 seems to have no adverse effects on a couple 
of simple tests and fixes the long delay.  Maybe 
this is a work-around for the problem?  But maybe
this introduces other errors.  What should I use 
as a definitive test?

An alternative is to build the cache, then delete
or rename etc/fonts/font.config.  This causes
font-building to abort early with an error message,
which also bypasses the timing problem.  This also 
seems to have no adverse effect on a couple of 
simple tests, but I don't use any unusual fonts. 

Trevor D

> Han-Wen Nienhuys - address@hidden - 

Trevor D