Does "-f tex --tex" influence char selection?

[Juergen Reuter]

Hi!

One of the most outstanding bugs in ancient notation which I am still 
trying to fix is that sometimes wrong characters are chosen within 
mensural ligatures.  However, on my machine this bug occurs only when 
building the documentation, but not when running lilypond manually from 
the command line.  This lets me assume that this bug may be related to 
some TeX font encoding problem rather than to the ancient notation code 
itself.

Specifically, I am talking about the first figure in section 5.17.9.1 of 
the manual (created from 
Documentation/user/out-www/lilypond/lily-833201559.ly, as of current CVS).  
When building the docu on my machine, I get the same wrong figure as on 
the lilypond.org website for both, html and pdf.  However, when I fetch 
lily-833201559.ly, comment out the "set! toplevel-score-handler" line 
(such that the example runs from the command line), and run lilypond upon 
it, that I get the expected result (see attachment).

So my questions are:

* Could someone else verify my observations by running lilypond from 
  the command line directly on lily-833201559.ly (with the 
  "toplevel-score-handler" line commented out or removed) and compare the 
  result with the figure in section 5.17.9.1 of the manual and/or with the 
  attachment below? (Yes, I know, there are a couple of "Infinity or NaN 
  encountered" warnings, but these warnings should be unrelated.)

* What are the major differences between the way the docu is built and 
  running lilypond from the command line without further options?  I can 
  see that the "-f tex" and "--tex" options are used when building the 
  docu; maybe this is the source of the error?

TIA,
Jürgen

mens-test.pdf
Description: Adobe PDF document