Here is the answer to my question from Juliusz Chroboczek:
ANK> I recently discovered that using your font server and/or
ANK> xfsft-based TrueType rasteriser with Lesstif causes Lesstif to
ANK> dump core. As it has turned aout, the problem is caused by the
ANK> fact, that TrueType fonts have min_bounds.width =
ANK> -max_bounds.width in XFontStruct.
This is correct.
ANK> Lesstif calculates average character width in font using formula
ANK> (min_width+max_width)/2 which obviously produces 0 in result.
This is not.
ANK> Is it legal to have characters with negative width in
ANK> left-to-right fonts?
Yes, unfortunately. I believe that this is an error in the X11
specifications, and most clients do not know how to deal with
negative-width glyphs (this is the case at least of Emacs/Mule and the
``Arabox'' Arabic version of Motif). I am thinking of making all
advance widths positive in a future version of XFree86.
ANK> If that is legal, what is the better way to calculate avarage
ANK> character width for font?
The safe way is to use the average width font property if present (but
even this might be zero for mixed-direction fonts), and to use the
width of `x' if not available; if the font contains no `x', use half
the maximum width. Please read the XLFD document, which you should
find in the directory `xc/doc/hardcopy' of your X11 distribution (make
sure you have the X11R6 version).
Note, however, that there is no good language-independent way of
determining the average width. Any method you use is an
approximation, and therefore the use of average widths should be
avoided whenever possible.
Sincerely,
J.
