> But I still mistrust things I can't repair with a text editor.
:-)
> According to Adobe, CFF fonts are "based on all of the same concepts
> for drawing paths and hinting" as Type 1 fonts, and "a Type 1 font
> can be converted into CFF/Type2 format, and back to Type 1 again,
> without any loss of quality".
Yeah, while correct for most fonts, this is not true in general.
Type 1 -> Type 2 charstrings (as used in CFF fonts) always works, but
Type 2 charstrings have some extensions like the `random' operator
which isn't supported in Type 1 fonts. Well, othersubr #28 returns a
random number (no idea where exactly this has been defined), but only
othersubr #0-#4 are defined for Type 1.
> With this in mind, I've attached a shell script that represents my
> "fix" of the lilypond file import issue. It works by completely
> removing the embedded fonts from the lilypond-generated Postscript
> file and replacing them by corresponding %%IncludeResource comments,
> which will later be resolved by grops's resource manager facility.
There is another Lilypond option to let the fonts be loaded by
GhostScript directly:
lilypond --ps -dbackend=eps -dgs-load-fonts ...
This uses GhostScript's `.loadfont' operator, making the section look
like this on my GNU/Linux box:
%%BeginProlog
/lilypond-datadir where
{pop}
{userdict
/lilypond-datadir (/usr/local/share/lilypond/2.17.11) put}
ifelse
%%BeginResource: font Emmentaler-11
lilypond-datadir
(/fonts/otf/emmentaler-11.otf)
concatstrings
(r)
file
.loadfont
%%EndResource
A nice side effect is that the EPS files become much smaller, but of
course they are no longer self-contained. Again, grops understands
this just fine, but you *must* use GhostScript afterwards to further
process the output file.
> * Lilypond outputs a space character at the top left. Since I have
> no experience with lilypond I don't know what generates this, and
> my solution is to simply edit it out.
IIRC, this is a bug which I haven't been able to catch yet (mostly due
to time constraints).
> * The rest of the script gets rid of the auto-positioning
> and calls ghostscript to find the bounding box. Make sure
> ghostscript also has access to the fonts.
Using -dbackend=eps, you don't need this since a bounding box gets
emitted.
Werner
