I've got two good news for you:
1) I was finally able to reproduce your issue on one of my machines
2) I think I found the culprit
Ad 1) I didn't realize that I still had another font installed, the
Opentype Cantarell font. I was so focussed on Truetype fonts and
gsfonts-x11 that I simply overlooked it. When I removed this one, I've
got the exact same symptoms that you described. Since this sounded a
bit too random, I suspected the culprit to lie in our fontconfig query.
Ad 2) In your cups-filters source copy, in the filters/texttopdf.c
file, there is line 62 which calls FcPatternAddInteger() and also has
a comment that this will "guide fc". However, this seems to have
puzzled the fontconfig pattern, instead.
Please comment out the line and rebuild the package following the
procedure I described in one of my previous mails. If you feel
unconfortable editing the file manually (you should), you might as
well apply the patch I attached to this email (again with "quilt
import /path/to/patch" and then "quilt push -a"). Please rebuild the
package and try again.
If this really turns out to be the culprit, I have to check with the
fontconfig maintainers what's wrong about this command. In any case,
it did not achieve what it was supposed to do, because most of the
fonts in the list suggested by fontconfig and printed out by my
fc-test program were non-monospaced, anyway.
- Fabian
--- cups-filters-1.0.16.orig/filter/texttopdf.c
+++ cups-filters-1.0.16/filter/texttopdf.c
@@ -59,7 +59,7 @@ EMB_PARAMS *font_load(const char *font)
} else {
FcInit ();
pattern = FcNameParse ((const FcChar8 *)font);
- FcPatternAddInteger (pattern, FC_SPACING, FC_MONO); // guide fc, in case substitution becomes necessary
+/* FcPatternAddInteger (pattern, FC_SPACING, FC_MONO); // Debian #670055 */
FcConfigSubstitute (0, pattern, FcMatchPattern);
FcDefaultSubstitute (pattern);
