At 2026-01-30T17:08:25+0100, Morten Bo Johansen wrote: > No, there has indeed been a regression between (in my case) > > GNU roff version 1.24.0.rc1.4-3a8f-modified > > and > > GNU roff version 1.24.0.rc2.26-48ff-modified
Why and how is your tree modified?
> Even if I enter the correct path to the urw fonts with the
> "--with-urw-fonts-dir=/home/mojo/.fonts" configure option, which is
> where my urw fonts are, this is ignored by ./configure and the path to
> the urw fonts in the configuration summary is set to the incorrect
> /usr/share/ghostscript/Resource/Font
That certainly sounds like incorrect behavior, but I also can't account
for it. My intent in the `GROFF_URW_FONTS_PATH` and
`GROFF_URW_FONTS_CHECK` macros in "m4/groff.m4" is to have that
"withval" defeat the search completely. This much is _not_ relatively
new code. Except for the `if test "$urwfontsupport" != no` guard, all
of the following dates back to 2022-2023. Except for the very place
your fonts were incorrectly found, added in October 2025 because of some
distributor (Alpine?). But that exception shouldn't matter because this
is a just a directory name inside a string whose value gets completely
clobbered by your `--with-urw-fonts-dir` selection.
That design is deliberate; if you specify `--with-urw-fonts-dir`, the
Autoconf macro `GROFF_URW_FONTS_CHECK` looks only in that one place for
fonts, and nowhere else, because it assumes you know what you mean.
That this is failing for you is inexplicable to me.
Maybe someone can spot my mistake. People can ignore the shouty
Autoconf macro names; if there's a logic error here, it's in POSIX
shell (or so I'd wager).
---snip---
AC_DEFUN([GROFF_URW_FONTS_PATH], [
AC_ARG_WITH([urw-fonts-dir],
[AS_HELP_STRING([--with-urw-fonts-dir=DIR],
[search for URW PostScript Type 1 fonts in DIR])],
[urwfontsdir="$withval"])
])
AC_DEFUN([GROFF_URW_FONTS_CHECK], [
AC_REQUIRE([GROFF_URW_FONTS_SUPPORT])
AC_REQUIRE([GROFF_URW_FONTS_PATH])
AC_REQUIRE([GROFF_GHOSTSCRIPT_PATH])
groff_have_urw_fonts=no
urwfontsdir=
if test "$urwfontsupport" != no
then
AC_MSG_CHECKING([for URW fonts in Type 1/PFB format])
dnl Keep this list in sync with font/devpdf/Foundry.in.
_list_paths="\
/usr/share/fonts/type1/gsfonts/ \
/usr/share/fonts/default/Type1/ \
/usr/share/fonts/default/Type1/adobestd35/ \
/usr/share/fonts/type1/urw-base35/ \
/usr/share/fonts/urw-base35 \
/usr/share/ghostscript/Resource/Font \
/opt/local/share/fonts/urw-fonts/ \
/usr/local/share/fonts/ghostscript/"
if test -n "$urwfontsdir"
then
_list_paths="$urwfontsdir"
fi
dnl Keep this list of font file names in sync with the corresponding
dnl entry in font/devpdf/util/BuildFoundries.pl.
for k in $_list_paths
do
for _font_file in \
URWGothic-Book \
URWGothic-Book.t1 \
URWGothic-Book.pfb \
URWGothicL-Book.pfb \
a010013l.pfb
do
if test -f $k/$_font_file
then
AC_MSG_RESULT([found in $k])
groff_have_urw_fonts=yes
urwfontsdir=$k
break 2
fi
done
done
if test $groff_have_urw_fonts = no
then
AC_MSG_RESULT([none found])
fi
fi
AC_SUBST([groff_have_urw_fonts])
AC_SUBST(urwfontsdir)
])
---end snip---
Observe that the test is crude; it checks only for the Bookman roman
font, under 5 different file names, and if that check fails, it decides
that no URW fonts are present at all. In principle, that could screw
you up.
> running autoreconf -fvi does not solve it. Running ./bootstrap gives
> me this fatal error:
>
> fatal: Unable to checkout 'dc9ad4ec31fb09d247036f6a031e4ec6efbaa26c' in
> submodule path 'gnulib'
Not sure this explains it, but for me, at least one of the FSF's Git
mirrors is stale by about two weeks. That's why I quoted the code above
instead of linking to https://cgit.git.savannah.gnu.org/cgit/groff.git/.
I wanted to offer a URL hitched to the "1.24.0.rc2" tag, but I can't,
because the mirror my Firefox session is stuck on doesn't even have the
rc*1* tag!
Also, restarting the browser doesn't help. Maybe there's an idiotic
caching proxy between me and gnu.org. And maybe for you too.
Try the rc2 archive file at <https://alpha.gnu.org/gnu/groff/>.
Regards,
Branden
signature.asc
Description: PGP signature
