There will be a delay in pushing this fix.

It seems to have exposed a way for the mom macro package (perhaps) to provoke an assertion failure in the formatter, which should certainly be impossible.

/usr/bin/mkdir -p `dirname contrib/mom/examples/mom-pdf.pdf` \
&& LC_ALL=C GROFF_COMMAND_PREFIX= GROFF_BIN_PATH="/home/branden/src/GIT/groff/build" PDFMOM_BIN_PATH="." /home/branden/src/GIT/groff/build/pdfmom -F/home/branden/src/GIT/groff/build/font -F/home/branden/src/GIT/groff/build/../font -M/home/branden/src/GIT/groff/build/tmac -M/home/branden/src/GIT/groff/build/../tmac -M../contrib/mom -K utf8 -p -e -t ../contrib/mom/examples/mom-pdf.mom >contrib/mom/examples/mom-pdf.pdf
troff: ../src/roff/troff/input.cpp:626: static int input_stack::finish_get(node**): Assertion `level == 0' failed.
groff: error: troff: Aborted (core dumped)

Alex Colomar reported to the groff list:

> here's [a warning] that I've only seen in -Tpdf.  It might be a bug in
> -Tpdf.
>
>       troff:man7/path_resolution.7:135: warning: macro 'and' not defined
> (possibly missing space after 'an')
>       make: *** [share/mk/build/pdf/troff.mk:22:
> .tmp/man/man7/path_resolution.7.pdf.set] Error 1
>       make: *** Deleting file '.tmp/man/man7/path_resolution.7.pdf.set'
>
> Here's a small reproducer:
>
>       alx@debian:~/tmp$ cat dot.man
>       .TH dot 7 2024-02-16 dots-in-SS
>       .SH Foo
>       .SS . and ..
>       \&. and .. are special files.
>       alx@debian:~/tmp$ groff -man -Tutf8 dot.man >/dev/null
>       alx@debian:~/tmp$ groff -man -Tps dot.man >/dev/null
>       alx@debian:~/tmp$ groff -man -Tpdf dot.man >/dev/null
>       troff:dot.man:3: warning: macro 'and' not defined (possibly missing
> space after 'an')

I can reproduce this with groff 1.23.0 and Git HEAD.

I believe the problem happens because this `nop` request is taking place inside a diversion (a "box" diversion, to be precise).  Consequently when the diversion is later interpolated, it can include macro calls or requests, so the formatter recognizes the control character at the beginning of the (interpolated) input line.  This causes the PDF bookmarks that are automatically constructed from the (sub)section heading texts to be susceptible to misinterpretation.  If you check the PDF generated from your reproducer, you will see that the corresponding bookmark for the subsection is absent.

The solution is to prevent a potential control character that starts the (sub)section heading text from appearing at the beginning of the (interpolated) input line.

Thus we need a dummy character prefix.

Proposed fix:

diff --git a/tmac/pdf.tmac b/tmac/pdf.tmac
index 54175434f..fe12d35da 100644
--- a/tmac/pdf.tmac
+++ b/tmac/pdf.tmac
@@ -253,7 +253,7 @@ .de pdfbookmark
 .         tr \[em]-
 .         nf
 .         box pdf:clean
-.         nop \\$*
+.         nop \&\\$*
 .         fl
 .         box
 .         chop pdf:clean