Since the .Ql macro action uses an output prefix of "'`" and an output
suffix of "`'", md_post_raw() would decrement the code_blocks state variable
even though md_pre_raw() had earlier neglected to increment it, hence
leaving the variable in an invalid negative state. That in turn could
result in corrupt output in a variety of ways.
Fix this by checking in md_pre_raw() whether the prefix *contains* a
backtick rather than only checking whether it *starts* with a backtick.
For consistency, apply the same change to md_post_raw() even though
there was no bug in that function: all *suffixes* containing a backtick
actually contain it in the leading position.
Thanks to job@ for reporting this bug. He noticed a particularly nasty
kind of output corruption: having .Ql in an input file would result
in ASCII_NBRSP (0x31) sneaking through into the output stream if later,
unrelated parts of the same input file directly or indirectly used
the \~ escape sequence, for example by using the .Ex macro.
-/* $OpenBSD: mdoc_markdown.c,v 1.36 2021/08/10 12:36:42 schwarze Exp $ */
+/* $OpenBSD: mdoc_markdown.c,v 1.37 2024/08/13 12:43:55 schwarze Exp $ */
/*
* Copyright (c) 2017, 2018, 2020 Ingo Schwarze <schwarze@openbsd.org>
*
if ((prefix = md_act(n->tok)->prefix) != NULL) {
md_rawword(prefix);
outflags &= ~MD_spc;
- if (*prefix == '`')
+ if (strchr(prefix, '`') != NULL)
code_blocks++;
}
return 1;
if ((suffix = md_act(n->tok)->suffix) != NULL) {
outflags &= ~(MD_spc | MD_nl);
md_rawword(suffix);
- if (*suffix == '`')
+ if (strchr(suffix, '`') != NULL)
code_blocks--;
}
}