https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123366
--- Comment #28 from Mark Millard <markmigm at gmail dot com> ---
(In reply to Rainer Orth from comment #27)
> I'm seeing those SEGVs myself, but only since I upgraded from FreeBSD 14.3 to
> 15.0, and even then only for 11000+ g++ tests where cc1plus SEGVs. The build
> itself works without issues.
>
> . . .
See comment 19 where it is not cc1plus and not posix_fallocate
that was involved but .build/gcc/f951 in linemap_check_ordinary
that SIGSEGV's when linemap_lookup returned a null pointer for
use as an argument, something it is designed to be able to do,
but the code using that linemap_lookup call does not try to
handle. So: Not the same SIGSEGV's.
But there is a lang/gcc16-devel internal configuration problem
as well where it matters if there is a prior installation of a
lang/gcc* vs. not --when there should be no requirement on
FreeBSD to have already installed a lang/gcc* in order to build
lang/gcc16-devel . This happens rather early in the overall
build, long before the gfortran related build activity. It
leads to variations like:
/* Define this macro if mbstowcs does not crash when its first argument is
NULL. */
#ifndef USED_FOR_TARGET
-/* #undef HAVE_WORKING_MBSTOWCS */
+#define HAVE_WORKING_MBSTOWCS 1
#endif
that just should not be occurring: both types of contexts should
have the "+" case. The "-" case is just wrong for the FreeBSD OS
involved in my testing.
Your cc1plus SIGSEGV problem is distinct from both the problems
that I identified for lang/gcc16-devel builds. No cc1plus
SIGSEGV has been observed for the lang/gcc16-devel builds.
