https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123366
--- Comment #27 from Rainer Orth <ro at gcc dot gnu.org> ---
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.
Looking at the failing invocation with truss, I see right before the end
69427:
openat(AT_FDCWD,"gcm.cache/has_internal_linkage-counter.gcm~",O_RDWR|O_CR
EAT|O_TRUNC|O_CLOEXEC,0666) = 4 (0x4)
69427: posix_fallocate(0x4,0x0,0x2000) ERR#45 'Operation not
supported
'
69427: posix_fallocate(0x4,0x0,0x1000) ERR#45 'Operation not
supported
'
69427: posix_fallocate(0x4,0x0,0x2000) ERR#45 'Operation not
supported
'
69427: posix_fallocate(0x4,0x0,0x1000) ERR#45 'Operation not
supported
'
69427: posix_fallocate(0x4,0x0,0x2000) ERR#45 'Operation not
supported
'
69427: posix_fallocate(0x4,0x0,0x1000) ERR#45 'Operation not
supported
'
69427: SIGNAL 11 (SIGSEGV) code=SEGV_MAPERR trapno=12 addr=0x10
69427: sigprocmask(SIG_SETMASK,{ SIGSEGV },0x0) = 0 (0x0)
ERR#45 isn't even documented as a possible error code for posix_fallocate ;-(
I wonder what filesystem type you are running your builds on?
In my case, it's zfs. I'm starting my bootstrap with a self-compiled
gcc 11.4.0 (which includes gdc).
Rather than comparing build logs and auto-host.h files, I think it would be way
more effective to investigate the SEGVs themselves to get a direct
understanding
of the failure.
