So since this is for the 32bit multilib and the test below doens't seem to
enforce LFS we get a large inode number in the build setup and fail the
stat(), mis-detecting the feature.

libgfortran configure.ac misses AC_SYS_LARGEFILE it seems.


dnl Check whether the st_ino and st_dev stat fields taken together uniquely
dnl identify the file within the system. This is should be true for POSIX
dnl systems; it is known to be false on mingw32.
AC_DEFUN([LIBGFOR_CHECK_WORKING_STAT], [
  AC_CACHE_CHECK([whether the target stat is reliable],
                  libgfor_cv_have_working_stat, [
  AC_RUN_IFELSE([AC_LANG_SOURCE([[
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>

int main ()
{ 
  FILE *f, *g;
  struct stat st1, st2;

  f = fopen ("foo", "w");
  g = fopen ("bar", "w");
  if (stat ("foo", &st1) != 0 || stat ("bar", &st2))
    return 1;
  if (st1.st_dev == st2.st_dev && st1.st_ino == st2.st_ino)
    return 1;
  fclose(f);
  fclose(g);
  return 0; 
}]])], libgfor_cv_have_working_stat=yes, libgfor_cv_have_working_stat=no, [
case "${target}" in
  *mingw*) libgfor_cv_have_working_stat=no ;;
  *) libgfor_cv_have_working_stat=yes;;
esac])])
if test x"$libgfor_cv_have_working_stat" = xyes; then
  AC_DEFINE(HAVE_WORKING_STAT, 1, [Define if target has a reliable stat.])
fi])

Hmm, no, that's there.

The logfile shows

[31010s] checking for special C compiler options needed for large files... no
[31010s] checking for _FILE_OFFSET_BITS value needed for large files... 64
...
[31035s] checking whether the target stat is reliable... no

so the individual test has to use -D_FILE_OFFSET_BITS=64?

That is really weird.
Because the
checking for _FILE_OFFSET_BITS value needed for large files... 64
test should add
#define _FILE_OFFSET_BITS 64
line to confdefs.h and I believe the working stat testing should use that:
  if test "$cross_compiling" = yes; then :

case "${target}" in
  *mingw*) libgfor_cv_have_working_stat=no ;;
  *) libgfor_cv_have_working_stat=yes;;
esac
else
  cat confdefs.h - <<_ACEOF >conftest.$ac_ext
/* end confdefs.h.  */
  
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
  
int main ()
{
  FILE *f, *g;
  struct stat st1, st2;
  
  f = fopen ("foo", "w");
  g = fopen ("bar", "w");
  if (stat ("foo", &st1) != 0 || stat ("bar", &st2))
    return 1;
  if (st1.st_dev == st2.st_dev && st1.st_ino == st2.st_ino)
    return 1;
  fclose(f);
  fclose(g);
  return 0;
}
_ACEOF

The cat should source in all of confdefs.h so far.

Created attachment 58075 [details]
conftest.c

I've hacked up my libgfortran/configure, added there sleep 60s after writing conftest.c for working stat check, and then rm -f config.cache; ./config.status --recheck in the
x86*/32/libgfortran directory and the above conftest.c is what I saw.

So besides maybe no way to reproduce the failure the source still has the issue
described.

(In reply to Richard Biener from comment #7)
> So besides maybe no way to reproduce the failure the source still has the
> issue
> described.

Sorry, not following -- jakub's attached one has #define _FILE_OFFSET_BITS 64, so we should be fine?

(In reply to Sam James from comment #8)
> (In reply to Richard Biener from comment #7)
> > So besides maybe no way to reproduce the failure the source still has the
> > issue
> > described.
> 
> Sorry, not following -- jakub's attached one has #define _FILE_OFFSET_BITS
> 64, so we should be fine?

Iff !HAVE_WORKING_STAT (for whatever reason) but also !defined(__MINGW32__)
then the source will fail to build.  It looks like the key is
!HAVE_WORKING_STAT, not sure why defined(__MINGW32__) is checked here.

Ah, sorry, you meant GCC sources, not configure test source. ack

So the trigger is Y2038, aka missing -D_TIME_BITS=64 - reproducible in a
qemu environment with -rtc base=2038-01-20T00:00:00

Re-confirmed with GCC 15.