On 02/06/2026 15:54, Bruno Haible via GNU coreutils General Discussion wrote:
The CI, this week, reports a test failure:
FAIL: tests/install/stdin.sh
on macOS and FreeBSD. Find attached the log from macOS.
The issue seems to be that copy.c tries to correlate
the device and inodes from stat() and open()+fstat().
On BSD the former gives the devfs generated inode value,
whereas the latter will resolve differently and give the fdesc value,
and copy_reg() bails out due to the difference.
We could avoid this in the test at least like:
test $(src/stat -L -c%i /dev/stdin) = \
$(src/stat -L -c%i - </dev/stdin) ||
skip_ '/dev/stdin correlation mismatch'
Note this is kind of like https://bugs.gnu.org/35713
and DEV_FD_MIGHT_BE_CHR is set on FreeBSD,
so perhaps we could do open()+fstat() in follow_fstatat()?
cheers,
Padraig
