Re: Build failures of coreutils-8.1 in Fedora build system "koji"

[Eric Blake]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

According to Ondřej Vašík on 11/24/2009 4:58 AM:
> utimensat(AT_FDCWD, "file", NULL, AT_SYMLINK_NOFOLLOW) = -1 ENOSYS (Function 
> not implemented)
> 
> Those two failing tests (touch/no-dereference and touch/trailing-slash) seems 
> to be new in coreutils test-suite (added in 8.1), so the issue was likely 
> there even before 8.1 and now shows in koji...

It looks like those are cases of an old kernel that can't set timestamps
on symlinks.  Does the kernel support lutimes?  The way the coreutils
tests are written is that they assume that if utimensat exists, then
touching symlink timestamps should be supported.  But obviously this is
not the case, when compiling against newer headers (utimensat exists) but
older kernel (AT_SYMLINK_NOFOLLOW not implemented).  touch/no-dereference
tries to skip the test if utimensat fails, but it is only skipping based
on the presence of utimensat; help in detecting ENOSYS to skip the test
would be appreciated.  touch/trailing-slash, on the other hand, should
work regardless of whether AT_SYMLINK_NOFOLLOW is implemented, since it
does not affect symlink timestamps.  So that seems to point to a bug in
gnulib's utimens.c.

Can you also run 'make -C gnulib-tests check' and see which gnulib unit
tests fail?  If test-utimens at also fails, the tail of the strace would
be nice; if it passes, then the test is not adequately exercising the
lutimens workaround for missing kernel support in AT_SYMLINK_NOFOLLOW.

- --
Don't work too hard, make some time for fun as well!

Eric Blake             address@hidden
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Public key at home.comcast.net/~ericblake/eblake.gpg
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAksL2ckACgkQ84KuGfSFAYAjRwCeMFnuBm6vyQmoRNY0qVEkrLAn
bR4AoMKY0/lyxEFQgwkQLOt4GI/8vS2Q
=nQDA
-----END PGP SIGNATURE-----