Re: chown --no-dereference broken in coreutils 6.3 ?

[Theodoros V. Kalamatianos]

On Tue, 3 Oct 2006, Jim Meyering wrote:

> "Theodoros V. Kalamatianos" <address@hidden> wrote:
> > I just downloaded coreutils-6.3 and the testsuite fails in chown/basic:
> ...
> > make[1]: Leaving directory `/usr/src/RPM/BUILD/coreutils-6.3/tests/chown'
> ...
> > ++ ls -n slink
> > + set _ lrwxrwxrwx 1 0 0 1 Oct 3 12:08 slink '->' f
> > + shift
> > + test 0:0 = 0:1
> > + fail=1
>
> Thank you for reporting that.
> This is not a bug in the coreutils.
> Rather, you may be using a buggy version of glibc.
> I reported a month ago that the underlying problem
> affects Ubuntu/edgy's version of glibc 2.4:

Yet another glibc bug I stumbled on... seems to be happening a lot 
lately...

>  Subject: fchownat ignores AT_SYMLINK_NOFOLLOW flag
>    https://lists.ubuntu.com/archives/ubuntu-users/2006-September/093218.html
>    https://launchpad.net/distros/ubuntu/+source/glibc/+bug/59083
>      Status: Unconfirmed
>      Importance: Undecided(!)
>
> > I verified by hand that chown --no-dereference 6.3 does not work, while
> > the one from coreutils-5.97 is fine. My system is a linux-2.6.17 system
> > with glibc-2.4.
>
> Please specify _which_ glibc-2.4 and linux-2.6.17, if you didn't
> build them from pristine sources.

Both were built from vanilla sources. glibc was slightly patched to fix 
some nscd issues and an irritating gconv bug - nothing that would affect 
one of the chown functions AFAICT.

> FYI, this doesn't affect upstream glibc, since the tests have passed
> on rawhide with both 2.6.17 and 2.6.18.

So glibc-2.5 does not have this issue ? If so then I'll try updating to 
that one...


Thanks for the fast reply,

Theodoros Kalamatianos