Unowned directories

Michael Schwendt mschwendt at gmail.com
Sat Nov 29 09:21:45 UTC 2008


On Sat, 29 Nov 2008 01:23:53 -0500, Ben Boeckel wrote:

> Hi,
> 
> I did the following command to find unowned directories (edited from my actual 
> command to get rid of stuff like tmp that came up with lots of false-positives—
> though it is quite possible there still are some here):
> 
> # rpm -q --whatprovides `find /{{,s}bin,boot,dev,etc,lib{,64},usr} -type d` | 
> grep -v -e .x86_64 -e noarch -e i386 -e i686 -e ppc
> 
> There were too many to consider doing a bug report for each,

Doesn't surprise me. The list has gotten longer. A lot. Reviewers don't
look for such issues, because reviewers themselves are packagers, and many
packagers either don't know or don't care about unowned dirs.

Some in your list most likely are false positives, however:

  $ rpm -qf /etc/selinux/targeted/contexts
  selinux-policy-targeted-3.3.1-107.fc9.noarch

My last check of unowned dirs didn't flag that one (note that I examine
the repo metadata under consideration of basic depsolving, but not
including any core group yet).

Other findings are false positives because a directory itself _is_ owned
by a package, but during package upgrade/removal RPM refuses to remove
the directory if it isn't empty:

> file /usr/lib64/firefox-3.0.2 is not owned by any package

That happens if (1) other packages store files in it without owning the
directory or without requiring a package that owns it, or (2) files in it
are created at run-time and don't belong into any package. Advanced packages
mark any such files as %ghost.

> emails to fedora-devel-announce and CC maintainers of suspected owners (looking 
> at owners of files in the directory, sibling files)?

Will give mixed results. Some package owners are very good at ignoring
private mail, because they insist on receiving problem reports in bugzilla.
Others cannot handle all their bz tickets. Also be prepared that some simply
don't understand the problem or refuse to think about it.

It would be good if we could use bugzilla for tracking these issue (and
not be disturbed by triagers), because recently I've simply gone into
cvs to fix aging issues myself in the devel tree.

Anyway, quite a lot are valid findings I think, and I believe there are
more, because my list seems longer+different. ;) I compared some with
my list, not limited to:

> file /etc/hotplug.d is not owned by any package
> file /etc/hotplug.d/usb is not owned by any package

=> gpsd-2.37-2.fc9.i386 (rawhide-development-i386)
/etc/hotplug.d
/etc/hotplug.d/usb

> file /usr/share/doc/jetty-5.1.14 is not owned by any package
> file /var/lib/jetty is not owned by any package

=> jetty-5.1.14-1.6.fc10.i386 (rawhide-development-i386)
/var/cache/jetty
/var/lib/jetty
/etc/logrotate.d
/usr/share/doc/jetty-5.1.14

=> jetty-javadoc-5.1.14-1.6.fc10.i386 (rawhide-development-i386)
/var/lib/jetty
/var/lib/jetty/webapps




More information about the devel mailing list