>And the package manifest is likely including /usr (which is perhaps
>unnecessary).
speaking of which, do directories installed into now have to
be owned by some package? this seems like a recent change,
LSB has a somewhat strange testbed where a minimal base LSB
conforming system is built, but is not managed by rpm.
It is, however, used to test the installation of LSB-conforming
packages and since we uplifted to 4.4.6 those packages that
/don't/ have package-specific directories in the manifest are
getting dependency failures on those directories. We can
easily solve this if that's the case, of course.
Rpm-list mailing list
Rpm-list (AT) redhat (DOT) com

Aug 21, 2006, at 8:18 PM, Wichmann, Mats D wrote:

>
>And the package manifest is likely including /usr (which is perhaps
>unnecessary).
>
speaking of which, do directories installed into now have to
be owned by some package? this seems like a recent change,
LSB has a somewhat strange testbed where a minimal base LSB
conforming system is built, but is not managed by rpm.
It is, however, used to test the installation of LSB-conforming
packages and since we uplifted to 4.4.6 those packages that
/don't/ have package-specific directories in the manifest are
getting dependency failures on those directories. We can
easily solve this if that's the case, of course.

Yes. rpm-4.4.6 and later creates a dependency for every parent
directory.
Another dependency is generated for the target of every symlink.

The rationale for adding is to insure that some package sets perms,
owners, and SELinux contexts from packaging, rather than having the
rpm implementation attempt to guess (root.root 0755 or 0644) what
was intended.

Add missing directory paths to /etc/rpm/sysinfo until vendors get their
packaging fixed.

The better fix is to identify a package to own the directories, and/
or to supply
targets for symlinks.

73 de Jeff

Rpm-list mailing list
Rpm-list (AT) redhat (DOT) com