Spec URL: http://people.redhat.com/rmccabe/conga/ricci.spec SRPM URL: http://people.redhat.com/rmccabe/conga/ricci-0.13.0-1.src.rpm Description: ricci is the agent that runs on locally on systems managed by conga. Conga is in RHEL4 and RHEL5, and more information about the project is at http://sourceware.org/cluster/conga/
Taking ownership to work on review.
rpmlint results: ricci.i386: E: non-standard-uid /var/lib/ricci/certs/cacert.config ricci ricci.i386: E: non-standard-gid /var/lib/ricci/certs/cacert.config ricci ricci.i386: E: non-standard-uid /var/lib/ricci/certs ricci ricci.i386: E: non-standard-gid /var/lib/ricci/certs ricci ricci.i386: E: non-standard-gid /usr/libexec/ricci ricci ricci.i386: E: non-standard-uid /var/lib/ricci/queue ricci ricci.i386: E: non-standard-gid /var/lib/ricci/queue ricci ricci.i386: E: non-standard-uid /var/lib/ricci/certs/clients ricci ricci.i386: E: non-standard-gid /var/lib/ricci/certs/clients ricci ricci.i386: E: non-standard-uid /var/lib/ricci ricci ricci.i386: E: non-standard-gid /var/lib/ricci ricci ricci.i386: E: non-standard-dir-perm /var/lib/ricci 0700 ricci.i386: E: non-standard-gid /usr/libexec/ricci/ricci-worker ricci ricci.i386: E: explicit-lib-dependency libcap ricci.i386: E: explicit-lib-dependency libxml2 The non-standard uid/gid errors are ok because a ricci user is created for the daemon to run under. The non-standard-dir-perm error is ok because only the ricci user should have access to that directory. I've fixed the explicit-lib-dependency errors in my patch that I'm attaching (as well as a minor fix to the BuildRoot location.
Created attachment 299231 [details] Patch to bring ricci specfile into line with Fedora requirements.
Ryan, if you make those changes and then post the locations of the fixed .spec & src.rpm I can approve this package.
Thanks. Made those changes and posted an updated src.rpm and spec file. Spec URL: http://people.redhat.com/rmccabe/conga/ricci.spec SRPM URL: http://people.redhat.com/rmccabe/conga/ricci-0.13.0-2.src.rpm
Looks good, package is ready to go.
It's still blocked by bug #433679 (clustermon review), as ricci depends on clustermon.
New Package CVS Request ======================= Package Name: ricci Short Description: Cluster and systems management agent Owners: rmccabe Branches: F-8 F-9 InitialCC: Cvsextras Commits: Yes
Humm... why do you HUP/kill dbus in post/postun? Thats very much a no no. dbus can't be killed without messing up the system and requiring a reboot.
SIGHUP doesn't kill dbus. from the man page: SIGHUP will cause the D-Bus daemon to PARTIALLY reload its configura- tion file and to flush its user/group information caches. Some configu- ration changes would require kicking all apps off the bus; so they will only take effect if you restart the daemon. Policy changes should take effect with SIGHUP. ricci installs new dbus (and oddjob) policy, so it needs dbus to reread config files. the reload action in the dbus init script does nothing, and restarting just does a stop and start, which is undesirable.
ah, quite right. I guess I misread that... sorry about that. ;( cvs done.
This package has been in for ages now, closing this review.
Not to rain on the parade here, but this bug was just mentioned in a discussion about package quality. This package review is basically a joke. Just posting rpmlint results is not enough: The review guidelines clearly define several items of which only one is mentioned in the review. A short look at the spec file makes me wonder if anything else then a cursory rpmlint and mock run was attempted. While it might be debateable if the .spec file is actually legible but the missing source url is a clear blocker: "MUST: The sources used to build the package must match the upstream source, as provided in the spec URL. Reviewers should use md5sum for this task. If no upstream URL can be specified for this package, please see the Source URL Guidelines for how to deal with this." What about the duplicate Summary line? What about duplicate Requires? What about unneeded Requires? Why is the main %description tag talking about Conga and a second %description tag with a name override of "ricci" (identical to the %name tag) talking about the real program? What happened to the %files section? Why is the directory %{_docdir}/ricci-%{version}/ included as regular files instead of being marked as docs? Why isn't the %doc macro used? After taking a closer look at the .spec file I have to revise my initial verdict: This is not only a glaring example of a major failure of our review process, the entire .spec file is made out of fail.
Re-opening because of fedora-review-.
Isn't stuff in %{_docdir} automatically marked %doc by RPM? The other complaints do sound valid though.
(In reply to comment #15) > Isn't stuff in %{_docdir} automatically marked %doc by RPM? This is correct: http://www.redhat.com/archives/fedora-maintainers/2005-May/msg00118.html Currently there are more directories under which all files/directories are automatically marked as %doc Note that I have not checked the spec file of this package at all.
Pulling myself of the assigned list, it appears as though this can be closed as this package is in fc11.
Chris: Can you address the items in comment #13? We should not close this while there are still outstanding issues to address. It doesn't matter if the package is "already in", it needs to meet guidelines...
I'm re-assigning this to rmccabe to fix the issues in comment #13. I believe he does all the package builds and can fix these issues.
Fixed a while ago.
ok, I guess we should ask Andreas to recheck things? Andreas?
Closing since ricci has been removed as of F17.