Description of problem:

When we wrote the original comps-serving code for cloned channels (bug 457478 and bug 452046), we recursed through all the cloned-from channels, not just did look at the immediate predecesor.

When the code was rewritted to Java, this was lost.

Version-Release number of selected component (if applicable):

Satellite 5.3.x.

How reproducible:

Looking at code. Also, reports by users.

Steps to Reproduce:
1. Have Satellite and client system with RHEL 5 registered to it. Have the RHEL 5 channel synced.
2. Verify that the client can do yum clan all ; yum grouplist Core > /dev/null without error and check that /var/cache/yum/rhel-i386-server-5/repomd.xml contains group data that point to comps file, and that the comps file was retrieved as well.
3. Clone channel from that RHEL 5 channel and change the base channel to this new cloned one.
4. Repeat step 2 and verify that it still works.
5. Clone channel from that cloned channel and change the base channel to this new cloned one.
5. Repease step 2.
  
Actual results:

It fails, if you have Java repodata generator (the default).

It passes if you use Python.

Expected results:

It should work in all cases.

Additional info:

The problem is that while we have

    def get_comps_file(self):
        """ Return a file-like object of the comps.xml for the channel. """
        if self.channel.comps:
            comps_view = view.CompsView(self.channel.comps)
            return comps_view.get_file()
        elif comps_mapping.has_key(self.channel.label):
            comps_view = view.CompsView(Comps(None, 
                os.path.join(CFG.mount_point, comps_mapping[self.channel.label])))
            return comps_view.get_file()
        else:
            if self.channel.cloned_from_id is not None:
                log_debug(1, "No comps and no comps_mapping for [%s] cloned from [%s] trying to get comps from the original one." \
                          % ( self.channel.id, self.channel.cloned_from_id ))
                cloned_from_channel = rhnChannel.Channel().load_by_id(self.channel.cloned_from_id)
                cloned_from_channel_label = cloned_from_channel._row['label']
                return Repository(rhnChannel.channel_info(cloned_from_channel_label)).get_comps_file()
        return None

in Python's Repository class, so we effectively recurse, in Java, this was only written as

        if (compsMapping.containsKey(channel.getLabel())) {
            compsPath = compsMapping.get(channel.getLabel());
        }
        else if (channel.isCloned()) {
            // If its a cloned channel see if we can get the comps
            // from the original channel.
            ClonedChannel clonedCh = (ClonedChannel) channel;
            Channel origChannel = clonedCh.getOriginal();
            compsPath = compsMapping.get(origChannel.getLabel());
        }

        return compsPath;

No recursion here.

If would also like to point out that in Python, by default we get the comps information from the database from that cloned-from channel, not just the static mapping.