Puppet

The latest version of yum that will run on RHEL 3 is yum- .0.8. Yum 2.0.8 does not have a yum python module that later versions do. I suspect that the new module based yum provider won’t work with old yum versions.

Can’t you just wrap the import in a try/except :

try:
    import yum
except ImportError:
    blah

where “blah” is returning something meaningful to yum.rb. Then yum.rb could potentially fall back to calling the yum executable directly like most of the other providers do. Given a python helper is used only for yum is it really worth keeping yum different to all the other package types?

This kind of shell-out if the yum libraries are missing seems like a good idea — then you at least get degenerate behaviour when nothing else is available.

My guess, though, is that it’d be impossible to have the shell-out produce the same output format as the yumhelper, which means, really, that the yum provider needs to support both the yumhelpler and yum directly, for when yumhelper fails, I guess.

Well, I’ve been working on this and while some discussions on IRC led me to think completely replacing yumhelper.py would result in better code this has not been the case. Parsing the output of the yum command correctly on all platforms and versions of yum IMHO is too error-prone. So I’m going to have a crack at just parsing the output of versions of yum (2.4 and earlier), and having everything else handled by yumhelper.py.

Until there are native ruby libraries for yum I doubt there will be a nicer way than handing off the work to either the executable or a python helper.

Can someone try the attached patch with both newer and older versions of RHEL to make sure it works?

If it makes any difference, I have tested it and am running all my machines with the patch provided and it works well. The configurations that make up my environment are: – RHEL 2.1 i386 – RHEL 3 i386 – RHEL 4 i386/x86_64 – RHEL 5 i386/x86_64

Of course if you need someone else to test then ignore this post.

ohookins wrote:

If it makes any difference, I have tested it and am running all my machines with the patch provided and it works well. The configurations that make up my environment are: – RHEL 2.1 i386 – RHEL 3 i386 – RHEL 4 i386/x86_64 – RHEL 5 i386/x86_64

Of course if you need someone else to test then ignore this post.

Ideally we’d like at least one other person to test. I’ll ask around.

I tested it on a couple Fedora machines (seem to have misplaced all my RHEL VM’s), but the patch looks good and at the very least doesn’t cause regressions.

Can you add a comment that specifically states that the shell_out business is only needed for RHEL3 and earlier (so we can kill it when RHEL3 is EOL’d)

Also, there’s a few minor things in shell_out:

• Instead of doing ‘rc >> 8’ use os.WEXITSTATUS, since that’s the documented way to get the child’s status in waitpid(2)
• Shouldn’t the regexp to skip a blank line be “^[ \t]*$” instead of “⁽ )?$” ? Would be more conservative

Other than that, looks good.