Initial refactoring of yumrepo.

[apenney]

[I've made this PR early, as I'd like feedback on the direction of the provider/type, and the tests. It seems to work for me, which is about as strongly tested as it is right now.]

This work strips out all of the provider code from the type, and
creates a new ruby provider for yumrepo.

While this code still uses inifile it's been rewritten to take advantage
of the modernization of Puppet. It's now a little easier to understand
and test.

[puppetcla]

CLA signed by all contributors.

[Sharpie]

Related tickets:

Redmine 8758
Redmine 9293

[Sharpie]

Also incorporates Redmine 22304

[apenney]

Added the stripping here to handle 'enabled = 1' cases.

[unux]

@apenney from the yum config parser: http://yum.baseurl.org/gitweb?p=yum.git;a=blob;f=yum/config.py#l387

class BoolOption(Option):
    """An option representing a boolean value.  The value can be one
        of 0, 1, yes, no, true, or false.
        """

Seems like yum's boolean options also take yes and no in addition to (0|1|false|true). This might be worth expanding into the rest of yumrepos options.

[apenney]

Ahhhh, nice, I didn't look in config.py, I had only got as far as repo.py. That'll teach me. I'll fix this globally.

[adrienthebo]

#instance_variable_get is generally evil, if this needs to be used by outside classes then can we make an attr_reader for entries?

[apenney]

I can take a look at this, I definitely don't like grubbing in inifile in this way.

[adrienthebo]

This can be rewritten as section(@resource[:name])[property.to_s] = value

[apenney]

I had troubles with this, and ended up using []= (which is actually defined in inifile special, just in case you thought I was relying on some standard ruby here.

[adrienthebo]

Could you add a bit more information to this? If this is actually necessary then it may indicate that something else is going on, and if it's not necessary then I would prefer the more common syntax.

[apenney]

I'll keep working on this. I'll make sure self.reposdir is covered as well.

[riton]

Hey, for #2268, you can now pull from 4599332 il you're okay with this.

Cheers

[apenney]

I rewrote reposdir and added find_conf_value for now to make this a little cleaner. Are you ok with this @adrienthebo ?

[adrienthebo]

+1, looks good

[joshcooper]

@apenney I noticed that travis failed on commits 866f5c5 and 537b6f4. Could you fix those up? That way it's easier to bisect in the future.

[adrienthebo]

This comment confuses me, because this method will be invoked after every resource has been evaluated, so we'll be calling self.class.store every time.

[adrienthebo]

This scares the bejezus out of me, providers should never have to look at Puppet[:noop] because destructive methods should never be called if Puppet is in noop mode. Why is this necessary?

[apenney]

My honest answer is "I don't know, it was in the original type". I suspect it predates whatever safety work is done in noop runs and can be removed. I'll do some experimentations with removing it and noop.

[adrienthebo]

To copy our conversation here, #section may return nil here and if we stub :[]= then incredibly horrible things happen.

[adrienthebo]

Merged into master in e638972; this should be released in 3.5.0. Thanks for all your hard work!

[kylog]

After seeing @peterhuene's fix up yesterday at a76d681, I took a look at this pull. There are some file api accesses via the native ruby api (e.g. here) and some via the Puppet::FileSystem abstraction (e.g. line 50). For consistency's sake, we should probably scrub this pull to use that abstraction consistently. @apenney @adrienthebo thoughts?

[adrienthebo]

I agree, I'll go though and do the needful shortly.