Fix for evaluation of distribution release and default provider.

[tabowling]

@mlichvar @larskarlitski I must admit I did not test this, but the prior logic did not seem to properly account for RHEL version >= 6.8.

[mlichvar]

What exactly did not work? The code was supposed to select chrony on 6.8 and later.
If I understand it correctly, the new code would select ntp instead.

[larskarlitski]

This fails on rhel-6 with No package matching 'ntp' found available, installed or updated and on rhel-7 with No package matching 'linuxptp' found available, installed or updated.

[mlichvar]

We have figured out with @pcahyna what's going on with the test failures. The problem seems to be in the rhel-6 and rhel-7 images. They don't have any repositories configured, so when the role tries to install a missing package, it fails. chrony is already installed in the rhel-7 image, but not in the rhel-6 one.

So, this is not a bug in the role, but in the system images used for testing.

[larskarlitski]

So, this is not a bug in the role, but in the system images used for testing.

I'm sorry, I don't understand. Shouldn't these roles work without errors on the released images?

[mlichvar]

The Fedora and CentOS images work.

If the RHEL image doesn't come with the linuxptp package and it's not possible to install packages from a repository, what should the role do if it is configured to synchronize to a PTP domain?

[larskarlitski]

The point of these roles is to make common tasks possible on different versions of operating systems. Surely there was a (different?) way to synchronize to a PTP domain on RHEL 6?

Why is not possible to install packages?

[mlichvar]

There are no configured repositories in /etc/yum.repos.d, so the ansible package module (which calls yum install) fails.

[larskarlitski]

Sorry, I misread that above. I guess we need to register these images to make repositories available.