UNIT_TESTS_FAIL: Fedora 18 support

[ecbypi]

Based on the discussion in #145, this enables iptables on Fedora 18 and disabled firewalld. Determining the persistence command has been changed to compare regexes against the operatingsystem and operatingsystemrelease instead of osfamily since the command is no longer solely dependent on osfamily. Special cases (Fedora 18, Fedora 15-17), go at the top of the case statement. These changes are structured in a way that these changes can easily be applied to RHEL 7 if desired.

[kbarber-jenkins-bot]

Can one of the admins verify this patch?

[kbarber]

@ecbypi the travis tests are failing on this patch: https://travis-ci.org/puppetlabs/puppetlabs-firewall/jobs/6772125

[kbarber]

ok to test

[kbarber-jenkins-bot]

Merged build triggered. (Status: PENDING, Details: null)

[kbarber-jenkins-bot]

Merged build started. (Status: PENDING, Details: http://box.bob.sh:8080/job/puppetlabs-firewall-system/158/)

[kbarber-jenkins-bot]

Merged build finished. (Status: SUCCESS, Details: http://box.bob.sh:8080/job/puppetlabs-firewall-system/158/)

[kbarber-jenkins-bot]

Can one of the admins verify this patch?

[kbarber]

@ecbypi btw - on the system testing front, although Jenkins reported 'Success' it was never really tested on Fedora 18, as I have it in a special exclusion list until it worked.

I did however run up a custom job here: http://box.bob.sh:8080/job/puppetlabs-firewall-system-fedora-18/1/RSPEC_SET=fedora-18-x64/consoleFull which shows how your code ran against Fedora 18. As you can see beyond the basic class we still have some parser bugs to solve for it to work 'out of the box', especially with existing rules.

Having said that, I think the class patch you have here looks pretty good ... this is a great start to getting Fedora 18 support.

[ecbypi]

@kbarber Ok. I'll look over this tonight. Is it possible to recreate the test run you linked to for fedora 18 on my local machine?

[kbarber]

@ecbypi system tests can be replicated by:

• Installing virtualbox 4.2.10+
• Installing vagrant 1.2.x+

Grabbing the gems using 'bundle update'

Then running:

rake spec:system

Fedora specifically can be tested with:

RSPEC_SET="fedora-18-x64" rake spec:system

The test framework is fairly new so let me know if you have issues. I've tested this on both Linux (Debian 7 in this case) and OS X 10.8.x.

[ecbypi]

Sweet. The RSPEC_SET is what I was looking for.

[kbarber]

Any progress on this @ecbypi ?

[ecbypi]

Unfortunately no. It's been a busy month and it keeps slipping my mind. I'll actually take a look at things tonight. Sorry for the delay.

[kbarber-jenkins-bot]

Can one of the admins verify this patch?

[ecbypi]

@kbarber I'm still in the process of debugging failures. I noticed that on in spec/system/class_spec.rb, the spec fails because changes are being made on Fedora 18 to install the package, stop firewalld and start iptables, resulting in an exit code of 2, instead of 0. It seems sufficient to require that the exit status not equal 1 (or 1 and 6).

What are your thoughts?

[kbarber]

I'm still in the process of debugging failures. I noticed that on in spec/system/class_spec.rb, the spec fails because changes are being made on Fedora 18 to install the package, stop firewalld and start iptables, resulting in an exit code of 2, instead of 0. It seems sufficient to require that the exit status not equal 1 (or 1 and 6).

Some tests have changed since you applied this patch, you might want to rebase. The current tests always test to make sure the exit_code first up is not 1 ... then the second tests for non zero. This is a pretty normal idempotence test. Unless of course, the current code you propose requires more than 1 puppet run to do everything?

[apenney]

I think this is now fixed upstream. I'm going to close this for now but feel fre to reopen and rebase if this is still a problem.