Impossible to set ipaddress6 fact

[olavmrk]

Describe the Bug

It is not possible to set the ipaddress6 fact to any other value than FE80:0000:0000:0000:AAAA:AAAA:AAAA.

Expected Behavior

The ipaddress6 fact should behave like other facts.

Steps to Reproduce

Create a class manifests/init.pp:

class puppet_ipaddress6_test {
  file { '/foo':
    content => $facts['ipaddress6'],
  }
}

And a spec test spec/classes/init_spec.rb:

require 'puppetlabs_spec_helper/module_spec_helper'

describe 'puppet_ipaddress6_test', :type => :class do
    let (:node) { 'server.example.org' }
    let (:facts) do {
        'ipaddress6' => '2001:db8::1',
    } end
    it {
        is_expected.to contain_file('/foo').
            with_content(/^2001:db8::1$/)
    }
end

And a Rakefile:

require 'puppetlabs_spec_helper/rake_tasks'

Then run the spec test:

rake spec

It will fail with the following error:

  1) puppet_ipaddress6_test is expected to contain File[/foo] with content =~ /^2001:db8::1$/
     Failure/Error:
       is_expected.to contain_file('/foo').
           with_content(/^2001:db8::1$/)
     
       expected that the catalogue would contain File[/foo] with content set to /^2001:db8::1$/ but it is set to "FE80:0000:0000:0000:AAAA:AAAA:AAAA"
     # ./spec/classes/init_spec.rb:9:in `block (2 levels) in <top (required)>'

Environment

• rspec-puppet version 2.10.0
• Gemfile
• Gemfile.lock

Additional Context

The hard-coded ipaddress6 fact was introduced in (#771) Set a default dummy ipaddress6 fact.

[ekohl]

In rodjek#772 (comment) I noted that it's the combination of setting both node and trying to override this fact. That may not be a valid workaround for you, but it helps to understand the problem.

[LukasAud]

Hi @olavmrk, can you confirm whether this issue is still present?

[olavmrk]

Hi @olavmrk, can you confirm whether this issue is still present?

Hi! I just retested with the rspec-puppet 3.0.0, and can confirm that the issue is still present.

This is the updated Gemfile.lock:

• Gemfile.lock

[ekohl]

It's this code that causes it:

rspec-puppet/lib/rspec-puppet/support.rb

Lines 271 to 276 in b478497

	# Merge node facts again on top of `let(:facts)` facts, but only if
	# a node name is given with `let(:node)`
	if respond_to?(:node) && RSpec.configuration.derive_node_facts_from_nodename
	result_facts.merge!(munge_facts(node_facts))
	(result_facts['networking'] ||= {}).merge!(networking_facts)
	end

So you can set the derive_node_facts_from_nodename setting to false and that should disable the behavior.

I still think it's wrong to hardcode an ipaddress6. 4ef2f36 looks like a workaround and now that we have a custom FacterImpl (48c5605) it's probably even redundant: no calls to the real facter should happen if you let it stub from facterdb. It's still optional to do that, so taking that into consideration I've written #61.

@olavmrk could you try out my patch?

[olavmrk]

@olavmrk could you try out my patch?

Thank you for looking into this! I tested your changes, but I am still seeing the same behavior. See my comment on your pull request for details.

---

In a way this issue has become less of a problem because ipaddress6 is a legacy fact. Since opening this issue I have moved most of the code away from $facts['ipaddress6'] and over to $facts['networking']['ip6'], which does not have the same problem.

I still think it is really weird that there is a fact that I cannot override though. Especially since the ipaddress fact can be overridden in the normal way.

[ekohl]

I still think it is really weird that there is a fact that I cannot override though. Especially since the ipaddress fact can be overridden in the normal way.

100% with you on that.