win32ole_vs_facter

Minimal puppet module to attempt reproducer for https://github.com/ISEexchange/issue/issues/475

Travis fails due to a facter bug with LoadError:

1) win32ole_vs_facter::windows
   Failure/Error: it { should contain_notify('windows') }
   LoadError:
     cannot load such file -- win32ole
   # ./spec/classes/windows_spec.rb:10:in `block (2 levels) in <top (required)>'

Expect

Mock windows kernel should not impact rspec outcome.
Given proper fixtures, there should be no difference in:

• bundle exec rake spec
• bundle exec ruby -S rspec spec/classes/windows_spec.rb

Actual

rake spec passes, but ruby -S rspec spec/classes/windows_spec.rb fails.

⚠️ The rspec fails when mocking :kernel => 'windows'.
This repo is a simple reproducer for a complex in-house puppet repo.

‼️ The fifth test runs git-bisect against upstream facter project and identifies the first bad commit as https://github.com/puppetlabs/facter/commit/0a8c231b4269. For a given travis build, you must drill down into job 5 to see the results.

The sixth test preserves facter cache between tests, which makes the symptom go away.

Tests 7 through 9 illustrate that version of puppetlabs_spec_helper gem has no effect on outcome.

ref: https://magnum.travis-ci.com/ISEexchange/win32ole_vs_facter/builds/

Simple module

For simplicity, the module classes are...

$ find . -regex '.*\.pp' -exec cat {} +
class win32ole_vs_facter {
  include "win32ole_vs_facter::${::kernel}"
}
class win32ole_vs_facter::windows {
  notify{'windows':}
}
class win32ole_vs_facter::linux {
  notify{'linux':}
}

...and the failing rspec is...

require 'spec_helper'

describe 'win32ole_vs_facter::windows' do
  let(:facts) do
    {
      kernel: 'windows'
    }
  end

  it { should contain_notify('windows') }
end