Calling "pkg update -n" on Solaris too frequently causes problems with zones

[alanhargreaves]

On a Solaris system, running "pkg update -n" in the global zone at the same time that it is running in a non-global zone, can cause the command to fail in one or the other.

If we have puppet running this command, it causes the puppet run to fail for that specific resource.

This patch removes some unnecessary calls to this command, and appropriately updates rspec.

Runs cleanly under rubocop.

[puppetlabs-jenkins]

Can one of the admins verify this patch?

[alanhargreaves]

Ref: https://github.com/puppetlabs/puppet/issues/9199

[joshcooper]

I'm a little confused by how the pkg provider worked before your PR. If I specify an exact match like:

package { 'git':
  ensure => '2.15.2-11.4.0.0.1.9.0:20180618T171726Z'
}

Then the should value (from the manifest) will match the regexp:

irb(main):003> /^[0-9.]+(,[0-9.]+)?(-[0-9.]+)?:[0-9]+T[0-9]+Z$/.match?("2.15.2-11.4.0.0.1.9.0:20180618T171726Z")
=> true

So we never call pkg list to detect that it's already installed and the insync? method returns false

puppet/lib/puppet/provider/package/pkg.rb

Line 188 in 5503fe2

false

That seems wrong to me... If an exact match is specified, I would expect the provider to compare if the current value (passed in as is) matches the desired value (should), something like:

def insync?(is)
  should = @resource[:ensure]
  if  /^[0-9.]+(,[0-9.]+)?(-[0-9.]+)?:[0-9]+T[0-9]+Z$/.match?(should)
    return is == should
  else
    # search for potential matches
  end
end

[alanhargreaves]

With regard to the recent comment about the call to insync?, The comments at the top of the function state:

  def insync?(is)
    # this is called after the generic version matching logic (insync? for the
    # type), so we only get here if should != is, and 'should' is a version
    # number. 'is' might not be, though.

which lead me to believe that the function is not called in this module if we have an exact match in the type call for it.

On checking lib/puppet/type/package.rb, I can see that the loop in that instance of insync? does exact version comparisons, justifying the comment in the provider inync?.