Allow runner specifications for local connections

[jerryaldrichiii]

This allows for overriding the default runner selection. For example, when running InSpec unit tests via AppVeyor on Windows the mock file loads would create too many named pipe processes thus exhausting available resources.

[adamleff]

We should probably raise an exception if a runner was supplied that we don't support.

[jerryaldrichiii]

Good catch, will add an exception.

[adamleff]

If we add the exception handling above, do we need this change? Would there ever be a case where we didn't have a valid @runner?

[jerryaldrichiii]

Yes, the first few commands run do not use GenericRunner, but instead just use ShellOut.

The difference being that GenericRunner attempts to do the CommandWrapper bits.

[adamleff]

Got it, I understand.

We can probably then just clean this up a wee bit.

def run_command_via_connection(cmd)
  # use the runner if it's defined
  return @runner.run_command(cmd) if defined?(@runner)

  # if we don't have a runner, such as at the beginning of setting up the transport
  # and while we're performing the first few steps of OS detection, fall back to
  # standard shelling out.
  res = Mixlib::ShellOut.new(cmd.
  ... etc ...
end

[jerryaldrichiii]

Love it. Much nicer! Will make it so.

[adamleff]

I also wonder if this should be :command_runner and not just :runner in case we have file runners, etc. in the future.

[jerryaldrichiii]

Don't see any reason not to do it. Will make it so.

[arlimus]

Fantastic fix thank you @jerryaldrichiii !