Killing a code deploy in the AWS console can dump a NilClass exception

[nikgregory]

I had a launch that somehow failed to have a successful code deployment #39 and to move things along, I stopped the code deploy via the AWS console.

it appears that

  def handle_deployment_failure(deployment_id)
    instances = cd_client.list_deployment_instances(deployment_id: deployment_id)
                         .instances_list.map do |instance_id|
      cd_client.get_deployment_instance(deployment_id: deployment_id,
                                        instance_id: instance_id)
    end

    instances.map(&:instance_summary).each do |inst_summary|
      next unless inst_summary.status == 'Failed'

      inst_summary.lifecycle_events.each do |event|
        next unless event.status == 'Failed'

        ilog.error(event.diagnostics.message)
        event.diagnostics.log_tail.each_line do |line|
          ilog.error(line)
        end
      end
    end

    raise 'Deployment was unsuccessful!'
  end

ilog.error(event.diagnostics.message) seems to be able to return a nil in diagnostics.

  E, [2016-11-30T17:29:59.362336 #29481] ERROR -- : Deployment �[0;34;49md-WI31N9YBJ�[0m failed with status 'Stopped'

  undefined method `message' for nil:NilClass (at /home/jenkins/jobs/workspace/sf-start-stage/apigems/ruby/2.3.0/gems/moonshot-1.0.0.rc1/lib/moonshot/deployment_mechanism/code_deploy.rb:304:in ```

And at this point we cannot really trap a decent error so that we might be able to retry 'something'