Comment 7 for bug 729431

Thanks for the thorough bug report, folks. The problem is that the manual plugin calls upon the shell plugin to run some tests that happen to have a command. However, this has no guarantee of setting the output nor the duration of the command in some cases. For example, when there is no command or when the command has already been run. So, I'm attaching a branch that removes this assumption which should address these exceptions from occuring.