Should arguments other than 'command' be checked for definedness?

[jkeenan]

In RT 41149 in 2008, checks were added to ensure that no undefined value was passed as the "command" element in the list of arguments to publicly available methods. However, at that time, Paul Fenwick wrote:

"I'd also like it to contain checks for any undefined elements being passed to IPC::System::Simple ...."

That was not completed at that time. However, in order to assess the benefits of warning or dieing on undefined values, we'd first have to have some plausible test cases where passing an undefined value other than to a command would be a bad thing, either because it caused a crash or bad result or because it generated a badly formatted error message or warning.

Until we get such test cases, this issue is stalled.