On 15 April 2014 00:54, Paul Gilmartin <[email protected]> wrote: > > I am not much of a fan of all these handlers: >
I have a 'signal on novalue' in all my REXX code because it can keep you busy for hours when there's a typo. But I don't have the actual handler there unless it's production code. Just get the traceback when it hits. And in production code I put another signal as Kris says to get the traceback. Experience shows that it's tedious in a pipeline stage to code the check on return code. It turns out to be sufficient to exit on non-zero return code from the subcommands, which is exactly what 'signal on error' does. And since we normally consider EOF a good exit, that's where 'return rc * (rc <> 12)' fits. When I have something that uses the return code for something else, I put it in a REXX procedure and handle it there. There's a lot of bonus value in keeping this compact so you can oversee a single routine. Rob
