> It is our own messages I'm talking about, not the invoked compilers, and > as > you imply, it's probably most relevant to warning messages, because things > like "can't open input dataset" are hard to ignore.
For functions (programs) running on our infrastructure, every function return passes through a common epilog routine. The epilog checks the return code from the exiting function and (if necessary) barks up diagnostic messages and a callback trace. We manage to catch an enormous percentage of errors this way because it happens without any programmer effort. It also allows us to control the "chattiness" of the interface via runtime-maintained state and a number of customer-visible parameters. For example, the presence of a //OSZWARN DD DUMMY causes messages to be issued for warning level exits (RC=4), whereas the default is to issue them for error and higher (RC>=8). If you already have such an environment yourself, or one you could elect to tweak (e.g. the SAS SPE runtime) then it is pretty easy. If you're grafting it onto something that doesn't do this sort of thing you've got a lot more work. We also use a common routine for issuing messages, which allows further tailoring and control within the software without requiring direct involvement or cooperation of the client product(s). That might also be an avenue to explore. CC ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
