Currently, we've got the warn and fail functions (which both call warn_or_fail behind the scenes). The latter will print to stdout and sys.exit(-1). IMO, this is a bad design. If the Argument Clinic tool exits with an error, we should print to stderr, as every well-behaved command line tool should do.
If we rewrote so fail raises a ClinicError exception and warn uses warnings.warn, we could...:
rework if __name__ == __main__ to catch ClinicError, print to stderr and exit
we could use assertRaisesRegex in the test suite (major simplification of the test code)
Currently, we've got the
warnandfailfunctions (which both callwarn_or_failbehind the scenes). The latter will print to stdout andsys.exit(-1). IMO, this is a bad design. If the Argument Clinic tool exits with an error, we should print to stderr, as every well-behaved command line tool should do.If we rewrote so
failraises a ClinicError exception andwarnuseswarnings.warn, we could...:if __name__ == __main__to catch ClinicError, print to stderr and exitassertRaisesRegexin the test suite (major simplification of the test code)