This is the mail archive of the egcs@egcs.cygnus.com mailing list for the EGCS project. See the EGCS home page for more information.
> > Many of the g++.robertl tests were things (even wrong things) that > > caused ICE's at one point. So fixing the code to be defined would > > avoid the ICE. Unfortunately, this type of data was lost when I > > boiled everything down to testsuite cases. > > > > This doesn't seem to be one of those cases. I suggest either fixing the > > testcase or removing it. It's entirely your decision. > > Well, if the original problem was an ICE (for invalid code), then the > test should only be compiled, but not executed. Sorry, I must have been unclear. This isn't one of those tests that ICEd. The original problem http://www.cygnus.com/ml/egcs-bugs/1997-Nov/0251.html in this case wasn't a link failure, a compilation error, or anything else so slippery. It was unexpected runtime behaviour which is exactly what you just described. Therefore, I think the test is bogus since it was relying on undefined behaviour. RJL