Hey, In <[EMAIL PROTECTED]> Jim Pick wrote: > We don't currently have a full QA test suite rigged up that a build > must pass before we do a release. I think that's something we should > put together before we do a production release.
well, we already have regressions. And apparently this is hard enough on some platforms already :) For some platforms "building" is already is already a challenge. Also, measuring the number of test cases is of limited measure. Of course "0" is what we strive for and "80 out of 147" means something is deeply broken. Once a few remain, they have to be taken with common sense. SOme stuff might not be implemented, otherwise even a single failure can indicate bugs in the GC or the threads... > The development releases do get quite a bit of testing (compile > testing on various platforms, mauve, gump, etc.) but it's done in a > distributed matter by various individuals in the projects. yes, this is very important, no test suite can replace "real" usage. Release platforms should have least possible known bug (we have two kinds of bugs: those deriving from the class library and those deriving from the JavaVM), build on as many platforms as we can and not depend on GNU make (once this was distinctive...) -R _______________________________________________ kaffe mailing list [email protected] http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
