> Hi, although I greatly appreciate the provided service (cpan smoking) it has > the disadvantage to give the result long after I need it which is before I > release to CPAN. > > I would like to see the possibility to run smoke tests as part of my pre- > release tests. I see no reason to release a module if it doesn't work on a > specific platform or at least I'd like to list that platform as unsupported. > Although most often the errors are simply found because the test smoking > machines have a different setup.
You can release a development version of a module to the CPAN. Smokers will pick it up and test it, but it'll be invisible to users unless they specifically request the dev version. > I also believe that the smoke test platforms should be available, on a > cpan/perl/smoke test related site, as virtual machines when possible. This isn't possible for some of the more interesting platforms because of licencing constraints. > This > would lower the demand on the smoke test infrastructure. There are VMs > available for download, IE on the OsZoo site, but those differ from the > "official" smoke boxes. The current bottleneck isn't on smoke boxes, but on the infrastructure for dealing with the reports they generate. > Running smoke tests in VMs would have the added advantage to add redundancy to > the smoking infrastructure. There's already plenty of redundancy for those platforms for which it would be practical to make VM images available - those being the common free x86 OSes. The ones where we only have limited coverage - eg Sparc, Alpha, MIPS, Windows, IBM - are those for which we can't make VM images available, either because of OS licencing constraints, or because there's no suitable virtualisation environment. > Can we get distributable VMs for non-free OSes? Not legally. -- David Cantrell | Reality Engineer, Ministry of Information Support terrierism! Adopt a dog today!