On Fri, Nov 27, 2009 at 2:54 AM, Andreas J. Koenig <andreas.koenig.7os6v...@franz.ak.mind.de> wrote: > Whenever you discover that you and another tester are doing the same > thing, go somewhere else. It's really boring to see two identical > results. It's like searching mushrooms on the same grid square that > somebody else has just visited before you.
For what it's worth, this is where I sort of disagree. If someone says "I'm testing X, Y, Z, is that useful?" my answer is "Yes!" We have nothing like a statistically valid sample even for common distributions and there are any number of odd ways for Perl 5.X.Y to be installed that are worth exploring. By not having a cookbook for how to set up a multi-perl smoker, everyone invents a slightly different way to do it and here is where we find useful diversity. On the other hand, if someone say "I want to test a lot of perls, what are some useful areas to explore", then I'm perfectly happy for us to point out new grid squares. And it's not just perl configurations, either. Here are some fun/evil ideas: * NOEXEC temp partition (or NOEXEC cpan build partition) at least * Readonly filesystem *except* for temp * statically compiled perl * *nix distro of your choice installed on a FAT32 filesystem * Offline (no internet) > If there is one thing you should avoid then it is to test outdated > distros that are superceded by a higher version. While it is good to > have old versions installed so you can study the impact on other modules > it is nothing but noise if you're testing Archive-Tar-0.23 when the > current version is 1.54[0] I *mostly* agree, with the caveat that I think anything listed in 02packages is fair game for testing. Anything there *can* be installed on a users system given the right prereq listing and modules removed from distros but not de-indexed are an authors problem. Smoking those is one (more) way to harass authors to clean things up. -- David
