Hey all, :) While working on the release of 5.23.8 I encountered something odd. I talked to rjbs about it and he suggested to contact the mailing list. It was on my backlog until today. One of the important tasks when working on a release is to review failures. The original instructions include a comment on "and if it fails, try to understand why and fix it"[1]. This becomes more complicated because some issues are recurring, some are new, but there is no way to know this. Even if I had no idea how to fix a Solaris compilation error (and trust me, I don't!), I won't even be able to raise a flag. Despite the fact that the porters are pretty hands on when such stuff happen, I would still be happy to have some form of responsibility and not assume others notice everything. That's at least part of the release manager's work. When a test report comes up, it is not connected to any previous report. Using just the platform is not good enough because you could have multiple smokers. Taking the submitter along doesn't help, because a submitter can run multiple smokers on the same platform or on a single machine. Eek. So, I'd like to ask: * Is there a way of which I'm unaware that allows me to correlate smoke reports and say "Uh oh, this is a *new* report from this smoker. This specific smoker didn't crash before." or "Yay, this smoker is now passing!" ? * If not, is it possible to introduce such a thing? I've heard of this QA Hackathon thing. It might be suitable. I understand Tux and Abe know a thing or two about smokers and tend to be present during the hackathon. Also, rjbs suggested this. :) * Once having this (whether we already do, or whether it was added), could we change the smoke reports to allow indicating a status *change* rather than just a status? Am I barking up the wrong tree? [1] Uh... yeah, sure. Just give me a moment to fix this Solaris error right here. :)Thread Next