On Wed, Feb 11, 2009 at 7:56 PM, Mark Mitchell <m...@codesourcery.com> wrote: > Paolo Bonzini wrote: > >>> I think the only reasonable release criteria is zero P1 regressions over >>> some period. 50 P2 regressions doesn't make a release blocker, neither >>> is 49 P2 regressions a clear sign for a non-blocked release. >> >> I agree. > > I mostly agree. P1 regressions are, by definition, Really Bad. In the > past, we've sometimes had to release in that state because it was clear > that there was no reasonable way to get the P1 regression fixed. But, > that's not what we want to do. > > However, I don't agree that P2 regressions aren't a factor. If we have > a ton of crashing on wrong-code, etc., regressions that adds up to a > release that won't work well for people.
In which case the important ones should be P1 ... I'd use P2 for bugs we are comfortable fixing later on the release branch or, if resources force it, not at all for the release. Which apart from exceptions would automatically make wrong-code or rejects-valid, ice-on-valid bugs P1. If they are not regressions or regressions that we knowingly released older versions with they IMHO qualify for P2. > I don't think there can be, or need to be, completely bright-line rules. > We need human beings to use their judgment about when the release is > good enough to go out the door. So, I'm all for guidelines, and I agree > that 49 vs. 50 isn't itself a big deal. But, I think that Paolo's > criterion should not be interpreted literally. Of course. Richard.
