Mike Kupfer <mike.kupfer at sun.com> writes: >>>>>> "Rich" == Richard Lowe <richlowe at richlowe.net> writes: > > Rich> It's also not correct in a multiple changesets per changegroup > Rich> world (many other things are suboptimal in that situation, too). > Rich> Do we still intend to encourage and support that, or do we intend > Rich> to behave as we do with current policy and disallow it? > > I thought we were following the current onnv policy. That is, either is > okay, but the fix for a single bug cannot span multiple changesets. > (And ideally, they should be sufficiently tested so that one or the > other can be individually backed out if necessary.)
At the time, there was much discussion about whether that *was* current ON policy. Partly due to confusion. I believe it came out that way though, yes. > Rich> I'm generally in favour of allowing it, if we can enforce sane > Rich> behaviour, but I'm starting to think that's harder than it first > Rich> seemed.) > > If current practice continues, we will rarely see multiple changesets > per changegroup. It's just too much of a hassle to construct them that > way. > It is, but "rarely" is not "never". That's the logic that would make me happy to forbid it, however. Though Danek would be most displeased. -- Rich
