On Tue, Jun 15, 2010 at 4:32 PM, nathan binkert <n...@binkert.org> wrote: > I think what happened is that we just weren't disciplined (or > dedicated) enough to do it. The problem is that usage is so bursty, > I'm not sure that a 1 month freeze would actually do much. Perhaps we > should just update stable every 3 months and actually apply bug fixes > to it as people find them. Would named branches help?
Our typical response when someone runs into a bug that's fixed in dev but not in stable is to say "use the dev repo", not to push the bug fix to stable. I'm fine with that, it's less work for us :-). If we update stable more frequently then the probability of someone hitting the window becomes even smaller. I think there are two questions that need to be answered: Q1. Is there an ideal frequency for updating the stable repo? Is it too daunting for people to work off a "stable" tree that could actually get updated weekly, even if those updates aren't buggy? Q2. How long does something need to sit in the dev repo before we consider it tested? Given that most of us use the dev repo, I don't think it's necessarily that long... maybe a month? If we know what the answers to these questions are, then I propose that every <Q1> we pull all the csets that are more than <Q2> old. (We probably want to do this manually to avoid the case where a buggy cset gets committed Q2+delta ago and fixes it Q2-delta ago.) If we don't know the answers to these questions, then we'll never come up with the right plan... Steve _______________________________________________ m5-dev mailing list m5-dev@m5sim.org http://m5sim.org/mailman/listinfo/m5-dev