On Tuesday, 19 August 2014 at 04:26:48 UTC, Andrei Alexandrescu wrote:
Well that's what happened - someone started 2.067. What's the advantage of doing this? Now we need to worry about master and 2.067 instead of just master. -- Andrei

Well, what you do at that point is just fix all of the regressions on the branch, and when it's ready you do another release. You don't put anything else on it. All of the normal dev work goes on master. And some point after the branch has been released as the next release, you branch again.

Now, unless we have enough regressions on master that it's going to take us over a month to fix them, I think that branching right after releasing is a bit much, though if some of the regressions are bad enough, maybe it would make sense to release faster. And given how long we've been trying to get 2.066 ready after branching it and how much work has been done on master since then, maybe it makes sense. I don't know.

I would have thought though that we'd aim to branch something like 2 to 4 weeks after releasing and then take about a month to make sure that all regressions are fixed so that we get a release about every two months. All the major dev work just continues on master, and it'll end up on a branch about every two months staggered from when that branch gets released as an official release.

Certainly, aiming for something along those lines would get us faster releases than we've been doing. We've been waiting way too long to branch and then been rather slow about getting through all of the regressions. By branching earlier, we should be able to release more quickly.

- Jonathan M Davis

Reply via email to