David Abrahams wrote:
on Fri Nov 23 2007, "Robert Ramey" <ramey-AT-rrsd.com> wrote:
I know it doesn't help right now, but I hope to co-manage the next release, and if that happens I will follow a procedure that will keep the trunk and the release branch much more closely synchronized. That should make it practical for you to keep your development branch much more closely synchronized with the trunk.
working the way I'm currently working has worked well for me. My branch is derived from 1.34.1. This has meant that I've been able to develope in an environment where all other variables other than the ones I explicitly introduce are stable. This is an imense improvement from the way I used to do it. It has resulted in some divirgence. But a lot of that is due to the fact that the changes I've been making are somewhat more ambitious than previous ones. Although I think my current version is more robust and complete than that on the trunk, introducing even one small error or inconsistency on one platform slows the release down by a week as things get sorted out. Also we do have a smal number of issues with the serialization library in the trunk. But things are made much simpler by the fact they're not mixed in with my current set. When these issues in the trunk are resolved, I'll merge them into my branch then merge the branch back to the trunk. Except for figuring out how to do the actual merge, I believe that this will result in a measurably higher quality product. Based in this experience and a better understanding of SVN, I expect to tweak my procedure for next time. Hopefully this will reduce the pain even further. Robert Ramey