On Mon, May 18, 2009 at 6:46 PM, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > > I don't think it's that simple. Some features won't be separate enough to > just be disabled, if they aren't in acceptable shape. Localization is a > perfect example -- it needs to be integrated with other kits and all > applications should use it. Unless everything is strictly #ifdef'ed and > built conditionally -- which will be ugly and quite a bit of work, I guess > -- disabling it won't be possible. That was my thought as well. There are probably more features like this than ones that can be easily disabled. > For large/invasive features I'd suggest to use features branches. This adds > a bit of extra work, but leaves us in control with respect to when to merge > the feature into a release. If we were using Git it would not be much work at all to have such branches :) On that note I do strongly suggest that we have the GSoC students working in Git repos, where they can easily make branches, make frequent local commits, and then easily upload to a server without disturbing the main Haiku SVN repo. I am currently updating a git-svn repo I got from Rene (thanks!) and will try to put it up somewhere within the next few days. > I suppose we should also come up with a roadmap to R1 in the near future. > The alpha 1 release process will be quite straight forward -- just create a > branch and only allow bug fixes in it -- but reverting to the current "put > everything into the trunk" strategy afterwards will certainly not be > helpful. Making a release will certainly add some complications. It would be worthwhile to explore what other projects do in this regard. Also again we have a situation where Git would be so much better than SVN. Git's branch management is 1000 times better. Regards, Ryan