on Sun Jan 01 2012, "Robert Ramey" <ramey-AT-rrsd.com> wrote:
So as things stand now, cmake build/test of boost is not part part of boost itself. It's not a required tool. So I can't see why a review would be necessary.
It might be necessary if Boost were going to adopt it as the official system. I certainly won't want to maintain it forever if Boost were _not_ going to adopt it.
What if it fails the review - would I be prevented from using it on my own machine?
How could anyone prevent that?
If things evolve and a significant number of people start to prefer it over the current system, we can look at the issue as to whether we have to choose and if so what should we choose.
I really don't want that to happen. When the system is completed I hope we'll have an evaluation period and make a decision quickly.
a totally separate issue - the GIT tree seems to clone the trunk - wouldn't it be better if it cloned the release branch?
I don't think so. The point of the project as we currently have it is to prove that we can port the latest in-development state of Boost to the new system without interruption when the system is complete. If we only worked on the release branch we would have massive disruption and breakage every time Boost made a new release and no activity at any other time. When the transition is finally completed, we'll preserve all Boost branches in the Git repository. You can see the release branch in an un-modularized Boost Git repo here: https://github.com/ryppl/boost-svn/tree/release -- Dave Abrahams BoostPro Computing http://www.boostpro.com