on Sun Jan 01 2012, "Vicente J. Botet Escriba" <vicente.botet-AT-wanadoo.fr> wrote:
Hi,
this should be considered as a new tool (and in addition it needs CMAKE build). While I consider the modularization useful, I find that adding a new build system to Boost will need some official maintainers of the CMake files for each one of the Boost libraries until the library authors have taken the time to be familiar with the new build system. Of course, this will imply that we need regular testers for both build systems which will be time consuming
I understand, but I disagree. There's no reason to make the (already high) hurdles to such a transition stratospherically high. If the Boost community decides to use any given new piece of infrastructure, there's no reason it has to be a protracted process; it can just happen.
Does this means that the library authors will need to maintain two build systems?
No, that would be bad, IMO. My point was that we should not let ourselves get into such an intermediate state.
Resuming, I think that we need a formal review for CMake build once it is able to build the whole Boost libraries.
Again I disagree. The review process is for libraries, and Boost has never formally reviewed tools like this. I am more than happy to have a discussion about it, and I believe that discussion should inform our decision, but I believe it is possible for the steering committee to make a decision as a matter of policy.
Oh, as lastly we have started to review also the tools
Oh, we have? Maybe I'm not paying enough attention. Could you clarify?
I thought it works like that also for a new build system, but you know better how Boost works. I think it is time to start this discussion. Please could the interested parties start a thread that results once for all in a decision on whether CMake based build is adopted or not?
It's not time to make such a decision, unless we're going to decide "no." The system isn't complete yet. -- Dave Abrahams BoostPro Computing http://www.boostpro.com