Subject: Re: Publishing code and VCSs
To: Adam Hamsik <email@example.com>
From: Johnny C. Lam <firstname.lastname@example.org>
Date: 07/10/2007 14:04:32
Adam Hamsik wrote:
> On Jul 10, 2007, at 4:00 PM, Julio M. Merino Vidal wrote:
>> Anyway, the boost surgery we are talking about is, I think, trivial.
>> The current Makefiles already allow building a library individually
>> (see boost-python), so adding packages for every binary library is a
>> fairly easy task. As a result, we'd get finer-grained dependencies,
>> which is a good thing in my eyes. (A while ago we discussed that such
>> things should be achieved by being able to generate many different
>> binary packages from a single source one, but we are not there yet.)
> Can we use option.mk framework for this ? or we are going to create ton
> of new packages :). e.g boost-python can be boost builded withoption -*
> python default option can be *. Does this make sense or not :)?
No, this is poor usage of the options framework. We want library
packages to not have these types of "options" or else one wouldn't be
able to depend on a library package having been built with the correct
set of options or dependencies.
Splitting the various shared libraries from Boost into smaller boost-*
packages is a worthwhile thing to do. The scope of Boost has certainly
broadened from the start of the project when it was mostly just a
repository for additional template code to augment the STL. Losing some
of the run-time dependencies for Boost would be helpful for packages
that use smaller parts of the Boost collection of libraries.
Julio, please go ahead and split the packages as you see fit.
-- Johnny Lam <email@example.com>