Beman Dawes wrote:
"bpm: installing module 'math'". Ouch! math is big. I guess it is a dependency of chrono. That isn't a bpm problem of course, so I'll post a separate message about that.
'math' is one of our most obvious problematic cases, where modules that only need fpclassify.hpp and sign.hpp for float and double get the whole math(35), as in lexical_cast. See http://www.pdimov.com/tmp/report-develop-3b5a9f4/module-weights.html I've been thinking of proposing a separate Boost.Float library, for consistency with Boost.Integer, that would contain those and other useful floating point utilities, but have no time right now to flesh that out. Other such cases are functional(24) for just hash (as were the case in, f.ex. array) and serialization(59) for just nvp.hpp and split_free.hpp (as in date_time). And of course the perennial type_traits(10).
"bpm: installing module 'numeric~conversion'" Note the '~'. Would it be a better reflection of conversion being in a sub-directory of numeric if it were a '/'?
boostdep deliberately replaces / with ~ in module names to make it easier for various tools (such as itself) that want to, f.ex. construct a filename or an URL from the module name, as they would otherwise need to watch for / and create directories. I can in principle replace it back to / in bpm for display purposes.
That's all for now. For alpha level code, the experience so far has been remarkably smooth!
Thanks.