Modularized Boost

[Klaim]

Just for information for people interested in this important change: Boost is finally

1. switching to git;
2. be officially hosted on github;
3. be "modularized", which mean that each boost library have it's own official repository - there is a super repository gathering all subrepos together for release purpose;
4. be less inter-dependent: there is too much inter dependency between each libraries but efforts have been made and are continuing to remove unnecessary dependencies;

It's happening right now, but it will take some days or weeks to fix issues.
Anyway it's a good thing as it will make possible to just pick a list of libraries you want from boost and get only the code from them (and their dependencies, but less than usual).

(as for Ogre, it might make possible later to rely on some of the useful libs, like stable_vector or flat_set/map or smart poitners, without having to force boost as a dependency, by simply putting a copy of these libraries into the ogre dependencies repo - but it might not happen, I'm just saying it will be made easier to consider).

[Faranwath]

After a couple of weeks busy with college, I return and see this. Finally!

[Klaim]

The switch is basically done but not all the process are in place, they are figuring issues and fixing them progressively. For example, the test bots don't run yet with git changes.
There is also some ne4w process being installed, like the Community Maintenance Team which will help maintaining the orphan libraries, by people from the community (validating PRs mostly): http://boost.2283326.n4.nabble.com/The- ... 56663.html

[Faranwath]

Well, it's well received. I remember something Dave Abrahams said (I may be mistaken, but I believe it's his quote) about Boost needing to devise a different library management/deployment system if it grew too much. I think the current version meets that definition, although much more could be done, of course.

[Klaim]

Yes it's Ryppl http://ryppl.org/

From what I understand, Dave Abrahams is currently too occupied with both working on clang (he was recently recruited by Apple to work on it) and also helping with the boost modularization which would help finishing Ryppl.
Meanwhile, some similar projects spawned but they are not as ambitious as Ryppl, but Rypple looks dead externally (even if they are working on it apparently slowly).

[Faranwath]

Rypple looks dead externally

A shame. CMake offers some help in this regard based on my personal experience, and careful design can result in quite beautiful systems (e.g. the one used by LLVM).

From what I understand, Dave Abrahams is currently too occupied with both working on clang (he was recently recruited by Apple to work on it)

Oh my, we'll have to prepare ourselves to deal with lots of esoteric template metaprogramming stuff when peeking at the source code

[Klaim]

Rypple looks dead externally

A shame. CMake offers some help in this regard based on my personal experience, and careful design can result in quite beautiful systems (e.g. the one used by LLVM).

As I was saying, it only "look" dead, they did a lot and if I understand correctly there is only a few issues to fix before it becomes usable, but boost have to be done first.
It's actually a very practical design, which IS based on CMake. I remember that one of the Ryppl team member also implemented improvement to very recent versions of CMake to fix some of the problems common to both projects, which resulted in interesting new features (which I didn't use yet so we'll see).

From what I understand, Dave Abrahams is currently too occupied with both working on clang (he was recently recruited by Apple to work on it)

Oh my, we'll have to prepare ourselves to deal with lots of esoteric template metaprogramming stuff when peeking at the source code

I don't know but I don't expect clang code to be very templated, they are very practical people with very simple code in the end (which is certainly why it's simple-to-follow code).