Re: [Boost-users] What's happened to Ryppl?

Subject: Re: [Boost-users] What's happened to Ryppl?
From: Robert Ramey (ramey_at_[hidden])
Date: 2011-01-25 12:06:19

Eric Niebler wrote:

A lot of work remains --- that is, if it's also our intention to
modularize boost and have a functioning cmake build system, too.

At least, modularization seems like it would be a good thing to do at the
same time. And nobody is working on a bjam system for modularized boost.

*** I don't believe there is any reason to couple modularization of boost to 
any particular build system.
I use bjam to build and test the serialization library on my local machine. 
I just set the current directory to libs/serialization/test and run bjam 
with some switches.  This builds/updates the prerequisites to the 
serialization library, builds the serialization library, then builds and 
runs the tests.  (and in my case builds a table of test results since i use 
library status.sh).  I'm would expect that I could the same with CTest.

The key issue is that the build system permit the building of just one 
"module" (and its necessary prerequisites). Bjam (and hopefully ctest) does 
this now.  Building of "all" of boost is just the building of each module. 
Building of some alternative "distribution" is just the building of each of 
the component modules (and their prequisites). There isn't even any reason 
why each module has to use the same build system.

<idle speculation>
Is it feasible to have both git and svn development going on
simultaneously? Two-way synchronization from non-modularized svn boost
to modularized git boost? Is that pure insanity?
</idle speculation>

*** By the same token, a "modularized" boost needn't require that all 
modules use the same source control system.  Ideally, the build for each 
module would use checkout/update the local copy of the module according to 
the "configuration file" (...v2 or ctest.?).

Once the procedure for for building a module is moved to the module rather 
than invoked "from the top", modularization can proceed incrementally.

Robert Ramey

-- 
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users