Re: [Boost-build] Boost.Build evolution

From: Alex Besogonov (cyberax_at_[hidden])
Date: 2006-05-02 10:28:17

Vladimir Prus wrote:
>> The rest of Jam can be ported without a lot of problems (Boost already
>> has a decent filesystem library, regexp engine, etc.) 
> I think we should not rush into rewriting everything. Ref-counted strings is 
> the most important change we'd want for bjam. 
And unfortunately it requires almost complete BJam rewrite. Jam 
interpreter should be definitely rewritten, dependency tracking in Jam 
is also very limited and somewhat hackish.

> Using Boost.Filesystem and Boost.Regex is not very reasonable. What we have 
> in bjam works, and requires considerably less code and considerably more 
 > portable. I'd suggest we don't touch those parts.
Well, Regex and filesystem parts in BJam are OK (though they require a 
little polishing, such as using 'const char*' instead of 'char*').

>> but if we're going 
>> to use SCons build engine it may be easier to write Jam-To-Python
>> translator.
> Well, we're not planning to use SCons as yet, because of performance concerns 
> and generally it's a big piece of functionality and we should not be using it 
> untill we fully understand what that will bring and what's the price.
SCons can be used as a _much_ _better_ dependency tracker. I'd love to 
do tricks like described in this message:
http://article.gmane.org/gmane.comp.lib.boost.build/10550

-- 
With respect,
             Alex Besogonov (cyberax_at_[hidden])