Re: [Boost-build] Future development

From: Vladimir Prus (ghost_at_[hidden])
Date: 2007-10-04 20:37:20

Douglas Gregor wrote:

> 
> On Oct 4, 2007, at 7:25 AM, Vladimir Prus wrote:
> 
>> Douglas Gregor wrote:
>>
>>>
>>> On Oct 1, 2007, at 12:51 AM, Vladimir Prus wrote:
>>>> In light of that, I think that the future of Boost.Build is a Python
>>>> implementation.
>>>
>>> When did it become acceptable for Boost's build tool to be written in
>>> Python?
>>
>> Sorry for picking nits, but Boost's build tool is Boost.Build M12.
>> When Python port is finished, we can separately discuss cost/
>> benefit of
>> using it for building Boost.
> 
> If the existence of a Python port means that the current bjam-based
> version will no longer be supported by the few people who understand
> it, then frankly Boost has no choice in the matter. We either move to
> the Python port and accept it, or maintain a system that very few of
> us understand, or we have to come up with a new build system for Boost.
> 
> Boost is completely dependent on Boost.Build, and decisions about
> Boost.Build have a huge effect on Boost, good or bad. Boost is also
> by far the largest and best-known user of Boost.Build, so it seems to
> me that Boost's requirements should have some influence on the
> direction of Boost.Build. No?

Except that it's not very reasonable to discuss Boost's requirements
at this point. About 10 different persons will immediately jump out,
suggesting that we "evaluate" alternative solution 1...N. About 10
other people will make various random complains about Python, and the
discussion will likely take a month, without any outcome.

Having concrete code will frame the discussion in a completely
different way, like "we have this Python port, that is 2x less code,
and is 10x faster and takes 10x less memory, where do we go from here".
There'll be no place for guesses and speculations.

>>> I love Python, but that's a rather large dependency for Boost to
>>> have, given that many Boost users will have to build Boost from
>>> sources.
>>
>> Err, larger that CMake?
> 
> Python + Boost.Build is larger than CMake, yes.

Cmake download is 2.5M, and Python is 10M. This is already
not very significant difference, it "just" 4 times. And if
we're going to use Python->EXE solution on Windows, I
suspect there will be no difference at all.

- Volodya