Re: [boost] Include only one Boost file

From: Daniel Frey (daniel.frey_at_[hidden])
Date: 2002-06-26 04:00:38

Maciej Sobczak wrote:
> 
> Would it be justified to make some minimal version of Boost (or at least
> some part of it) for decent compilers? (kidding)

Why you say 'kidding'? I think it is a valid point for a very subtle
reason: Boost adds lots of work-around for broken compilers. This makes
the code not what it should be, as things are not as direct as they
should be. We write

  BOOST_STATIC_CONSTANT( int, value = 42 * sizeof( T ) );

instead of

  enum { value = 42 * sizeof( T ) };

thus adding layers just for broken compilers. Besides the effect on
boost itself, there is also an effect on the compilers: Why should the
vendor fix its compiler if the boost compiler status pages say "OK, it
works"? I think it was already mentioned that a lot of compiler vendors
are paying attention to what boost is doing, so why not put a reasonable
amount of pressure on them? I imaging something like a tri-state to
indicate the compiler status:

  OK  /  Fails (but workaround is provided, thus it's OK) / Fails

> I know that once the whole Boost gets included in C++0x (:-)), the
> problem will vanish - today nobody cares about the dependencies between
> standard headers - but at the moment it bothers me a bit.

Problems will vanish if compilers get fixed and the old versions are
history.

At that time, Boost 2.0 will remove all ballast from the past. (kidding?
:)

Regards, Daniel

--
Daniel Frey
aixigo AG - financial training, research and technology
Schloß-Rahe-Straße 15, 52072 Aachen, Germany
fon: [phone], fax: [phone]
eMail: daniel.frey_at_[hidden], web: http://www.aixigo.de