improved config setup

From: John Maddock (John_Maddock_at_[hidden])
Date: 2001-06-23 06:57:23

I've posted a first draft of an improved boost config setup to the boost
files section as:

http://groups.yahoo.com/group/boost/files/config/config_03.zip

Main features:

Splits config up into separate files and separate sections.  One file per
compiler etc, and split into logically orthogonal
compiler/library/platform/user sections.
Comprehensive regression tests.
Optional autoconf script runs regression tests and compares existing config
with what it thinks you need (this is not infallible at the moment).
Test program split in two: config_test.cpp actually includes the regression
tests (and so may not compile on incorrectly configured systems), while
config_info.cpp should always build and run and prints out more information
about your system/compiler/standard lib than most people will ever need,
but it helps a lot when tracking down config problems.

The aim is to satisfy the requirements that Vesa Karvonen set out a while
ago (no breakage of dependencies when the config for a compiler you're not
using is changed, better overall stability of the config files, the ability
to "short circuit" the default behaviour and specify your own config).  It
should also be much easier to setup and maintain than the existing system.

Known issues:

There are no docs yet (I will try and rectify this next week).
Needs testing on more platforms (only win32 so far, again I aim to do more
testing next week).
The regression test for BOOST_NO_OPERATORS_IN_NAMESPACE doesn't fail when
it should - this has me stumped as the test code is taken right out of
operators.hpp - I don't see why this builds cleanly when operators.hpp
doesn't (with gcc).
I haven't added tests for the new macros added in the last day or so
(function lib), Doug do you have test cases for these?

- John Maddock
http://ourworld.compuserve.com/homepages/john_maddock/