From: Douglas Gregor (gregod_at_[hidden])
Date: 2002-12-05 09:22:29
On Thursday 05 December 2002 08:51 am, David Abrahams wrote:
> Douglas Gregor <gregod_at_[hidden]> writes:
> F> On Wednesday 04 December 2002 08:53 am, David Abrahams wrote:
> >> It looks like some people (ahem! <wink) have been using
> >> BOOST_NO_CONFIG where they should be using BOOST_STRICT_CONFIG. See
> >> boost/function/function_base.hpp.
> > Oops. Fixed now.
> Well, now John Maddock may be quarrelling with you over the correct
> approach. We need to decide how this will work.
It seems like places that need BOOST_(STRICT|NO)_CONFIG in the source don't
fit the documentation for either macro. BOOST_STRICT_CONFIG just tells the
config system how pessimistic to be w.r.t. new compiler versions.
BOOST_NO_CONFIG tells the config system not to include any of its
configuration headers (i.e., don't define any defect/extension macros).
As it stands now, I think the right way to write
compiler/library/platform-dependent workarounds is to check for _both_
BOOST_STRICT_CONFIG and BOOST_NO_CONFIG. BOOST_STRICT_CONFIG disables the
workaround when dealing with a new compiler version, whereas BOOST_NO_CONFIG
disables the workaround when the user has asked not to configure for the
compiler. Maybe BOOST_NO_CONFIG (actually, BOOST_NO_COMPILER_CONFIG) should
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk