Re: [boost] A comment and a question about the current config macros

From: John Maddock (jm_at_[hidden])
Date: 2002-08-19 06:45:29

> I've seen that boost has currently BOOST_NO_STRINGSTREAM; why isn't
> there a BOOST_NO_NEW_IOSTREAMS macro instead?

Good question, no ones asked for it (until now!).

> Moreover, I find the section 'Macros that describe defects' in
>
> http://www.boost.org/libs/config/config.htm
>
>
> misleading because it describes the BOOST_NO_xxx macros as "the
> implementation lacks the feature xxx".
> As far as I've seen by working at dynamic_bitset this is not true in
> general: for instance some of the compilers I've tried had problems
> with member template friends (BOOST_NO_MEMBER_TEMPLATE_FRIENDS), but
> only in some circumstances and, often, not for the syntax of the
> declaration per se. So I would like the description to say that boost
> renounces the usage of the feature xxx, for the reason that the
> configured compiler has problems 'related' to that feature, not that
> the feature isn't supported at all. In practice this means that one
> has to intend the macros as
>
> BOOST_DOESNT_USE_featurexxx
>
> It's only a different point of view and it affects the documentation
> only but I think it's important anyway. It also avoid us to make
> incorrect assertions about compiler bugs (we only say that there are
> bugs associated with a certain feature, and that therefore the code of
> the library takes them into account - in practice, it doesn't use the
> feature, but that in theory is not the only option).

Yes, that's what we mean.  The problem is that the descriptions are pretty
thin for some of the macros, but I probably should add a global caveat to
the docs to make it clear that we are testing for particular failures,
rather than a feature that is completely missing.  I guess ultimately it is
the test cases that are the real docs:-)

John Maddock
http://ourworld.compuserve.com/homepages/john_maddock/index.htm