Use of the __GMP_CC and __GMP_CFLAGS macros

[Gabriel Dos Reis]

On Fri, Oct 5, 2012 at 10:29 AM, Marc Glisse <marc.glisse at inria.fr> wrote:
> On Fri, 5 Oct 2012, Steve M. Robbins wrote:
>
>> For one thing, supposing a distribution does choose a non-default ABI one
>> day
>
>
> I am not sure it would make much sense for a distribution to ship a library
> for a different ABI. Whatever ABI they choose, they will make it the default
> and compile everything for it. And if there are multiple ABIs, the reverse
> (what we already have in debian) where changing the ABI (with -m32 for
> instance) makes the compiler include a different gmp.h file is more logical
> than having the gmp.h file tell you: "oh, I guess you must have used -m32 or
> I wouldn't have been included" ;-)

I fear that may be too a naive simplification.

I don't expect major distributions to duplicate copies of packages based on
the selection of language flavour (say C++03 vs. C++11) which may affect
ABI.  Rather, I expect them to continue to do what most of them have already
been doing: conditional code elision/inclusion based on macro feature tests.

Consequently, Steve's suggestion of ABI-affecting flags make sense.

-- Gaby