Re: Use of config.h: summary of responses.

[Russ Allbery]

Daniel Reed <address@hidden> writes:

> My point is that an interface can be written that depends on a
> particular environment, being a set of requirements. In fact, all
> interface headers depend on particular environments to some degree: My
> contention is that the scope of the environment need not be a lowest
> common denominator of all underlying system environments out there.

> Whenever an interface uses the type "void" it is relying on a
> non-guaranteed (non-universal) aspect of its environment.

Personally, I've found that most of these sorts of problems have actually
gone away.  The really difficult bits of config.h to deal with as an
installed header file for me are the ones that define or undefine standard
C keywords for portability to older systems, and I've stopped needing that
for current software.  In practice, const is implemented universally now.
In practice, void is implemented universally now.  There are only a very
small number of software packages that actually need that degree of
portability (although there are a much larger set of software packages
that give lip service to it but wouldn't actually work on a system missing
const or void for entirely different reasons).

It may be that this is just cyclical, and in a few more years we'll all be
dealing with the same issue for "restrict" and the like, but at least for
right now I tend to only have to use the HAVE_* defines, which are far,
far easier to deal with.

-- 
Russ Allbery (address@hidden)             <http://www.eyrie.org/~eagle/>