Re: declaring only the functions that are defined

[Bruno Haible]

Paul Eggert wrote:
> I'm a bit dubious about this one, as it adds to the .h maintenance
> burden and I'm not sure the benefit is worth the cost.
> ...
> The other parts I'm ambivalent about.

So let's drop the idea. I was hesitating too.

> For the special case of nanosleep, one might want nanosleep declared
> even if the nanosleep module isn't being used, for the benefit of
> hosts that define nanosleep but don't declare it.

Huh? If the program uses nanosleep(), it needs the gnulib module,
for those hosts that don't have the function at all.

> Also, in the future, I'm leaning towards modifying gnulib so that
> nanosleep and struct timespec are declared by <time.h> instead of
> having a special timespec.h include file.  This is more compatible
> with POSIX and would be simpler for programs to use.

Yes, I agree this is the approach to follow. We just did the same thing
successfully with <wctype.h> and <dirent.h>.

Bruno