Re: [bug-gnulib] human, hard-locale changes to assume <locale.h>

[Stepan Kasal]

Hello Paul,

On Sat, Dec 04, 2004 at 10:52:35PM -0800, Paul Eggert wrote:
> coreutils had been assuming <locale.h> for some time (since coreutils
> 5.0, anyway, and probably earlier) without anybody noticing or caring,
> so I took the liberty of simplifying human and hard-locale to assume
> <locale.h>.

maybe I don't understand the context, but IMHO this reasoning is
somewhat flawed:

I think coreutils are not compiled on non-GNU hosts as frequently as
other utils.  It is quite usual that people need the GNU extensions
of a program, so they compile it there.  But with core utils, they feel
somewhat oblidged to stay with the vendor version.

Imagine that a maintainer uses the gnulib in hope that it will
_increase_ the portability of the package.  And it breaks the build on
non-locale hosts.

This could undermine the trust in gnulib.

> If this turns into a real problem on older or
> freestanding hosts we can supply a substitute <locale.h> that does
> nothing.

Please do it.  Perhaps it should be a new module, called ``locale''.

I apologize, but I don't have enough knowledge to write the module.

If I were to do it, I would have to go the old way, with all the #ifdef's.
In other words, I cannot write a better solution then to revert your
patch, or I would have to invest too much time into it.

Have a nice day,
        Stepan