Re: <signal.h> broken with _POSIX_SOURCE=1 defined

[Zack Weinberg <zack@rabi.columbia.edu>]

On Tue, 03 Nov 1998 08:08:07 -0500, Zack Weinberg wrote:
>
>As an experiment I wrote a simple test which just includes every
>single public header we provide with _GNU_SOURCE defined.

Followup:  Doing the same thing with _POSIX_SOURCE instead produces
some spurious errors.  They come from the internal prototypes in the
wrapper headers.  

In file included from t-standards/all-headers.c:8:
../include/aio.h:5: warning: `struct aioinit' declared inside parameter list
../include/aio.h:5: warning: its scope is only this definition or declaration,
../include/aio.h:5: warning: which is probably not what you want.
In file included from t-standards/all-headers.c:137:
../include/search.h:13: parse error before `__free_fn_t'
../include/search.h:13: warning: function declaration isn't a prototype

In both cases the problem is that the internal protos expect a type
definition which is predicated on _GNU_SOURCE.  That's safe when
building libc, but not when testing it.

The internal prototypes should all be shielded somehow.  #ifdef _LIBC
would work except that if I don't define that for the tests then
features.h will go look for gnu/stubs.h which doesn't exist.  Suggestions?

zw