On Wed, 9 Dec 1998, Jason Thorpe wrote:

:  > Curiously:  Are we driving toward userland code not #including
:  > <sys/*> at all, or what?  I'm getting that impression from the tone
:  > the thread is taking.
:  > 
:  > Probably delusional as usual.
: 
: It's an issue of what standards say which headers to include to get
: a certain interface.

Yes, unfortunately.

We could do a very tricky thing for standards-conscious code with some cpp
foolery.  In pseudo-C:

===== <sys/signal.h>

#if !_POSIX_C_SOURCE && !... && !defined(__SIGNAL_PROTOS_DEFINED__)
#define __SIGNAL_PROTOS_DEFINED__
int sigblahfunc(...) __RENAME(__sigblahfunc14);
#endif

===== <signal.h>

#if !defined(__SIGNAL_PROTOS_DEFINED__)
#define __SIGNAL_PROTOS_DEFINED__
int sigblahfunc(...) __RENAME(__sigblahfunc14);
#endif

=====

In any case, we need to do something - because of the way signal renaming
works, `old' code needs to see the __RENAME()s in <sys/signal.h>.

-- 
-- Todd Vierling (Personal tv@pobox.com; Bus. todd_vierling@xn.xerox.com)