sa_sigaction [was: snapshot in preparation for m4 1.4.12]

[Eric Blake]

According to Tom G. Christensen on 8/13/2008 2:45 PM:
> > Weird.  So Irix 5.3 has sigaction, but not sa_sigaction?  Can you please
> > post the definition of struct sigaction from <signal.h>?
> >
> > From <sys/signal.h>:
> typedef struct sigaction {
>         int sa_flags;                   /* see below for values         */
>         __sigret_t (*sa_handler)(_sigargs);     /* SIG_DFL, SIG_IGN, or *fn */
>         sigset_t sa_mask;               /* additional set of sigs to be */
>                                         /* blocked during handler execution */
>         int sa_resv[2];
> } sigaction_t;

Let's try resolving this one first.  Does this program compile 
successfully (although it probably issues a warning)?

#include <signal.h>
#define sa_sigaction sa_handler

static void
handler (int i, siginfo_t *s, void *v)
{
}

int main()
{
  struct sigaction sa;
  sa.sa_sigaction = handler;
  sa.sa_flags = SA_SIGINFO;
  return 0;
}

If so, then we only need to adjust the docs (the presence of SA_SIGINFO is 
no longer a reliable flag for having sa_sigaction), and use 
HAVE_STRUCT_SIGACTION_SA_SIGACTION in signal.h.in.  Everywhere except 
c-stack.c only cares about the name resolving for compilation to work, 
without actually trying to use it, and c-stack can easily be taught to 
avoid SA_SIGINFO if sa_sigaction doesn't work, and compiler warnings can 
be ignored.  If not, then there is more work to do, but I think it is 
still possible.

Then, since your latest logs show:
libsigsegv: mips-sgi-irix5.3 | yes | yes | 2.5
it looks like c-stack will work with libsigsegv, once we downgrade the 
compiler error on sa_sigaction to a warning.

--
Don't work too hard, make some time for fun as well!

Eric Blake             address@hidden

signature.asc
Description: OpenPGP digital signature