Is it always safe to do this?

        ...
        old_handler = signal(SIGFOO, my_handler);
        ...
        old_handler();    /* <---- suspicious... */
        ...

The reason I'm suspicious is that I cheated and looked at some
header files, and found that SIG_IGN and SIG_DFL are *not* pointers
to actual functions. Instead, they are typecasted ints!

--

Louisiana Tech University BSEE/Sr  * All rights reserved. No warranty, etc

        A signal can be generated by a program only by calling the raise
function.  Such signals, however, work more or less like ordinary function
calls - thus it is permissible to call the standard library functions from a
signal occurring because of a raise function call.

-----
/* Teach me to feel another's woe,
   To hide the fault I see;
   That mercy I to others show,
   That mercy show to me                - Alexander Pope */

/* Ajoy Krishnan T,
   Senior S/W Engineer., Hughes Software Systems,
   New Delhi - 19, India.

   The macros defined are SIG_DFL, SIG_ERR, SIG_IGN which expand to
   constant expressions with distinct values that have type compatible
   with the second argument to and the return value of the signal
   function, and whose value compares unequal to the address of any
   declarable function.

So, provided your implementation defines suitable sematics for casting
integer constants to pointers and obeys certain a suitable linkage
convention, defining these macros as suitable cast integers is
perfectly valid.  Consequently, here is no guarantee that either of
these are pointers to real functions.

--
Alan Watson                        | Users of double-precision complex will  

Department of Astronomy            |                     --  Brian Kernighan
University of Wisconsin -- Madison |