: It is "undefined" for signals not generated by kill(2) et 
: al., so strictly speaking any of those would be ok.  (Same 
: for SIGBUS, SIGILL, and SIGSEGV.)

Is it possible to determine the difference between a signal which
has been generated by kill(2) vs. one which has been simply incurred
as a runtime result of stepping out of bounds or dividing by zero?

[I remember under 4.3BSD having a C program which kept on recursing
until it generated a SEGV, which I trapped into a diagnostic routine.
It then kept reiterating the diagnostic routine (it didn't exit) until
I interrupted the program.  I was still in school, in larval stage,
poking around and going, "gee, I wonder what if..." with several
things...  Never got around to running kill(0,SIGSEGV), though...]

--*greywolf;

....:....1....:....2....:....3....:....4....:....5....:....6....:....7..