thomson (10/25/82)
In 4.1BSD we have more than once encountered programs which
for(i = 1; i <= NSIG; i++)
signal(i, SIG_IGN);
which has some painful consequences:
1) If you manage to stop such a program (eg. with the unignorable SIGSTOP
signal) you can't restart it, since SIGCONT is ignored.
2) If the program gets a segmentation fault SIGSEG, it will ignore
the trap and continually attempt to re-execute the offending instruction.
Clearly, you can always kill a program in one of these states, but I believe
that ignoring these signals is both dangerous and not useful and should
be disallowed.