signal/bsd_signal

[Roland McGrath]

> Roland McGrath <roland@frob.com> writes:
> 
> > The only observable difference between having it or not is in what
> > value (sigaction(sig, 0, &oact), oact.sa_mask) has.
> 
> Right.  So adding the sigaddset() seems correct.

Eh?  Well, it is superfluous and so I don't see why you say it "seems
correct" (as opposed to its lack).  If you want signal to be anally
compatible with BSD, then one will have to check whether the sa_mask value
reported by sigaction after a signal call contains the signal in question
or not.  I have to stretch to contemplate ways a program could manage to
care one way or the other.