proper recovery from EINTR

[phil at rice.ARPA]

> On 4.2BSD, you could block signals during a read() or write() without
> losing them.  This would be MUCH better than character-at-a-time
> I/O system calls!

Ahhh, but sometimes that does not produce the desired effect.  What
about the case where you want the signal to do something (like set a
flag) but you don't want the system call to be interrupted?  I can't
remember for sure, but I think the message that brought this whole
issue up talked about trying to do something like this.

                                William LeFebvre
				Department of Computer Science
				Rice University
                                <phil at Rice.arpa>