Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!att!cbnewsd!tainter
From: tainter@cbnewsd.ATT.COM (johnathan.tainter)
Newsgroups: comp.lang.c
Subject: Re: ReadKey like Function in C
Message-ID: <1330@cbnewsd.ATT.COM>
Date: 23 Aug 89 19:47:37 GMT
References: <1255@calvin.EE.CORNELL.EDU>
Reply-To: tainter@cbnewsd.ATT.COM (johnathan.tainter,ih,)
Organization: AT&T Bell Laboratories
Lines: 18

In article <13653@bloom-beacon.MIT.EDU> scs@adam.pika.mit.edu (Steve Summit) writes:
>I don't know why control-C's are polled under DOS; the low-level
>keyboard scan codes do come in as true interrupts, so a proper,
>asynchronous control-C interrupt could have been generated.

The reasoning I have seen is that the BIOS is so fragile it cannot stand
to have interrupts that do nontrivial things (like the DOS equivalent of
tearing down a process) so they have to insure that ^C is only checked
in known states of the BIOS.  Thus this abomination before G*d of having it
checked on i/o operations only.

This might be of interrest.  Of what I have seen of UNIX implementations
they avoid this problem by mapping the 'hardware' interrupt into a
'software' interrupt (a signal) and checking for software interrupts when
the process gets restarted after a suspend (i.e. in a known state).

--johnathan.a.tainter--
   att!ihlpb!tainter