Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!njin!princeton!udel!mmdf
From: jnall%FSU.BITNET@cornellc.cit.cornell.edu (John Nall 904-644-5241)
Newsgroups: comp.os.minix
Subject: Re: Re: Gettin' there with Evans stuff...
Message-ID: <17671@louie.udel.EDU>
Date: 15 Jun 89 00:09:20 GMT
Sender: mmdf@udel.EDU
Lines: 26

I wrote:
>>  (1)  If I put /usr/bin `readclock` </dev/tty in my /etc/rc file
>>       it locks the system up tight....

Bruce Evans replied:
>  Did you recompile readclock?

Yes, I did recompile readclock, with the readclock.c.cd changes in it.
The problem (on my system, anyway) turns out to be the following:

  Almost the first statement in main() of readclock has a statement of
the form  if(peek(CPU_TYPE_SEGMENT, CPU_TYPE,OFFSET) != PC_AT), and if
the statement is true (it does not pull FC from location 0xffffe) then
it outputs a -q and does an exit(1).  Unfortunately, the routine peek
returns an integer (not unsigned) and the FC becomes FFFC, which does
not compare with PC_AT (defined as 0xFC).  If this happens when the
statement is as given in my original comment, it locks the system up.
If one does a "readclock" later on, however, it just outputs the -q and
runs OK.  So there is something sinister in having it happen in /etc/rc.
I leave that to others of greater talent and curiousity.

It is easily cured by making the statement:

  if ( (unsigned) peek(CPU_TYPE_SEGMENT, CPU_TYPE_OFFSET) != PC_AT)

John Nall