eric (11/22/82)
I take exception to the statement "All of the run-time arithmetic modules
were bug-fixed for System III ..." in sri-unix.4288.
The System III we are running arrived with the same old bug in the long
division and remainder routines. The following routine should print:
20000000000 1 20000000000
On an 11/23 the result is 20000000000 1 0.
On an 11/40 the result is 20000000000 1 37777700001.
Admittedly these are not typical UNIX machines, but the Processor Handbooks all
say that the values left in R and Rv1 are unpredictable if overflow occurs.
I would be interested in finding out if there are ANY machines on which the
'vanilla' ldiv and lrem routines work properly.
I will post our fixup if interest warrants.
Eric Norum
(...!decvax!harpo!utah-cs!sask!hssg40!adec23!eric)
/*
* Check for bug in PDP-11 long division routines
*/
long int i, j, k;
main ()
{
i = 020000000000;
j = 1;
k = i / j;
printf ("%lo %lo %lo\n", i, j, k);
}