Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site vaxine.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!talcott!panda!genrad!decvax!wanginst!vaxine!jwf
From: jwf@vaxine.UUCP (Jim Franklin)
Newsgroups: net.unix-wizards
Subject: bug related to "cc -p" on SUN's
Message-ID: <612@vaxine.UUCP>
Date: Thu, 1-Aug-85 11:28:06 EDT
Article-I.D.: vaxine.612
Posted: Thu Aug  1 11:28:06 1985
Date-Received: Sat, 3-Aug-85 05:46:53 EDT
Organization: Automatix, Inc., Billerica, MA
Lines: 27

There is a bug related to the C profiler switch (i.e., cc -p) on
SUN Microsystems 4.2BSD systems.  Take the following C program
and call it "junk.c".

main ()
{
        int i, j;

        i = -2147483648;
        j = i / 10;
        printf ("%d\n", j);
}

Then do the following on your SUN:

        > cc junk.c -o junk
        > junk
        -214748364                      (right)
        > cc -p junk.c -o junk
        > junk
        214748364                       (oops)

We are running 4.2BSD, SUN version 1.2.  This works fine on our VAXen.
I looked at the .s files for the profiled and normal compile, the only
difference is the mcount call.  The SUN compiler calls ldivt() to do the
integer divide.  I suspect that the problem is in the ldivt() you get
from the profiling version of libc.a (/usr/lib/libc_p.a).