Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!TRANSARC.COM!cfe
From: cfe@TRANSARC.COM (Craig Everhart)
Newsgroups: comp.soft-sys.andrew
Subject: Re: Simple fix to core dump bug in AMS
Message-ID: <wZpj16r0BwwOEzFUcS@transarc.com>
Date: 12 Feb 90 17:20:06 GMT
References: <cZphB=q0M2Yt0U4XE=@thumper.bellcore.com>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 42

I'd make a better fix to the problem that Nathaniel discusses.

Yes, the changes he suggests will prevent a core dump.  I would argue
that they don't necessarily do the right thing in the cases that would
have dumped core.

There are several callers of this code, not just the place in rawfile.c
where zero-length files are handled.  All other callers, however, always
ensure that the ``domain'' argument is non-null, and the call in
rawfile.c at present will always pass a ``domain'' argument when the
file in question is being read from AFS.

Nathaniel's fix is fine as a point of defense, but I'd suggest the
following to get correct semantics.  In ams/libs/ms/rawfile.c, line 132
in my copy, is a block of code that looks like the following:

	if (IsOnVice(fdtemp) && NewMessage->AuthCell) {
	    p = getcpwuid(statbuf.st_uid, NewMessage->AuthCell);
	} else {
	    p = getpwuid(statbuf.st_uid);
	}
	if (p) {
	    GetNameFromGecos(p->pw_gecos, p->pw_name, NewMessage->AuthCell, &uname);
	}

The fix I'd suggest is to replace that with:

	if (IsOnVice(fdtemp) && NewMessage->AuthCell) {
	    p = getcpwuid(statbuf.st_uid, NewMessage->AuthCell);
	    if (p) {
		GetNameFromGecos(p->pw_gecos, p->pw_name, NewMessage->AuthCell, &uname);
	    }
	} else {
	    p = getpwuid(statbuf.st_uid);
	    if (p) {
		GetNameFromGecos(p->pw_gecos, p->pw_name, MyMailDomain, &uname);
	    }
	}



		Craig