Ticket History #2016: fake-addrinfo.h calls sprintf() without including stdio.h

From jlb@twu.net Wed Nov 19 16:39:57 2003
Received: from pacific-carrier-annex.mit.edu (PACIFIC-CARRIER-ANNEX.MIT.EDU [18.7.21.83]) by krbdev.mit.edu (8.9.3p2) with ESMTP
id QAA01556; Wed, 19 Nov 2003 16:39:57 -0500 (EST)
Received: from twu.net (bilbo.twu.net [64.246.24.15])
by pacific-carrier-annex.mit.edu (8.12.4/8.9.2) with ESMTP id hAJLdt7O007427
for <krb5-bugs@mit.edu>; Wed, 19 Nov 2003 16:39:56 -0500 (EST)
Received: from twu.net (localhost.localdomain [127.0.0.1])
by twu.net (8.12.9/8.12.9) with ESMTP id hAJLaRHI018890
for <krb5-bugs@mit.edu>; Wed, 19 Nov 2003 16:36:27 -0500
Received: from localhost (jlb@localhost)
by twu.net (8.12.9/8.12.8/Submit) with ESMTP id hAJLaRGb018887
for <krb5-bugs@mit.edu>; Wed, 19 Nov 2003 16:36:27 -0500
Date: Wed, 19 Nov 2003 16:36:27 -0500 (EST)
From: JLB <jlb@twu.net>
To: krb5-bugs@mit.edu
Subject: Kerberos 5 will not compile on Solaris 7/GCC or 8/GCC
Message-ID: <Pine.LNX.4.53.0311191500440.18834@twu.net>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII



http://jlb.twu.net/buildproblems/krb5/solaris7-gcc2.95.3.txt


--
Jessica L. Blank, Systems Administrator, twu.net

I'm sort of surprised you are seeing this problem. We seem to build
fine on Solaris using Gcc. It may be Solaris 9 but I would be mildly
surprised at a change that mattered.

Ok, there does appear to be a problem with fake-addrinfo.h -- if it's
included prior to any inclusion of stdio.h, it compiles code calling out
to sprintf() without a prototype. This probably is only seen on
platforms where we actually need to use our own getaddrinfo()
implementation.

This break the build of 1.3.x on Solaris 7. ASST is believed to
consider this problem critical.


In addition:

1) We seem to have compiler and CPU tests in get-addrinfo.h. That's
wrong; use a feature test macro even if the only way of setting the
feature test macro you know is based on cpu and compiler. The
separation serves as documentation and abstraction and is useful
for porting.

2) || 1 is bad form in committed code. If it really is what you
always want, then remove the other fork of the preprocessor macro.


3) IT seems that we got a thread related pull up to the 1.3.x branch based on this.