From S.R.Kilvington@soton.ac.uk  Mon Oct 30 08:45:50 1995
Received: from beech.soton.ac.uk  for S.R.Kilvington@soton.ac.uk
	by www.ccl.net (8.6.10/950822.1) id IAA09935; Mon, 30 Oct 1995 08:41:57 -0500
Received: from willow.soton.ac.uk (willow.soton.ac.uk [152.78.128.20])
   by beech.soton.ac.uk (8.6.12/hub-8.5a) with ESMTP id NAA09625
   for <CHEMISTRY@www.ccl.net>; Mon, 30 Oct 1995 13:41:36 GMT
Received: (from srk@localhost) by willow.soton.ac.uk (8.6.10/client-8.8)
   id NAA26423 for CHEMISTRY@www.ccl.net; Mon, 30 Oct 1995 13:37:43 GMT
Message-Id: <199510301337.NAA26423@willow.soton.ac.uk>
Subject: bug in IRIX 4 C compiler
To: CHEMISTRY@www.ccl.net
Date: Mon, 30 Oct 1995 13:37:43 +0000 (GMT)
From: "Simon Kilvington" <S.R.Kilvington@soton.ac.uk>
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Content-Length: 1412      


hello,

	this is probably not the right place to post this - so, sorry in
advance. 

	anyway, onto the problem - it's not serious - the correct code is
generated - but you get a warning message cuz the compiler gets confused
about function argument types.

	the shortest program to demonstrate it is...

/* bug.c */

char proc1(char);
char proc2(char);

void
warning(int choice)
{
   char (*proc)(char);

   proc = (choice) ? proc1 : proc2;
}

	when you compile it "cc -c bug.c" the following warnings are
generated...

accom: Warning 294: bug.c, line 11: Function has prototyped and
non-prototyped declarations.  Default promotion of non-prototype form of
argument 1 does not match prototype: prototype: char is different from
actual: int
          proc = (choice) ? proc1 : proc2;
       ----------------------------------^
accom: Warning 196: bug.c, line 11: function prototypes are not
equivalent; function pointer assignment invalid
          proc = (choice) ? proc1 : proc2;
       ----------------------------------^

	for some reason it thinks the last argument to "proc1" and
"proc2" should be an int. 

	if you write...

if(choice)
   proc = proc1;
else
   proc = proc2;

	instead, all is well.

	the compilers with IRIX 5 and IRIX 6 don't have this trouble,
and as I said before, the correct code is generated - so it's more
annoying rather than anything else. 

---
Simon Kilvington, srk@soton.ac.uk