From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jan Hubicka <jh@suse.cz>
To: "Joseph S. Myers" <jsm28@cam.ac.uk>
Cc: Jan Hubicka <jh@suse.cz>, gcc-patches@gcc.gnu.org
Subject: Re: spec2000 regression
Date: Wed, 15 Aug 2001 07:36:00 -0000
Message-id: <20010815163652.R19872@atrey.karlin.mff.cuni.cz>
References: <20010815144127.V19872@atrey.karlin.mff.cuni.cz> <Pine.LNX.4.33.0108151432240.8933-100000@kern.srcf.societies.cam.ac.uk>
X-SW-Source: 2001-08/msg00922.html

> On Wed, 15 Aug 2001, Jan Hubicka wrote:
> 
> > I am now testing following patch. OK to install if it succeeds?
> 
> You should add a testcase to the testsuite for this bug.  You can do this
> by calling a non-existent function such that the call will be optimised
> away if the optimisation succeeds.  (Look e.g. at
> gcc.c-torture/execute/builtin-abs-1.c - in this case you'll want to use
> 'set additional_flags "-fstrict-aliasing"' in the .x file so that the
> optimisation works at -O1 (which doesn't include -fstrict-aliasing) and
> #ifndef __OPTIMIZE__ is the only test needed for the dummy version of the
> function that aborts.)
OK,
Would be following testsuite OK?

long long a1, *a=&a1;
short b1, *b=&b1;
main()
{
#ifdef __OPTIMIZE__
  *a = 1;
  *b = 2;
  if (*a != 1)
	  undefined ();
#endif
  return 0;
}