IPCP testcase for PR29122

Razya Ladelsky RAZYA@il.ibm.com
Tue Nov 21 07:53:00 GMT 2006

Janis Johnson <janis187@us.ibm.com> wrote on 20/11/2006 23:45:09:

> On Mon, Nov 20, 2006 at 04:39:35PM +0200, Razya Ladelsky wrote:
> > The fix for PR29122 was approved by Diego and will be committed 
> > I want to add this testcase to the IPA testuite.
> > 
> > I wasn't sure whether  { target lp64 } is aiming correctly to the 
> > that enable -m64.
> > Is it ?
> { target lp64 } says to run the test only if it will be compiled as
> LP64.  This include testing a compiler that always generates LP64 code,
> and biarch compilers with options to generate LP64 code.  Most (I hope)
> people who test biarch compilers run the tests with both -m32 and -m64
> (or whatever is appropriate for that system), so tests should not use
> those options explicitly.
> Unless there is a reason why the test should be restricted to LP64
> code, remove both the restriction and the -m64.
> Leave in other options that were necessary to trigger the ICE.

The ICE occured only when -m64 was enabled.
What can I do in this case?

> > O.K for GCC4.3?
> > 
> > 2006-11-20 Razya Ladelsky  <razya@il.ibm.com>
> > 
> >         * testsuite/gcc.dg/ipa/ipa-6.c: New.
> If the -fdump-ipa-cp option was necessary for the ICE, leave it in but
> add a dg-final directive to remove the dump file; look at other tests
> that use that option to see what's needed.
> Is the call to printf necessary, or can it be replaced with something
> else?  We try to avoid references to outside libraries.
> Janis

More information about the Gcc-patches mailing list