This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH][4/5] Handle internal_fn in operand_equal_p
- From: Richard Biener <rguenther at suse dot de>
- To: Tom de Vries <Tom_deVries at mentor dot com>
- Cc: Jakub Jelinek <jakub at redhat dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>, Michael Matz <matz at suse dot de>
- Date: Fri, 20 Feb 2015 12:54:24 +0100 (CET)
- Subject: Re: [PATCH][4/5] Handle internal_fn in operand_equal_p
- Authentication-results: sourceware.org; auth=none
- References: <54E5BB06 dot 2080102 at mentor dot com> <54E5C403 dot 7050606 at mentor dot com> <alpine dot LSU dot 2 dot 11 dot 1502191343240 dot 28824 at zhemvz dot fhfr dot qr> <20150219125122 dot GF1746 at tucnak dot redhat dot com> <alpine dot LSU dot 2 dot 11 dot 1502191406130 dot 28824 at zhemvz dot fhfr dot qr> <54E5FB07 dot 1010307 at mentor dot com>
On Thu, 19 Feb 2015, Tom de Vries wrote:
> On 19-02-15 14:07, Richard Biener wrote:
> > On Thu, 19 Feb 2015, Jakub Jelinek wrote:
> >
> > > On Thu, Feb 19, 2015 at 01:44:46PM +0100, Richard Biener wrote:
> > > > I'd call it a bug though, and we do have internal fns in
> > > > generic already thus the issue is latent (with ubsan at least).
> > > >
> > > > Which means ok for trunk now.
> > >
> > > But the patch should better handle the internal calls right.
> > > I.e. return 0 only if only one, not both CALL_EXPR_FNs are NULL,
> > > or if both are NULL and CALL_EXPR_IFN is different, or if
> > > call_expr_nargs is different.
> >
> > The question is whether generic call handling works (esp. call_expr_flags
> > works correctly - the argument compare should work fine already).
> >
> > Tom - care to update the patch?
> >
>
> I agree, the current patch is conservative and we can do better.
> But I think it's wiser to do that as a stage1 follow-up, and commit this
> conservative patch for stage4. Is that acceptable?
Then just defer it for stage1 completely. If a problem pops up with
GCC 5 we can backport the proper patch together with a testcase.
Richard.