From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
	id 9B2873858C52; Sun, 19 Feb 2023 03:55:40 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 9B2873858C52
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1676778940;
	bh=D96cVz9tDAHZL1I5kS+GgUqD52vsbtl9d16bq9cJalk=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=T3EzMaTA+yeUKgta0vQ7xpn2QJv6I9DD4H77LmKk50phv4kklx7R8OPuOpe5zxzT1
	 QjPea/PR67bMeqohsVq/rjeTbnHT0UfxHKdniYX/Kaim/aIqb7MvkSGD7ikbY0q/hk
	 SGwz6Nx/0NmLtCdO7pRGaGBYC8DyQnp/CJuQSwt8=
From: "qrzhang at gatech dot edu" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug sanitizer/108845] Unnecessary signed integer overflow checks
Date: Sun, 19 Feb 2023 03:55:39 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: sanitizer
X-Bugzilla-Version: unknown
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: qrzhang at gatech dot edu
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-108845-4-yTJALabnGN@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-108845-4@http.gcc.gnu.org/bugzilla/>
References: <bug-108845-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
List-Id: <gcc-bugs.sourceware.org>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D108845

--- Comment #4 from Qirun Zhang <qrzhang at gatech dot edu> ---
(In reply to Qirun Zhang from comment #3)
> (In reply to Jakub Jelinek from comment #2)
> > I'm not convinced it is a good idea.
> > Sure, in the above case it is obvious it will never trigger, but if we =
say
> > use ranger to decide if the operation can or can't overflow, then VRP i=
s in
> > many cases based on assumptions which only hold for valid code, but
> > sanitizers actually want to diagnose invalid code.
>=20
>=20
> Thanks!
>=20
> Here is another (similar) example.  Earlier versions of GCC will not inje=
ct
> UBSAN_CHECK_ADD. However, the latest version of GCC will.
>=20
> the code example:
> =3D=3D=3D=3D=3D=3D
> void main() {
>   int a =3D 0;
>   for (; a !=3D 2; a++)
>     ;
> }
> =3D=3D=3D=3D=3D=3D
>=20
> Compile with "gcc-11 -fsanitize=3Dsigned-integer-overflow -O3=20
> -fdump-tree-optimized", we got no UBSAN checks:
>=20
> =3D=3D=3D=3D=3D=3D
> void main ()
> {
>   int a;
>=20
>   <bb 2> [local count: 118111600]:
>=20
>   <bb 3> [local count: 955630225]:
>   # a_6 =3D PHI <1(3), 0(2)>
>   a_3 =3D a_6 + 1;
>   if (a_3 !=3D 2)
>     goto <bb 3>; [87.64%]
>   else
>     goto <bb 4>; [12.36%]
>=20
>   <bb 4> [local count: 118111600]:
>   return;
>=20
> }
> =3D=3D=3D=3D=3D=3D
>=20
> Compile with "gcc-trunk -fsanitize=3Dsigned-integer-overflow -O3=20
> -fdump-tree-optimized", we got one:
>=20
> =3D=3D=3D=3D=3D=3D
> void main ()
> {
>   int a;
>=20
>   <bb 2> [local count: 118111600]:
>=20
>   <bb 3> [local count: 955630225]:
>   # a_5 =3D PHI <a_3(3), 0(2)>
>   a_3 =3D .UBSAN_CHECK_ADD (a_5, 1);
>   if (a_3 !=3D 2)
>     goto <bb 3>; [89.00%]
>   else
>     goto <bb 4>; [11.00%]
>=20
>   <bb 4> [local count: 118111600]:
>   return;
>=20
> }
> =3D=3D=3D=3D=3D=3D
>=20
> $ gcc-trunk -v
> gcc version 13.0.1 20230218 (experimental) [master r13-6132-g32b5875c911]
> (GCC)

I did a bisect. For the testcase provided in comment #3, the behavior was
introduced in the following commit:


commit 502ffb1f389011b28ee51815242c7397790802d5
Author: Andrew MacLeod <amacleod@redhat.com>
Date:   Tue Nov 2 21:26:44 2021 -0400

    Switch vrp2 to ranger.

    This patch flips the default for the VRP2 pass to execute ranger vrp ra=
ther
    than the assert_expr version of VRP.

            * params.opt (param_vrp2_mode): Make ranger the default for VRP=
2.=