[Bug c++/112767] New: [14 regression] spurious -Waggressive-loop-optimizations

[Bug c++/112767] New: [14 regression] spurious -Waggressive-loop-optimizations

[sss@li-snyder.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112767

            Bug ID: 112767
           Summary: [14 regression] spurious
                    -Waggressive-loop-optimizations
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: sss@li-snyder.org
  Target Milestone: ---

With a recent checkout of gcc14 (20231129), on a x86_64-pc-linux-gnu host,
the following source gives a seemingly bogus -Waggressive-loop-optimizations
warning with -O3:

--------------------------------------------------------------
double reg_dict[32];

void foo(int);

void initialize()
{
  int i=8;
  for (int phi=0; phi<8; ++phi) {
    reg_dict[i]=0;
    int sn = 0;
    if (i < 16) sn = 20;
    foo(sn);
    ++i;
  }
}
--------------------------------------------------------------

$ g++ -c -O3 x.cc
x.cc: In function ‘void initialize()’:
x.cc:9:16: warning: iteration 16 invokes undefined behavior
[-Waggressive-loop-optimizations]
    9 |     reg_dict[i]=0;
      |     ~~~~~~~~~~~^~
x.cc:8:22: note: within this loop
    8 |   for (int phi=0; phi<8; ++phi) {
      |                   ~~~^~


This is a confusing warning, as the containing loop is executed only 8 times.
Exactly which undefined behavior is meant is also not spelled out.
From the context, i'd guess it to be an out-of-bounds array write,
but that's a false positive.

The warning happens with -O3 but not with lower optimization levels.
It also does not occur with gcc 13.2.1 20230728.

[Bug tree-optimization/112767] [14 regression] spurious -Waggressive-loop-optimizations

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112767

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-11-29
     Ever confirmed|0                           |1
   Target Milestone|---                         |14.0
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed. Looks like loop split is messing up ...

[Bug tree-optimization/112767] [14 regression] spurious -Waggressive-loop-optimizations

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112767

Patrick Palka <ppalka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ppalka at gcc dot gnu.org

--- Comment #2 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Started with r14-2944-g3d48c11ad082de

[Bug tree-optimization/112767] [14 regression] spurious -Waggressive-loop-optimizations since r14-2944-g3d48c11ad082de

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112767

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.  Now with SCCP after loop splitting that's the pass
exposing enough info but not eliding the loop before IVCANON comes and
diagnoses this.

SCCP does

   <bb 17> [local count: 119292719]:
-  # i_23 = PHI <i_8(3)>
-  # phi_24 = PHI <phi_9(3)>
+  i_23 = 16;
+  phi_24 = 8;
   if (phi_24 != 8)

and only copyprop3 removes the condition.

[Bug tree-optimization/112767] [14 regression] spurious -Waggressive-loop-optimizations since r14-2944-g3d48c11ad082de

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112767

--- Comment #4 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:2dde9f326ded84814a78c3044294b535c1f97b41

commit r14-6010-g2dde9f326ded84814a78c3044294b535c1f97b41
Author: Richard Biener <rguenther@suse.de>
Date:   Thu Nov 30 12:38:53 2023 +0100

    tree-optimization/112767 - spurious diagnostic after sccp/loop-split swap

    We are diagnosing an unreachable loop which we only manage to elide
    after the copyprop pass after sccp which leaves the code open for
    diagnosing by the intermittent ivcanon pass.  The following makes sure
    to clean things up a bit earlier, propagating constant final values
    to uses immediately.

            PR tree-optimization/112767
            * tree-scalar-evolution.cc (final_value_replacement_loop):
            Propagate constants to immediate uses immediately.

            * gcc.dg/tree-ssa/pr112767.c: New testcase.
            * gcc.dg/graphite/pr83255.c: Disable SCCP.

[Bug tree-optimization/112767] [14 regression] spurious -Waggressive-loop-optimizations since r14-2944-g3d48c11ad082de

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112767

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.