[Bug c++/100925] New: tree check fail in make_range_step, at fold-const.c:5061

[Bug c++/100925] New: tree check fail in make_range_step, at fold-const.c:5061

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

            Bug ID: 100925
           Summary: tree check fail in make_range_step, at
                    fold-const.c:5061
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dcb314 at hotmail dot com
  Target Milestone: ---

For this C++ code:

struct QScopedPointerDeleter {
  static void cleanup(int *);
};
class QScopedPointer {
  typedef int *QScopedPointer::*RestrictedBool;

public:
  operator RestrictedBool() { return d ? nullptr : &QScopedPointer::d; }
  void reset() {
    if (d)
      QScopedPointerDeleter::cleanup(d);
  }
  int *d;
};
class DOpenGLPaintDevicePrivate {
public:
  QScopedPointer fbo;
} DOpenGLPaintDeviceresize_d;
void DOpenGLPaintDeviceresize() {
  if (DOpenGLPaintDeviceresize_d.fbo)
    DOpenGLPaintDeviceresize_d.fbo.reset();
}

compiled with recent gcc trunk, does this:

$ /home/dcb/gcc/results.20210605//bin/g++ -c -w -O1 bug722.cc 2>&1 | more
during GIMPLE pass: reassoc
bug722.cc: In function ‘void DOpenGLPaintDeviceresize()’:
bug722.cc:19:6: internal compiler error: tree check: expected boolean_type or 
enumeral_type or integer_type, have offset_type in make_range_step, at fold-co
nst.c:5061
   19 | void DOpenGLPaintDeviceresize() {
      |      ^~~~~~~~~~~~~~~~~~~~~~~~
0x8a9565 tree_check_failed(tree_node const*, char const*, int, char const*, ..
.)
        ../../trunk.git/gcc/tree.c:8684
0x7a0a73 any_integral_type_check(tree_node*, char const*, int, char const*)
        ../../trunk.git/gcc/tree.h:3617
0x7ad465 any_integral_type_check(tree_node*, char const*, int, char const*)
        ../../trunk.git/gcc/tree.h:3485
0x7ad465 make_range_step(unsigned int, tree_code, tree_node*, tree_node*, tree
_node*, tree_node**, tree_node**, int*, bool*)
        ../../trunk.git/gcc/fold-const.c:5061

The problem first seems to occur sometime between date 20210527 and 20210604.
Git hashes 6c67afaf524a5e0e and ad3f0ad4bafe3770.

I will have my usual go at refining the range of the git hashes.

[Bug c++/100925] tree check fail in make_range_step, at fold-const.c:5061

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

--- Comment #1 from David Binderman <dcb314 at hotmail dot com> ---
Seems ok at ea418485c700494c. Trying 7ed1cd9665d8ca0f

[Bug c++/100925] tree check fail in make_range_step, at fold-const.c:5061

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

--- Comment #2 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #1)
> Seems ok at ea418485c700494c. Trying 7ed1cd9665d8ca0f

Tried it and it seems bad. Range seems to be (ea418485c700494c,
7ed1cd9665d8ca0f).
Trying d2d74c9fc0cf46f6.

[Bug c++/100925] tree check fail in make_range_step, at fold-const.c:5061

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

--- Comment #3 from David Binderman <dcb314 at hotmail dot com> ---
That one seems bad, so current range is (ea418485c700494c..d2d74c9fc0cf46f6),
so 15 revisions left.

[Bug tree-optimization/100925] [12 Regression] tree check fail in make_range_step with -O1 in reassoc

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-06-05
                 CC|                            |pinskia at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
It was exposed by r12-1152.

Phi-opt produces this now:
  _9 = _3 != 0B;
  _10 = (<<< Unknown tree: offset_type >>>) _9;
  _11 = -_10;
/// The above is new

  _6 = _3 != 0B;
  _7 = _11 != -1;
  _8 = _6 & _7;

----- CUT ----
before reassoc we have:
  _9 = _3 != 0B;
  _10 = (<<< Unknown tree: offset_type >>>) _9;
  _11 = -_10;
  _7 = _11 != -1;
  _8 = _7 & _9;
---- CUT ---

But _9 and _7 will always have the same value even.

[Bug tree-optimization/100925] [12 Regression] tree check fail in make_range_step with -O1 in reassoc

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
   Target Milestone|---                         |12.0
           Assignee|unassigned at gcc dot gnu.org      |pinskia at gcc dot gnu.org

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #4)
> It was exposed by r12-1152.
> 
> Phi-opt produces this now:
>   _9 = _3 != 0B;
>   _10 = (<<< Unknown tree: offset_type >>>) _9;
>   _11 = -_10;
> /// The above is new
> 
>   _6 = _3 != 0B;
>   _7 = _11 != -1;
>   _8 = _6 & _7;
> 
> ----- CUT ----
> before reassoc we have:
>   _9 = _3 != 0B;
>   _10 = (<<< Unknown tree: offset_type >>>) _9;
>   _11 = -_10;
>   _7 = _11 != -1;
>   _8 = _7 & _9;
> ---- CUT ---
> 
> But _9 and _7 will always have the same value even.

The problem is r12-1055 really which causes the offset type to used here.  
I need a check for INTEGRAL_TYPE_P to avoid this.

[Bug tree-optimization/100925] [12 Regression] tree check fail in make_range_step with -O1 in reassoc

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=53336

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Related to the older PR 53336 bug.

[Bug tree-optimization/100925] [12 Regression] tree check fail in make_range_step with -O1 in reassoc

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |https://gcc.gnu.org/piperma
                   |                            |il/gcc-patches/2021-June/57
                   |                            |2317.html
           Keywords|                            |patch

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Patch submitted:
https://gcc.gnu.org/pipermail/gcc-patches/2021-June/572317.html

[Bug tree-optimization/100925] [12 Regression] tree check fail in make_range_step with -O1 in reassoc

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:d4faa36e7540c573c5dc17850bcd938d0900b2e9

commit r12-1349-gd4faa36e7540c573c5dc17850bcd938d0900b2e9
Author: Andrew Pinski <apinski@marvell.com>
Date:   Sat Jun 5 21:25:58 2021 -0700

    Fix PR 100925: Limit some a?CST1:CST2 optimizations to intergal types only

    The problem here is with offset (and pointer) types is we produce
    a negative expression when this optimization hits.
    It is easier to disable this optimization for all non-integeral types
    instead of finding an integer type which is the same precission as the
    type to do the negative expression on it.

    OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.

    gcc/ChangeLog:

            PR tree-optimization/100925
            * match.pd (a ? CST1 : CST2): Limit transformations
            that would produce a negative to integeral types only.
            Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.

    gcc/testsuite/ChangeLog:

            * g++.dg/torture/pr100925.C: New test.

[Bug tree-optimization/100925] [12 Regression] tree check fail in make_range_step with -O1 in reassoc

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

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

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

--- Comment #9 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed.

[Bug tree-optimization/100925] [12 Regression] tree check fail in make_range_step with -O1 in reassoc

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100925

--- Comment #10 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:2f1686ff70b25fceb04ca2ffc0a450fb682913ef

commit r12-1608-g2f1686ff70b25fceb04ca2ffc0a450fb682913ef
Author: Andrew Pinski <apinski@marvell.com>
Date:   Sat Jun 5 19:03:06 2021 -0700

    Disallow pointer and offset types on some gimple

    While debugging PR 100925, I found that the gimple verifiers
    don't reject NEGATE on pointer or offset type.
    This patch adds the check on some unary and binary gimple which
    should not have operated on pointer/offset types.

    OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.

    Thanks,
    Andrew Pinski

    gcc/ChangeLog:

            * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
            types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and
CNONJ_EXPR.
            (verify_gimple_assign_binary): Reject point and offset types on
            MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
            FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
            FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.