[Bug tree-optimization/103596] New: [11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88

[Bug tree-optimization/103596] New: [11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

            Bug ID: 103596
           Summary: [11/12 Regression] ICE: tree check: expected class
                    'type', have 'exceptional' (error_mark) in
                    useless_type_conversion_p, at gimple-expr.c:88
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: asolokha at gmx dot com
  Target Milestone: ---

gcc 12.0.0 20211205 snapshot (g:c9419faef0bfaf31e6a6f744baa064892e0d105c) ICEs
when compiling the following testcase w/ -O3 --param case-values-threshold=1:

int n;

void
qux (int a)
{
}

int
baz (void)
{
  return -1;
}

__attribute__ ((returns_twice)) int
bar (int b)
{
  if (n != 0)
    {
      if (b != 2)
        if (b != 0)
          return n + b;

      if (n == 2)
        return 0;
    }
}

void
foo (void)
{
  qux (n);
  bar (baz ());
}

% gcc-12.0.0 -O3 --param case-values-threshold=1 -c ubzlb8tx.c
during GIMPLE pass: fre
ubzlb8tx.c: In function 'foo':
ubzlb8tx.c:33:1: internal compiler error: tree check: expected class 'type',
have 'exceptional' (error_mark) in useless_type_conversion_p, at
gimple-expr.c:88
   33 | }
      | ^
0x78363b tree_class_check_failed(tree_node const*, tree_code_class, char
const*, int, char const*)
       
/var/tmp/portage/sys-devel/gcc-12.0.0_p20211205/work/gcc-12-20211205/gcc/tree.c:8752
0x6b6df4 tree_class_check(tree_node*, tree_code_class, char const*, int, char
const*)
       
/var/tmp/portage/sys-devel/gcc-12.0.0_p20211205/work/gcc-12-20211205/gcc/tree.h:3564
0x6b6df4 useless_type_conversion_p(tree_node*, tree_node*)
       
/var/tmp/portage/sys-devel/gcc-12.0.0_p20211205/work/gcc-12-20211205/gcc/gimple-expr.c:88
0xee4e5a verify_gimple_comparison
       
/var/tmp/portage/sys-devel/gcc-12.0.0_p20211205/work/gcc-12-20211205/gcc/tree-cfg.c:3550
0xef9fe9 verify_gimple_in_cfg(function*, bool)
       
/var/tmp/portage/sys-devel/gcc-12.0.0_p20211205/work/gcc-12-20211205/gcc/tree-cfg.c:5514
0xdbbecf execute_function_todo
       
/var/tmp/portage/sys-devel/gcc-12.0.0_p20211205/work/gcc-12-20211205/gcc/passes.c:2084
0xdbc40c execute_todo
       
/var/tmp/portage/sys-devel/gcc-12.0.0_p20211205/work/gcc-12-20211205/gcc/passes.c:2138

[Bug tree-optimization/103596] [9/10/11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[11/12 Regression] ICE:     |[9/10/11/12 Regression]
                   |tree check: expected class  |ICE: tree check: expected
                   |'type', have 'exceptional'  |class 'type', have
                   |(error_mark) in             |'exceptional' (error_mark)
                   |useless_type_conversion_p,  |in
                   |at gimple-expr.c:88         |useless_type_conversion_p,
                   |                            |at gimple-expr.c:88
   Target Milestone|---                         |9.5

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Testcase which shows this is an older bug:
int n;

void
qux (int a)
{
}

int
baz (void)
{
  return -1;
}

__attribute__ ((returns_twice)) int
bar (int b)
{
  if (n != 0)
    {
      switch (b) {
          default:  return n + b;
          case 0:
          case 2:;
      }
      if (n == 2)
        return 0;
    }
}

void
foo (void)
{
  qux (n);
  bar (baz ());
}

[Bug tree-optimization/103596] [9/10/11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |7.5.0
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-12-07
      Known to fail|                            |8.1.0

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.
Reduced testcase:
int n;

void qux (int a){}
int baz (void)
{
  return -1;
}

__attribute__ ((returns_twice)) int
bar (int b)
{
  if (n != 0)
    {
      switch (b) {
          default:  return n + b;
          case 0:
          case 2:;
      }
      if (n == 2)
        return 0;
    }
    __builtin_unreachable();
}
void
foo (void)
{
  qux (n);
  bar (baz ());
}
-------- CUT -----
Only -O3 is needed to produce the ICE with the above testcase.
And it was working in GCC 7.5.0 with the above one and started to fail in GCC
8.1.0.

[Bug tree-optimization/103596] [9/10/11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88 since r8-565-g7581ce9a1ad6df9c

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |marxin at gcc dot gnu.org,
                   |                            |rguenth at gcc dot gnu.org
           Keywords|needs-bisection             |
            Summary|[9/10/11/12 Regression]     |[9/10/11/12 Regression]
                   |ICE: tree check: expected   |ICE: tree check: expected
                   |class 'type', have          |class 'type', have
                   |'exceptional' (error_mark)  |'exceptional' (error_mark)
                   |in                          |in
                   |useless_type_conversion_p,  |useless_type_conversion_p,
                   |at gimple-expr.c:88         |at gimple-expr.c:88 since
                   |                            |r8-565-g7581ce9a1ad6df9c

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
Started with r8-565-g7581ce9a1ad6df9c.

[Bug tree-optimization/103596] [9/10/11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88 since r8-565-g7581ce9a1ad6df9c

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
           Priority|P3                          |P2

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
We have a released SSA name in the IL.

[Bug tree-optimization/103596] [9/10/11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88 since r8-565-g7581ce9a1ad6df9c

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
So the issue is that we value number

 _3(ab) = -1;
 switch (_3(ab))
  {
  }

in a way to determine the known executable edge but later do not force
propagation into the switch and may_propagate_copy disallows propagating
into the abnormal SSA name.  This is because of

  /* Similarly if DEST flows in from an abnormal edge then the copy cannot be
     propagated.  */
  else if (TREE_CODE (dest) == SSA_NAME
           && SSA_NAME_OCCURS_IN_ABNORMAL_PHI (dest))
    return false;

which is overly restrictive if the propagation does not happen into a PHI
argument on an abnormal edge.

Testing a fix.

[Bug tree-optimization/103596] [9/10/11/12 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88 since r8-565-g7581ce9a1ad6df9c

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

--- Comment #6 from CVS 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:6e8a31275fda445fb3e8d98e53f5e1541f4727af

commit r12-5821-g6e8a31275fda445fb3e8d98e53f5e1541f4727af
Author: Richard Biener <rguenther@suse.de>
Date:   Tue Dec 7 11:13:39 2021 +0100

    tree-optimization/103596 - fix missed propagation into switches

    may_propagate_copy unnecessarily restricts propagating non-abnormals
    into places that currently contain an abnormal SSA name but are
    not the PHI argument for an abnormal edge.  This causes VN to
    not elide a CFG path that it assumes is elided, resulting in
    released SSA names in the IL.

    The fix is to enhance the may_propagate_copy API to specify the
    destination is _not_ a PHI argument.  I chose to not update only
    the relevant caller in VN and the may_propagate_copy_into_stmt API
    at this point because this is a regression and needs backporting.

    2021-12-07  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/103596
            * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
            Note we are not propagating into a PHI argument to
may_propagate_copy.
            * tree-ssa-propagate.h (may_propagate_copy): Add
            argument specifying whether we propagate into a PHI arg.
            * tree-ssa-propagate.c (may_propagate_copy): Likewise.
            When not doing so we can replace an abnormal with
            something else.
            (may_propagate_into_stmt): Update may_propagate_copy calls.
            (replace_exp_1): Move propagation checking code to
            propagate_value and rename to ...
            (replace_exp): ... this and elide previous wrapper.
            (propagate_value): Perform checking with adjusted
            may_propagate_copy call and dispatch to replace_exp.

            * gcc.dg/torture/pr103596.c: New testcase.

[Bug tree-optimization/103596] [9/10/11 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88 since r8-565-g7581ce9a1ad6df9c

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by Richard Biener
<rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:08f594465fa9b6a4ea1e3816cfecf256e883ff53

commit r11-9524-g08f594465fa9b6a4ea1e3816cfecf256e883ff53
Author: Richard Biener <rguenther@suse.de>
Date:   Tue Dec 7 11:13:39 2021 +0100

    tree-optimization/103596 - fix missed propagation into switches

    may_propagate_copy unnecessarily restricts propagating non-abnormals
    into places that currently contain an abnormal SSA name but are
    not the PHI argument for an abnormal edge.  This causes VN to
    not elide a CFG path that it assumes is elided, resulting in
    released SSA names in the IL.

    The fix is to enhance the may_propagate_copy API to specify the
    destination is _not_ a PHI argument.  I chose to not update only
    the relevant caller in VN and the may_propagate_copy_into_stmt API
    at this point because this is a regression and needs backporting.

    2021-12-07  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/103596
            * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
            Note we are not propagating into a PHI argument to
may_propagate_copy.
            * tree-ssa-propagate.h (may_propagate_copy): Add
            argument specifying whether we propagate into a PHI arg.
            * tree-ssa-propagate.c (may_propagate_copy): Likewise.
            When not doing so we can replace an abnormal with
            something else.
            (may_propagate_into_stmt): Update may_propagate_copy calls.
            (replace_exp_1): Move propagation checking code to
            propagate_value and rename to ...
            (replace_exp): ... this and elide previous wrapper.
            (propagate_value): Perform checking with adjusted
            may_propagate_copy call and dispatch to replace_exp.

            * gcc.dg/torture/pr103596.c: New testcase.

    (cherry picked from commit 6e8a31275fda445fb3e8d98e53f5e1541f4727af)

[Bug tree-optimization/103596] [9/10 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88 since r8-565-g7581ce9a1ad6df9c

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Richard Biener
<rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:8a7dd8900d0a183d843e9d8ec91e091dbe4fbd5b

commit r10-10547-g8a7dd8900d0a183d843e9d8ec91e091dbe4fbd5b
Author: Richard Biener <rguenther@suse.de>
Date:   Tue Dec 7 11:13:39 2021 +0100

    tree-optimization/103596 - fix missed propagation into switches

    may_propagate_copy unnecessarily restricts propagating non-abnormals
    into places that currently contain an abnormal SSA name but are
    not the PHI argument for an abnormal edge.  This causes VN to
    not elide a CFG path that it assumes is elided, resulting in
    released SSA names in the IL.

    The fix is to enhance the may_propagate_copy API to specify the
    destination is _not_ a PHI argument.  I chose to not update only
    the relevant caller in VN and the may_propagate_copy_into_stmt API
    at this point because this is a regression and needs backporting.

    2021-12-07  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/103596
            * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
            Note we are not propagating into a PHI argument to
may_propagate_copy.
            * tree-ssa-propagate.h (may_propagate_copy): Add
            argument specifying whether we propagate into a PHI arg.
            * tree-ssa-propagate.c (may_propagate_copy): Likewise.
            When not doing so we can replace an abnormal with
            something else.
            (may_propagate_into_stmt): Update may_propagate_copy calls.
            (replace_exp_1): Move propagation checking code to
            propagate_value and rename to ...
            (replace_exp): ... this and elide previous wrapper.
            (propagate_value): Perform checking with adjusted
            may_propagate_copy call and dispatch to replace_exp.

            * gcc.dg/torture/pr103596.c: New testcase.

    (cherry picked from commit 6e8a31275fda445fb3e8d98e53f5e1541f4727af)

[Bug tree-optimization/103596] [9 Regression] ICE: tree check: expected class 'type', have 'exceptional' (error_mark) in useless_type_conversion_p, at gimple-expr.c:88 since r8-565-g7581ce9a1ad6df9c

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103596

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|9.5                         |10.4
      Known to fail|                            |9.5.0
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed for GCC 10.4.