[Bug target/113083] New: [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[Bug target/113083] New: [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[mjires at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

            Bug ID: 113083
           Summary: [14 Regression][arm] ICE in fold_convert_loc, at
                    fold-const.cc:2602 since r14-5979-g99d114c15523e0
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: mjires at suse dot cz
                CC: polacek at redhat dot com
  Target Milestone: ---
            Target: arm

Compiling reduced testcase g++.dg/cpp0x/constexpr-98295.C results in ICE since
r14-5979-g99d114c15523e0.

$ cat constexpr-98295.C
struct A { constexpr A(); };

void f() {
  A b;
}

constexpr A::A() {}


$ arm-linux-gnueabi-g++ constexpr-98295.C -Os
constexpr-98295.C: In constructor ‘constexpr A::A()’:
constexpr-98295.C:7:19: internal compiler error: in fold_convert_loc, at
fold-const.cc:2602
    7 | constexpr A::A() {}
      |                   ^
0x1517af9 fold_convert_loc(unsigned int, tree_node*, tree_node*)
        /home/mjires/git/GCC/master/gcc/fold-const.cc:2602
0x1613814 gimplify_modify_expr
        /home/mjires/git/GCC/master/gcc/gimplify.cc:6353
0x16471a1 gimplify_expr(tree_node**, gimple**, gimple**, bool (*)(tree_node*),
int)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:17640
0x1617f67 gimplify_stmt(tree_node**, gimple**)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:7477
0x15ff631 gimplify_and_add(tree_node*, gimple**)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:493
0x16045a5 gimplify_return_expr
        /home/mjires/git/GCC/master/gcc/gimplify.cc:1883
0x16482da gimplify_expr(tree_node**, gimple**, gimple**, bool (*)(tree_node*),
int)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:17902
0x1617f67 gimplify_stmt(tree_node**, gimple**)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:7477
0x1605a4d gimplify_statement_list
        /home/mjires/git/GCC/master/gcc/gimplify.cc:2222
0x1649002 gimplify_expr(tree_node**, gimple**, gimple**, bool (*)(tree_node*),
int)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:18085
0x1617f67 gimplify_stmt(tree_node**, gimple**)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:7477
0x1603668 gimplify_bind_expr
        /home/mjires/git/GCC/master/gcc/gimplify.cc:1614
0x1647f13 gimplify_expr(tree_node**, gimple**, gimple**, bool (*)(tree_node*),
int)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:17841
0x1617f67 gimplify_stmt(tree_node**, gimple**)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:7477
0x164bdb4 gimplify_body(tree_node*, bool)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:18907
0x164c6a9 gimplify_function_tree(tree_node*)
        /home/mjires/git/GCC/master/gcc/gimplify.cc:19106
0x13a0cb1 cgraph_node::analyze()
        /home/mjires/git/GCC/master/gcc/cgraphunit.cc:685
0x13a2d6d analyze_functions
        /home/mjires/git/GCC/master/gcc/cgraphunit.cc:1248
0x13a62f3 symbol_table::finalize_compilation_unit()
        /home/mjires/git/GCC/master/gcc/cgraphunit.cc:2555
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.


$ arm-linux-gnueabi-g++ -v
Using built-in specs.
COLLECT_GCC=arm-linux-gnueabi-g++
COLLECT_LTO_WRAPPER=/home/mjires/built/master/libexec/gcc/arm-linux-gnueabi/14.0.0/lto-wrapper
Target: arm-linux-gnueabi
Configured with: /home/mjires/git/GCC/master/configure
--prefix=/home/mjires/built/master --target=arm-linux-gnueabi
--disable-bootstrap --enable-languages=c,c++,fortran --disable-multilib
--disable-libsanitizer --enable-checking
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.0 20231219 (experimental) (GCC)

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|target                      |c++
   Target Milestone|---                         |14.0

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Maybe targetm.cxx.cdtor_returns_this () related?

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mpolacek at gcc dot gnu.org
   Last reconfirmed|                            |2023-12-19
           Assignee|unassigned at gcc dot gnu.org      |mpolacek at gcc dot gnu.org
             Status|UNCONFIRMED                 |ASSIGNED
           Priority|P3                          |P1
     Ever confirmed|0                           |1

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

--- Comment #2 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Previously we had:
{
  return A::A (this);
}

now:
{
  return *this = {};
}

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

--- Comment #3 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
The ICE can be reproduced with -O -fdeclone-ctor-dtor as well.

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

--- Comment #4 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
The problem occurs only when we declone cdtors and are on a
targetm.cxx.cdtor_returns_this target like ARM.

Decloning causes us to create a thunk calling the "main" ctor:

A*
A::A (A *const this)
{
  return A::A (this);
}

'this' is now accepted in constexpr so we evaluate the call into {} and end up
with

{
  return *this = {};
}

which is gimplified into

{
  A *D.4937;
  D.4937 = *this = {};
}

but that means there's a discrepancy: we're converting A to A* and that
crashes.

I wonder if we should refuse to evaluate A::A (this) (returning a pointer) into
{} (not a pointer).

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

Jason Merrill <jason at gcc dot gnu.org> changed:

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

--- Comment #5 from Jason Merrill <jason at gcc dot gnu.org> ---
(In reply to Marek Polacek from comment #4)
> I wonder if we should refuse to evaluate A::A (this) (returning a pointer)
> into {} (not a pointer).

If the constructor returns a pointer, instead of maybe_constant_value
evaluating A::A (this) to (*this = {}), perhaps it should become (*this = {},
this)?

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
What about:
2024-02-22  Jakub Jelinek  <jakub@redhat.com>

        PR c++/113083
        * cp-gimplify.cc (cp_fold): For targetm.cxx.cdtor_returns_this ()
        wrap r into a COMPOUND_EXPR and return folded CALL_EXPR_ARG (x, 0).

        * g++.dg/cpp0x/constexpr-113083.C: New test.

--- gcc/cp/cp-gimplify.cc.jj    2024-02-22 21:45:09.663430066 +0100
+++ gcc/cp/cp-gimplify.cc       2024-02-22 22:16:41.816591451 +0100
@@ -3415,6 +3415,10 @@ cp_fold (tree x, fold_flags_t flags)
                tree s = build_fold_indirect_ref_loc (loc,
                                                      CALL_EXPR_ARG (x, 0));
                r = cp_build_init_expr (s, r);
+               if (targetm.cxx.cdtor_returns_this ())
+                 r = build2 (COMPOUND_EXPR, TREE_TYPE (x), r,
+                             fold_convert (TREE_TYPE (x),
+                                           CALL_EXPR_ARG (x, 0)));
              }
            x = r;
            break;
--- gcc/testsuite/g++.dg/cpp0x/constexpr-113083.C.jj    2024-01-13
00:05:00.077372302 +0100
+++ gcc/testsuite/g++.dg/cpp0x/constexpr-113083.C       2024-02-22
22:20:20.622618992 +0100
@@ -0,0 +1,16 @@
+// PR c++/113083
+// { dg-do compile { target c++11 } }
+// { dg-options "-Os" }
+
+struct A { constexpr A (); };
+
+void
+foo ()
+{
+  A b;
+}
+
+constexpr
+A::A ()
+{
+}

Though, perhaps we need cp_save_expr in that case to make sure if there were
side-effects in CALL_EXPR (x, 0) that we wouldn't evaluate them twice.

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
As in:
2024-02-22  Jakub Jelinek  <jakub@redhat.com>

        PR c++/113083
        * cp-gimplify.cc (cp_fold): For targetm.cxx.cdtor_returns_this ()
        wrap r into a COMPOUND_EXPR and return folded CALL_EXPR_ARG (x, 0).

        * g++.dg/cpp0x/constexpr-113083.C: New test.

--- gcc/cp/cp-gimplify.cc.jj    2024-02-22 21:45:09.663430066 +0100
+++ gcc/cp/cp-gimplify.cc       2024-02-22 22:30:23.481428242 +0100
@@ -3412,9 +3412,15 @@ cp_fold (tree x, fold_flags_t flags)
            if (DECL_CONSTRUCTOR_P (callee))
              {
                loc = EXPR_LOCATION (x);
-               tree s = build_fold_indirect_ref_loc (loc,
-                                                     CALL_EXPR_ARG (x, 0));
+               tree a = CALL_EXPR_ARG (x, 0);
+               bool return_this = targetm.cxx.cdtor_returns_this ();
+               if (return_this)
+                 a = cp_save_expr (a);
+               tree s = build_fold_indirect_ref_loc (loc, a);
                r = cp_build_init_expr (s, r);
+               if (return_this)
+                 r = build2_loc (loc, COMPOUND_EXPR, TREE_TYPE (x), r,
+                                 fold_convert_loc (loc, TREE_TYPE (x), a));
              }
            x = r;
            break;
--- gcc/testsuite/g++.dg/cpp0x/constexpr-113083.C.jj    2024-01-13
00:05:00.077372302 +0100
+++ gcc/testsuite/g++.dg/cpp0x/constexpr-113083.C       2024-02-22
22:20:20.622618992 +0100
@@ -0,0 +1,16 @@
+// PR c++/113083
+// { dg-do compile { target c++11 } }
+// { dg-options "-Os" }
+
+struct A { constexpr A (); };
+
+void
+foo ()
+{
+  A b;
+}
+
+constexpr
+A::A ()
+{
+}

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

--- Comment #8 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:fdf9df9d55802e1d8ff0bd14585ea61b2bb9d798

commit r14-9158-gfdf9df9d55802e1d8ff0bd14585ea61b2bb9d798
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Fri Feb 23 18:55:12 2024 +0100

    c++: Fix ICE due to folding a call to constructor on cdtor_returns_this
arches (aka arm32) [PR113083]

    When targetm.cxx.cdtor_returns_this () (aka on arm32 TARGET_AAPCS_BASED)
    constructor is supposed to return this pointer, but when we cp_fold such
    a call, we don't take that into account and just INIT_EXPR the object,
    so we can later ICE during gimplification, because the expression doesn't
    have the right type.

    2024-02-23  Jakub Jelinek  <jakub@redhat.com>

            PR c++/113083
            * cp-gimplify.cc (cp_fold): For targetm.cxx.cdtor_returns_this ()
            wrap r into a COMPOUND_EXPR and return folded CALL_EXPR_ARG (x, 0).

            * g++.dg/cpp0x/constexpr-113083.C: New test.

[Bug c++/113083] [14 Regression][arm] ICE in fold_convert_loc, at fold-const.cc:2602 since r14-5979-g99d114c15523e0

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113083

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.