[Bug c++/109160] New: [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[Bug c++/109160] New: [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[vincent_saulue at hotmail dot fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

            Bug ID: 109160
           Summary: [Valid code] Constraint on deduced NTTP from method
                    call causes ICE/Segfault.
           Product: gcc
           Version: 12.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: vincent_saulue at hotmail dot fr
  Target Milestone: ---

To my knowledge, the following code is well-formed c++20, and compiles with
clang and MSVC.
- g++ 11.x & 12.x: this code causes a segmentation fault.
- g++ 10.x & 13.0(godbolt trunk): this code causes an ICE.

The issue is reproducible on godbolt: https://godbolt.org/z/bxbeTa69f

Code:
```
struct Traits {
    static constexpr bool value = true;
};

template<typename T, typename Traits>
concept cBarOf = Traits::value;

template<typename Traits, auto bar>
struct Foo {
    template<cBarOf<Traits> auto rhsBar> // <-- this concept causes ICE.
    void doNothing(Foo<Traits, rhsBar> const&);
};

void someFunction() {
    Foo<Traits, 0> lhs{};
    lhs.doNothing(lhs);
}
```

Compiler output on my local machine:
```
# g++ --version
g++-12 (Ubuntu 12.1.0-2ubuntu1~22.04) 12.1.0

# g++ -std=c++20 source.cpp
source.cpp: In substitution of ‘template<auto [requires ::cBarOf<<placeholder>,
Traits>] rhsBar> void Foo<Traits, 0>::doNothing(const Foo<Traits, rhsBar>&)
[with auto [requires ::cBarOf<<placeholder>, Traits>] rhsBar = <missing>]’:
source.cpp:16:18:   required from here
source.cpp:16:18: internal compiler error: Segmentation fault
   16 |     lhs.doNothing(lhs);
      |     ~~~~~~~~~~~~~^~~~~
0xd910e3 crash_signal
        ../../src/gcc/toplev.cc:322
0x7f546f77251f ???
        ./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0
0x801af0 iterative_hash_template_arg(tree_node*, unsigned int)
        ../../src/gcc/cp/pt.cc:1786
0x6e9d99 sat_hasher::hash(sat_entry*)
        ../../src/gcc/cp/constraint.cc:2491
0x6e9d99 hash_table<sat_hasher, false, xcallocator>::find_slot(sat_entry*
const&, insert_option)
        ../../src/gcc/hash-table.h:435
0x6e9d99 satisfaction_cache::satisfaction_cache(tree_node*, tree_node*,
sat_info)
        ../../src/gcc/cp/constraint.cc:2584
0x6ec446 satisfy_atom
        ../../src/gcc/cp/constraint.cc:2914
0x6ec446 satisfy_constraint_r
        ../../src/gcc/cp/constraint.cc:3023
0x6ecf33 satisfy_normalized_constraints
        ../../src/gcc/cp/constraint.cc:3048
0x6eaa53 satisfy_nondeclaration_constraints
        ../../src/gcc/cp/constraint.cc:3130
0x6eaa53 constraint_satisfaction_value
        ../../src/gcc/cp/constraint.cc:3285
0x6ecfbb constraints_satisfied_p(tree_node*, tree_node*)
        ../../src/gcc/cp/constraint.cc:3317
0x80c2c2 do_auto_deduction(tree_node*, tree_node*, tree_node*, int,
auto_deduction_context, tree_node*, int)
        ../../src/gcc/cp/pt.cc:30349
0x8166d9 unify
        ../../src/gcc/cp/pt.cc:24272
0x816791 unify
        ../../src/gcc/cp/pt.cc:24476
0x82d48e try_class_unification
        ../../src/gcc/cp/pt.cc:23437
0x81633b unify
        ../../src/gcc/cp/pt.cc:24513
0x82bfa3 unify_one_argument
        ../../src/gcc/cp/pt.cc:22650
0x814938 type_unification_real
        ../../src/gcc/cp/pt.cc:22773
0x831b0a fn_type_unification(tree_node*, tree_node*, tree_node*, tree_node*
const*, unsigned int, tree_node*, unification_kind_t, int, conversion**, bool,
bool)
        ../../src/gcc/cp/pt.cc:22096
```

Output of godbolt's gcc(trunk):
```
# g++ --version
g++
(Compiler-Explorer-Build-gcc-0c061da91a3657afdb3fac68e4595af685909a1a-binutils-2.38)
13.0.1 20230316 (experimental)

# g++ -std=c++20
<source>: In substitution of 'template<auto [requires ::cBarOf<<placeholder>,
Traits>] rhsBar> void Foo<Traits, 0>::doNothing(const Foo<Traits, rhsBar>&)
[with auto [requires ::cBarOf<<placeholder>, Traits>] rhsBar = <missing>]':
<source>:16:18:   required from here
<source>:16:18: internal compiler error: tree check: accessed elt 1 of
'tree_vec' with 0 elts in hash, at cp/constraint.cc:2571
   16 |     lhs.doNothing(lhs);
      |     ~~~~~~~~~~~~~^~~~~
0x23b52ae internal_error(char const*, ...)
        ???:0
0x94819b tree_vec_elt_check_failed(int, int, char const*, int, char const*)
        ???:0
0xafc7c4 sat_hasher::hash(sat_entry*)
        ???:0
0xaf7adf satisfaction_cache::satisfaction_cache(tree_node*, tree_node*,
sat_info)
        ???:0
0xafb922 constraints_satisfied_p(tree_node*, tree_node*)
        ???:0
0xc79882 do_auto_deduction(tree_node*, tree_node*, tree_node*, int,
auto_deduction_context, tree_node*, int)
        ???:0
0xcb0831 fn_type_unification(tree_node*, tree_node*, tree_node*, tree_node*
const*, unsigned int, tree_node*, unification_kind_t, int, conversion**, bool,
bool)
        ???:0
0xaac75a build_new_method_call(tree_node*, tree_node*, vec<tree_node*, va_gc,
vl_embed>**, tree_node*, int, tree_node**, int)
        ???:0
0xc5e0df c_parse_file()
        ???:0
0xd9ac69 c_common_parse_file()
        ???:0
```

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ppalka at gcc dot gnu.org
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2023-03-17
             Blocks|                            |67491
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |ppalka at gcc dot gnu.org

--- Comment #1 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Confirmed, this never worked properly.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67491
[Bug 67491] [meta-bug] concepts issues

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

--- Comment #2 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Patrick Palka <ppalka@gcc.gnu.org>:

https://gcc.gnu.org/g:a5de246535db1b4fdc61287f27de0fdd074fc4b3

commit r13-6971-ga5de246535db1b4fdc61287f27de0fdd074fc4b3
Author: Patrick Palka <ppalka@redhat.com>
Date:   Sat Apr 1 12:01:30 2023 -0400

    c++: NTTP constraint depending on outer parms [PR109160]

    Here we're crashing during satisfaction for the NTTP 'C<B> auto V'
    ultimately because convert_template_argument / unify don't pass all
    outer template arguments to do_auto_deduction, and during satisfaction
    we need to know all arguments.  While these callers do pass some outer
    arguments, they are only sufficient to properly substitute the
    (level-lowered) 'auto' and are not necessarily the entire set.

    Fortunately it seems these callers have access to the full set of outer
    arguments via convert_template_argument's 'in_decl' parameter and
    unify's 'tparms' parameter.  So this patch adds a new parameter to
    do_auto_deduction, used only during adc_unify deduction, through which
    these callers can pass the enclosing (partially instantiated) template
    and from which do_auto_deduction can obtain _all_ outer template
    arguments for sake of satisfaction.

    This patch also ensures that the 'in_decl' argument passed to
    coerce_template_parms is always a TEMPLATE_DECL, which in turn allows us
    to pass it as-is to do_auto_deduction; the only coerce_template_parms
    caller that needed adjustment was tsubst_decl it seems.

            PR c++/109160

    gcc/cp/ChangeLog:

            * cp-tree.h (do_auto_deduction): Add defaulted tmpl parameter.
            * pt.cc (convert_template_argument): Pass 'in_decl' as 'tmpl' to
            do_auto_deduction.
            (tsubst_decl) <case VAR_/TYPE_DECL>: Pass 'tmpl' instead of 't' as
            'in_decl' to coerce_template_parms.
            (unify) <case TEMPLATE_PARM_INDEX>: Pass TPARMS_PRIMARY_TEMPLATE
            as 'tmpl' to do_auto_deduction.
            (do_auto_deduction): Document default arguments.  Rename local
            variable 'tmpl' to 'ctmpl'.  Use 'tmpl' to obtain a full set of
            template arguments for satisfaction in the adc_unify case.

    gcc/testsuite/ChangeLog:

            * g++.dg/cpp2a/concepts-placeholder12.C: New test.

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |12.3

--- Comment #3 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Fixed for GCC 12 so far.

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

--- Comment #4 from Patrick Palka <ppalka at gcc dot gnu.org> ---
(In reply to Patrick Palka from comment #3)
> Fixed for GCC 12 so far.

GCC 13*

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[vincent_saulue at hotmail dot fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

--- Comment #5 from Vincent Saulue-Laborde <vincent_saulue at hotmail dot fr> ---
(In reply to Patrick Palka from comment #4)
> (In reply to Patrick Palka from comment #3)
> > Fixed for GCC 12 so far.
> 
> GCC 13*

I confirm that gcc trunk works fine from my side.

Thanks for the fix !

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|12.3                        |12.4

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 12.3 is being released, retargeting bugs to GCC 12.4.

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by Patrick Palka
<ppalka@gcc.gnu.org>:

https://gcc.gnu.org/g:56da68414b5b5f0124cac29756c88710e5752762

commit r12-9525-g56da68414b5b5f0124cac29756c88710e5752762
Author: Patrick Palka <ppalka@redhat.com>
Date:   Sat Apr 1 12:01:30 2023 -0400

    c++: NTTP constraint depending on outer parms [PR109160]

    Here we're crashing during satisfaction for the NTTP 'C<B> auto V'
    ultimately because convert_template_argument / unify don't pass all
    outer template arguments to do_auto_deduction, and during satisfaction
    we need to know all arguments.  While these callers do pass some outer
    arguments, they are only sufficient to properly substitute the
    (level-lowered) 'auto' and are not necessarily the entire set.

    Fortunately it seems these callers have access to the full set of outer
    arguments via convert_template_argument's 'in_decl' parameter and
    unify's 'tparms' parameter.  So this patch adds a new parameter to
    do_auto_deduction, used only during adc_unify deduction, through which
    these callers can pass the enclosing (partially instantiated) template
    and from which do_auto_deduction can obtain _all_ outer template
    arguments for sake of satisfaction.

    This patch also ensures that the 'in_decl' argument passed to
    coerce_template_parms is always a TEMPLATE_DECL, which in turn allows us
    to pass it as-is to do_auto_deduction; the only coerce_template_parms
    caller that needed adjustment was tsubst_decl it seems.

            PR c++/109160

    gcc/cp/ChangeLog:

            * cp-tree.h (do_auto_deduction): Add defaulted tmpl parameter.
            * pt.cc (convert_template_argument): Pass 'in_decl' as 'tmpl' to
            do_auto_deduction.
            (tsubst_decl) <case VAR_/TYPE_DECL>: Pass 'tmpl' instead of 't' as
            'in_decl' to coerce_template_parms.
            (unify) <case TEMPLATE_PARM_INDEX>: Pass TPARMS_PRIMARY_TEMPLATE
            as 'tmpl' to do_auto_deduction.
            (do_auto_deduction): Document default arguments.  Rename local
            variable 'tmpl' to 'ctmpl'.  Use 'tmpl' to obtain a full set of
            template arguments for satisfaction in the adc_unify case.

    gcc/testsuite/ChangeLog:

            * g++.dg/cpp2a/concepts-placeholder12.C: New test.

    (cherry picked from commit a5de246535db1b4fdc61287f27de0fdd074fc4b3)

[Bug c++/109160] [Valid code] Constraint on deduced NTTP from method call causes ICE/Segfault.

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109160

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

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

--- Comment #8 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Fixed for GCC 12.4+