[Bug c++/96821] New: [concepts] Incorrect evaluation of concept with ill-formed expression

[Bug c++/96821] New: [concepts] Incorrect evaluation of concept with ill-formed expression

[dudkindaniilm2 at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

            Bug ID: 96821
           Summary: [concepts] Incorrect evaluation of concept with
                    ill-formed expression
           Product: gcc
           Version: 10.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dudkindaniilm2 at yandex dot ru
  Target Milestone: ---

[Bug c++/96821] [concepts] Incorrect evaluation of concept with ill-formed expression

[dudkindaniilm2 at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

--- Comment #1 from Daniil Dudkin <dudkindaniilm2 at yandex dot ru> ---
Considering the code below the correct output would be (without the
indentation): 
  Called with constant value
  Called with non-const value
But the actual output is:
  Called with constant value
  Called with constant value

The code:
#include <cstdio>

struct WithConstant {
    static constexpr auto value = true;
};

struct WithNotConstant {
    static inline auto value = false;
};

template <auto>
concept constant_expression = true;

template <typename T>
concept with_value_constant = constant_expression<T::value>;

template <with_value_constant T>
void foo() {
    std::puts("Called with constant value");
}

template <typename T>
void foo() {
    std::puts("Called with non-const value");
}

int main() {
    foo<WithConstant>();
    foo<WithNotConstant>();
}


Compile and run logs:

$ g++-10 -v
Using built-in specs.
COLLECT_GCC=g++-10
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/10/lto-wrapper
OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa:hsa
OFFLOAD_TARGET_DEFAULT=1
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu
10-20200411-0ubuntu1' --with-bugurl=file:///usr/share/doc/gcc-10/README.Bugs
--enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --prefix=/usr
--with-gcc-major-version-only --program-suffix=-10
--program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id
--libexecdir=/usr/lib --without-included-gettext --enable-threads=posix
--libdir=/usr/lib --enable-nls --enable-clocale=gnu --enable-libstdcxx-debug
--enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new
--enable-gnu-unique-object --disable-vtable-verify --enable-plugin
--enable-default-pie --with-system-zlib --enable-libphobos-checking=release
--with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch
--disable-werror --with-arch-32=i686 --with-abi=m64
--with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic
--enable-offload-targets=nvptx-none,amdgcn-amdhsa,hsa --without-cuda-driver
--enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu
--target=x86_64-linux-gnu
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 10.0.1 20200411 (experimental) [master revision
bb87d5cc77d:75961caccb7:f883c46b4877f637e0fa5025b4d6b5c9040ec566] (Ubuntu
10-20200411-0ubuntu1)

$ g++-10 bug.cpp -Wall -Wextra -std=c++20
$ ./a.out
Called with constant value
Called with constant value

The godbolt link: https://godbolt.org/z/e3zvfc

The problem is reproducible with gcc 10.1, gcc 10.2 and gcc trunk also

[Bug c++/96821] [concepts] Incorrect evaluation of concept with ill-formed expression

[dudkindaniilm2 at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

--- Comment #2 from Daniil Dudkin <dudkindaniilm2 at yandex dot ru> ---
Basically it described in 4th paragraph in [temp.constr.decl] of the latest C++
standard draft: 
http://eel.is/c++draft/temp.constr.decl#4

[Bug c++/96821] [concepts] Incorrect evaluation of concept with ill-formed expression

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ppalka at gcc dot gnu.org
      Known to fail|                            |10.2.0, 11.0, 9.3.0
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2020-09-21

--- Comment #3 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Confirmed.  One workaround is to define with_value_constant as

  template <typename T>
  concept with_value_constant
    = requires { constant_expression<T::value>; };

[Bug c++/96821] [concepts] Incorrect evaluation of concept with ill-formed expression

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |INVALID

--- Comment #4 from Patrick Palka <ppalka at gcc dot gnu.org> ---
GCC's behaviour is correct, I think.  Since the concept constant_expression
doesn't use its template parameter, the normal form of foo's associated
constraint is just 'true (with an empty parameter mapping)', so the
satisfaction value of with_value_constant<T> for any T is trivially true and
independent of T.

Another workaround to have the with_value_constant concept work as you expect
is to make the constant_expression concept depend on its template parameter in
a trivial way, e.g. define it as:

  template <auto V>
  concept constant_expression = requires { V; };

[Bug c++/96821] [concepts] Incorrect evaluation of concept with ill-formed expression

[dudkindaniilm2 at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

Daniil Dudkin <dudkindaniilm2 at yandex dot ru> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|INVALID                     |---
             Status|RESOLVED                    |REOPENED

--- Comment #5 from Daniil Dudkin <dudkindaniilm2 at yandex dot ru> ---

(In reply to Patrick Palka from comment #4)
> GCC's behaviour is correct, I think.  Since the concept constant_expression
> doesn't use its template parameter, the normal form of foo's associated
> constraint is just 'true (with an empty parameter mapping)', so the
> satisfaction value of with_value_constant<T> for any T is trivially true and
> independent of T.


No, I don't think that GCC behaviour is correct. 

http://eel.is/c++draft/temp.constr#atomic-3

> If substitution results in an invalid type or expression, the constraint is not satisfied.

The substitution with_value_constant<WithNotConstant::value> results in an
invalid expression because WithNotConstant::value is not a constant expression,
so it cannot be used as a non-type template argument. That means
with_value_constant<WithNotConstant::value> should be false regardless how
with_value_constant is defined.

[Bug c++/96821] [concepts] Incorrect evaluation of concept with ill-formed expression

[zamazan4ik at tut dot by]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

Alexander Zaitsev <zamazan4ik at tut dot by> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |zamazan4ik at tut dot by

--- Comment #6 from Alexander Zaitsev <zamazan4ik at tut dot by> ---
Any updates on the issue? Such behaviour is strange too since Clang and MSVC
have a different opinion from GCC for the code.

[Bug c++/96821] [concepts] Incorrect evaluation of concept with ill-formed expression

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96821

--- Comment #7 from Patrick Palka <ppalka at gcc dot gnu.org> ---
(In reply to Daniil Dudkin from comment #5)
> (In reply to Patrick Palka from comment #4)
> > GCC's behaviour is correct, I think.  Since the concept constant_expression
> > doesn't use its template parameter, the normal form of foo's associated
> > constraint is just 'true (with an empty parameter mapping)', so the
> > satisfaction value of with_value_constant<T> for any T is trivially true and
> > independent of T.
> 
> 
> No, I don't think that GCC behaviour is correct. 
> 
> http://eel.is/c++draft/temp.constr#atomic-3
> 
> > If substitution results in an invalid type or expression, the constraint is not satisfied.

'Substitution' here refers to substitution into the atomic constraints of a
constraint-expression, not into the constraint-expression directly.  Atomic
constraints are formed via constraint normalization.  In the case of
with_value_constant<T>, normalization of this constraint-expression yields the
single atomic constraint 'true (with an empty parameter mapping)' as per
[temp.constr.atomic]/1 and [temp.constr.normal].  The substitution
'T=WithNonConstant' into this atomic constraint will never fail, because the
atomic constraint doesn't depend on any template parameters.

> 
> The substitution with_value_constant<WithNotConstant::value> results in an
> invalid expression because WithNotConstant::value is not a constant
> expression, so it cannot be used as a non-type template argument. That means
> with_value_constant<WithNotConstant::value> should be false regardless how
> with_value_constant is defined.

As mentioned before, you need to make the concept constant_expression depend on
its template parameter so that normalization doesn't throw away the 'T::value'
template argument inside the definition of with_value_constant.

(In reply to Alexander Zaitsev from comment #6)
> Any updates on the issue? Such behaviour is strange too since Clang and MSVC
> have a different opinion from GCC for the code.

I think GCC is behaving correctly here.