[Bug c++/53181] New: static_assert sees as non constant the comparison between a constexpr and a template argument

[Bug c++/53181] New: static_assert sees as non constant the comparison between a constexpr and a template argument

[blaffablaffa at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53181

             Bug #: 53181
           Summary: static_assert sees as non constant the comparison
                    between a constexpr and a template argument
    Classification: Unclassified
           Product: gcc
           Version: 4.6.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: blaffablaffa@gmail.com


Created attachment 27278
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27278
test case

As per title.
The weird is that it seems that the constexpr-ness seems to changes if the
expression inside static_assert evaluates to true (compiles, expression seen as
constexpr) or false (fails to compile, but not due to the assertion failing but
to the described error).

[Bug c++/53181] static_assert sees as non constant the comparison between a constexpr and a template argument

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53181

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-05-12
     Ever Confirmed|0                           |1

--- Comment #1 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-05-12 22:16:36 UTC ---
Confirmed.

[Bug c++/53181] static_assert sees as non constant the comparison between a constexpr and a template argument

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53181

--- Comment #2 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-10-18 14:45:00 UTC ---
In the case of wrong_string, fold_comparison (called from cp_build_binary_op
via fold_if_not_in_template) cannot fold the comparison to a constant: base0 !=
base1. Most likely the issue is that we shouldn't end up passing such pointers.

[Bug c++/53181] static_assert sees as non constant the comparison between a constexpr and a template argument

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53181

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

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

--- Comment #3 from Jason Merrill <jason at gcc dot gnu.org> ---
The difference between the namespace- and class-scope cases is that the global
declarations are known to be defined, while the static members don't have a
visible out-of-class definition, and the middle end thinks it can't assume that
two symbols defined elsewhere won't turn out to have the same address.

Adding definitions of the static members causes the two cases to get the same
result.

constexpr char Test::wrong_string[];
constexpr char Test::right_string[];

We should probably override the middle end in this case.

[Bug c++/53181] static_assert sees as non constant the comparison between a constexpr and a template argument

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53181

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2019-02-11 00:00:00         |2021-8-4

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
So this is dependent on which standard you use.
For C++11 and C++14, trunk is still rejecting it.
For C++17, starting in GCC 7, the code is accepted and is correct.
C++20 on trunk accepts it.

It would be interesting to see why GCC 7+ acts this way.

(In reply to Martin Sebor from comment #5)
> No change in GCC 9.

Actually there was :).

[Bug c++/53181] static_assert sees as non constant the comparison between a constexpr and a template argument

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53181

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> ---
GCC 7 accepts it in C++17 mode because of r241137

    Implement P0386R2 - C++17 inline variables

That's because an out-of-class definition is not needed in C++17, the compiler
will generate one from the in-class declaration.