[Bug c++/99086] New: including <iterator> and defaulting spaceship operator causes compiler segfault

[Bug c++/99086] New: including <iterator> and defaulting spaceship operator causes compiler segfault

[crillion at tiscali dot it]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99086

            Bug ID: 99086
           Summary: including <iterator> and defaulting spaceship operator
                    causes compiler segfault
           Product: gcc
           Version: 10.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: crillion at tiscali dot it
  Target Milestone: ---

Created attachment 50178
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50178&action=edit
an archive containing .ii and .s files for windows and linux

Hi,

  a little program with a tiny class defaulting the spaceship operator and the
inclusion of <iterator> header causes compiler segfault (the compiler gives
segmentation error while compiling, no executable is produced).

I could see this behaviour on these systems :

(windows 10 64bit)
(it's the g++ 10.2 from mingw-w64 distribution obtained by msys2)
g++ --version
g++ (Rev6, Built by MSYS2 project) 10.2.0

(linux opensuse 15.2 64bit, hosted on virtual box 6.1.16 r140961 (Qt5.6.2),
running on my above windows system)
> g++ --version
g++ (SUSE Linux) 10.2.1 20200825 [revision
c0746a1beb1ba073c7981eb09f55b3d993b32e5c]

==============
program.cpp :
------------------


#include <iterator>

class value_wrapper
{
        public :

                bool operator<=>(const value_wrapper& rhs) const = default;

        private :

                unsigned short value_;
};

int main(int, char**)
{
        value_wrapper lhs, rhs;

        auto x = lhs < rhs;
}

=================================
batch file used to compile it (already including your recommended switches in
bug reporting instructions, which turn out not to mitigate the segfault error)

windows :

g++ -std=c++20 -pedantic -Wall -Wextra -Werror=return-type -Wshadow=local
-Wempty-body -fdiagnostics-color -s -Os -fno-strict-aliasing -fwrapv
-fno-aggressive-loop-optimizations program.cpp -save-temps -o program_gpp.exe

linux :
g++ -std=c++20 -pedantic -Wall -Wextra -Werror=return-type -Wshadow=local
-Wempty-body -fdiagnostics-color -s -Os -fno-strict-aliasing -fwrapv
-fno-aggressive-loop-optimizations program.cpp -save-temps -o program_gpp


I add as attachment a 7zip archive containing the requested .ii and .s files
respectively for the windows and linux compilation:

program.ii.linux
program.ii.windows
program.s.linux
program.s.windows

I was originally using a field of type boost gregorian date, which in turn
causes inclusion of <iterator>. I could reduce boost include files just to the
inclusion of <iterator>
Sorry but I didn't reduce, nor inspect the <iterator> header, since I consider
standard library headers too complicated for me to analyze.

Thanks,
           Marco

[Bug c++/99086] including <iterator> and defaulting spaceship operator causes compiler segfault

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99086

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
           Keywords|                            |ice-on-invalid-code
      Known to work|                            |11.0
      Known to fail|                            |10.2.1
   Last reconfirmed|                            |2021-02-14
                 CC|                            |jason at gcc dot gnu.org

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to crillion from comment #0)
> #include <iterator>
> 
> class value_wrapper
> {
> 	public :
> 		
> 		bool operator<=>(const value_wrapper& rhs) const = default;

This is a bug, the spaceship operator has to return one of
std::strong_ordering, std::weak_ordering or std::partial_ordering.

GCC's crash on the invalid code has already been fixed in r11-5866:

    c++: Fix defaulted <=> fallback to < and == [PR96299]

I don't know if there are plans to backport that.

[Bug c++/99086] including <iterator> and defaulting spaceship operator causes compiler segfault

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99086

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #1)
> (In reply to crillion from comment #0)
> > #include <iterator>
> > 
> > class value_wrapper
> > {
> > 	public :
> > 		
> > 		bool operator<=>(const value_wrapper& rhs) const = default;
> 
> This is a bug, the spaceship operator has to return one of
> std::strong_ordering, std::weak_ordering or std::partial_ordering.

That's not quite true, it can be something that std::strong_ordering is
convertible to. But it can't be bool.

If you're defaulting it, you probably want to just return auto and let the
compiler get the type right.

[Bug c++/99086] including <iterator> and defaulting spaceship operator causes compiler segfault

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99086

Jonathan Wakely <redi at gcc dot gnu.org> changed:

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

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #1)
> GCC's crash on the invalid code has already been fixed in r11-5866:
> 
>     c++: Fix defaulted <=> fallback to < and == [PR96299]

And we already have multiple duplicates of this.

*** This bug has been marked as a duplicate of bug 97121 ***