[Bug c++/95158] New: Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[Bug c++/95158] New: Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[sudgylacmoe at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

            Bug ID: 95158
           Summary: Templates + Diamond Inheritance + Final = Pure Virtual
                    Function Call
           Product: gcc
           Version: 10.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: sudgylacmoe at gmail dot com
  Target Milestone: ---

With this code:

class Base {
    public:
        virtual void foo()=0;
};

template <typename T>
class MiddleA : virtual public Base {
    public:
        virtual void foo() override {}
};

class MiddleB : virtual public Base {};

template <typename T>
class Derived final : public MiddleA<T>, public MiddleB {
    public:
        void bar()
        {
            this->foo();
        }
};

int main()
{
    auto a = Derived<void>();
    a.bar(); // Instantiate the template
}

Compiling it with gcc 10.1 with just "g++ test.cpp" (and many other ways to
compile, such as optimization levels and warnings) causes gcc to try to call
Base::foo(), a pure virtual function:

/usr/bin/ld: /tmp/cce0CW5N.o: in function `Derived<void>::bar()':
test.cpp:(.text._ZN7DerivedIvE3barEv[_ZN7DerivedIvE3barEv]+0x14): undefined
reference to `Base::foo()'
collect2: error: ld returned 1 exit status

Trying a few different versions, with gcc 9.3 and before this code compiles
correctly.  On any other compiler I tried (the most recent versions of clang,
msvc and icc on godbolt) it compiled fine.  Removing the template from either
MiddleA or Derived fixes it, removing the inheriting from MiddleB fixes it, and
removing final from Derived fixes it.  If a definition of Base::foo() is added,
the code compiles and calls it.  A workaround is to call MiddleA::foo()
directly.

Output of g++ -v:

Using built-in specs.
COLLECT_GCC=g++
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-pc-linux-gnu/10.1.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /build/gcc/src/gcc/configure --prefix=/usr --libdir=/usr/lib
--libexecdir=/usr/lib --mandir=/usr/share/man --infodir=/usr/share/info
--with-bugurl=https://bugs.archlinux.org/
--enable-languages=c,c++,ada,fortran,go,lto,objc,obj-c++,d --with-isl
--with-linker-hash-style=gnu --with-system-zlib --enable-__cxa_atexit
--enable-cet=auto --enable-checking=release --enable-clocale=gnu
--enable-default-pie --enable-default-ssp --enable-gnu-indirect-function
--enable-gnu-unique-object --enable-install-libiberty --enable-linker-build-id
--enable-lto --enable-multilib --enable-plugin --enable-shared
--enable-threads=posix --disable-libssp --disable-libstdcxx-pch
--disable-libunwind-exceptions --disable-werror
gdc_include_dir=/usr/include/dlang/gdc
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 10.1.0 (GCC)

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |9.3.0
           Keywords|                            |wrong-code
   Target Milestone|---                         |10.2
            Summary|Templates + Diamond         |[10/11 Regression]
                   |Inheritance + Final = Pure  |Templates + Diamond
                   |Virtual Function Call       |Inheritance + Final = Pure
                   |                            |Virtual Function Call

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2020-05-26
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jason at gcc dot gnu.org
                 CC|                            |jason at gcc dot gnu.org
     Ever confirmed|0                           |1

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |paolo.carlini at oracle dot com
             Status|ASSIGNED                    |NEW
           Assignee|jason at gcc dot gnu.org           |unassigned at gcc dot gnu.org
         Depends on|                            |67184

--- Comment #1 from Jason Merrill <jason at gcc dot gnu.org> ---
Much like PR 90909, this was introduced by Paolo's patch for PR 67184 in
r10-1665-g26f8363d85fd386430ddb612a2e70624715c2ac3


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67184
[Bug 67184] Missed optimization with C++11 final specifier

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[paolo.carlini at oracle dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

--- Comment #2 from Paolo Carlini <paolo.carlini at oracle dot com> ---
That the issue goes away when templates are not involved seems an useful hint:
are we trying to optimize too early?

Sorry, for the time being I don't feel like assigning the bug to me, I'm in the
middle of too many other things...

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |jason at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

--- Comment #3 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jason Merrill <jason@gcc.gnu.org>:

https://gcc.gnu.org/g:0ddb93ce77374004c49cdfbd748ba35867620cf1

commit r11-954-g0ddb93ce77374004c49cdfbd748ba35867620cf1
Author: Jason Merrill <jason@redhat.com>
Date:   Wed Jun 3 23:50:06 2020 -0400

    c++: Fix FE devirt with diamond inheritance [PR95158]

    This started breaking in GCC 8 because of the fix for PR15272; after that
    change, we (correctly) remember the lookup from template parsing time that
    found Base::foo through the non-dependent MiddleB base, and so we overlook
    the overrider in MiddleA.  But given that, the devirtualization condition
    from the fix for PR59031 is insufficient; we know that d has to be a
    Derived, and we found Base::foo in Base, but forcing a non-virtual call
    gets the wrong function.

    Fixed by removing the PR59031 code that the PR67184 patch moved to
    build_over_call, and instead looking up the overrider in BINFO_VIRTUALS.

    gcc/cp/ChangeLog:

            PR c++/95158
            * class.c (lookup_vfn_in_binfo): New.
            * call.c (build_over_call): Use it.
            * cp-tree.h (resolves_to_fixed_type_p): Add default argument.
            (lookup_vfn_in_binfo): Declare.

    gcc/testsuite/ChangeLog:

            PR c++/95158
            * g++.dg/template/virtual5.C: New test.

[Bug c++/95158] [10/11 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Jason Merrill
<jason@gcc.gnu.org>:

https://gcc.gnu.org/g:e244b0acf3111c95bb1559e7610f84740b90589f

commit r10-8242-ge244b0acf3111c95bb1559e7610f84740b90589f
Author: Jason Merrill <jason@redhat.com>
Date:   Wed Jun 3 23:50:06 2020 -0400

    c++: Fix FE devirt with diamond inheritance [PR95158]

    This started breaking in GCC 8 because of the fix for PR15272; after that
    change, we (correctly) remember the lookup from template parsing time that
    found Base::foo through the non-dependent MiddleB base, and so we overlook
    the overrider in MiddleA.  But given that, the devirtualization condition
    from the fix for PR59031 is insufficient; we know that d has to be a
    Derived, and we found Base::foo in Base, but forcing a non-virtual call
    gets the wrong function.

    Fixed by removing the PR59031 code that the PR67184 patch moved to
    build_over_call, and instead looking up the overrider in BINFO_VIRTUALS.

    gcc/cp/ChangeLog:

            PR c++/95158
            * class.c (lookup_vfn_in_binfo): New.
            * call.c (build_over_call): Use it.
            * cp-tree.h (resolves_to_fixed_type_p): Add default argument.
            (lookup_vfn_in_binfo): Declare.

    gcc/testsuite/ChangeLog:

            PR c++/95158
            * g++.dg/template/virtual5.C: New test.

[Bug c++/95158] [8/9 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[10/11 Regression]          |[8/9 Regression] Templates
                   |Templates + Diamond         |+ Diamond Inheritance +
                   |Inheritance + Final = Pure  |Final = Pure Virtual
                   |Virtual Function Call       |Function Call
   Target Milestone|10.2                        |8.5
      Known to work|9.3.0                       |

--- Comment #5 from Jason Merrill <jason at gcc dot gnu.org> ---
If I change bar to call foo on a local Derived variable instead of 'this', it
breaks back to GCC 8.  Fixed so far for 10.2/11.

[Bug c++/95158] [8/9 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-8 branch has been updated by Jason Merrill
<jason@gcc.gnu.org>:

https://gcc.gnu.org/g:a2bdff4f24d9065791e6d8820004772b9fe0c4c1

commit r8-10637-ga2bdff4f24d9065791e6d8820004772b9fe0c4c1
Author: Jason Merrill <jason@redhat.com>
Date:   Wed Jun 3 23:50:06 2020 -0400

    c++: Fix FE devirt with diamond inheritance [PR95158]

    This started breaking in GCC 8 because of the fix for PR15272; after that
    change, we (correctly) remember the lookup from template parsing time that
    found Base::foo through the non-dependent MiddleB base, and so we overlook
    the overrider in MiddleA.  But given that, the devirtualization condition
    from the fix for PR59031 is insufficient; we know that d has to be a
    Derived, and we found Base::foo in Base, but forcing a non-virtual call
    gets the wrong function.

    Fixed by removing the PR59031 code, and instead looking up the overrider in
    BINFO_VIRTUALS.

    gcc/cp/ChangeLog:

            PR c++/95158
            * class.c (lookup_vfn_in_binfo): New.
            * call.c (build_over_call): Use it.
            (build_new_method_call_1): Don't set LOOKUP_NONVIRTUAL.
            * cp-tree.h (resolves_to_fixed_type_p): Add default argument.
            (lookup_vfn_in_binfo): Declare.

    gcc/testsuite/ChangeLog:

            PR c++/95158
            * g++.dg/template/virtual5.C: New test.

[Bug c++/95158] [8/9 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by Jason Merrill
<jason@gcc.gnu.org>:

https://gcc.gnu.org/g:19323ea3e9344eb773f8fe872eadbe4f1ac6461f

commit r9-9066-g19323ea3e9344eb773f8fe872eadbe4f1ac6461f
Author: Jason Merrill <jason@redhat.com>
Date:   Wed Jun 3 23:50:06 2020 -0400

    c++: Fix FE devirt with diamond inheritance [PR95158]

    This started breaking in GCC 8 because of the fix for PR15272; after that
    change, we (correctly) remember the lookup from template parsing time that
    found Base::foo through the non-dependent MiddleB base, and so we overlook
    the overrider in MiddleA.  But given that, the devirtualization condition
    from the fix for PR59031 is insufficient; we know that d has to be a
    Derived, and we found Base::foo in Base, but forcing a non-virtual call
    gets the wrong function.

    Fixed by removing the PR59031 code, and instead looking up the overrider in
    BINFO_VIRTUALS.

    gcc/cp/ChangeLog:

            PR c++/95158
            * class.c (lookup_vfn_in_binfo): New.
            * call.c (build_over_call): Use it.
            (build_new_method_call_1): Don't set LOOKUP_NONVIRTUAL.
            * cp-tree.h (resolves_to_fixed_type_p): Add default argument.
            (lookup_vfn_in_binfo): Declare.

    gcc/testsuite/ChangeLog:

            PR c++/95158
            * g++.dg/template/virtual5.C: New test.

[Bug c++/95158] [8/9 Regression] Templates + Diamond Inheritance + Final = Pure Virtual Function Call

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95158

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

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

--- Comment #8 from Jason Merrill <jason at gcc dot gnu.org> ---
Also fixed for 8.5/9.4.