[Bug other/103681] New: Unusual behavior for tail padding with different c++ standards

[Bug other/103681] New: Unusual behavior for tail padding with different c++ standards

[n.deshmukh at samsung dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

            Bug ID: 103681
           Summary: Unusual behavior for tail padding with different c++
                    standards
           Product: gcc
           Version: 8.3.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: other
          Assignee: unassigned at gcc dot gnu.org
          Reporter: n.deshmukh at samsung dot com
  Target Milestone: ---

When I tried to link two modules which were compiled with different c++
standards, I observed that the offset of some fields of struct were different
when the same struct was accessed from both the modules. The issue is due to
the use of tail padding to allocate member variables in some standards (c++03,
c++11) and not with others (c++14, c++17). I created a small program to
reproduce the behaviour. 
-----------------------------padding_error.cpp----------------------------
#include <iostream>
#include <stdint.h>
#include <cstddef>
struct A {
  int a;
  uint64_t b;
  int c = -1;
};
struct B : public A {
    int d;
};
int main() {
    std::cout << "sizeof(A): " << sizeof(A);
    std::cout << ", sizeof(B): " << sizeof(B) << std::endl;
    std::cout << "offset of d in B: " << (int)offsetof(B, d) << std::endl; 
}
--------------------------------------------------------------------------

The output of this program depends on the -std flag.
~$ /opt/rh/devtoolset-8/root/bin/g++ -std=c++03 padding_error.cpp -o c03
~$ ./c03
sizeof(A): 24, sizeof(B): 24
offset of d in B: 20
~$ /opt/rh/devtoolset-8/root/bin/g++ -std=c++11 padding_error.cpp -o c11
~$ ./c11
sizeof(A): 24, sizeof(B): 24
offset of d in B: 20
~$ /opt/rh/devtoolset-8/root/bin/g++ -std=c++14 padding_error.cpp -o c14
~$ ./c14
sizeof(A): 24, sizeof(B): 32
offset of d in B: 24
~$ /opt/rh/devtoolset-8/root/bin/g++ -std=c++17 padding_error.cpp -o c17
~$ ./c17
sizeof(A): 24, sizeof(B): 32
offset of d in B: 24


I tried with different versions of gcc but the output is the same. 
Because of the difference in the offset of the fields, I cannot link files
which were compiled with different -std flags. 

The issue can be reproduced with all the gcc compilers which support c++14.

[Bug other/103681] Unusual behavior for tail padding with different c++ standards

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |blocker
           Keywords|                            |ABI, wrong-code
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-12-13
      Known to fail|                            |12.0
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

Some more information from 
Jonathan Wakely:
The C++14 behaviour started with r216750

   Implement N3653 (Member initializers and aggregates) and fix
references to 'this' in constexpr cons
tructors.

           Implement N3653 (Member initializers and aggregates) and fix
           references to 'this' in constexpr constructors.
           * class.c (check_field_decls): In C++14 an NSDMI does not make the
           class non-aggregate.

But I think the layout should be using "POD for the purpose of
layout", which means the C++98 POD rules. Please do report it to
bugzilla.

[Bug c++/103681] [9/10/11/12 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Unusual behavior for tail   |[9/10/11/12 Regression]
                   |padding with different c++  |Unusual behavior for tail
                   |standards and NSDMI         |padding with different c++
                   |                            |standards and NSDMI
   Target Milestone|---                         |9.5

[Bug c++/103681] [9/10/11/12 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

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
           Priority|P3                          |P1

[Bug c++/103681] [9/10/11/12 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

--- Comment #2 from Jason Merrill <jason at gcc dot gnu.org> ---
compile-time testcase

struct A {
  long l;
  char c = -1;
};
struct B : public A {
  char d;
};

#define SA(X) static_assert(X,#X)
SA(sizeof (B) == sizeof (A) || alignof (long) < 2*sizeof(char));

[Bug c++/103681] [9/10/11/12 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P1                          |P2

[Bug c++/103681] [9/10/11/12 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

--- 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:a37e8ce3b66325f0c6de55c80d50ac1664c3d0eb

commit r12-6028-ga37e8ce3b66325f0c6de55c80d50ac1664c3d0eb
Author: Jason Merrill <jason@redhat.com>
Date:   Tue Dec 14 17:00:40 2021 -0500

    c++: layout of aggregate base with DMI [PR103681]

    C++14 changed the definition of 'aggregate' to allow default member
    initializers, but such classes still need to be considered "non-POD for the
    purpose of layout" for ABI compatibility with C++11 code.  It seems rare to
    derive from such a class, as evidenced by how long this bug has
    survived (since r216750 in 2014), but it's certainly worth fixing.

    We only warn when we were failing to allocate another field into the
    tail padding of the newly aggregate class; this is the only ABI impact.

    This also changes end_of_class to consider all data members, not just empty
    data members; that used to be an additional flag, removed in r9-5710, but I
    don't see any reason not to always include them.  This makes the result of
    the function correspond to the ABI nvsize term and its nameless counterpart
    that does include virtual bases.

    When looking closely at other users of end_of_class, I realized that we
were
    assuming that the latter corresponded to the ABI dsize term, but it doesn't
    if the class ends with an empty virtual base (in the rare case that the
    empty base can't be assigned offset 0), and this matters for layout of
    [[no_unique_address]].  So I added another mode that returns the desired
    value for that case.  I'm not adding a warning for this ABI fix because
it's
    a C++20 feature.

            PR c++/103681

    gcc/ChangeLog:

            * common.opt (fabi-version): Add v17.

    gcc/cp/ChangeLog:

            * cp-tree.h (struct lang_type): Add non_pod_aggregate.
            (CLASSTYPE_NON_POD_AGGREGATE): New.
            * class.c (check_field_decls): Set it.
            (check_bases_and_members): Check it.
            (check_non_pod_aggregate): New.
            (enum eoc_mode): New.
            (end_of_class): Always include non-empty fields.
            Add eoc_nv_or_dsize mode.
            (include_empty_classes, layout_class_type): Adjust.

    gcc/c-family/ChangeLog:

            * c-opts.c (c_common_post_options): Update defaults.

    gcc/testsuite/ChangeLog:

            * g++.dg/abi/macro0.C: Update value.
            * g++.dg/abi/no_unique_address6.C: New test.
            * g++.dg/abi/nsdmi-aggr1.C: New test.
            * g++.dg/abi/nsdmi-aggr1a.C: New test.

[Bug c++/103681] [9/10/11 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[9/10/11/12 Regression]     |[9/10/11 Regression]
                   |Unusual behavior for tail   |Unusual behavior for tail
                   |padding with different c++  |padding with different c++
                   |standards and NSDMI         |standards and NSDMI
      Known to fail|12.0                        |
      Known to work|                            |12.0
           Severity|blocker                     |major

--- Comment #4 from Jason Merrill <jason at gcc dot gnu.org> ---
Fixed for GCC 12 so far.

[Bug c++/103681] [9/10/11 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

--- Comment #5 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:23be9f83bbd2c6f03580757adbfe599de6bf702b

commit r12-6800-g23be9f83bbd2c6f03580757adbfe599de6bf702b
Author: Jason Merrill <jason@redhat.com>
Date:   Fri Jan 21 11:16:49 2022 -0500

    c++: [[no_unique_address]] and virtual base [PR104139]

    Fixing a thinko in my patch for 103681: when computing the size of a
virtual
    base, it would help to use its binfo instead of the one for the derived
    class.

            PR c++/104139
            PR c++/103681

    gcc/cp/ChangeLog:

            * class.cc (end_of_class): Use base_binfo.

    gcc/testsuite/ChangeLog:

            * g++.dg/abi/no_unique_address2.C: Adjust to detect this on x86-64.

[Bug c++/103681] [9/10/11 Regression] Unusual behavior for tail padding with different c++ standards and NSDMI

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103681

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
   Target Milestone|9.5                         |12.0
         Resolution|---                         |FIXED

--- Comment #6 from Jason Merrill <jason at gcc dot gnu.org> ---
Not backporting layout change.