[Bug c++/94951] New: dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[Bug c++/94951] New: dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jorgen.lind at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

            Bug ID: 94951
           Summary: dereferencing type-punned pointer will break
                    strict-aliasing rules when using super class for a
                    template type
           Product: gcc
           Version: 9.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jorgen.lind at gmail dot com
  Target Milestone: ---

The following program issues the warning when compiled with -O2 or -O3.

The workaround is to static_cast usage of members to super class.
In the godbolt example its sufficient to add brackets around the class instance
before member access, but was not sufficient in a more complex example. 

The behavior is present from 7.1 until trunk.
Godbolt link:https://godbolt.org/z/Kfp-ac

#include <stdint.h>

struct A
{
    uint32_t someData[32];
};

template<int N>
struct B : public A
{
    static B<N> createB(uint32_t (&otherData)[32])
    {
        B<N> toReturn;
        toReturn.someData[2] = 4;
        return toReturn;
    }
};

[Bug c++/94951] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
           Keywords|                            |diagnostic
   Last reconfirmed|                            |2020-05-05

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
We're warning via

#0  warning_at (location=426247, opt=664, 
    gmsgid=0x2a36f08 "dereferencing type-punned pointer will break
strict-aliasing rules") at ../../src/trunk/gcc/diagnostic.c:1442
#1  0x0000000000e324c9 in strict_aliasing_warning (loc=426247, 
    type=<pointer_type 0x7ffff6d9fbd0>, expr=<addr_expr 0x7ffff6d97480>)
    at ../../src/trunk/gcc/c-family/c-warn.c:729
#2  0x0000000000d45def in cp_build_indirect_ref_1 (loc=426535, 
    ptr=<nop_expr 0x7ffff6d974a0>, errorstring=RO_NULL, complain=3, 
    do_fold=true) at ../../src/trunk/gcc/cp/typeck.c:3364
#3  0x0000000000d46191 in cp_build_fold_indirect_ref (
    pointer=<nop_expr 0x7ffff6d974a0>) at ../../src/trunk/gcc/cp/typeck.c:3444
#4  0x00000000009ecb7a in build_base_path (code=PLUS_EXPR, 
    expr=<nop_expr 0x7ffff6d974a0>, binfo=<tree_binfo 0x7ffff6da0060>, 
    nonnull=1, complain=3) at ../../src/trunk/gcc/cp/class.c:513
#5  0x0000000000d4309f in build_class_member_access_expr (object=..., 
    member=<field_decl 0x7ffff6d9eab0 someData>, 
    access_path=<record_type 0x7ffff6d9f5e8 B>, preserve_reference=false, 
    complain=3) at ../../src/trunk/gcc/cp/typeck.c:2603
#6  0x0000000000d45377 in finish_class_member_access_expr (object=..., 
    name=<identifier_node 0x7ffff6d9b980 someData>, template_p=false, 
    complain=3) at ../../src/trunk/gcc/cp/typeck.c:3209

[Bug c++/94951] [8/9/10/11 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |8.5
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |jason at gcc dot gnu.org
            Summary|dereferencing type-punned   |[8/9/10/11 Regression]
                   |pointer will break          |dereferencing type-punned
                   |strict-aliasing rules when  |pointer will break
                   |using super class for a     |strict-aliasing rules when
                   |template type               |using super class for a
                   |                            |template type

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r7-755-g23cb72663051cd3f5a8952d4aa2186d50243b7d0

[Bug c++/94951] [8/9/10/11 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Slightly simplified testcase:
struct A { unsigned a[32]; };
template <int N>
struct B : public A
{
  static B<N> foo () { B<N> t; t.a[0] = 4; return t; }
};

[Bug c++/94951] [8/9/10/11 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
And the array isn't needed either:
struct A { int a; };
template <int N>
struct B : public A
{
  static B<N> foo () { B<N> t; t.a = 4; return t; }
};

[Bug c++/94951] [8/9/10/11 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug c++/94951] [8/9/10/11 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 48456
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48456&action=edit
gcc11-pr94951.patch

Untested fix.  It seems just wrong to call the c-family strict_aliasing_warning
for dependent types, we shouldn't query alias sets of dependent types and do
any decisions based on what that gives.

[Bug c++/94951] [8/9/10/11 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:46fcef99f49cc2d9f28d98f8ecdbf8263e9e0a87

commit r11-153-g46fcef99f49cc2d9f28d98f8ecdbf8263e9e0a87
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed May 6 23:38:13 2020 +0200

    c++: Avoid strict_aliasing_warning on dependent types or expressions
[PR94951]

    The following testcase gets a bogus warning during build_base_path,
    when cp_build_indirect_ref* calls strict_aliasing_warning with a dependent
    expression.  IMHO calling get_alias_set etc. on dependent types feels wrong
    to me, we should just defer the warnings in those cases until instantiation
    and only handle the cases where neither type nor expr are dependent.

    2020-05-06  Jakub Jelinek  <jakub@redhat.com>

            PR c++/94951
            * typeck.c (cp_strict_aliasing_warning): New function.
            (cp_build_indirect_ref_1, build_reinterpret_cast_1): Use
            it instead of strict_aliasing_warning.

            * g++.dg/warn/Wstrict-aliasing-bogus-tmpl.C: New test.

[Bug c++/94951] [8/9/10/11 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Jakub Jelinek
<jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:d82414ebcf7716ea24688510594a2c464a105908

commit r10-8115-gd82414ebcf7716ea24688510594a2c464a105908
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed May 6 23:38:13 2020 +0200

    c++: Avoid strict_aliasing_warning on dependent types or expressions
[PR94951]

    The following testcase gets a bogus warning during build_base_path,
    when cp_build_indirect_ref* calls strict_aliasing_warning with a dependent
    expression.  IMHO calling get_alias_set etc. on dependent types feels wrong
    to me, we should just defer the warnings in those cases until instantiation
    and only handle the cases where neither type nor expr are dependent.

    2020-05-06  Jakub Jelinek  <jakub@redhat.com>

            PR c++/94951
            * typeck.c (cp_strict_aliasing_warning): New function.
            (cp_build_indirect_ref_1, build_reinterpret_cast_1): Use
            it instead of strict_aliasing_warning.

            * g++.dg/warn/Wstrict-aliasing-bogus-tmpl.C: New test.

[Bug c++/94951] [8/9 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[8/9/10/11 Regression]      |[8/9 Regression]
                   |dereferencing type-punned   |dereferencing type-punned
                   |pointer will break          |pointer will break
                   |strict-aliasing rules when  |strict-aliasing rules when
                   |using super class for a     |using super class for a
                   |template type               |template type

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed for 10.2+ and 11+ so far.

[Bug c++/94951] [8/9 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by Jakub Jelinek
<jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:eca1dbea010ba4c18ccec90bacbad5d8ec72c09f

commit r9-8897-geca1dbea010ba4c18ccec90bacbad5d8ec72c09f
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Wed May 6 23:38:13 2020 +0200

    c++: Avoid strict_aliasing_warning on dependent types or expressions
[PR94951]

    The following testcase gets a bogus warning during build_base_path,
    when cp_build_indirect_ref* calls strict_aliasing_warning with a dependent
    expression.  IMHO calling get_alias_set etc. on dependent types feels wrong
    to me, we should just defer the warnings in those cases until instantiation
    and only handle the cases where neither type nor expr are dependent.

    2020-05-06  Jakub Jelinek  <jakub@redhat.com>

            PR c++/94951
            * typeck.c (cp_strict_aliasing_warning): New function.
            (cp_build_indirect_ref_1, build_reinterpret_cast_1): Use
            it instead of strict_aliasing_warning.

            * g++.dg/warn/Wstrict-aliasing-bogus-tmpl.C: New test.

    (cherry picked from commit d82414ebcf7716ea24688510594a2c464a105908)

[Bug c++/94951] [8/9 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed for 8.5 in r8-10503-g79a11b9833baa91cef7c881532e01470fa25047f and by the
above commit for 9.4+ too.

[Bug c++/94951] [8/9 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rrrlasse at hotmail dot com

--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 81152 has been marked as a duplicate of this bug. ***

[Bug c++/94951] [8/9 Regression] dereferencing type-punned pointer will break strict-aliasing rules when using super class for a template type

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94951

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |a-yee at u dot northwestern.edu

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 91343 has been marked as a duplicate of this bug. ***