[Bug c++/99801] New: Address sanitizer false positive with pointer to member function.

[Bug c++/99801] New: Address sanitizer false positive with pointer to member function.

[fsb4000 at yandex dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

            Bug ID: 99801
           Summary: Address sanitizer false positive with pointer to
                    member function.
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: fsb4000 at yandex dot ru
  Target Milestone: ---

Hi. 

Problematic piece of code: https://gcc.godbolt.org/z/s8x3KqaMb

If we change `auto` to `void (Curses_table::*)()` then it works:
https://gcc.godbolt.org/z/bezacsTPa

clang works in both cases:

https://gcc.godbolt.org/z/1e719Maa6

https://gcc.godbolt.org/z/8eqWzGzaW

[Bug sanitizer/99801] Address sanitizer false positive with pointer to member function.

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-03-30
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
           Assignee|unassigned at gcc dot gnu.org      |marxin at gcc dot gnu.org

--- Comment #1 from Martin Liška <marxin at gcc dot gnu.org> ---
I can take a look.

[Bug sanitizer/99801] Address sanitizer false positive with pointer to member function.

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

Martin Liška <marxin at gcc dot gnu.org> changed:

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

--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
So let's consider the following example:

$ cat pr99801.C
class Curses_table {
public:
        void draw_row() {};
        void draw_table();
};

void Curses_table::draw_table()
{
  void (Curses_table::*draw_fn) (void) = &Curses_table::draw_row;
        auto Curses_table::*draw_fn2 = &Curses_table::draw_row;
        (this->*draw_fn)();
        (this->*draw_fn2)();
}

int main() {
        Curses_table t;
        t.draw_table();
}

What happens is that we wrongly assign DECL_SIZE_UNIT for draw_fn2 in:

Breakpoint 5, layout_decl (decl=<var_decl 0x7ffff7fbef30 draw_fn>,
known_align=0) at /home/marxin/Programming/gcc/gcc/stor-layout.c:634
(gdb) p debug_tree(type)
 <offset_type 0x7ffff777d690
    type <template_type_parm 0x7ffff777d5e8 auto VOID
        align:8 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff777d5e8
       index 0 level 1 orig_level 1
        chain <type_decl 0x7ffff760ee40 auto>>
    DI
    size <integer_cst 0x7ffff75e1eb8 type <integer_type 0x7ffff75ff0a8
bitsizetype> constant 64>
    unit-size <integer_cst 0x7ffff75e1ed0 type <integer_type 0x7ffff75ff000
sizetype> constant 8>
    align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff777d690 basetype <record_type 0x7ffff777d888 Curses_table>>

while draw_fn is called with:

(gdb) p debug_tree(type)
 <record_type 0x7ffff777d3f0 type_2 TI
    size <integer_cst 0x7ffff75e1f00 type <integer_type 0x7ffff75ff0a8
bitsizetype> constant 128>
    unit-size <integer_cst 0x7ffff75e1f18 type <integer_type 0x7ffff75ff000
sizetype> constant 16>
    align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff777d498
    fields <field_decl 0x7ffff760eab0 __pfn
        type <pointer_type 0x7ffff777d2a0 type <method_type 0x7ffff777d0a8>
            unsigned DI
            size <integer_cst 0x7ffff75e1eb8 constant 64>
            unit-size <integer_cst 0x7ffff75e1ed0 constant 8>
            align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff777d348>
        unsigned nonaddressable DI
/home/marxin/Programming/testcases/pr99801.C:9:38 size <integer_cst
0x7ffff75e1eb8 64> unit-size <integer_cst 0x7ffff75e1ed0 8>
        align:64 warn_if_not_align:0 offset_align 128
        offset <integer_cst 0x7ffff75e1ee8 constant 0>
        bit-offset <integer_cst 0x7ffff75e1f30 constant 0> context <record_type
0x7ffff777d3f0>
        chain <field_decl 0x7ffff760eb48 __delta type <integer_type
0x7ffff75ff738 long int>
            nonaddressable DI /home/marxin/Programming/testcases/pr99801.C:9:38
size <integer_cst 0x7ffff75e1eb8 64> unit-size <integer_cst 0x7ffff75e1ed0 8>
            align:64 warn_if_not_align:0 offset_align 128 offset <integer_cst
0x7ffff75e1ee8 0> bit-offset <integer_cst 0x7ffff75e1eb8 64> context
<record_type 0x7ffff777d3f0>>> ptrmemfunc fn type <pointer_type 0x7ffff777d2a0>
    chain <type_decl 0x7ffff760ebe0 __ptrmemfunc_type>>

later then the corresponding record_type is set as TREE_TYPE of draw_fn2:

(gdb) bt
#0  0x0000000000a16cd0 in cp_finish_decl (decl=decl@entry=<var_decl
0x7ffff7794510 draw_fn2>, init=<optimized out>, init@entry=<ptrmem_cst
0x7ffff7774b20>, init_const_expr_p=true, asmspec_tree=<optimized out>,
asmspec_tree@entry=<tree 0x0>, flags=5)
    at /home/marxin/Programming/gcc/gcc/cp/decl.c:7744
#1  0x0000000000b06168 in cp_parser_init_declarator (parser=0x7ffff760e7b8,
flags=<optimized out>, decl_specifiers=0x7fffffffd2b0, checks=0x0,
function_definition_allowed_p=<optimized out>, member_p=<optimized out>,
declares_class_or_enum=0, 
    function_definition_p=0x7fffffffd350, maybe_range_for_decl=0x0,
init_loc=0x7fffffffd29c, auto_result=0x7fffffffd3f0) at
/home/marxin/Programming/gcc/gcc/cp/parser.c:21872
#2  0x0000000000ae166d in cp_parser_simple_declaration (parser=0x7ffff760e7b8,
function_definition_allowed_p=<optimized out>, maybe_range_for_decl=0x0) at
/home/marxin/Programming/gcc/gcc/cp/parser.c:14442
#3  0x0000000000ae385b in cp_parser_declaration_statement
(parser=parser@entry=0x7ffff760e7b8) at
/home/marxin/Programming/gcc/gcc/cp/parser.c:13577
#4  0x0000000000ae3ea4 in cp_parser_statement (parser=0x7ffff760e7b8,
in_statement_expr=<tree 0x0>, in_compound=<optimized out>, if_p=<optimized
out>, chain=<optimized out>, loc_after_labels=0x0) at
/home/marxin/Programming/gcc/gcc/cp/parser.c:11808
#5  0x0000000000ae5983 in cp_parser_statement_seq_opt
(parser=parser@entry=0x7ffff760e7b8,
in_statement_expr=in_statement_expr@entry=<tree 0x0>) at
/home/marxin/Programming/gcc/gcc/cp/parser.c:12174

but the DECL_SIZE_UNIT is not adjusted.
@Jason: Can you please take a look?

[Bug c++/99801] Address sanitizer false positive with pointer to member function.

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
@Jason: May I please ping this?

[Bug c++/99801] Address sanitizer false positive with pointer to member function.

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |WAITING

[Bug c++/99801] Address sanitizer false positive with pointer to member function.

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW
           Assignee|marxin at gcc dot gnu.org          |unassigned at gcc dot gnu.org

[Bug c++/99801] Address sanitizer false positive with pointer to member function.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 54155
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54155&action=edit
Original testcase

[Bug c++/99801] Address sanitizer false positive with pointer to member function.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99801

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2021-03-30 00:00:00         |2022-12-24

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
-O0 -fsanitize=address