[Bug gdb/29220] New: GDB prints wrong value at -O2

[Bug gdb/29220] New: GDB prints wrong value at -O2

[yangyibiao at outlook dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29220

            Bug ID: 29220
           Summary: GDB prints wrong value at -O2
           Product: gdb
           Version: HEAD
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: gdb
          Assignee: unassigned at sourceware dot org
          Reporter: yangyibiao at outlook dot com
  Target Milestone: ---

$ gdb --version
GNU gdb (GDB) 13.0.50.20220531-git

$ gcc --version
gcc (GCC) 13.0.0 20220530 (experimental)

$ cat -n small.c
     1  void foo(int *a)
     2  {
     3    (*a)++;
     4  }
     5
     6  int main ()
     7  {
     8    int a = 0;
     9    foo (&a);
    10    return 0;
    11  }

$ gcc -O2 -g small.c; gdb -q a.out
Reading symbols from a.out...
(gdb) b main
Breakpoint 1 at 0x401020: file small.c, line 10.
(gdb) r
Starting program: /root/DeVIL/a.out

Breakpoint 1, main () at small.c:10
10        return 0;
(gdb) info locals
a = 0
(gdb) 

### The expected value is "a = 1". However, gdb prints "a = 0". I know that
this function is optimized out. I was wondering that a better solution for this
is that gdb report the value of a is optmized out as well.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug gdb/29220] GDB prints wrong value at -O2

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29220

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tromey at sourceware dot org

--- Comment #1 from Tom Tromey <tromey at sourceware dot org> ---
I didn't try your test program, but normally gdb just reports
whatever the compiler tells it here.  Most likely this is a compiler
issue, and not a gdb issue.  However we'd need to examine the DWARF
to be really sure.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug gdb/29220] GDB prints wrong value at -O2

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29220

--- Comment #2 from Tom Tromey <tromey at sourceware dot org> ---
readelf -wi shows 'a' defined as:

 <2><4c>: Abbrev Number: 3 (DW_TAG_variable)
    <4d>   DW_AT_name        : a
    <4f>   DW_AT_decl_file   : 1
    <50>   DW_AT_decl_line   : 8
    <51>   DW_AT_decl_column : 8
    <52>   DW_AT_type        : <0x5f>
    <56>   DW_AT_location    : 0x10 (location list)
    <5a>   DW_AT_GNU_locviews: 0xc

However the location lists say:

    00000010 v000000000000002 v000000000000003 views at 0000000c for:
             0000000000401020 0000000000401020 (DW_OP_lit0; DW_OP_stack_value)
    0000001d v000000000000003 v000000000000004 views at 0000000e for:
             0000000000401020 0000000000401020 (DW_OP_lit1; DW_OP_stack_value)


So on the one hand, I think gcc is setting the default here
and gdb is printing 0 correctly.

On the other hand, maybe locviews mean that gdb could sometimes show '1'.
I don't know though, and I can't find documentation on this attribute.

Anyway I tend to think this is at worst a missing feature from gdb.

-- 
You are receiving this mail because:
You are on the CC list for the bug.