[Bug debug/44126] New: wrong location description for DW_AT_vtable_elem_location

[Bug debug/44126] New: wrong location description for DW_AT_vtable_elem_location

[tromey at gcc dot gnu dot org]

Consider this simple example:

class K {
public:
  virtual int x() { return 23; }
};

K k;  


This yields this attribute in the dwarf:

<7a>   DW_AT_vtable_elem_location: 2 byte block: 10 0       (DW_OP_constu: 0)

However, this is incorrect.
The location description ought to compute the address of the vtable slot.


-- 
           Summary: wrong location description for
                    DW_AT_vtable_elem_location
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: debug
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: tromey at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

[Bug debug/44126] wrong location description for DW_AT_vtable_elem_location

[jason at gcc dot gnu dot org]

------- Comment #1 from jason at gcc dot gnu dot org  2010-05-13 20:42 -------
"An entry for a virtual function also has a DW_AT_vtable_elem_location
attribute whose value contains a location description yielding the address of
the slot for the function within the virtual function table for the enclosing
class. The address of an object of the enclosing type is pushed onto the
expression stack before the location description is evaluated."

Dodji, want to look at this?


-- 

jason at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at gcc dot gnu dot
                   |                            |org, dodji at gcc dot gnu
                   |                            |dot org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2010-05-13 20:42:01
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

[Bug debug/44126] wrong location description for DW_AT_vtable_elem_location

[jakub at gcc dot gnu dot org]

------- Comment #2 from jakub at gcc dot gnu dot org  2010-05-13 21:48 -------
If I understand it well, we should:
--- dwarf2out.c 2010-05-13 23:36:24.000000000 +0200
+++ dwarf2out.c    2010-05-13 23:55:07.422464196 +0200     
@@ -17094,10 +17094,19 @@ add_pure_or_virtual_attribute (dw_die_re
       add_AT_unsigned (die, DW_AT_virtuality, DW_VIRTUALITY_virtual);

       if (host_integerp (DECL_VINDEX (func_decl), 0))
-       add_AT_loc (die, DW_AT_vtable_elem_location,
-                   new_loc_descr (DW_OP_constu,
-                                  tree_low_cst (DECL_VINDEX (func_decl), 0),
-                                  0));
+       {
+         dw_loc_descr_ref loc;
+         HOST_WIDE_INT offset 
+           = tree_low_cst (DECL_VINDEX (func_decl), 0)
+             * (POINTER_SIZE / BITS_PER_UNIT);
+         if (DWARF2_ADDR_SIZE == (POINTER_SIZE / BITS_PER_UNIT))
+           loc = new_loc_descr (DW_OP_deref, 0, 0);
+         else
+           loc = new_loc_descr (DW_OP_deref_size,
+                                POINTER_SIZE / BITS_PER_UNIT, 0);
+         add_loc_descr (&loc, new_loc_descr (DW_OP_plus_uconst, offset, 0));
+         add_AT_loc (die, DW_AT_vtable_elem_location, loc);
+       }

       /* GNU extension: Record what type this method came from originally.
        * */
       if (debug_info_level > DINFO_LEVEL_TERSE

The consumers should be pushing address of object, so we need to dereference
its first pointer to get at vtable + 8 resp. 16 and DECL_VINDEX is number of
pointers after that spot (seems to be the case even on ia64).

The problem is that gdb won't grok that, so we have a chicken-and-egg issue.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

[Bug debug/44126] wrong location description for DW_AT_vtable_elem_location

[dodji at redhat dot com]

------- Comment #3 from dodji at gcc dot gnu dot org  2010-05-14 06:41 -------
Subject: Re:  wrong location description for
 DW_AT_vtable_elem_location

> Dodji, want to look at this?
Sure.

Like, Jakub said, we need to synchronize with GDB. I'll test Jakub's
patch ASAP and push the change when GDB is ready to consume it then.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

[Bug debug/44126] wrong location description for DW_AT_vtable_elem_location

[tromey at gcc dot gnu dot org]

------- Comment #4 from tromey at gcc dot gnu dot org  2010-06-11 14:53 -------
I think the problem with this patch is that it leaves gdb no way
to determine which approach it should use.  This is important because
there is a lot of existing code compiled with the incorrect approach.

Currently gdb has a heuristic where it uses the incorrect style
if it sees DW_AT_containing_type in the virtual function's DIE.
This appears to be a GNU extension (you can see it at the end of the
patch, in the context -- but there is another instance in dwarf2out.c).
Maybe removing that would help.  I'm not sure if it is needed and/or
useful for some other reason, maybe a little exploration of the history
would help.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

[Bug debug/44126] wrong location description for DW_AT_vtable_elem_location

[tromey at gcc dot gnu dot org]

------- Comment #5 from tromey at gcc dot gnu dot org  2010-06-11 15:07 -------
Jakub pointed out that gdb can just look for an isolated
DW_OP_constu to fall back to the old code.
I will write a gdb patch.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

[Bug debug/44126] wrong location description for DW_AT_vtable_elem_location

[tromey at gcc dot gnu dot org]

------- Comment #6 from tromey at gcc dot gnu dot org  2010-06-11 20:02 -------
Ok, I committed the gdb change:
http://sourceware.org/ml/gdb-patches/2010-06/msg00287.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

[Bug debug/44126] wrong location description for DW_AT_vtable_elem_location

[tromey at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=44126

--- Comment #7 from Tom Tromey <tromey at gcc dot gnu.org> ---
I happened to be looking in this area and I see that gcc still
generates the old, incorrect form.