[Bug default/24823] [dwz] Bad DW_OP_GNU_variable_value DIE

[Bug default/24823] [dwz] Bad DW_OP_GNU_variable_value DIE

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=24823

--- Comment #3 from Tom de Vries <vries at gcc dot gnu.org> ---
*** Bug 24728 has been marked as a duplicate of this bug. ***

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

[Bug default/24823] [dwz] Bad DW_OP_GNU_variable_value DIE

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=24823

Tom de Vries <vries at gcc dot gnu.org> changed:

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

--- Comment #2 from Tom de Vries <vries at gcc dot gnu.org> ---
https://sourceware.org/git/?p=dwz.git;a=commit;h=5f3ba3ae6054153ea9126d8ec39979e16a430e69

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

[Bug default/24823] New: [dwz] Bad DW_OP_GNU_variable_value DIE

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=24823

            Bug ID: 24823
           Summary: [dwz] Bad DW_OP_GNU_variable_value DIE
           Product: dwz
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: default
          Assignee: nobody at sourceware dot org
          Reporter: vries at gcc dot gnu.org
                CC: dwz at sourceware dot org
  Target Milestone: ---

[ Spinoff of PR24728 - "varval FAILs with cc-with-dwz-m" ]

With the gdb.dwarf2/varval.exp test-case of gdb, we have:
...
$ gdb -q -batch varval -ex start -ex "print constval"
Temporary breakpoint 1 at 0x4004ab

Temporary breakpoint 1, 0x00000000004004ab in main ()
Bad DW_OP_GNU_variable_value DIE.
...

In the original exec we have a DIE with DW_OP_GNU_variable_value ref to DIE
0x112:
...
 <1><112>: Abbrev Number: 8 (DW_TAG_variable)
    <113>   DW_AT_name        : var_c
    <119>   DW_AT_type        : <0xd4>
    <11d>   DW_AT_external    : 1
    <11e>   DW_AT_const_value : 53
 ...
 <2><1f7>: Abbrev Number: 25 (DW_TAG_variable)
    <1f8>   DW_AT_name        : constval
    <201>   DW_AT_type        : <0xd4>
    <205>   DW_AT_location    : 6 byte block: fd 12 1 0 0 9f   
(DW_OP_GNU_variable_value: <0x112>; DW_OP_stack_value)
...
but after dwz -m we have moved var_c to the multifile:
...
 <1><86>: Abbrev Number: 1 (DW_TAG_variable)
    <87>   DW_AT_name        : var_c
    <8d>   DW_AT_type        : <0x7a>
    <8e>   DW_AT_external    : 1
    <8f>   DW_AT_const_value : 53
...
while constval has not, and now has an invalid DW_OP_GNU_variable_value ref:
...
 <2><159>: Abbrev Number: 18 (DW_TAG_variable)
    <15a>   DW_AT_name        : constval
    <163>   DW_AT_type        : <alt 0x7a>
    <167>   DW_AT_location    : 6 byte block: fd 4 0 0 0 9f    
(DW_OP_GNU_variable_value: <0x4>; DW_OP_stack_value)
...

So this is a dwz bug rather than a gdb bug.

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

[Bug default/24823] [dwz] Bad DW_OP_GNU_variable_value DIE

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=24823

Tom de Vries <vries at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at redhat dot com,
                   |                            |mark at klomp dot org

--- Comment #1 from Tom de Vries <vries at gcc dot gnu.org> ---
Given that the DW_OP_GNU_variable_value is similar to DW_FORM_ref_addr, and
does normally not point into other files ( the exception being some old ada
compiler, see http://eagercon.com/dwarf/issues/010322-2.htm ), there's no way
to reach the multifile from the original file.

So the fix is to forbid DW_OP_GNU_variable_value-referenced DIEs to move to the
multifile:
...
diff --git a/dwz.c b/dwz.c
index e5de164..577ea2c 100644
--- a/dwz.c
+++ b/dwz.c
@@ -1596,6 +1596,7 @@ read_exprloc (DSO *dso, dw_die_ref die, unsigned char
*ptr, size_t len,
                     dso->filename, get_DW_OP_str (op));
              return 1;
            }
+         ref->die_no_multifile = 1;
          if (unlikely (low_mem))
            {
              ref->die_referenced = 1;
...

[ AFAIU, the DIEs could move together to the multifile, but I don't know of any
current mechanism to enforce two DIEs to be either both moved to the multifile,
or both not moved to the multifile. ]

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