[Bug debug/40012] New: Bad debug info for local variables

[Bug debug/40012] New: Bad debug info for local variables

[gcc at magfr dot user dot lysator dot liu dot se]

Given the attached program compiled with

gcc -ggdb3 -o dlmod dlmod.i

the DW_AT_location for the variable 'problem' seems to be off by 16 bytes, so
when I try to examine the variable from the debugger that returns the wrong
value.

Below is some information about my compiler:

svnversion reports that I am using revision 147075.

$ gcc --version --verbose
Using built-in specs.
gcc (GCC) 4.5.0 20090503 (experimental)
Copyright (C) 2009 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.


Target: i686-pc-linux-gnu
Configured with: ../trunk/configure --enable-languages=c,c++
--prefix=/usr/local
/gcc-head --enable-shared --with-system-zlib --without-included-gettext
--enable
-threads --disable-nls --enable-__cxa_atexit --enable-clocale=gnu
--enable-libst
dcxx-debug --with-arch=core2 --with-tune=core2 --enable-checking=release
--enabl
e-concept-checks --disable-libgomp --disable-libmudflap --disable-libssp
Thread model: posix


-- 
           Summary: Bad debug info for local variables
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: debug
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: gcc at magfr dot user dot lysator dot liu dot se
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] Bad debug info for local variables

[gcc at magfr dot user dot lysator dot liu dot se]

------- Comment #1 from gcc at magfr dot user dot lysator dot liu dot se  2009-05-03 22:05 -------
Created an attachment (id=17794)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17794&action=view)
Test case demonstrating the problem.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Bad debug info for local variables

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-debug
            Summary|Bad debug info for local    |[4.5 Regression] Bad debug
                   |variables                   |info for local variables
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Bad debug info for local variables

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2009-05-28 14:52 -------
Confirmed.

(gdb) start
Temporary breakpoint 2, main () at dlmod.i:24
24        list a3 = { 0 };
(gdb) p &a3
$4 = (list *) 0xffffd05c
(gdb) n
25        f0(0, 0, 0, &a3);
(gdb) s
f0 (a=0x0, b=0x0, c=0x0, d=0xffffd05c) at dlmod.i:15
15          for(problem = d; problem; problem = problem->next) {
(gdb) n
16            int variable = 0;
(gdb) p problem
$5 = (list *) 0x0
(gdb) n
17            f1(c, problem);
(gdb) s
f1 (a=0x0, b=0xffffd05c) at dlmod.i:29
29      void f1(void*a, list*b) { }
(gdb) p b
$6 = (list *) 0xffffd05c
(gdb) up
#1  0x0804840b in f0 (a=0x0, b=0x0, c=0x0, d=0xffffd05c) at dlmod.i:17
17            f1(c, problem);
(gdb) p problem
$7 = (list *) 0x0

works with 4.4.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
      Known to fail|                            |4.5.0
      Known to work|                            |4.4.0
           Priority|P3                          |P1
   Last reconfirmed|0000-00-00 00:00:00         |2009-05-28 14:52:09
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[hjl dot tools at gmail dot com]

------- Comment #3 from hjl dot tools at gmail dot com  2009-05-29 03:22 -------
Revision 146817:

http://gcc.gnu.org/ml/gcc-cvs/2009-04/msg01459.html

is the cause.


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hjl dot tools at gmail dot
                   |                            |com, matz at suse dot de
            Summary|[4.5 Regression] Bad debug  |[4.5 Regression] Revision
                   |info for local variables    |146817 generated bad debug
                   |                            |info for local variables


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[dje at gcc dot gnu dot org]

------- Comment #4 from dje at gcc dot gnu dot org  2009-06-01 21:49 -------
Local variable debugging is broken for AIX stabs as well.

On AIX, the DECL and the insn contain different offsets from virtual-stack-vars
for the same variable:

(mem/c/i:SI (plus:SI (reg/f:SI 115 virtual-stack-vars)
        (const_int 8 [0x8])) [0 a+0 S4 A128])

versus

(insn 11 10 12 3 ad.c:11 (set (reg:SI 123)
        (mem/c/i:SI (reg/f:SI 115 virtual-stack-vars) [0 a+0 S4 A32])) -1
(nil))


-- 

dje at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dje at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[dje at gcc dot gnu dot org]

------- Comment #5 from dje at gcc dot gnu dot org  2009-06-02 12:51 -------
cfgexpand.c:expand_used_vars() expands and instantiates virtual-stack-vars for
INSNs when it walks the partion map:

  for (i = 0; i < SA.map->num_partitions; i++)
    {
      tree var = partition_to_var (SA.map, i);

      gcc_assert (is_gimple_reg (var));
      if (TREE_CODE (SSA_NAME_VAR (var)) == VAR_DECL)
        expand_one_var (var, true, true);

It then instantiates the same virtual-stack-vars in debug DECLs when it calls
expand_used_vars_for_block (outer_block, true), which calls

  /* Expand all variables at this level.  */
  for (t = BLOCK_VARS (block); t ; t = TREE_CHAIN (t))
    if (TREE_USED (t))
      expand_one_var (t, toplevel, true);

The INSN var is an SSA_NAME and expand_one_stack_var_at() uses the alignment in
the DECL; the BLOCK var is not an SSA_NAME and the alignment is computed as the
STACK_BOUNDARY.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[ppluzhnikov at google dot com]

------- Comment #6 from ppluzhnikov at google dot com  2009-06-02 23:02 -------
Also a problem on x86_64-linux:
http://gcc.gnu.org/ml/gcc-patches/2009-06/msg00190.html


-- 

ppluzhnikov at google dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ppluzhnikov at google dot
                   |                            |com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[matz at gcc dot gnu dot org]

------- Comment #7 from matz at gcc dot gnu dot org  2009-06-03 09:51 -------
I'm looking at this now.


-- 

matz at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |matz at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2009-05-28 14:52:09         |2009-06-03 09:51:23
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[ramana at gcc dot gnu dot org]

------- Comment #8 from ramana at gcc dot gnu dot org  2009-06-03 13:17 -------
Also broken on arm-eabi cross. Using gdb and the exact steps as in Comment #0
shows me problem as 0x100.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[matz at gcc dot gnu dot org]

------- Comment #9 from matz at gcc dot gnu dot org  2009-06-03 15:57 -------
Bah, simply not expanding those variables (whose SSA partitions were already)
by clearing TREE_USED on them fixes the problem, but makes those variables
have no debug info at all, as the DECLs then have no DECL_RTL_SET anymore.

Which indeed we can't easily use in case one base variable got split into
several SSA partitions: we have several places but only one DECL.  This is
no problem in -g -O0 mode, so I can hack around this.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[matz at gcc dot gnu dot org]

------- Comment #10 from matz at gcc dot gnu dot org  2009-06-08 11:37 -------
Subject: Bug 40012

Author: matz
Date: Mon Jun  8 11:37:12 2009
New Revision: 148273

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=148273
Log:
        PR debug/40012

        * cfgexpand.c (set_rtl): Store place also in DECL_RTL, if all
        partitions use the same.
        (expand_one_var): Deal with DECL_RTL sometimes begin set also
        for basevars of SSA_NAMEs.
        (expand_used_vars): Reset TREE_USED for basevars of SSA_NAMEs,
        to not expand them twice.
        (gimple_expand_cfg): Clear DECL_RTL for those decls that have
        multiple places.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/cfgexpand.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[matz at gcc dot gnu dot org]

------- Comment #11 from matz at gcc dot gnu dot org  2009-06-08 11:38 -------
Fixed.


-- 

matz at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[ppluzhnikov at google dot com]

------- Comment #12 from ppluzhnikov at google dot com  2009-06-27 00:49 -------
This appears to still be broken in 32-bit mode.

I just built GCC @148996 on x86_64, and compiled the attached test case with
'-m32':

(gdb) start
Breakpoint 1 at 0x80483d9: file gcc-pr40012-test.c, line 24.
main () at gcc-pr40012-test.c:24
24        list a3 = { 0 };
(gdb) n
25        f0(0, 0, 0, &a3);
(gdb) p &a3
$1 = (list *) 0xffffccb4
(gdb) s
f0 (a=0x0, b=0x0, c=0x0, d=0xffffcc8c) at gcc-pr40012-test.c:15
15          for(problem = d; problem; problem = problem->next) {

Clearly GDB and GCC are in disagreement about the location of a3.

Disassebly shows:

0x080483d0 <main+0>:    push   %ebp
0x080483d1 <main+1>:    mov    %esp,%ebp
0x080483d3 <main+3>:    and    $0xfffffff0,%esp
0x080483d6 <main+6>:    sub    $0x20,%esp
0x080483d9 <main+9>:    movl   $0x0,0x1c(%esp)
0x080483e1 <main+17>:   lea    0x1c(%esp),%eax
0x080483e5 <main+21>:   mov    %eax,0xc(%esp)
0x080483e9 <main+25>:   movl   $0x0,0x8(%esp)
0x080483f1 <main+33>:   movl   $0x0,0x4(%esp)
0x080483f9 <main+41>:   movl   $0x0,(%esp)
0x08048400 <main+48>:   call   0x8048394 <f0>
0x08048405 <main+53>:   mov    $0x0,%eax
0x0804840a <main+58>:   leave  
0x0804840b <main+59>:   ret    

readelf -w:

 <1><1ee>: Abbrev Number: 12 (DW_TAG_subprogram)
    <1ef>   DW_AT_external    : 1       
    <1f0>   DW_AT_name        : (indirect string, offset: 0xf5): main   
    <1f4>   DW_AT_decl_file   : 1       
    <1f5>   DW_AT_decl_line   : 22      
    <1f6>   DW_AT_type        : <0x1df> 
    <1fa>   DW_AT_low_pc      : 0x80483d0       
    <1fe>   DW_AT_high_pc     : 0x804840c       
    <202>   DW_AT_frame_base  : 0x38    (location list)
    <206>   DW_AT_sibling     : <0x218> 
 <2><20a>: Abbrev Number: 13 (DW_TAG_variable)
    <20b>   DW_AT_name        : a3      
    <20e>   DW_AT_decl_file   : 1       
    <20f>   DW_AT_decl_line   : 24      
    <210>   DW_AT_type        : <0x161> 
    <214>   DW_AT_location    : 2 byte block: 91 14     (DW_OP_fbreg: 20)

And

    00000038 080483d0 080483d1 (DW_OP_breg4: 4)
    00000038 080483d1 080483d3 (DW_OP_breg4: 8)
    00000038 080483d3 0804840b (DW_OP_breg5: 8)
    00000038 0804840b 0804840c (DW_OP_breg4: 4)
    00000038 <End of list>


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[ebotcazou at gcc dot gnu dot org]

------- Comment #13 from ebotcazou at gcc dot gnu dot org  2009-06-27 20:18 -------
> This appears to still be broken in 32-bit mode.

Yes, I've seen similar problems with 4.4.0 and 4.5.0 on x86.  Probably the
stack realignment stuff.  I'd suggest opening a new PR.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012

[Bug debug/40012] [4.5 Regression] Revision 146817 generated bad debug info for local variables

[ppluzhnikov at google dot com]

------- Comment #14 from ppluzhnikov at google dot com  2009-06-29 20:11 -------
I verified that @146817 is not causing the current 32-bit breakage; I'll open a
separate PR.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40012