buglet in binutils/debug.c

buglet in binutils/debug.c

[Aaron J. Grier]

--- binutils-2.9.5.0.37/binutils/debug.c.orig   Fri Jun 16 17:18:56 2000
+++ binutils-2.9.5.0.37/binutils/debug.c        Fri Jun 16 17:28:16 2000
@@ -2198,6 +2198,12 @@
                   debug_get_type_name (handle, type));
          return NULL;
        }
+      if (l->next == l)
+        {
+          fprintf (stderr,
+                   _("debug_get_real_type: l->next == l, I'm in a loop\n"));
+          break;
+        }
     }
 
   rl.next = list;

this is probably not an optimal solution, but I was getting into an
endless loop when running m68k-rtems-objdump --debugging --disassemble
and needed a quick fix.

-- 
  Aaron J. Grier   |   Frye Electronics, Tigard, OR   |   aaron@frye.com

Re: buglet in binutils/debug.c

[Nick Clifton]

Hi Aaron,

: --- binutils-2.9.5.0.37/binutils/debug.c.orig   Fri Jun 16 17:18:56 2000
: +++ binutils-2.9.5.0.37/binutils/debug.c        Fri Jun 16 17:28:16 2000
: @@ -2198,6 +2198,12 @@
:                    debug_get_type_name (handle, type));
:           return NULL;
:         }
: +      if (l->next == l)
: +        {
: +          fprintf (stderr,
: +                   _("debug_get_real_type: l->next == l, I'm in a loop\n"));
: +          break;
: +        }
:      }
:  
:    rl.next = list;
: 
: this is probably not an optimal solution, but I was getting into an
: endless loop when running m68k-rtems-objdump --debugging --disassemble
: and needed a quick fix.

You are right, this is not really the right way to solve this
problem.  You ought to track down just why the list is getting
corrupted.  (Even with the patch above applied you will probably find
that you still run into other probekms, all due to the underlying
cause of the list corruption).

Hence I cannot accept this patch.  Sorry.

Cheers
	Nick

Re: buglet in binutils/debug.c

[Aaron J. Grier]

On Fri, Jun 16, 2000 at 05:41:49PM -0700, Aaron J. Grier wrote:
> --- binutils-2.9.5.0.37/binutils/debug.c.orig   Fri Jun 16 17:18:56 2000
> +++ binutils-2.9.5.0.37/binutils/debug.c        Fri Jun 16 17:28:16 2000
> @@ -2198,6 +2198,12 @@
>                    debug_get_type_name (handle, type));
>           return NULL;
>         }
> +      if (l->next == l)
> +        {
> +          fprintf (stderr,
> +                   _("debug_get_real_type: l->next == l, I'm in a loop\n"));
> +          break;
> +        }
>      }
>  
>    rl.next = list;
> 
> this is probably not an optimal solution, but I was getting into an
> endless loop when running m68k-rtems-objdump --debugging --disassemble
> and needed a quick fix.

I noticed this a year ago.  this problem still exists in
binutils-2.12.2, and in the current version of binutils, apparent for
m68k-rtems-elf when trying to objcopy from elf to ieee with --debugging.

any ideas where to start hunting this bug down?  it's certainly been
there long enough...

-- 
  Aaron J. Grier   |    Frye Electronics, Tigard, OR   |   aaron@frye.com
                       Of course unix clones exist...
                  How else are unix supposed to reproduce?

Re: buglet in binutils/debug.c

[Aaron J. Grier]

On Wed, Jul 18, 2001 at 04:22:02PM -0700, Aaron J. Grier wrote:

> any ideas where to start hunting this bug down?  it's certainly been
> there long enough...

looks like it might be a problem with the compiler, as the problem goes
away when I turn off optimization.  my build machine:

aaron2$ uname -a
Linux aaron2.internal 2.2.16 #1 Tue Jun 20 11:04:37 PDT 2000 i586 unknown
aaron2$ gcc --version
egcs-2.91.66

time to upgrade my compiler, I guess.  :)

-- 
  Aaron J. Grier   |    Frye Electronics, Tigard, OR   |   aaron@frye.com
                       Of course unix clones exist...
                  How else are unix supposed to reproduce?