[RFA]: Fix bug gdb/321

[RFA]: Fix bug gdb/321

[Corinna Vinschen]

Hi,

the following patch seems to fix the bug described in gdb/321:

http://sources.redhat.com/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gdb&pr=321

Basically the problem is that gdb's backtrace reports the wrong calling
function if a function ends in a call to abort() (or another noreturn func)
under some conditions.  The bug report suggests to use a by one decremented
pc when evaluating the function.  The following patch does exactly this by
using the return value of frame_address_in_block() instead of frame->pc in
calls which evaluate symbols.

Corinna

ChangeLog:

        * stack.c (print_frame): Use result of frame_address_in_block()
        instead of fi->pc when evaluating symbols.
        (backtrace_command_1): Ditto.

Index: stack.c
===================================================================
RCS file: /cvs/src/src/gdb/stack.c,v
retrieving revision 1.39
diff -u -p -r1.39 stack.c
--- stack.c	10 Jun 2002 23:25:50 -0000	1.39
+++ stack.c	11 Jul 2002 18:42:47 -0000
@@ -461,7 +461,7 @@ print_frame (struct frame_info *fi, 
   stb = ui_out_stream_new (uiout);
   old_chain = make_cleanup_ui_out_stream_delete (stb);
 
-  func = find_pc_function (fi->pc);
+  func = find_pc_function (frame_address_in_block (fi));
   if (func)
     {
       /* In certain pathological cases, the symtabs give the wrong
@@ -480,7 +480,7 @@ print_frame (struct frame_info *fi, 
          ever changed many parts of GDB will need to be changed (and we'll
          create a find_pc_minimal_function or some such).  */
 
-      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (fi->pc);
+      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (frame_address_in_block (fi));
       if (msymbol != NULL
 	  && (SYMBOL_VALUE_ADDRESS (msymbol)
 	      > BLOCK_START (SYMBOL_BLOCK_VALUE (func))))
@@ -528,7 +528,7 @@ print_frame (struct frame_info *fi, 
     }
   else
     {
-      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (fi->pc);
+      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (frame_address_in_block (fi));
       if (msymbol != NULL)
 	{
 	  funname = SYMBOL_NAME (msymbol);
@@ -1111,7 +1111,7 @@ backtrace_command_1 (char *count_exp, in
 	   fi = get_prev_frame (fi))
 	{
 	  QUIT;
-	  ps = find_pc_psymtab (fi->pc);
+	  ps = find_pc_psymtab (frame_address_in_block (fi));
 	  if (ps)
 	    PSYMTAB_TO_SYMTAB (ps);	/* Force syms to come in */
 	}

-- 
Corinna Vinschen
Cygwin Developer
Red Hat, Inc.
mailto:vinschen@redhat.com

Re: [RFA]: Fix bug gdb/321

[Daniel Jacobowitz]

On Thu, Jul 11, 2002 at 08:52:08PM +0200, Corinna Vinschen wrote:
> Hi,
> 
> the following patch seems to fix the bug described in gdb/321:
> 
> http://sources.redhat.com/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gdb&pr=321
> 
> Basically the problem is that gdb's backtrace reports the wrong calling
> function if a function ends in a call to abort() (or another noreturn func)
> under some conditions.  The bug report suggests to use a by one decremented
> pc when evaluating the function.  The following patch does exactly this by
> using the return value of frame_address_in_block() instead of frame->pc in
> calls which evaluate symbols.

FWIW, I think this is perfect - thanks!

> 
> Corinna
> 
> ChangeLog:
> 
>         * stack.c (print_frame): Use result of frame_address_in_block()
>         instead of fi->pc when evaluating symbols.
>         (backtrace_command_1): Ditto.
> 
> Index: stack.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/stack.c,v
> retrieving revision 1.39
> diff -u -p -r1.39 stack.c
> --- stack.c	10 Jun 2002 23:25:50 -0000	1.39
> +++ stack.c	11 Jul 2002 18:42:47 -0000
> @@ -461,7 +461,7 @@ print_frame (struct frame_info *fi, 
>    stb = ui_out_stream_new (uiout);
>    old_chain = make_cleanup_ui_out_stream_delete (stb);
>  
> -  func = find_pc_function (fi->pc);
> +  func = find_pc_function (frame_address_in_block (fi));
>    if (func)
>      {
>        /* In certain pathological cases, the symtabs give the wrong
> @@ -480,7 +480,7 @@ print_frame (struct frame_info *fi, 
>           ever changed many parts of GDB will need to be changed (and we'll
>           create a find_pc_minimal_function or some such).  */
>  
> -      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (fi->pc);
> +      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (frame_address_in_block (fi));
>        if (msymbol != NULL
>  	  && (SYMBOL_VALUE_ADDRESS (msymbol)
>  	      > BLOCK_START (SYMBOL_BLOCK_VALUE (func))))
> @@ -528,7 +528,7 @@ print_frame (struct frame_info *fi, 
>      }
>    else
>      {
> -      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (fi->pc);
> +      struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (frame_address_in_block (fi));
>        if (msymbol != NULL)
>  	{
>  	  funname = SYMBOL_NAME (msymbol);
> @@ -1111,7 +1111,7 @@ backtrace_command_1 (char *count_exp, in
>  	   fi = get_prev_frame (fi))
>  	{
>  	  QUIT;
> -	  ps = find_pc_psymtab (fi->pc);
> +	  ps = find_pc_psymtab (frame_address_in_block (fi));
>  	  if (ps)
>  	    PSYMTAB_TO_SYMTAB (ps);	/* Force syms to come in */
>  	}
> 
> -- 
> Corinna Vinschen
> Cygwin Developer
> Red Hat, Inc.
> mailto:vinschen@redhat.com
> 

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: [RFA]: Fix bug gdb/321

[Corinna Vinschen]

On Thu, Jul 11, 2002 at 02:56:03PM -0400, Daniel Jacobowitz wrote:
> On Thu, Jul 11, 2002 at 08:52:08PM +0200, Corinna Vinschen wrote:
> > Hi,
> > 
> > the following patch seems to fix the bug described in gdb/321:
> > 
> > http://sources.redhat.com/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gdb&pr=321
> > 
> > Basically the problem is that gdb's backtrace reports the wrong calling
> > function if a function ends in a call to abort() (or another noreturn func)
> > under some conditions.  The bug report suggests to use a by one decremented
> > pc when evaluating the function.  The following patch does exactly this by
> > using the return value of frame_address_in_block() instead of frame->pc in
> > calls which evaluate symbols.
> 
> FWIW, I think this is perfect - thanks!

Applied.

Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Developer
Red Hat, Inc.
mailto:vinschen@redhat.com

Re: [RFA]: Fix bug gdb/321

[Daniel Jacobowitz]

On Thu, Jul 11, 2002 at 09:30:10PM +0200, Corinna Vinschen wrote:
> On Thu, Jul 11, 2002 at 02:56:03PM -0400, Daniel Jacobowitz wrote:
> > On Thu, Jul 11, 2002 at 08:52:08PM +0200, Corinna Vinschen wrote:
> > > Hi,
> > > 
> > > the following patch seems to fix the bug described in gdb/321:
> > > 
> > > http://sources.redhat.com/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gdb&pr=321
> > > 
> > > Basically the problem is that gdb's backtrace reports the wrong calling
> > > function if a function ends in a call to abort() (or another noreturn func)
> > > under some conditions.  The bug report suggests to use a by one decremented
> > > pc when evaluating the function.  The following patch does exactly this by
> > > using the return value of frame_address_in_block() instead of frame->pc in
> > > calls which evaluate symbols.
> > 
> > FWIW, I think this is perfect - thanks!
> 
> Applied.

For the record, I can't approve patches - although this one was fairly
obvious, so I don't think anyone will complain.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: [RFA]: Fix bug gdb/321

[Corinna Vinschen]

On Thu, Jul 11, 2002 at 03:33:08PM -0400, Daniel Jacobowitz wrote:
> On Thu, Jul 11, 2002 at 09:30:10PM +0200, Corinna Vinschen wrote:
> > On Thu, Jul 11, 2002 at 02:56:03PM -0400, Daniel Jacobowitz wrote:
> > > FWIW, I think this is perfect - thanks!
> > 
> > Applied.
> 
> For the record, I can't approve patches - although this one was fairly
> obvious, so I don't think anyone will complain.

Oops.

Corinna