[RFA] Avoid segfault in decode_line_2

[RFA] Avoid segfault in decode_line_2

[Michal Ludvig]

[-- Attachment #1: Type: text/plain, Size: 717 bytes --]

Hi all,
this patch fixes a problem that I met while debugging a testsuite 
failure on amd64:

 > Running gdb-head/gdb/testsuite/gdb.c++/templates.exp ...
 > FAIL: gdb.c++/templates.exp: constructor breakpoint (timeout)

(gdb) break T5<int>::T5
-> Segfault in linespec.c:486 [decode_line_2()] because
    values.sals[i].symtab is NULL and dereferencing of
    values.sals[i].symtab->filename crashes.

After some investigation I found out that .debug_line section of the 
input file was broken (reported to binutils@ list).
However broken debug info is not an excuse for GDB to crash.

OK to apply to head and branch?

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* (+420) 296.545.373 * http://www.suse.cz

[-- Attachment #2: linespec-safety-1.diff --]
[-- Type: text/plain, Size: 1123 bytes --]

2003-07-11  Michal Ludvig  <mludvig@suse.cz>

	* linespec.c (decode_line_2): Avoid crash if 
	find_function_start_sal() returns empty record.

Index: linespec.c
===================================================================
RCS file: /cvs/src/src/gdb/linespec.c,v
retrieving revision 1.49
diff -u -p -r1.49 linespec.c
--- linespec.c	8 Jun 2003 18:27:13 -0000	1.49
+++ linespec.c	11 Jul 2003 06:59:01 -0000
@@ -483,11 +483,16 @@ decode_line_2 (struct symbol *sym_arr[],
       if (sym_arr[i] && SYMBOL_CLASS (sym_arr[i]) == LOC_BLOCK)
 	{
 	  values.sals[i] = find_function_start_sal (sym_arr[i], funfirstline);
-	  printf_unfiltered ("[%d] %s at %s:%d\n",
+	  printf_unfiltered ("[%d] %s at %s:%d ",
 			     (i + 2),
 			     SYMBOL_PRINT_NAME (sym_arr[i]),
-			     values.sals[i].symtab->filename,
+			     values.sals[i].symtab ? 
+			       values.sals[i].symtab->filename :
+			       "?FILE",
 			     values.sals[i].line);
+	  if (! values.sals[i].symtab)
+	    printf_unfiltered ("[No symtab? Probably a broken debug info...]" );
+	  printf_unfiltered ("\n");
 	}
       else
 	printf_unfiltered ("?HERE\n");

Re: [RFA] Avoid segfault in decode_line_2

[Elena Zannoni]

Michal Ludvig writes:
 > Hi all,
 > this patch fixes a problem that I met while debugging a testsuite 
 > failure on amd64:
 > 
 >  > Running gdb-head/gdb/testsuite/gdb.c++/templates.exp ...
 >  > FAIL: gdb.c++/templates.exp: constructor breakpoint (timeout)
 > 
 > (gdb) break T5<int>::T5
 > -> Segfault in linespec.c:486 [decode_line_2()] because
 >     values.sals[i].symtab is NULL and dereferencing of
 >     values.sals[i].symtab->filename crashes.
 > 
 > After some investigation I found out that .debug_line section of the 
 > input file was broken (reported to binutils@ list).
 > However broken debug info is not an excuse for GDB to crash.
 > 
 > OK to apply to head and branch?
 > 
 > Michal Ludvig
 > -- 
 > * SuSE CR, s.r.o     * mludvig@suse.cz
 > * (+420) 296.545.373 * http://www.suse.cz
 > 2003-07-11  Michal Ludvig  <mludvig@suse.cz>
 > 
 > 	* linespec.c (decode_line_2): Avoid crash if 
 > 	find_function_start_sal() returns empty record.
 > 
 > Index: linespec.c
 > ===================================================================
 > RCS file: /cvs/src/src/gdb/linespec.c,v
 > retrieving revision 1.49
 > diff -u -p -r1.49 linespec.c
 > --- linespec.c	8 Jun 2003 18:27:13 -0000	1.49
 > +++ linespec.c	11 Jul 2003 06:59:01 -0000
 > @@ -483,11 +483,16 @@ decode_line_2 (struct symbol *sym_arr[],
 >        if (sym_arr[i] && SYMBOL_CLASS (sym_arr[i]) == LOC_BLOCK)
 >  	{
 >  	  values.sals[i] = find_function_start_sal (sym_arr[i], funfirstline);
 > -	  printf_unfiltered ("[%d] %s at %s:%d\n",
 > +	  printf_unfiltered ("[%d] %s at %s:%d ",
 >  			     (i + 2),
 >  			     SYMBOL_PRINT_NAME (sym_arr[i]),
 > -			     values.sals[i].symtab->filename,
 > +			     values.sals[i].symtab ? 
 > +			       values.sals[i].symtab->filename :
 > +			       "?FILE",
 >  			     values.sals[i].line);
 > +	  if (! values.sals[i].symtab)
 > +	    printf_unfiltered ("[No symtab? Probably a broken debug info...]" );
 > +	  printf_unfiltered ("\n");
 >  	}
 >        else
 >  	printf_unfiltered ("?HERE\n");

Yes, but, could you change this to use an if (values.sals[i].symtab)
before the printf_filtered and avoid the conditional expression? You
will end up getting rid of the if() for the No symtab?.. case as well,
since that can be folded into a single printf.

elena

Re: [RFA] Avoid segfault in decode_line_2

[Elena Zannoni]

Elena Zannoni writes:
 > Michal Ludvig writes:
 >  > Hi all,
 >  > this patch fixes a problem that I met while debugging a testsuite 
 >  > failure on amd64:
 >  > 
 >  >  > Running gdb-head/gdb/testsuite/gdb.c++/templates.exp ...
 >  >  > FAIL: gdb.c++/templates.exp: constructor breakpoint (timeout)
 >  > 
 >  > (gdb) break T5<int>::T5
 >  > -> Segfault in linespec.c:486 [decode_line_2()] because
 >  >     values.sals[i].symtab is NULL and dereferencing of
 >  >     values.sals[i].symtab->filename crashes.
 >  > 
 >  > After some investigation I found out that .debug_line section of the 
 >  > input file was broken (reported to binutils@ list).
 >  > However broken debug info is not an excuse for GDB to crash.
 >  > 
 >  > OK to apply to head and branch?
 >  > 
 >  > Michal Ludvig
 >  > -- 
 >  > * SuSE CR, s.r.o     * mludvig@suse.cz
 >  > * (+420) 296.545.373 * http://www.suse.cz
 >  > 2003-07-11  Michal Ludvig  <mludvig@suse.cz>
 >  > 
 >  > 	* linespec.c (decode_line_2): Avoid crash if 
 >  > 	find_function_start_sal() returns empty record.
 >  > 
 >  > Index: linespec.c
 >  > ===================================================================
 >  > RCS file: /cvs/src/src/gdb/linespec.c,v
 >  > retrieving revision 1.49
 >  > diff -u -p -r1.49 linespec.c
 >  > --- linespec.c	8 Jun 2003 18:27:13 -0000	1.49
 >  > +++ linespec.c	11 Jul 2003 06:59:01 -0000
 >  > @@ -483,11 +483,16 @@ decode_line_2 (struct symbol *sym_arr[],
 >  >        if (sym_arr[i] && SYMBOL_CLASS (sym_arr[i]) == LOC_BLOCK)
 >  >  	{
 >  >  	  values.sals[i] = find_function_start_sal (sym_arr[i], funfirstline);
 >  > -	  printf_unfiltered ("[%d] %s at %s:%d\n",
 >  > +	  printf_unfiltered ("[%d] %s at %s:%d ",
 >  >  			     (i + 2),
 >  >  			     SYMBOL_PRINT_NAME (sym_arr[i]),
 >  > -			     values.sals[i].symtab->filename,
 >  > +			     values.sals[i].symtab ? 
 >  > +			       values.sals[i].symtab->filename :
 >  > +			       "?FILE",
 >  >  			     values.sals[i].line);
 >  > +	  if (! values.sals[i].symtab)
 >  > +	    printf_unfiltered ("[No symtab? Probably a broken debug info...]" );
 >  > +	  printf_unfiltered ("\n");
 >  >  	}
 >  >        else
 >  >  	printf_unfiltered ("?HERE\n");
 > 
 > Yes, but, could you change this to use an if (values.sals[i].symtab)
 > before the printf_filtered and avoid the conditional expression? You
 > will end up getting rid of the if() for the No symtab?.. case as well,
 > since that can be folded into a single printf.
 > 
 > elena


I checked in the following:

2003-07-23  Michal Ludvig  <mludvig@suse.cz>
	    Elena Zannoni  <ezannoni@redhat.com>

 	* linespec.c (decode_line_2): Avoid crash if 
 	find_function_start_sal() returns empty record.

Index: linespec.c
===================================================================
RCS file: /cvs/uberbaum/gdb/linespec.c,v
retrieving revision 1.49
diff -u -p -r1.49 linespec.c
--- linespec.c	8 Jun 2003 18:27:13 -0000	1.49
+++ linespec.c	23 Jul 2003 19:20:14 -0000
@@ -483,11 +483,18 @@ decode_line_2 (struct symbol *sym_arr[],
       if (sym_arr[i] && SYMBOL_CLASS (sym_arr[i]) == LOC_BLOCK)
 	{
 	  values.sals[i] = find_function_start_sal (sym_arr[i], funfirstline);
-	  printf_unfiltered ("[%d] %s at %s:%d\n",
-			     (i + 2),
-			     SYMBOL_PRINT_NAME (sym_arr[i]),
-			     values.sals[i].symtab->filename,
-			     values.sals[i].line);
+	  if (values.sals[i].symtab)
+	    printf_unfiltered ("[%d] %s at %s:%d\n",
+			       (i + 2),
+			       SYMBOL_PRINT_NAME (sym_arr[i]),
+			       values.sals[i].symtab->filename,
+			       values.sals[i].line);
+	  else
+	    printf_unfiltered ("[%d] %s at ?FILE:%d [No symtab? Probably broken debug info...]\n",
+			       (i + 2),
+			       SYMBOL_PRINT_NAME (sym_arr[i]),
+			       values.sals[i].line);
+
 	}
       else
 	printf_unfiltered ("?HERE\n");

Re: [RFA] Avoid segfault in decode_line_2

[Michal Ludvig]

Elena Zannoni told me that:
> Elena Zannoni writes:
>  > Michal Ludvig writes:
>  >  > Hi all,
>  >  > this patch fixes a problem that I met while debugging a testsuite 
>  >  > failure on amd64:
>  > 
>  > Yes, but, could you change this to use an if (values.sals[i].symtab)
>  > before the printf_filtered and avoid the conditional expression? You
>  > will end up getting rid of the if() for the No symtab?.. case as well,
>  > since that can be folded into a single printf.
>  > 
>  > elena
> 
> 
> I checked in the following:
> 
> 2003-07-23  Michal Ludvig  <mludvig@suse.cz>
> 	    Elena Zannoni  <ezannoni@redhat.com>
> 
>  	* linespec.c (decode_line_2): Avoid crash if 
>  	find_function_start_sal() returns empty record.

Thanks!
(And sorry, I forgot to do it... :-)

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* (+420) 296.545.373 * http://www.suse.cz