[patch] Fix a buglet in elfread.c

[patch] Fix a buglet in elfread.c

[Paul Pluzhnikov]

Greetings,

GDB appears to make unwarranted assumptions about presence of .data,
.rodata in and ELF image.

In http://sourceware.org/ml/gdb/2009-04/msg00228.html you can see an image
on Solaris which has .rodata, but no .data, and this causes

  elfread.c:366: internal-error: sect_index_data not initialized

when using /usr/lib/libXau.so.6 from Solaris snv103 (SunOS 5.11) for SPARC.

Attached patch fixes that.

OK?

--
Paul Pluzhnikov

2009-04-30  Paul Pluzhnikov  <ppluzhnikov@google.com>

	* elfread.c (elf_symtab_read): Don't assume .data and .rodata
	are present.


Index: elfread.c
===================================================================
RCS file: /cvs/src/src/gdb/elfread.c,v
retrieving revision 1.76
diff -u -p -u -r1.76 elfread.c
--- elfread.c	3 Jan 2009 05:57:51 -0000	1.76
+++ elfread.c	30 Apr 2009 15:03:41 -0000
@@ -427,10 +427,11 @@ elf_symtab_read (struct objfile *objfile
 			  int max_index;
 			  size_t size;
 
-			  max_index 
-			    = max (SECT_OFF_BSS (objfile),
-				   max (SECT_OFF_DATA (objfile),
-					SECT_OFF_RODATA (objfile)));
+			  max_index = SECT_OFF_BSS (objfile);
+			  if (objfile->sect_index_data > max_index)
+			    max_index = objfile->sect_index_data;
+			  if (objfile->sect_index_rodata > max_index)
+			    max_index = objfile->sect_index_rodata;
 
 			  /* max_index is the largest index we'll
 			     use into this array, so we must

Re: [patch] Fix a buglet in elfread.c

[Joel Brobecker]

> 2009-04-30  Paul Pluzhnikov  <ppluzhnikov@google.com>
> 
> 	* elfread.c (elf_symtab_read): Don't assume .data and .rodata
> 	are present.

I remember facing this exact same type of issue on Tru64. I haven't
had a chance to look at the patch yet, but I'll try to get to it soon.
I also want to test it on x86 & sparc solaris.

(although, theoretically, it should be sufficient to test it on
a system using ELF such as GNU/Linux)

-- 
Joel

Re: [patch] Fix a buglet in elfread.c

[Joel Brobecker]

> 2009-04-30  Paul Pluzhnikov  <ppluzhnikov@google.com>
> 
> 	* elfread.c (elf_symtab_read): Don't assume .data and .rodata
> 	are present.

Looks good to me. Please go ahead and apply.

I tested this patch on sparc-solaris 2.8 with -gstabs+, since this
seems to be related to stabs. I don't think we can really test
this part of the code with a GCC compiler, but perhaps it's is
still exercised throught the loading of some of the Solaris shared
libraries.  We did our best, and we had 2 pairs of eyes looking at it,
so it should be fine.

-- 
Joel

Re: [patch] Fix a buglet in elfread.c

[Paul Pluzhnikov]

On Thu, Apr 30, 2009 at 2:54 PM, Joel Brobecker <brobecker@adacore.com> wrote:

> Looks good to me. Please go ahead and apply.

Thanks. So committed.
-- 
Paul Pluzhnikov

Re: [patch] Fix a buglet in elfread.c

[Frank Middleton]

On 04/30/09 17:54, Joel Brobecker wrote:
>> 2009-04-30  Paul Pluzhnikov<ppluzhnikov@google.com>
>>
>> 	* elfread.c (elf_symtab_read): Don't assume .data and .rodata
>> 	are present.
>
> Looks good to me. Please go ahead and apply.
>
> I tested this patch on sparc-solaris 2.8 with -gstabs+, since this
> seems to be related to stabs. I don't think we can really test
> this part of the code with a GCC compiler, but perhaps it's is
> still exercised throught the loading of some of the Solaris shared
> libraries.  We did our best, and we had 2 pairs of eyes looking at it,
> so it should be fine.

FWIW this wasn't a problem until at or after Solaris 10 (SunOS 5.10),
so your testing would presumably show that this patch didn't break
anything. I've been using the patched version all day on SunOS 5.11
(snv103) and it seems to be working perfectly, so it doesn't seem
like anything got broken and it certainly seems fixed. Thanks again!