Re: bfd_get_full_section_contents memory leak, plus

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Sat, Oct 20, 2012 at 5:21 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Sat, Oct 20, 2012 at 3:17 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Sat, Oct 20, 2012 at 1:28 AM, Alan Modra <amodra@gmail.com> wrote:
>>> On Sat, Oct 20, 2012 at 10:42:21AM +1030, Alan Modra wrote:
>>>> This on top of Tom's patch ought to cure the failures.  Testing
>>>> in progress.
>>>>
>>>>       * compress.c: Reinstate 2012-10-19 change.
>>>>       (bfd_get_full_section_contents): Don't free unless we alloc.
>>>>       Use proper decompress size.  Delete some vars, rename others.
>>>
>>> Testing successful.  Committed.
>>>
>>
>> It breaks objdump.  I checked in this tectase to show the
>> error.
>>
>> --
>> H.J.
>> ---
>> 2012-10-20  H.J. Lu  <hongjiu.lu@intel.com>
>>
>>         * binutils-all/compress.exp: Add an objdump test.
>>         * binutils-all/dw2-1.W: New file.
>>
>
> This patch works for me:
>
>
> diff --git a/binutils/objdump.c b/binutils/objdump.c
> index 44e857a..b1a5be6 100644
> --- a/binutils/objdump.c
> +++ b/binutils/objdump.c
> @@ -2277,6 +2277,7 @@ load_specific_debug_section (enum
> dwarf_section_display_enum debug,
>           the bfd_section, and tell it that the contents are
>           already in memory.  */
>        sec->contents = section->start;
> +      sec->compress_status = COMPRESS_SECTION_NONE;
>        sec->flags |= SEC_IN_MEMORY;
>        sec->size = section->size;
>
>

This one may be better,


-- 
H.J.
---
diff --git a/bfd/compress.c b/bfd/compress.c
index 294bfd3..2ffa6f0 100644
--- a/bfd/compress.c
+++ b/bfd/compress.c
@@ -166,6 +166,9 @@ bfd_get_full_section_contents (bfd *abfd, sec_ptr
sec, bfd_byte **ptr)
   bfd_byte *compressed_buffer;
 #endif

+  if (p != NULL && p == sec->contents)
+    return TRUE;
+
   if (abfd->direction != write_direction && sec->rawsize != 0)
     sz = sec->rawsize;
   else