On 01/17/2013 01:17 PM, Florian Weimer wrote:
On 01/16/2013 06:35 PM, Mark Wielaard wrote:
> On Wed, Jan 16, 2013 at 05:16:06PM +0100, Jan Kratochvil wrote:
>> On Wed, 16 Jan 2013 17:12:56 +0100, Florian Weimer wrote:
>>> Not sure if this a toolchain bug or elfutils bug. readelf from
>>> binutils prints GNU_UNIQUE.
>>
>> If not obvious binutils/readelf.c contains:
>>
>> if (binding == STB_GNU_UNIQUE
>> && (elf_header.e_ident[EI_OSABI] == ELFOSABI_GNU
>> /* GNU is still using the default value 0. */
>> || elf_header.e_ident[EI_OSABI] == ELFOSABI_NONE))
>> return "UNIQUE";
>
> So I think this is a toolchain bug. It is nice of binutils to
> accept STB_GNU_UNIQUE with the wrong EI_OSABI, but I don't think
> you can rely on all tools being so lenient.
Okay, I'll try to figure what's special about the affected DSOs and
isolate the toolchain bug.
I think I have found the binutils bug which is responsible:
<
http://sourceware.org/ml/binutils/2013-01/msg00362.html>
--
Florian Weimer / Red Hat Product Security Team