this is a forward from gcc bugzilla:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36568
afaics the glibc-2.8 is affected and there's
a patch provided in debian/ubuntu distro.
could you please verify and commit linked patch?

------- Additional Comment #1 From Ulrich Drepper 2008-08-13 05:21 -------

I see no problem with the current code:

$ ./a threaded
0x804886d handler+0x1a
0x11f400 __kernel_sigreturn+0x0
0x804888b crash+0x10
0x1ff20e clone+0x5e

------- Additional Comment #2 From Pawel Sikora 2008-08-13 07:30 -------

(In reply to comment #1)
> I see no problem with the current code:
>
> $ ./a threaded
> 0x804886d handler+0x1a
> 0x11f400 __kernel_sigreturn+0x0
> 0x804888b crash+0x10
> 0x1ff20e clone+0x5e

which kernel/glibc/gcc/distribution you are using?

------- Additional Comment #3 From Pawel Sikora 2008-11-18 08:24 -------

(In reply to comment #1)
> I see no problem with the current code:
> 
> $ ./a threaded
> 0x804886d handler+0x1a
> 0x11f400 __kernel_sigreturn+0x0
> 0x804888b crash+0x10
> 0x1ff20e clone+0x5e

did you test it on the x86_64 architecture?
i'm asking because your logs shows 32-bits addresses
and the problem exist only on x86_64.