[OE-core] [dizzy][PATCH] libunwind: backport patch to link against libgcc_s intead of libgcc

[Jonathan Liu]

On 11 March 2015 at 14:19, Khem Raj <raj.khem at gmail.com> wrote:
>
>> On Mar 10, 2015, at 8:05 PM, Jonathan Liu <net147 at gmail.com> wrote:
>>
>> On 11 March 2015 at 13:43, Khem Raj <raj.khem at gmail.com> wrote:
>>>
>>>> On Mar 10, 2015, at 7:33 PM, Jonathan Liu <net147 at gmail.com> wrote:
>>>>
>>>> Signed-off-by: Jonathan Liu <net147 at gmail.com>
>>>> ---
>>>> ...-libunwind-to-libgcc_s-rather-than-libgcc.patch | 42 ++++++++++++++++++++++
>>>> meta/recipes-support/libunwind/libunwind_1.1.bb    |  1 +
>>>> 2 files changed, 43 insertions(+)
>>>> create mode 100644 meta/recipes-support/libunwind/libunwind-1.1/Link-libunwind-to-libgcc_s-rather-than-libgcc.patch
>>>>
>>>> diff --git a/meta/recipes-support/libunwind/libunwind-1.1/Link-libunwind-to-libgcc_s-rather-than-libgcc.patch b/meta/recipes-support/libunwind/libunwind-1.1/Link-libunwind-to-libgcc_s-rather-than-libgcc.patch
>>>> new file mode 100644
>>>> index 0000000..0e55c91
>>>> --- /dev/null
>>>> +++ b/meta/recipes-support/libunwind/libunwind-1.1/Link-libunwind-to-libgcc_s-rather-than-libgcc.patch
>>>> @@ -0,0 +1,42 @@
>>>> +From 508ca17a7be01d1cc960d9a07d0af4513948fb8d Mon Sep 17 00:00:00 2001
>>>> +From: Thierry Reding <treding at nvidia.com>
>>>> +Date: Fri, 27 Jun 2014 08:40:33 +0200
>>>> +Subject: [PATCH] Link libunwind to libgcc_s rather than libgcc
>>>> +
>>>> +For some architectures, -lgcc and -lgcc_s are not equivalent. On ARM for
>>>> +example, libgcc_s.so.1 contains some symbols needed by libunwind which
>>>> +are not present in libgcc.
>>>> +
>>>> +This causes the following link error when building the X.Org X server
>>>> +with libunwind support:
>>>> +
>>>> +       CCLD     Xorg
>>>> +     /usr/lib/libunwind.so: undefined reference to `__aeabi_unwind_cpp_pr0'
>>>> +     /usr/lib/libunwind.so: undefined reference to `__aeabi_unwind_cpp_pr1'
>>>> +
>>>> +Linking against libgcc_s explicitly solves this problem.
>>>> +
>>>
>>> have you tried it at runtime too ?
>>>
>> strace 4.9 with libunwind PACKAGECONFIG wouldn't build on master
>> without the patch.
>> If I add strace 4.9 recipe to dizzy branch, using the libunwind
>> PACKAGECONFIG also fails to build without the patch. It gives the same
>> errors during linking.
>>
>> I tested at runtime on master branch which has libunwind 1.1 running
>> strace 4.9 on the target and didn't notice any issues. I haven't
>> tested on dizzy branch at runtime though.
>
> OK thanks, it would be interesting to see if exception handling in some binaries and shared objects is able to unwind through
> to make sure this all works
>

Seems to work on dizzy branch too with strace 4.9 + libunwind 1.1 +
libunwind patch:
# strace -k uname 2>&1 | sed -n '/^write(1/,$p'
write(1, "Linux\n", 6Linux
)                  = 6
> /lib/libc-2.20.so(__write+0x1c) [0xbf8cc]
> /lib/libc-2.20.so(_IO_file_write+0x20) [0x69c10]
> /lib/libc-2.20.so(_IO_file_setbuf+0xd0) [0x690b8]
> /lib/libc-2.20.so(_IO_do_write+0x1c) [0x6ab20]
> /lib/libc-2.20.so(_IO_file_sync+0xc0) [0x68fb0]
> /lib/libc-2.20.so(fflush+0x8c) [0x5e1f4]
> /bin/busybox.nosuid(+0x0) [0x67178]
exit_group(0)                           = ?
+++ exited with 0 +++
> /lib/libc-2.20.so(_exit+0x10) [0x9a930]
> /lib/libc-2.20.so(__libc_secure_getenv+0xf8) [0x30dc8]
> /bin/busybox.nosuid(+0x0) [0x67194]

Regards,
Jonathan