[OE-core] [PATCH v2] libunwind: add aarch64 support

[Fathi Boudra]

On 22 September 2016 at 11:56, Fathi Boudra <fathi.boudra at linaro.org> wrote:
> On 20 September 2016 at 17:21, Randy MacLeod
> <randy.macleod at windriver.com> wrote:
>> On 2016-09-20 07:57 AM, Fathi Boudra wrote:
>>>
>>> Hi,
>>>
>>> On 20 September 2016 at 05:12, ChenQi <Qi.Chen at windriver.com
>>> <mailto:Qi.Chen at windriver.com>> wrote:
>>>>
>>>> On 09/26/2014 05:59 PM, Fathi Boudra wrote:
>>>>>
>>>>>
>>>>> * pass --enable-debug-frame on aarch64 architecture
>>>>
>>>>
>>>>
>>>> Hi Fathi,
>>>>
>>>> I'm now dealing with a problem related to libunwind support on aarch64.
>>>> Could you please tell me why '--enable-debug-frame' is needed for
>>>> aarch64?
>>>
>>>
>>> libunwind needs to be configured with --enable-debug-frame to prevent a
>>> linkage error.
>>> Note: --enable-debug-frame is automatically selected on ARM32, not on
>>> ARM64.
>>>
>
> Admittedly, I haven't checked recently and things may have progressed upstream.
> I'm going to double check with our toolchain guys.

After asking to the TC guys, AArch64 doesn't default to debug-frame usage yet.
But as it is the only unwinding mechanism implemented in it, it would
make sense to make it the default.
They're going to give a deeper look to it, and propose a patch if needed.

>> Thanks.
>> We dropped the configuration flag and don't see any linking/linkage
>> problems yet. Can you be more specific about the problems encountered?
>>
>> Our goal is to make tcmalloc from gperftools work on ARM64.

tcmalloc >= 2.4 has support for arm64. We've upstreamed support a year ago.
If you use latest release (2.5), it should work without issues.

>> ../Randy