[PATCH 1/8] Use gcc attribute ifunc in libc_ifunc macro instead of inline assembly due to false debuginfo.

[Stefan Liebler]

On 08/05/2016 05:11 PM, Paul E. Murphy wrote:
>
>
> On 08/05/2016 03:34 AM, Florian Weimer wrote:
>> On 08/05/2016 10:31 AM, Andreas Schwab wrote:
>>> On Fr, Aug 05 2016, Stefan Liebler <stli@linux.vnet.ibm.com> wrote:
>>>
>>>> Requiring a specific gcc / binutils version does not work. E.g. on s390x a
>>>> gcc 4.7.4 has attribute ifunc support.
>>>> According to <gcc-src>/gcc/configure.ac there is an option
>>>> --enable-gnu-indirect-function. If it is not specified it defaults to
>>>> default_gnu_indirect_function variable. This is set to yes in
>>>> <gcc-src>/gcc/config.gcc for intel / s390. Otherwise it is no.
>>>
>>> That should probably be updated, then.
>>
>> I agree.  Can we require that GCC is configured with --enable-gnu-indirect-function for future glibc releases, please?
>>
>> Thanks,
>> Florian
>>
>
> Predicated on the desire to enable multiarch support.  Otherwise,
> I agree.
>
>
> Stefan, would you mind splitting out the configury changes into a
> separate patch to enforce this stricter toolchain requirement when
> testing for multiarch support?
>
> Likewise, I would prefer disabling by default, and throwing
> an error at configure time if --enable-multi-arch is requested
> on such a toolchain.  Then, libc-symbols.h would not need to
> maintain an old version of those macros which require less
> rigid symbol mucking.
>
Okay. I've added an extra patch for the configure check
and posted the whole series again. See follow-up thread:

"[PATCH v2 1/9] Add configure check to test if gcc supports attribute 
ifunc."
https://www.sourceware.org/ml/libc-alpha/2016-08/msg00258.html