On 07/11/2018 06:24 PM, Adhemerval Zanella wrote:
> 
> 
> On 11/07/2018 12:49, Szabolcs Nagy wrote:
>> On 10/07/18 20:26, Adhemerval Zanella wrote:
>>>
>>>
>>> On 06/07/2018 09:46, Florian Weimer wrote:
>>>> On 02/02/2018 06:04 PM, Adhemerval Zanella wrote:
>>>>> @@ -139,7 +139,9 @@ libpthread-routines = nptl-init vars events version pt-interp \
>>>>>                   pthread_mutex_getprioceiling \
>>>>>                   pthread_mutex_setprioceiling \
>>>>>                   pthread_setname pthread_getname \
>>>>> -              pthread_setattr_default_np pthread_getattr_default_np
>>>>> +              pthread_setattr_default_np pthread_getattr_default_np \
>>>>> +              thrd_create thrd_current thrd_detach thrd_equal \
>>>>
>>>> thrd_current should live in libc, not in libpthread.  Same for thrd_equal, I suppose.  You should just duplicate the comparison, seeing that we haven't done that particular cleanup for pthread_equal yet.
>>>>
>>>> Not sure about thrd_sleep and thrd_yield, but I lean towards libc there as well.
>>>
>>> Right, I added all four symbols (thrd_current, thrd_equal, thrd_sleep, and
>>> thrd_yield) to libc on my personal branch [1] (the change is just to move
>>> the objects from libpthread-routines to routines rule in nptl/Makefile).
>>> I also adjusted the abilist patch with the change.
>>>
>>
>> the abilist patch seems to add
>> sysdeps/unix/sysv/linux/tile/tilegx64/libpthread.abilist
>> i think you can drop that (and update the ChangeLog).
> 
> Thanks for catching this up, I fixed on my personal c11 thread branch.

I still see the attached ABI diffs.

There is also a new linknamespace failure on hppa:

[initial] thrd_create -> [libpthread.a(thrd_create.o)] 
__pthread_create_2_1 -> [libpthread.a(pthread_create.o)] mprotect

That should be easy to fix.  I'll post a separate patch for that.

Thanks,
Florian