Do we think the __need_* interface is a good idea?

[Carlos O'Donell]

On 03/03/2017 08:19 PM, Joseph Myers wrote:
> On Fri, 3 Mar 2017, Carlos O'Donell wrote:
> 
>> Joseph,
>>
>> In the upstream kernel their signal.h has exactly the same
>> issues that ours does, for example they need a definition of size_t:
>> https://lkml.org/lkml/2017/2/25/191
>>
>> I mention here how we solve this:
>> https://lkml.org/lkml/2017/3/2/364
>>
>> Is our use of special "hooks" like __need_size_t something we want
>> to keep using or has someone thought of a better solution?
> 
> Zack's bits/types/*.h is a better solution.  (For types from <stddef.h> I 
> think it would be reasonable still to use bits/types/*.h, with the headers 
> there serving to wrap the GCC header with __need_* defined.)

Thanks. I've followed up with upstream linux kernel making a suggestion
to match what we do in glibc as best practice.

-- 
Cheers,
Carlos.