[PATCH] memalign: move alignment adjustment to __libc_memalign

[Adhemerval Zanella Netto]

On 15/01/26 12:37, Florian Weimer wrote:
> * Siddhesh Poyarekar:
> 
>> On 2026-01-15 08:50, Adhemerval Zanella Netto wrote:
>>>> alignment > 2 is redundant. If you want to use __glibc_unlikely, this is the place
>>>> to use it since one uses memalign typically only with powers of 2.
>>>>
>>>> +    {
>>>> +      alignment = (size_t) 2 << (__builtin_clzl (1)
>>>> +                                - __builtin_clzl (alignment - 1));
>>> As a side note, I think it would be better to use stdbit.h
>>> definitions here.
>>>
>>
>> You mean stdc_count_zeros_ul?  I considered it, but decided against it
>> because AFAICT it's not used anywhere within glibc at the moment.
>> Maybe we leave this as is for now and do a mass refactoring later?
> 
> We can't use it for the backports because <stdbit.h> is quite new.
> So your proposed approach sounds fine to me.

Right, we can replace it later with stdbit then.