[PATCH v2] assert: Refactor assert/assert_perror

[Florian Weimer]

* Adhemerval Zanella Netto:

> On 19/08/25 18:32, Florian Weimer wrote:
>> * Adhemerval Zanella Netto:
>> 
>>>> Why do we need __libc_message_nargs when things go into a varargs
>>>> function without an explicit argument count in the end anyway?
>>>
>>> This is the same logic from sysdep/unix/syscall.h, used on INLINE_SYSCALL_CALL,
>>> and it is required to correctly paste the argument is the expected order.  I
>>> think maybe there is a better way to accomplish it.
>> 
>> This is used for INLINE_SYSCALL_CALL etc. because the final target is
>> not variadic, but has a fixed number of arguments (loading only the
>> registers that are really required).  This isn't the case here as far as
>> I can see: the final function call is still variadic.
>
> I used the same strategy to try catch misuse with more arguments than
> supported at compile time, but since eeb7b079d5d8785e760ca08c3692ffa579bbb737
> we also have a runtime check so this is not strictly required. I still
> think this is a nice think to have.

You could use __builtin_va_arg_pack_len for that, at least when building
with GCC.  I think it's much clearer.

Thanks,
Florian