Re: bind(2): Missing [[gnu::nonnull]]

[Alejandro Colomar via Libc-alpha <libc-alpha@sourceware.org>]

[-- Attachment #1.1: Type: text/plain, Size: 1600 bytes --]

Hi Xi,

On 12/3/22 16:55, Xi Ruoyao wrote:
> On Sat, 2022-12-03 at 16:33 +0100, Alejandro Colomar via Libc-alpha
> wrote:
>> Hi!
>>
>> I'm documenting NULLness of parameters in the Linux man-pages.  While doing
>> that, I noticed bind(2) is not prototyped with nonnull, but I don't think it
>> makes sense to accept NULL.  Is it a mistake?  Should I send a patch for adding it?
> 
> Hi Alejandro,
> 
> Currently the man page says:
> 
> EFAULT: addr points outside the user's accessible address space.
> 
> And bind(2) indeed sets errno to EFAULT and return -1 when NULL is
> passed as addr.
> 
> gnu::nonnull is not only a diagnostic attribute: it also allows the
> compiler to assume addr is never NULL.  i. e. if addr was gnu::nonnull
> and bind(2) is called with addr == NULL, the behavior would be
> undefined.
> 
> So this will be an API change.  Yes I agree calling bind with NULL does
> not make any sense, but I guess we still need to keep API
> "compatibility" with those nonsense code...
> 

Hmmm, makes sense.  What I'll do (unless someone opposes and suggests that I do 
otherwise), though, is ignore that problem in the man-pages.  I'm documenting 
_Nullable instead of _Nonnull, since there are much less places where it needs 
to be added, so what I'll do is be silent in the case of bind(2).  I'm not 
specifying _Nonnull, so it's not a lie, but I'm not suggesting programmers that 
they should be careless about NULL, which I wouldn't want to do.

Thanks for the prompt answer!

Cheers,

Alex


-- 
<http://www.alejandro-colomar.es/>

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]