[BUG] mq_[gs]etattr(3): NULL passed to __nonnull argument

[BUG] mq_[gs]etattr(3): NULL passed to __nonnull argument

[Alejandro Colomar]

[-- Attachment #1.1: Type: text/plain, Size: 1019 bytes --]

Hi!

Using __nonnull internally uncovered something:

../sysdeps/unix/sysv/linux/mq_getattr.c: In function ‘__mq_getattr’:
../sysdeps/unix/sysv/linux/mq_getattr.c:28:10: error: argument 2 null where non-null expected [-Werror=nonnull]
   28 |   return mq_setattr (mqdes, NULL, mqstat);
      |          ^~~~~~~~~~

This is not a bug in the use of mq_setattr(3), since it accepts NULL
with the meaning of mq_getattr(3), per the manual page.

I'm not sure about if this is a bug in the declaration of mq_setattr(3),
though.  If we want __nonnull because we want to disallow users from
using it in that way, then we need to use a #pragma in that specific
use.

Or maybe it was only added by accident, and we want to allow users to
call it with NULL.  In that case, we should remove __nonnull from the
declaration.  I guess this is more likely.

What do you think about it?

Thanks,
Alex

-- 
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5

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

Re: [BUG] mq_[gs]etattr(3): NULL passed to __nonnull argument

[Alejandro Colomar]

[-- Attachment #1.1: Type: text/plain, Size: 1463 bytes --]

On 2023-07-11 15:40, Alejandro Colomar wrote:
> Hi!
> 
> Using __nonnull internally uncovered something:
> 
> ../sysdeps/unix/sysv/linux/mq_getattr.c: In function ‘__mq_getattr’:
> ../sysdeps/unix/sysv/linux/mq_getattr.c:28:10: error: argument 2 null where non-null expected [-Werror=nonnull]
>    28 |   return mq_setattr (mqdes, NULL, mqstat);
>       |          ^~~~~~~~~~
> 
> This is not a bug in the use of mq_setattr(3), since it accepts NULL
> with the meaning of mq_getattr(3), per the manual page.
> 
> I'm not sure about if this is a bug in the declaration of mq_setattr(3),
> though.  If we want __nonnull because we want to disallow users from
> using it in that way, then we need to use a #pragma in that specific
> use.

Ah, no, I misunderstood the man page.  It doesn't seem to be allowed to
be NULL, but it seems glibc does use it for internal purposes only.  If
that's the case, we should maybe use an internal function that accepts
NULL, which would be used for the implementation of mq_getattr(3), and
then the public wrapper would have __nonnull.

> 
> Or maybe it was only added by accident, and we want to allow users to
> call it with NULL.  In that case, we should remove __nonnull from the
> declaration.  I guess this is more likely.
> 
> What do you think about it?
> 
> Thanks,
> Alex
> 

-- 
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5


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