Re: openat wrapper assumes that O_RDONLY is disjoint from R_RDWR

[Paul Eggert]

On 3/4/20 9:24 AM, Dan Gohman wrote:
> > Would
> >     if (flags & ((O_CREAT | O_WRONLY | O_RDWR) & ~O_RDONLY))
> > be correct?
> That would fix the problem for systems that define O_RDONLY as Hurd does,
> while not breaking any known systems.
>
> It wouldn't be correct on a hypothetical system

Thanks for mentioning this. We might as well fix the code for the hypothetical 
systems while we're at it, since we can use O_ACCMODE here. The issue also 
occurs in the 'open' wrapper. I installed the attached patch.

0001-open-openat-port-to-O_RDWR-O_RDONLY-0.patch
Description: Text Data