Re: [PATCH 2/6] Open a directory with the usual flags

Corinna Vinschen <[email protected]> Wed, 10 Oct 2018 15:21:13 +0200
Newsgroups	gmane.comp.lib.newlib
Message-ID	<[email protected]>
On Oct 10 13:55, Sebastian Huber wrote:
> On 10/10/2018 13:35, Corinna Vinschen wrote:
> > On Oct 10 13:16, Sebastian Huber wrote:
> > > On 10/10/2018 11:50, Corinna Vinschen wrote:
> > > > On Oct  8 15:38, Sebastian Huber wrote:
> > > > > Use O_RDONLY since you are not supposed to write to a directory.
> > > > > 
> > > > > Use O_DIRECTORY as mandated by POSIX (The Open Group Base Specifications
> > > > > Issue 7, 2018 edition IEEE Std 1003.1-2017):
> > > > > 
> > > > > "If the type DIR is implemented using a file descriptor, the descriptor
> > > > > shall be obtained as if the O_DIRECTORY flag was passed to open()."
> > > > > 
> > > > > Use O_CLOEXEC as mandated by POSIX:
> > > > > 
> > > > > "When a file descriptor is used to implement the directory stream, it
> > > > > behaves as if the FD_CLOEXEC had been set for the file descriptor."
> > > > > 
> > > > > Drop the fcntl() call in favour of O_CLOEXEC.
> > > > Yeah, that really makes sense, but what about targets not (yet)
> > > > implementing O_CLOEXEC???  I'm not sure how to handle this, if
> > > > we have to handle that at all.
> > > In RTEMS there are no processes (fork(), exec(), etc. don't work). So, it is
> > > trivial to support the O_CLOEXEC. The situation in which this flag is
> > > relevant simply cannot happen.
> > > 
> > > Are Cygwin and Linux the only Newlib systems which supports processes?
> > I really don't know, but the other problem is to use a flag in a target
> > dependent call which might be unsupported, but tested.  This is backed
> > by POSIX allowing open to return EINVAL for invalid flags.
> > 
> > We should probably define _FNOINHERIT only on targets known to support
> > it (even if trivially) and to call open(O_CLOEXEC) or fcntl() depending
> > on that test, no?
> 
> POSIX says there is an O_CLOEXEC, so I think this is a valid oflag value. An
> unsupported close on execute would be more an ENOSYS which is not mentioned
> in the error list:
> 
> http://pubs.opengroup.org/onlinepubs/9699919799/functions/open.html
> 
> I would define the POSIX flags unconditionally and in case errors show up on
> a particular system, then we can find a solution, e.g. not define it on
> system X.

Ok, ACK.


Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
Red Hat
signature.asc (application/pgp-signature, 833 B)