This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: aio_fsync() a directory ?
- From: Roland McGrath <roland at hack dot frob dot com>
- To: Xavier Roche <roche+kml2 at exalead dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Thu, 11 Apr 2013 15:50:58 -0700 (PDT)
- Subject: Re: aio_fsync() a directory ?
- References: <513EEF07 dot 5090901 at 3ds dot com> <20130408212729 dot ED7592C085 at topped-with-meat dot com> <5163BA06 dot 30600 at exalead dot com>
> - 0000671: aio_fsync() EBADF and "open for writing"
> <http://austingroupbugs.net/view.php?id=671>
The resolution indicates the standard will be changed not to require the
O_RDONLY check. Please file a bug in bugzilla and refer to both this link
and a mailing-list archive link for the discussion we've had here. We'd
then accept a patch to remove the check (listing that BZ# in its log entry).
> - 0000672: Necessary step(s) to synchronize filename operations on disk
> <http://austingroupbugs.net/view.php?id=672>
Thankfully this is outside the scope of anything libc needs to worry about.
If it's resolved in a way that requires libc changes, we can revisit it.
(The possible resolutions that seem likely are either to bless the de facto
method, which is only a requirement on kernel behavior that libc can't do
anything about; or to add a new interface of some sort, which we would
probably not implement until a new version of the standard specifying such
new interfaces is reasonably close to being finalized.)
Thanks,
Roland