[Bug-tar] Re: [patch] O_NOATIME support

[Paul Eggert]

Ian Turner <address@hidden> writes:

> Why don't we put a patch now for --preserve-atime=system, and we can
> add --preserve-atime=heuristic when kernel support exists, since
> that might take some time?

(no pun intended :-).  Yes, I guess that makes sense for now.

>> Or -- better yet -- why not have open(...O_NOATIME) fail with errno ==
>> EINVAL when invoked on a file system that does not support O_NOATIME?
>> That would be similar to how open(...O_SYNC) is supposed to behave, so
>> it makes a lot of sense to do it that way.

> I'd agree that that makes sense, except that O_NOATIME is a GNU
> extension, and the GNU interface does not specify the possibility of
> EINVAL. And we have kernels out there already that silently ignore
> O_NOATIME.

I don't know what you mean about "possibility of EINVAL", but clearly
this is a kernel bug and should get fixed.  Can you please report it
to people who can fix it?  (If they'd rather return some other errno
value I won't argue.)