[00:28:49] --- dev-zero@jabber.org has left
[00:47:49] --- Russ has left: Disconnected
[01:01:25] --- abo has left
[01:06:10] --- abo has become available
[01:12:04] --- dev-zero@jabber.org has become available
[01:40:55] --- Simon Wilkinson has become available
[02:05:22] --- haba has left
[02:44:25] --- haba has become available
[06:29:13] --- Simon Wilkinson has left
[07:12:39] --- deason has become available
[07:24:53] --- reuteras has left
[08:21:53] --- Simon Wilkinson has become available
[08:28:11] --- abo has left
[09:26:03] --- Rrrrred has become available
[10:11:20] --- Russ has become available
[10:16:08] --- haba has left
[10:33:49] --- dev-zero@jabber.org has left
[11:12:00] --- haba has become available
[12:12:16] --- dev-zero@jabber.org has become available
[12:14:21] <deason> hm, back to that discussion on SAFSS_Rename we had a few days ago; I think I/we were missing something in the non-vDirectory case...
[12:14:46] <deason> we only break callbacks on the target fid if it was a directory, and it was renamed to a different directory
[12:15:04] <jhutz@jis.mit.edu/owl> Right, because that's the only case where the target changes
[12:15:12] <deason> but if we rename a normal file to a different directory, fileptr->disk.parent changes
[12:15:27] <deason> and the ParentVnode, ParentUnique members in AFSFetchStatus would change
[12:17:25] <deason> oh, wait, are those actually utilized anywhere, client-side?
[12:17:37] <jhutz@jis.mit.edu/owl> Interesting.
The answer is "yes, sometimes".
[12:18:03] <jhutz@jis.mit.edu/owl> Perhaps more importantly, if the parent changes, the effective ACL
changes, and we need clients to discard axs cache entries.
[12:18:32] <jhutz@jis.mit.edu/owl> I bet there is an easily-reproducible bug here
[12:19:32] <jhutz@jis.mit.edu/owl> - Put a file in a directory you have read access to.
- Read it.
- From a different client, rename it into a directory to which the
  first client doesn't have read access
- Read it again.
[12:19:56] <jhutz@jis.mit.edu/owl> Expected result: second read returns EACCESS (on open)
Actual result, if we have a bug: second read succeeds.
[12:25:51] <deason> can't produce that result; I wonder why
[12:26:56] <jhutz@jis.mit.edu/owl> Oh.  The first client needs 'l' but not 'r' on the second directory.
[12:27:07] <deason> yes, I have that
[12:27:33] <deason> I can even 'fs getfid' it
[12:30:31] <jhutz@jis.mit.edu/owl> What error do you get, and when?
[12:32:57] <deason> well, userspace is just the open O_RDONLY; lemme get this on a solaris box; probably easier to dtrace and get the specific failing call
[12:37:03] <jhutz@jis.mit.edu/owl> You get EACCESS from open(2) ?
That's what should happen, but I'm not clear on where.
[12:39:17] <jhutz@jis.mit.edu/owl> Oh, duh.
[12:39:52] <jhutz@jis.mit.edu/owl> It works correctly because the client knows that on a non-directory that
is not CForeign, some access bits from from the containing directory. 
[12:40:17] <jhutz@jis.mit.edu/owl> ... and so it actually uses the cached access rights on that directory.
[12:42:35] <jhutz@jis.mit.edu/owl> And, the CM's idea of a vnode's parent is updated as a result of a lookup.
[12:43:17] <jhutz@jis.mit.edu/owl> ... which must always be done, because the directory changed
[12:43:47] <deason> that's why I was wondering how much it was used... that member of FetchStatus doesn't appear to be referenced nearly at all in the 1.4 unix client
[12:43:59] <deason> but should we break the callback on it?
[12:44:39] <deason> and set changed_newTime = 1, in that case
[12:46:24] --- mmeffie has become available
[12:46:27] <jhutz@jis.mit.edu/owl> changed_newTime does get set, as it should.
that's to cause it to get a new server-side timestamp, so it gets
included in incremental volume dumps.
[12:48:11] <jhutz@jis.mit.edu/owl> I don't see a need to add a callback break in that case.
[12:48:39] <deason> okay
[13:46:51] --- Simon Wilkinson has left
[14:22:39] --- Simon Wilkinson has become available
[14:35:13] --- dev-zero@jabber.org has left
[15:05:04] --- mmeffie has left
[15:32:31] --- Simon Wilkinson has left
[16:55:20] --- matt has become available
[16:55:29] --- matt has left
[18:41:35] --- Russ has left: Disconnected
[19:42:01] --- Jeffrey Altman has left: Replaced by new connection
[21:29:29] --- Russ has become available
[21:45:56] --- dev-zero@jabber.org has become available
[21:55:20] --- deason has left
[22:12:08] --- deason has become available
[22:12:12] --- deason has left
[22:13:18] --- reuteras has become available
[22:15:35] --- andersk@mit.edu/yaz-pistachio has become available
[22:15:47] --- andersk@mit.edu/yaz-pistachio has left
[23:43:46] --- dev-zero@jabber.org has left