[05:11:25] --- wiesand has become available
[05:12:06] <wiesand> test
[06:39:05] --- meffie has become available
[06:45:40] --- meffie has left
[06:45:41] --- meffie has become available
[07:00:04] --- deason has become available
[07:01:29] <wiesand> Hello
[07:01:46] <kaduk@jabber.openafs.org/barnowl> 'mornin.
[07:02:35] --- Marc Dionne has become available
[07:02:47] <meffie> hello
[07:02:57] <Marc Dionne> hi Mike
[07:03:11] <wiesand> Hi Marc. BY any chance, could you try 3.15-rc1 yet?
[07:03:39] <Marc Dionne> yes, still looks OK.  only tried on master though, not 1.6
[07:04:02] <deason> hi
[07:04:36] <wiesand> There wasn't anything merged on master lately that would affect afsd on Linux, right?
[07:05:09] <Marc Dionne> are you seeing something unusual?
[07:05:12] <shadow@gmail.com/barnowl7628413F> define "affect"
[07:05:39] <wiesand> "make it work with 3.15-rc1 for master but not the current head of 1.6.x"
[07:07:09] <wiesand> Or, to ask the actual question: Should we delay 1.6.8pre2 until we have smoke test results with that kernel?
[07:07:33] <kaduk@jabber.openafs.org/barnowl> I doubt it.
[07:07:40] <Marc Dionne> ah, well if you give me a little bit of time I will do the test with 1.6  (like, before this meeting ends)
[07:08:01] <wiesand> That would be great, thanks.
[07:08:32] <wiesand> The same question applies to the stack overflow issue.
[07:08:33] <Marc Dionne> already have a machine running, just a matter of compiling 16
[07:09:23] <kaduk@jabber.openafs.org/barnowl> I didn't yet get a chance to look at Mike's stack-size-reduction
changes.
[07:10:25] <wiesand> Is the set of changes for reducing stack usage so close to ready that we should wait with pre2 for them?
[07:10:48] <wiesand> I must confess the last week was very busy, and I'm not quite up to date.
[07:11:51] <wiesand> I see Mike just pushed new versions.
[07:12:44] <meffie> yeah, this set avoids adding a lot of new asserts.
[07:14:11] <wiesand> Delay 1.6.8pre2 for it?
[07:15:11] <wiesand> What's the story on RT #131852?
[07:15:20] <shadow@gmail.com/barnowl7628413F> given it hasn't hit master yet that might be a bit premature; it's
minimally but not noninvaive
[07:15:21] <meffie> i dont know. it's meant to be straightforward, but is a lot of changes to review.
[07:15:27] <kaduk@jabber.openafs.org/barnowl> I'm sort of thinking that we may not get to address stack size for
1.6.8.
[07:15:53] <wiesand> If Ben is right, no point indeed.
[07:15:56] <shadow@gmail.com/barnowl7628413F> 131852 is a heimdal bug
[07:16:15] <kaduk@jabber.openafs.org/barnowl> It sounds like we could work around it, though.
[07:16:22] <wiesand> Would anyone object to releasing 1.6.8 without addressing stack usage?
[07:16:39] <shadow@gmail.com/barnowl7628413F> namely if the keytab is missing, it blows up. yes, we could probably
work around it by ensuring at least an empty file exists
[07:17:09] <deason> we should work around it, since that crash seems like quite a regression
[07:17:25] <kaduk@jabber.openafs.org/barnowl> That patch would be 1.6-only.
[07:17:27] <deason> and is guaranteed to occur if you actually follow the k5/kdf procedures as written
[07:17:53] <shadow@gmail.com/barnowl7628413F> to be clear, stephan, we are talking about the RT thing, not the stack
issue
[07:17:59] <kaduk@jabber.openafs.org/barnowl> "Who wants to write it?"
[07:18:03] <wiesand> Got it, but thanks.
[07:18:03] <deason> (assuming the reason for this occurring is correct; I haven't looked)
[07:18:10] <deason> why would it be 1.6-only?
[07:18:20] <kaduk@jabber.openafs.org/barnowl> We don't read from keytabs on master.
[07:18:28] <shadow@gmail.com/barnowl7628413F> testing showed that touching a keytab made the issue go away. on
master we use KeyFileExt
[07:18:30] <deason> right right, okay
[07:19:20] <wiesand> Volunteers?
[07:19:48] <deason> I would probably be able to this weekend
[07:21:09] <Marc Dionne> btw current 1.6 branch compiles and works fine on 3.15-rc1
[07:21:18] <wiesand> Andrew: Great. If you managed that, I think that should be in pre2.
[07:21:26] <wiesand> Marc: Thanks!
[07:23:08] <wiesand> Next one: RT #131847
[07:23:55] <kaduk@jabber.openafs.org/barnowl> Someone needs to get and look at a kernel crash dump.
[07:24:19] <deason> I thought that's been true forever; or at least the behavior has been a little strange
[07:24:21] <kaduk@jabber.openafs.org/barnowl> I didn't reply to the requestor noting that because I'm not equipped
to handle debugging linux kernel crash dumps.
[07:25:10] <wiesand> Sounds like "would be good to address eventually, but not in 1.6.8"?
[07:25:12] <kaduk@jabber.openafs.org/barnowl> It mostly sounds like it should be easy for a developer to reproduce,
though.  But this sort of thing tends to take a decent chunk of time
to diagnose
[07:25:18] <kaduk@jabber.openafs.org/barnowl> Yes, that.
[07:25:43] <wiesand> Ok, no blocker for 1.6.8.
[07:25:51] <deason> I assumed it's something that could be fixed by maybe improving the general startup/shutdown procedures, to make them not so awkward
[07:26:48] <wiesand> Anything else we'd urgently want in 1.6.8pre2?
[07:27:58] <wiesand> So the plan is to issue pre2 once we have the missing keytab workaround. Or as soon as it's clear that this will take much longer.
[07:28:18] <kaduk@jabber.openafs.org/barnowl> Sounds like a plan.
[07:29:19] <wiesand> Next topic: testing
[07:29:50] <wiesand> I received a mail from Stephen. They'd be willing to run tests if it's really easy.
[07:30:06] <shadow@gmail.com/barnowl7628413F> i don't think it will take long to work around; you need to establish
if the keytab is a file type, and if it is, stat the file.
[07:31:00] <deason> I currently have 3 sites that have expressed interest; I'll be sending a solicitation to the lists today
[07:31:18] <wiesand> We still need the code. Andrew volunteered to sacrifice part of this weekend (huge thanks), but if things go differently, I won't blame hime.
[07:31:42] <wiesand> Ok. Any other news on the topic?
[07:32:16] <deason> the dafileserver heimdal crashing thing is a good example of why such things are useful, even if the site can't provide much load :)
[07:33:18] <wiesand> I fully agree that a standard test suite would be really good to have.
[07:33:40] <deason> well, that too
[07:34:33] <wiesand> Last topic: next stable branches
[07:35:48] <wiesand> There was some feedback on the list, in preference of not using odd numbers for releases meant to be stable.
[07:36:19] <wiesand> Anything else I missed? New thoughts?
[07:36:22] <deason> specifically, x.y, where y is not odd
[07:37:23] <deason> I still think it'd be useful to mention it to -info; if just to get peoples' input
[07:37:55] <wiesand> Weren't you about to do that anyway?
[07:38:27] <shadow@gmail.com/barnowl7628413F> there. have 11075. needs testing yet. pushed for review of the
concept.
[07:39:57] <wiesand> Thanks :)
[07:40:14] <kaduk@jabber.openafs.org/barnowl> I am trying to remember if there are any places where we might have
krb5_kt_default_name return an WRFILE: type.
[07:41:48] <deason> if 'we' means 'mit', it wouldn't seem to matter for the immediate purpose :)
[07:42:06] <shadow@gmail.com/barnowl7628413F> well, it means the code needs ifdefs for heimdal only
[07:42:08] <kaduk@jabber.openafs.org/barnowl> Yes, MIT.
[07:42:58] <deason> I don't see why; if it's wrfile, the test would just be skipped and since mit doesn't crash for this case 'who cares'
[07:43:50] <shadow@gmail.com/barnowl7628413F> yeah. but i guess since a user can set WRFILE: for heimdal and it
would work, i need to handle it
[07:44:07] <shadow@gmail.com/barnowl7628413F> at least for the user-set case.
[07:44:28] <deason> further discussion deferred to gerrit... ?
[07:44:47] <shadow@gmail.com/barnowl7628413F> sure. but patchset 2 is landing now with WRFILE support
[07:45:06] <shadow@gmail.com/barnowl7628413F> no need to hold us up here. but yes, i think this is doable for pre2
;)
[07:45:45] <wiesand> Please review & discuss in gerrit.
[07:46:28] <wiesand> BTW I got the commit message in 10987 :)
[07:47:47] <wiesand> And I take it that there is no more to say about the "new stable branch(es)" topic?
[07:48:06] <wiesand> I'm getting the feeling that this will drag on forever :-(
[07:48:21] <kaduk@jabber.openafs.org/barnowl> :(
[07:49:59] <wiesand> It's not quite clear to me how the "OpenAFS for Windows" and "Portable OpenAFS" proposal would look in detail.
[07:50:09] <deason> > (09:37:25) deason: I still think it'd be useful to mention it to -info; if just to get peoples' input
> (09:37:57) wiesand: Weren't you about to do that anyway?
did you mean on stable version numbers? I didn't think I was doing anything about that
[07:50:23] <shadow@gmail.com/barnowl7628413F> someone(s) will basically make a decision at some point and make some
subset of people unhappy. and that's how life goes
[07:50:31] <kaduk@jabber.openafs.org/barnowl> Andrew: Stephan was talking about the testing stuff.
[07:51:05] <wiesand> Right, that was the testing stuff.
[07:51:16] <deason> okay, my immediately preceding comment was about the version numbers
[07:51:40] <wiesand> Daria: The problem is, it's not happening.
[07:53:16] <deason> but I could be the one to send something to -info about it, I guess... at least mention it, if not describe in detail all of the possible approaches
[07:54:20] <wiesand> It may help drive this.
[07:55:06] <wiesand> Anything else to discuss today?
[07:55:41] <shadow@gmail.com/barnowl7628413F> i got nothing
[07:56:28] <wiesand> So, thanks a lot everyone. I'll write the minutes later.
[07:56:36] <meffie> thank you
[07:56:43] <wiesand> cu
[07:56:45] --- wiesand has left
[07:56:49] --- deason has left
[07:56:56] --- meffie has left
[08:05:25] --- Marc Dionne has left
[14:10:13] --- deason has become available
[14:10:19] --- deason has left