- Wednesday, February 25, 2015

release-team@conference.openafs.org

Wednesday, February 25, 2015< ^ >

Room Configuration

Room Occupants

GMT+0
[06:03:12] shadow@gmail.com/barnowlE5B64A04 leaves the room
[06:03:21] shadow@gmail.com/barnowlE5B64A04 joins the room
[06:06:24] shadow@gmail.com/barnowlE5B64A04 leaves the room
[06:07:01] shadow@gmail.com/barnowlE5B64A04 joins the room
[08:15:51] Jeffrey Altman leaves the room
[14:42:59] meffie joins the room
[14:46:59] wiesand joins the room
[14:51:51] Jeffrey Altman joins the room
[14:57:43] kaduk joins the room
[15:00:03] <kaduk> $timeofday
[15:00:10] <wiesand> Hi
[15:01:43] <wiesand> Let’s start...
[15:02:08] <wiesand> Has anyone tested pre2 or received feedback from a 3rd party yet?
[15:02:45] <kaduk> It's in debian experimental and I'm running it on my main debian machine.
[15:03:02] <wiesand> Ah, good. Which kernel?
[15:03:18] <kaduk> 3.16.0-4 is what they call it, I think.
[15:03:49] <kaduk> or maybe 3.16.7-ckt4-2; I haven't figured out exactly which version number corresponds to what
[15:04:09] <wiesand> ;-) At least it works.
[15:04:25] <kaduk> Yup :)
[15:04:39] <wiesand> I’m inclined to release rather sooner than later.
[15:04:51] <kaduk> As am I.
[15:05:12] <wiesand> Because it’s about time, and that’s when we’ll get actual testing ;-)
[15:05:47] <wiesand> This Friday? Next Friday?
[15:06:01] <wiesand> Thoughts?
[15:06:14] <kaduk> Any reason not to do this friday?
[15:06:25] <wiesand> Or is anyone in favour of a pre3 with the 4.0 changes?
[15:06:45] <wiesand> I’d rather release 1.6.11 soon and have a 1.6.11.1 for Linux 4.
[15:06:49] <shadow@gmail.com/barnowlE5B64A04> pre2 works on osx.
[15:07:01] <wiesand> Yosemite?
[15:07:17] <shadow@gmail.com/barnowlE5B64A04> this friday i will probably be only vaguely online, so if you need me,
it should wait til monday
[15:07:24] <shadow@gmail.com/barnowlE5B64A04> and mavericks
[15:07:28] kaduk leaves the room
[15:07:28] kaduk joins the room
[15:07:39] <wiesand> Great.
[15:08:16] <wiesand> Monday is fine too I think.
[15:08:28] <kaduk> Sounds like a plan.
[15:08:54] <wiesand> Great. Monday it is.
[15:09:37] <meffie> thanks
[15:10:11] <wiesand> I’ll try to get a “make 1.6.11” change merged a.s.a.p. Binary builders not depending on the tag could then start.
[15:11:05] <wiesand> Should manage to push it later today.
[15:11:23] <shadow@gmail.com/barnowlE5B64A04> if you do it today i can push the tag today
[15:11:32] <Jeffrey Altman> Marc says there are no other changes required for linux 4.0 at this time and since rc1 has been shipped it is unlikely but not impossible that further breakage will be introduced
[15:11:35] <meffie> (running pre2 on my main debian host now too)
[15:12:25] <wiesand> We should give folks at least a whole week between announcing pre2 and releasing 1.6.11 ;-)
[15:13:30] <wiesand> Linux 4.0: Thanks Jeff. Still, I’d rather get out 1.6.11 w/o another delay.
[15:14:10] <kaduk> It's still in the rc stage; we don't need to support it immediately, I think
[15:15:12] <wiesand> Right. The only problem is that then we either stall merges for 1.6.12 or need a branch for 1.6.11.1.
[15:15:26] <kaduk> It's git; branches are cheap ;)
[15:15:35] <shadow@gmail.com/barnowlE5B64A04> 1.6.11.1
[15:15:58] <wiesand> Unless the other changes going into 1.6.12 are so minor that we don’t care and just release 1.6.12 as the first kernel supporting 4.0.
[15:16:21] <wiesand> I’ts gerrit. Last time, branches required action by Simon...
[15:16:55] <wiesand> Daria: 1.6.11.1?
[15:18:11] <shadow@gmail.com/barnowlE5B64A04> if we need to release something for 4.0 before 1.6.12 is ready
[15:18:51] <wiesand> Yes, but do we want to not merge anything for 1.6.12 for ~6 weeks, or is the branch indeed cheap if we need it?
[15:19:14] <shadow@gmail.com/barnowlE5B64A04> branch is very cheap
[15:19:35] <wiesand> Fine then.
[15:20:10] <wiesand> Move on to “1.8”?
[15:20:41] <wiesand> Ben?
[15:20:51] <kaduk> I've been fighting with libtool and not getting much done
[15:21:41] <kaduk> The short form: we're cheating and not letting libtool know that libtool was used to build roken and hcrypto, and if we want to stop cheating and do things libtool's way, we have to drop 'make dest'.
[15:21:57] <kaduk> I don't think we are willing to drop 'make dest' for 1.8, though.
[15:22:05] <kaduk> (We can do it for 2.0?  Please?)
[15:22:42] <shadow@gmail.com/barnowlE5B64A04> chaskiel will apparently be sad. i have no love for make dest
[15:23:18] <kaduk> So, probably I am going to abandon most of the changes I have staged in that area and either just link librokenafs.a directly into compile_et, or find some other way to avoid a fixed-size buffer in compile_et (this is 9545)
[15:24:17] <kaduk> Mike, have you been looking at shake-loose-vcaches at all?
[15:24:41] <meffie> yes
[15:24:49] <kaduk> There's also a handful of -1'd changes in the 'external log rotation' topic
[15:25:11] <kaduk> Do you want help with any of those things?
[15:25:11] <meffie> yes, thank you for the reviews!
[15:26:34] <meffie> what is the situation with softsig? it seems to me we really dont need to still carry support for the old linux threads.
[15:26:53] <kaduk> I agree; we probably don't need to carry support for old LinuxThreads.
[15:27:14] <wiesand> Hopefully not.
[15:27:22] <meffie> in which case we can assign sigusr1 for the log reopen?
[15:27:30] <kaduk> That's relevant for 6947, too, I think.
[15:29:08] <kaduk> I guess my proposal then is to desupport systems using LinuxThreads, freshen up 6947 under that assumption, and allow SIGUSR1 to be repurposed for log rotation
[15:29:26] <meffie> yes, that would be nice.
[15:30:20] <wiesand> No objection from this user. I don’t even remember when we last had LinuxThreads.
[15:30:27] <kaduk> Any gatekeeper buy-in?
[15:33:10] <kaduk> Maybe we should send mail anyway, I guess.
[15:34:04] <kaduk> Moving on to 11691 (the linux lock order thing Has-Werner started us looking at), any thoughts about my proposal to change the afs_lockctl() contract to requre the caller to hold a write lock around it?
[15:34:43] <Jeffrey Altman> There is or was patch in Gerrit to remove LinuxThreads support that has/had a lot of discussion on the subject.
[15:35:07] <Jeffrey Altman> I believe it was from Andrew
[15:35:22] <kaduk> I can look for it after the meeting.
[15:37:31] <wiesand> “make 1.6.11” is 11762
[15:39:08] <Jeffrey Altman> its 6947 from Simon.   The underlying issue is whether or not Linux 2.4 is still supported.
[15:39:18] <Jeffrey Altman> in 2012 the answer was yet
[15:39:23] <Jeffrey Altman> "yes".
[15:39:40] <Jeffrey Altman> If in 2015 the answer is "no", then LinuxThreads support can go away
[15:41:54] <kaduk> I don't really see any obvious claim that linux 2.4 is still supported in the comments on 6947
[15:42:45] <meffie> i dont think someone can reasonably be running linux 2.4.x and want to install openafs 1.8.0 on it.
[15:43:12] <kaduk> Do we need to send mail to -info announcing this fact?
[15:43:43] <meffie> yes, i think os
[15:43:48] <meffie> *so
[15:44:16] <kaduk> Okay.
[15:44:25] <Jeffrey Altman> The very first comment from Andrew is the one that highlights that Simon's change breaks 2.4 support
[15:44:34] <kaduk> I will put that on my list of things to do.
[15:44:56] <wiesand> EL3 already had NPTL
[15:45:41] <kaduk> My position is roughly that if you insist on running some ancient software, you can deal with not running the most recent version of other softwares on the same machine.
[15:46:52] <meffie> yes
[15:47:19] <kaduk> Okay, so I will send mail; I think we can move on to the afs_lockctl() contract?
[15:49:39] <shadow@gmail.com/barnowlE5B64A04> i already told you i was down with pushing linuxthreads out of the
fore
[15:51:25] <Jeffrey Altman> 1.6.5 support for rxkad_k5 is available for linux 2.4 so I do not much care.   I do not think that linux 2.4 support is required to be supported by rxgk
[15:52:01] <Jeffrey Altman> We will poll our customers to see if there is anyone that cares.   SNA should do the same.
[15:52:16] <meffie> yes, makes sense.
[15:56:18] <kaduk> If no one cares about afs_lockctl(), how about 10789 (don't retry timed-out RW operations forever)?
[15:57:23] <kaduk> It seems that we should let idledead do its thing, to me.
[15:57:44] <meffie> (wondering, should these conversations be in the other jabber room?)
[15:57:52] <wiesand> 01f42cad663cc860e451f694f973ca46f03289ff was a fix for a very serious issue.
[16:02:12] <Jeffrey Altman> I don't have an opinion on afs_lockctl() as I am not familiar enough with the unix cm locking hierarchy.
[16:02:31] <Jeffrey Altman> Marc would probably be the best to comment on it.
[16:03:01] <kaduk> *nods*
[16:05:46] <kaduk> I updated 7896 (ptuser prname NUL terminators) to add the checking in viced/host.c and libadmin/pts, but did not do anything about logging.  Does anyone object to that?
[16:07:01] <Jeffrey Altman> I have no objection.
[16:07:09] <shadow@gmail.com/barnowlE5B64A04> do it
[16:07:38] <kaduk> Oh, and jhutz says in zephyr "by all means, merge them both [11689+11738, the ubik restart stuff]"
[16:08:22] <Jeffrey Altman> it would be nice if he would +1 the changes in gerrit
[16:08:51] <kaduk> Somehow I thought he did.  Sigh.
[16:09:15] <Jeffrey Altman> he didn't +1 11689
[16:09:39] <meffie> i think he did on patchset 1, if i'm reading this correctly
[16:10:44] <meffie> so he should look at ps2
[16:10:50] <kaduk> Right.
[16:11:36] <kaduk> Oh, but he sent mail.
[16:13:32] <kaduk> More review of 9985 would be nice, since the log-rotation stuff should probably get rebased on top of it.
[16:14:50] <kaduk> 11734 is a small patch that is a release blocker; Perry and Chas have +1'd it.
[16:16:22] <Jeffrey Altman> there is a typo in the #if construction.
[16:16:34] <meffie> yay twisty mazes of ifdefs, all look alike.
[16:16:51] <kaduk> Indeed.
[16:17:48] <kaduk> I have a few things to rebase that could otherwise be merged, I think.
[16:18:03] <kaduk> There's probably not too much left to discuss.
[16:18:58] <Jeffrey Altman> lets get 1.6.11 out on Monday.   1.6.12 can be Linux 4.0 plus ubik fixes
[16:19:04] <kaduk> I forget, were YFSI folks going to remove vos -stayonline and also investigate "lockless path through d_revalidate" or just the former?
[16:19:09] <Jeffrey Altman> yes
[16:19:18] <Jeffrey Altman> the former yes
[16:19:48] <Jeffrey Altman> ask Marc to review lockless path in Gerrit
[16:19:54] <kaduk> Okay.
[16:21:03] <kaduk> I think we're done for the meeting, then.
[16:21:45] <meffie> thank you
[16:21:54] <wiesand> Fine.
[16:22:08] <wiesand> A +1 or two on 11762 would be nice.
[16:22:46] <wiesand> Once that’s merged I’d build preliminary tarballs.
[16:23:02] <wiesand> But now I have to run. Thanks a lot everyone!
[16:23:06] wiesand leaves the room
[16:24:35] meffie leaves the room
[19:57:57] kaduk leaves the room
[20:54:24] kaduk joins the room