Home
release-team@conference.openafs.org
Friday, January 12, 2018< ^ >
Room Configuration
Room Occupants

GMT+0
[13:35:39] meffie joins the room
[13:48:35] wiesand joins the room
[13:59:41] <meffie> good afternoon wiesand
[14:01:42] <wiesand> Hello Mike
[14:02:01] <kaduk@jabber.openafs.org/barnowl> Greetings
[14:02:06] <wiesand> Hi Ben
[14:02:43] <wiesand> There's not that much to discuss from my p.o.v. today
[14:03:32] <kaduk@jabber.openafs.org/barnowl> Do you think you got enough review on everything linked from last
week?
[14:04:09] <wiesand> It could be more, but I'd merge the three stacks w/o further review
[14:04:22] <wiesand> If they pass smoke test on my systems ;-)
[14:04:51] <kaduk@jabber.openafs.org/barnowl> :)
[14:04:52] <meffie> Mark was going to review the 1.6.x ubik fixes
[14:05:25] <wiesand> That would help getting them in…
[14:05:34] <wiesand> As I said, they frighten me
[14:05:35] <meffie> those fixes have been running in production for a while now, as patches against 1.6.
[14:05:50] <kaduk@jabber.openafs.org/barnowl> Nothing wrong with a little healthy fear...
[14:06:07] <wiesand> I'll also have some questions
[14:06:23] <meffie> agreed, just saying they are not just theoretical
[14:07:08] <meffie> ubik is wickly complicated, so yeah it it healthy to have fear of changes there
[14:07:13] <wiesand> from my notes (these may be out of date):
[14:07:23] <wiesand> what about 12283 ?
[14:08:03] <wiesand> 12803: there's a standing objection to this one
[14:08:28] <wiesand> my notes also suggest 12803 would have needed 4490, but that seems not to be in the stack
[14:08:54] <meffie> i think marcios fixes are an alternate to 12803
[14:09:26] <kaduk@jabber.openafs.org/barnowl> There were a couple variant options floating around, yes, though I
don't remember the details off the top of my head.
[14:09:47] <wiesand> 12803 was pulled up to 1.6 by Marcio
[14:10:18] <meffie> oh, i am misremembering then.
[14:11:12] <wiesand> 12283 was never merged on master, I don't know why
[14:11:33] <meffie> i recall there are several ubik patches that are lower priority but have not been merged on master yet.
[14:11:41] <wiesand> This state contributes significantly to making me nervous about the whole topic
[14:11:50] <meffie> i think marcio said there are 3 of them.
[14:12:13] <kaduk@jabber.openafs.org/barnowl> 12283 is just some "internal bookkeeping" to make what's going on in
the code a little more clear to the reader.  It shouldn't have any
impact on the wire.
[14:12:39] <meffie> he said they would be nice to have, but we have not been able to get them merged yet.
[14:13:51] <wiesand> Anyway, review of the whole topic should really start now if they're meant to make 1.6.23
[14:14:43] <wiesand> Ben, is anything coming up that should prevent me from starting to merge the changes foreseen for 1.6.22.1 soon?
[14:15:00] <kaduk@jabber.openafs.org/barnowl> I don't think so.
[14:15:46] <wiesand> Great. I think I'll wait for the buildbot results after 4.15-rc8 and then start
[14:16:08] <wiesand> Unless an rc9 is expected
[14:17:02] <wiesand> I also drafted a NEWS change today: 12869
[14:17:28] <wiesand> If you have a better description for the EL7.4 getcwd thing, I'm all ears
[14:19:03] <kaduk@jabber.openafs.org/barnowl> For 1.8.0pre4, I had that as:
+    * Use a more correct (less aggressive) scheme to react to downward
+      pressure on cache usage, avoiding d_invalidate(), which can cause
+      getcwd() failures on RHEL 7.4.
[14:19:05] Marcio Barbosa joins the room
[14:19:16] <Marcio Barbosa> good morning all
[14:19:22] <kaduk@jabber.openafs.org/barnowl> good morning!
[14:19:49] <Marcio Barbosa> reading
[14:19:59] <wiesand> Ben: ah, I should have remembered. I'll just  steal that part then ;-)
[14:19:59] <meffie> i figured im a terrible messenger, marcio can explain directly :)
[14:24:38] meffie leaves the room
[14:24:44] meffie joins the room
[14:25:24] <meffie> (ping. did i drop?)
[14:25:52] meffie leaves the room
[14:25:59] <wiesand> now you did
[14:26:36] <wiesand> or did I?
[14:26:40] meffie joins the room
[14:27:07] <Marcio Barbosa> > 12283 was never merged on master, I don't know why
it seems we have to update the commit message. in my point of view, this patch is important and should be taken into consideration.
[14:27:13] <kaduk@jabber.openafs.org/barnowl> In the blank space I'll note that we haven't gotten any debian bug
reports about 1.8.0pre4 yet, but did have like four people open bugs
noting that the spectre/meltdown kernel update broke our module build
on stable and oldstable.
[14:27:53] <wiesand> How did that happen?
[14:28:04] <wiesand> I saw no issues on Ubuntu and EL
[14:30:29] <kaduk@jabber.openafs.org/barnowl> > How did [12283 not get merged]?
It probably dropped off my radar
[14:31:16] <Marcio Barbosa> concerning ubik patches: 12716 and 12640 were not merged on master yet.
[14:32:10] <meffie> sounds like we should abandon any ubik patches we dont care about and update the commit messages on the others, and make sure they have a topic.
[14:32:12] <wiesand> my "how did?" question was related to the module build break
[14:33:00] <Marcio Barbosa> the problem fixed by 12716 was masked by the other bugs. but now that the other fixes were merged i think we should take this patch into consideration.
[14:34:13] <kaduk@jabber.openafs.org/barnowl> > module build break
The inode_change_ok/setattr_prepare thing.  Debian stable and
oldstable are stuck on laughably old openafs versions
[14:34:34] <wiesand> Ah, thanks.
[14:34:51] <wiesand> And also thanks on the comment on 12716.
[14:35:32] <wiesand> It does not inspire confidence in merging anything ubik into the stable series anytime soon though…
[14:35:45] <wiesand> sorry mike..
[14:36:36] <kaduk@jabber.openafs.org/barnowl> I mean, I think part of this is that I share the heaklthy fear of
touching ubik, and require an absurd level of review even before
merging to master
[14:37:37] <meffie> understandable, but some of these are already merged on master, and we need them in production, or we cant use openafs :)
[14:38:26] <Marcio Barbosa> do we need more reviews? what do we need?
[14:39:00] <meffie> esp 12592
[14:39:47] <kaduk@jabber.openafs.org/barnowl> 12592 is merged?
[14:40:10] <meffie> yes, on master
[14:40:36] <meffie> i believe that is 12803 on 1.6, right marcio?
[14:41:34] <Marcio Barbosa> not merged on 1.6 yet
[14:41:51] <meffie> yes, but the gerrit for 1.6 is 12803?
[14:42:08] <Marcio Barbosa> right
[14:42:09] <wiesand> It's been up there for a month, and isn't carrying a single +1, from anyone.
[14:42:14] <wiesand> Of course it's not merged
[14:42:33] <wiesand> Answering your question what you need:
[14:43:00] <Marcio Barbosa> in my point of view 12803 is harmless
[14:43:32] <wiesand> You need to make me confident that the whole set of Ubik changes queued for 1.6 will not introduce any new problems. And right now, you're far from that.
[14:43:37] <meffie> but, marcio, we need to put +1's in gerrit.
[14:43:58] <meffie> and we need to get other developers to agree
[14:44:35] <meffie> by putting +1s in gerrit
[14:45:41] <meffie> i feel we should focus on the one already merged on master/1.8.
[14:45:52] <meffie> s/one/ones/
[14:46:58] <kaduk@jabber.openafs.org/barnowl> Probably.
[14:47:00] <Marcio Barbosa> > You need to make me confident that the whole set of Ubik changes queued for 1.6 will not introduce any new problems. And right now, you're far from that.
Understood. How can I do that? I can write a summary about each patch if necessary.
[14:48:07] <wiesand> You can beg for review on openafs-devel
[14:48:28] <kaduk@jabber.openafs.org/barnowl> Er, is it correct that confidence is instilled in the form of +1
reviews in gerrit?
[14:49:19] <wiesand> Additional information on why the changes are important but harmless would probably help get others review the changes.
[14:49:40] <wiesand> Ben: yes.
[14:51:02] <wiesand> Especially in ubik space, I don't understand a whole lot of what's going on. I need the wizards' verdicts on those changes.
[14:51:47] <meffie> yes, that is reasonable
[14:52:44] <wiesand> +1s from Mark, Mike, Marcio, yourself, and preferably someone like jhutz too would help a lot to make me confident that merging these is a good idea
[14:52:56] <Marcio Barbosa> iirc, some of those patches were review by Jeffrey Hutzelman
[14:53:08] <Marcio Barbosa> 12609
[14:54:48] <meffie> yes, i would expect patches were vetted on master already, so those same developers are not starting from scratch.
[14:55:23] <meffie> they would just need to verify the backport does not introduce bugs or regressions
[14:55:34] <meffie> right wiesand?
[14:56:06] <wiesand> right.
[14:56:46] <wiesand> and they should express their opinion that a change is suitable for the current stable series, not just the next one
[14:56:48] <meffie> excellent. so marcio, we should contact the same reviewers and plead for backport reviews.
[14:57:32] <meffie> at a minimum that is.
[14:58:09] <Marcio Barbosa> understood. reviewing the backports should be simpler (since we do not have to deal with the locks)
[14:59:15] <meffie> also, if you review other peoples patches you get review karma, and they are more willing to look at your patches ;)
[14:59:52] <meffie> did i get that right wiesand?
[15:00:33] <wiesand> I think that mechanism exists to some extent, yes.
[15:00:49] <meffie> (sorry to derail the flow of your meeting...)
[15:01:45] <wiesand> NP. Looks like there was an actual need for this discussion.
[15:02:04] <meffie> yes, i feel it was helpful.
[15:02:25] <wiesand> I'm probably not communicating these things well enough.
[15:02:53] <wiesand> Ben, want to move on to 1.8/master?
[15:03:42] <kaduk@jabber.openafs.org/barnowl> Okay
[15:04:27] <kaduk@jabber.openafs.org/barnowl> Andrew submitted a change for this deadlock in master/1.8 on old
kernels that have the splice alias race, which sounds like something
we want for 1.8.0 final.
[15:04:51] <kaduk@jabber.openafs.org/barnowl> But I only skimmed the patch so far and will think a little bit about
whether there is an alternative that might be better.
[15:05:11] <kaduk@jabber.openafs.org/barnowl> (Is there value in having a bug report in RT when the issue is known
but before the patch is ready?)
[15:05:39] <kaduk@jabber.openafs.org/barnowl> But other than that, I don't think we've gotten any negative feedback
on 1.8.0pre4.  A couple folks on IRC seemed to say it was working okay
for them.
[15:07:29] <wiesand> It seems unlikely that folks still running EL6 will jump on 1.8
[15:07:30] <kaduk@jabber.openafs.org/barnowl> So we're probably looking at a couple (?) weeks more of settling time,
letting us get in a fix to the aforementioned locking issue, then a
release candidate.
[15:07:49] <wiesand> So I wouldn't care too much about performance on pre-2.6.34 kernels
[15:07:59] <wiesand> Unless it's really bad
[15:08:12] <wiesand> Sounds good
[15:09:05] <kaduk@jabber.openafs.org/barnowl> Not too much news on master; not enough time to review the backlog.
[15:09:55] <kaduk@jabber.openafs.org/barnowl> Andrew did raise a question of a meomry leak in the rxgk crypto code
if krb5_init_context could sporadically fail, and I think I've come
around to the position of embedding a krb5_context in the rxgk_key
structure as the least-bad option for now.
[15:11:00] <kaduk@jabber.openafs.org/barnowl> But I guess there's not much to discuss for this week.  Unless people
want to bump the priority of patches sitting in gerrit for a long
time, like those ubik ones we were talking about earlier.
[15:12:35] <wiesand> Well, if there really are some that fix issues unmasked by the accepted fixes, those should probably get some attention...
[15:12:59] <meffie> ok, good point
[15:13:16] <kaduk@jabber.openafs.org/barnowl> But mostly I am stuck using gerrit's default sort, which is "most
recent activity" ... so a comment on old-but-important changes will
"ping" them up to the top of the list.
[15:13:54] <wiesand> right. not too hard or tedious IMHO…
[15:15:15] <wiesand> Looks like we're finished for today?
[15:15:41] <kaduk@jabber.openafs.org/barnowl> Sounds like it
[15:15:43] <meffie> for what it's worth, i will have updates for marks 'solaris-deadlock' topic soon, but not in time for 1.6.23pre1 it seems
[15:16:30] <meffie> thanks, that's all i got.
[15:16:45] <wiesand> As discussed last week, those have a much lower bar for (late) inclusion than the ubik ones ;-)
[15:16:45] <kaduk@jabber.openafs.org/barnowl> thanks everyone!
[15:17:06] <wiesand> Thanks a lot everyone!
[15:17:16] <wiesand> And review review review…
[15:17:46] wiesand leaves the room
[15:24:44] meffie leaves the room
[15:29:16] <mvita> sorry to miss the meeting this week
[15:57:06] meffie joins the room
[16:00:25] meffie leaves the room
[16:00:26] meffie joins the room
[19:36:49] meffie leaves the room
[19:36:49] meffie joins the room
[21:03:52] meffie leaves the room
[22:41:06] Marcio Barbosa leaves the room