Home
release-team@conference.openafs.org
Friday, October 25, 2019< ^ >
Room Configuration
Room Occupants

GMT+0
[03:25:43] mvita leaves the room
[11:58:11] mbarbosa joins the room
[13:20:34] meffie joins the room
[14:44:33] <kaduk@jabber.openafs.org/barnowl> [the perennial game of "has my client silently disconnected during the
week?"]
[14:53:58] <meffie> seems like it is still connected
[14:54:28] <kaduk@jabber.openafs.org/barnowl> Indeed.
[15:00:52] wiesand joins the room
[15:01:17] <wiesand> Hi
[15:01:33] <meffie> good afternoon
[15:02:00] <wiesand> > I'm quite optimistic that next week will be a lot more interesting than this one.
[15:02:07] <wiesand> and indeed it was
[15:02:28] <kaduk@jabber.openafs.org/barnowl> Let's hope it was the good "interesting" and not the bad one, though?
[15:02:51] <wiesand> Ben: Great job! Thanks a lot!
[15:02:58] <meffie> yes, thank you.
[15:03:45] <kaduk@jabber.openafs.org/barnowl> You're welcome!  Also, thanks to Andrew for doing most of the legwork
on the investigation/fixes
[15:04:30] <wiesand> Well there's always an up- and a downside to my beloved security releases. And yes, Kudos to Andrew!
[15:05:38] <kaduk@jabber.openafs.org/barnowl> I did try out your idea about running all the changes through
gerrit/buildbot.
[15:05:49] <kaduk@jabber.openafs.org/barnowl> The solaris and windows builders felt kind of slow to wait for
[15:06:07] <wiesand> Re 1.8.x, we were lucky enough that we had a clean HEAD at this point. And thank you *so*much* for taking my preferred route!
[15:07:06] <kaduk@jabber.openafs.org/barnowl> Indeed, it was handy that we could time it this way.
[15:07:30] <wiesand> It's really helpful that for once gerrit and git (yes I know it's actually part of gerrit) agree on the course of history.
[15:07:41] <kaduk@jabber.openafs.org/barnowl> BTW, for 1.6.x I had the changes run through gerrit to land on
openafs-stable-1_6_x, then manually cherry-picked them to a temporary
branch based on 1.6.23, and tagged that temporary branch.
[15:09:33] <wiesand> Re 1.6.x, this of course (for my simple mind) adds to the mess we already had (which could very well be fallout of the 1.6.23 security release - I remember Ben sent a mail about some cleanup which actually nevenr happened).
[15:10:18] <kaduk@jabber.openafs.org/barnowl> Maybe I am remembering different cleanup, since I think it happened.
[15:10:43] <kaduk@jabber.openafs.org/barnowl> (There's a merge commit for merging openafs-stable-1_6_23 into
openafs-stable-1_6_x, and another one from this week merging
openafs-stable-1_6_24 into openafs-stable-1_6_x)
[15:12:43] <wiesand> My current inclination is that if we want to have future non-security 1.6.x releases, we should probably reset the HEAD of stable-1_6_x to the stable-1_6_24 tag, discard what's in the 1.6.x queue right now, and then start cherry-picking again from 1.8.x.
[15:13:19] <wiesand> In order, taking care of path conflicts while pulling up
[15:13:41] <kaduk@jabber.openafs.org/barnowl> I'm not super-keen on force-pushing to rewind history, especially
given those merge commits, but maybe you mean something else
[15:14:13] <meffie> maybe he means to abandon gerrits?
[15:14:20] <wiesand> It's not rewinding/rewriting history.
[15:16:32] <wiesand> It's just getting back into a clean state.  Mike is right: let's abandon those gerrits never committed and take a fresh start, on top of the 1.6.24 release.
[15:16:38] <meffie> there are 30 open gerrits on  refs/for/openafs-stable-1_6_x ?
[15:17:20] <meffie> i can just rebase those on top of 304a589767 (scm/openafs-stable-1_6_x, origin/openafs-stable-1_6_x) Merge tag 'openafs-stable-1_6_24' into openafs-stable-1_6_x
[15:17:30] <meffie> as new patchsets.
[15:17:43] <wiesand> Yes, and those as well as what we would abandon should be a guideline of what to cherry-pick from 1.8.x on top of 1.6.24 to get to a 1.6.25pre1.
[15:17:43] <kaduk@jabber.openafs.org/barnowl> Ah, sure.
[15:18:37] <wiesand> Rebase, cherry-pick on top of some new tag, there's no real difference.
[15:18:44] <meffie> and any we dont want we can abandon.
[15:20:40] <wiesand> As I mentioned last week, some rebases on top fo the current HEAD, which would be required to go forward, are not trivial at all.
[15:21:33] <wiesand> And as I mentioned too, it's probably my fault.
[15:21:34] <meffie> ok, i can ask for help if i get stuck.
[15:22:50] <wiesand> If you fell you can get the 1_6_x branch back into a state where it can be worked on without a significant initial effort, please just submit a stack of changes getting that done.
[15:24:18] <kaduk@jabber.openafs.org/barnowl> Do we think there are many changes that we want to have on 1.6.x that
are not already in gerrit?
[15:24:43] <kaduk@jabber.openafs.org/barnowl> (specifically, ones that would be "interleaved" with ones already in
gerrit, when considered in order of appearance on the 1.8.x branch, if
that makes sense)
[15:25:13] <wiesand> It's very possible that you have better ideas how to get that done. My best idea is to reset to the 1_6_24 tag and then (re-start) picking the cherries from the 1_8_x tree. In order, with no path conflicts and no merge commits obscuring things for dummies like me.
[15:26:37] <wiesand> What we have in gerrit for 1.6.x can't be merged right now. It cannot even be rebased without major editing. That's the actual problem.
[15:26:57] <meffie> ok, let me take a look.
[15:27:28] <meffie> i can do some work locally to see.
[15:28:32] <kaduk@jabber.openafs.org/barnowl> <tiny>and if it's too hard, 1.6 can go into "security fixes only"
mode</tiny>
[15:29:03] <wiesand> That's the alternative, yes.
[15:29:30] <wiesand> And at this point in time, I wouldn't exclude that.
[15:30:08] <meffie> i'm fetching them now to see what can be done.
[15:30:21] <wiesand> That's why I prefixed my whole moaning with "if we want to have non-security 1.6.x releases"
[15:31:28] <wiesand> Mike: thanks a lot! Just don't sink a work-week into this.
[15:31:36] <meffie> ok
[15:32:15] <meffie> i like rebasing/cherry-picking commits! i am weird that way.
[15:33:30] <meffie> (got all 30 now, will look after the meeting)
[15:34:30] <kaduk@jabber.openafs.org/barnowl> Thanks Mike!
[15:34:50] <wiesand> That's probably enough on 1.6 for today? Re 1.8, plenty of changes lined up already - I haven't had a chance yet to check whether the security release actually got in the way.
[15:34:56] <wiesand> Any news on Linux 5.4?
[15:35:06] <kaduk@jabber.openafs.org/barnowl> I am pretty sure the security release did not get in the way
[15:35:28] <kaduk@jabber.openafs.org/barnowl> (and hopefully I did not hold up too much actual work by asking you to
delay!)
[15:35:46] <meffie> (quick peek of them all in gitk; doesnt look too bad)
[15:38:23] <meffie> linux 5.4 seems ok. but ubuntu has problems with building *any* kernel module at the moment.
[15:38:39] <kaduk@jabber.openafs.org/barnowl> I don't have any 5.4 news; my debian machine is still on 5.2.
(This ends up making it a bit annoying when I'm trying to test
security releases, since the bleeding-edge linux support is not always
merged to the old branch.)
[15:38:41] <meffie> so the buildbots (which are ubuntu based) are failing.
[15:38:58] <meffie> cwills has opened tickets with ubuntu folks.
[15:39:12] <meffie> and continues his local testing on gentoo
[15:39:53] <meffie> marcio has patches to install on maco catalan and has been talking to yadav about them.
[15:40:16] <meffie> the changes seem to be confined to the installation process.
[15:40:29] <meffie> catalina that is.
[15:40:55] <meffie> marcio will also be creating a macos catalan buildbot for me.
[15:41:11] <meffie> for me to add to the buildbot.openafs.org that is.
[15:43:33] <wiesand> Ben: No, as I said the timing was quite perfect. We basically shut down our whole site over this weekend due to major network work. I've been busy preparing that anyway (my last night was ~3h).  OTOH hands, we have the security updates on all our servers already and will have it on all important clients after the shutdown :)
[15:45:04] <kaduk@jabber.openafs.org/barnowl> somebody (Mike?) should probably send me email so I actually remember
to merge the catalan bits
[15:45:42] <meffie> i think it still a work in progess, not pushed to gerrit yet.
[15:46:07] <wiesand> Catalina - probably the most buggy macOS update ever?
[15:46:28] <kaduk@jabber.openafs.org/barnowl> Ah, okay, I feel less bad now
[15:46:40] <meffie> :)
[15:48:13] <wiesand> NB I have two dumb questions regarding the security update…
[15:49:20] <kaduk@jabber.openafs.org/barnowl> go ahead
[15:49:41] <wiesand> 1) The commit messages indicate to me that the information leaks affect servers only, while the security advisories sound like they affect both clients and servers. What's correct?
[15:50:17] <wiesand> 2) If clients are affected, does this hold true for the Windows ones too?
[15:50:17] <kaduk@jabber.openafs.org/barnowl> they affect the server side of the RPC interaction, but cache managers
act as servers for the AFSCB RPCs
[15:51:00] <kaduk@jabber.openafs.org/barnowl> I have no reason to think that the windows clients would be
unaffected, but I also haven't specifically looked.
[15:51:10] <wiesand> Ok, thanks. So the client is affected though the commit messages are about "serrver RPCs".
[15:51:32] <kaduk@jabber.openafs.org/barnowl> right
[15:51:40] <wiesand> :)
[15:52:57] <wiesand> I'm done ranting and wining for today ;-) On to master/1.9/anything?
[15:53:41] <kaduk@jabber.openafs.org/barnowl> Nothing too exciting, I think -- things are getting reviewed and a few
are almost ready to go
[15:55:24] <kaduk@jabber.openafs.org/barnowl> I guess I should actually put out 1.9.0 now that the security releases
are out and I'm not distracted by that
[15:57:11] <meffie> andrew is still making progress on rxgk-phase2 patches.
[15:58:52] <kaduk@jabber.openafs.org/barnowl> good to know
[15:59:18] <kaduk@jabber.openafs.org/barnowl> does anyone have anything else interesting going on for master to
mention?
[15:59:41] <meffie> cwills is changing the ivp6-prep patches again.
[15:59:56] <kaduk@jabber.openafs.org/barnowl> Ah, I just commented on the first one in that stack
[16:00:04] <meffie> yes, thank you!
[16:00:48] <meffie> after those are in, we can get back to afs_int32 to sockaddr/saddr conversions :)
[16:02:08] <wiesand> piece of cake, compared to the 1.6.x state ;-)
[16:03:06] <meffie> heh, i wish
[16:04:02] <kaduk@jabber.openafs.org/barnowl> That reminds me of the looming need for vldb format updates...
[16:04:35] <meffie> yes...
[16:04:43] <wiesand> why?
[16:04:52] <meffie> that is becoming very urgent now.
[16:05:18] <meffie> a number of reasons. one is the hard limit on the number of volumes.
[16:05:22] <kaduk@jabber.openafs.org/barnowl> We need to have space to store IPv6 addresses for fileservers, and to
store per-fileserver rxgk keys
[16:05:49] <meffie> another is ipv6, another will be rxgk.
[16:06:14] <wiesand> ok got it…
[16:07:08] <kaduk@jabber.openafs.org/barnowl> How does the limit on number of volumes take effect?
[16:07:53] <meffie> current limit is 14.5 x 10^6 volumes, limited by the 32-bit record indexing
[16:08:01] <wiesand> though the last thing the site admin personality of me is format changes, be it the vldb or /vicep ondisk
[16:09:19] <kaduk@jabber.openafs.org/barnowl> Indeed, it's never something to make one feel comfortable...
[16:09:20] <wiesand> *wants* is format changes
[16:09:37] <kaduk@jabber.openafs.org/barnowl> Anyway, we're past the hour and are maybe winding down.  Any other
topics for today?
[16:09:46] <wiesand> not here
[16:10:13] <meffie> thanks everyone!
[16:10:23] <kaduk@jabber.openafs.org/barnowl> Thanks everyone!
[16:11:00] <wiesand> Thanks  Ben & Mike! Have a nice weekend.
[16:11:02] <meffie> (as an aside, i pushed a new version of https://github.com/meffie/git-gerrit)
[16:11:29] <meffie> it has a new command to make it easier to "cherry-pick" gerrits to the stable branches
[16:11:56] <meffie> have a good weekend
[16:12:29] <wiesand> I should have a closer look at this already. CU next week.
[16:12:32] wiesand leaves the room
[16:14:24] meffie leaves the room
[17:47:26] meffie joins the room
[18:17:51] mbarbosa leaves the room
[18:17:52] mbarbosa joins the room
[18:38:15] meffie leaves the room
[21:41:55] mbarbosa leaves the room