20:28:23 <prometheanfire> #startmeeting requirements
20:28:23 <openstack> Meeting started Wed Jul 18 20:28:23 2018 UTC and is due to finish in 60 minutes.  The chair is prometheanfire. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:28:24 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:28:27 <openstack> The meeting name has been set to 'requirements'
20:28:31 <prometheanfire> #topic rollcall
20:28:34 <prometheanfire> tonyb, prometheanfire, number80, dirk, coolsvap, toabctl, smcginnis, dhellmann ping
20:28:37 <prometheanfire> o/
20:28:42 <dhellmann> o/
20:28:52 <prometheanfire> dirk said he couldn't be here iirc
20:30:19 <smcginnis> o/
20:31:07 <prometheanfire> gonna wait a couple min for tony if able
20:32:05 <tonyb> \o
20:32:37 * tonyb wonders if Octonauts references would work with this group?
20:33:01 * dhellmann googles octonauts
20:33:10 * prometheanfire googles octonauts
20:33:17 <prometheanfire> that the switch thing?
20:33:37 <tonyb> that was pretty much what I thought :(
20:33:42 <tonyb> It's a kids TV show
20:33:46 <smcginnis> I'm well past that stage. :)
20:33:58 <tonyb> anyway let's do this meeting thing ;P
20:33:58 <dhellmann> so, nothing to do with october
20:34:03 <prometheanfire> #topic Any controversies in the Queue?
20:34:17 <smcginnis> Is it being broken controversial?
20:34:50 <prometheanfire> lol
20:34:51 <dhellmann> is the pypi mirror still causing trouble, or is there something else?
20:34:59 <prometheanfire> just the mirror, ya
20:35:01 <tonyb> hehe I dont' see any controversies, but clearly it's broken and we wanted to talk about post releases
20:35:19 <prometheanfire> ok
20:35:29 <prometheanfire> #topic brokenness
20:35:33 <tonyb> prometheanfire: are infra on it? or should I monitor that during my day
20:35:47 <prometheanfire> they are
20:35:50 <smcginnis> There's a patch in gate right now to fix it.
20:35:51 <tonyb> \o/
20:36:04 <smcginnis> Or at least work around it for now.
20:36:15 <prometheanfire> just pinged mordred about it
20:36:27 <prometheanfire> still bad
20:36:38 <smcginnis> Oh, looks like the one they hoped would get around it merged: https://review.openstack.org/#/c/583571/
20:36:42 <tonyb> Ahh okay, that kind of broken I thought it was just the normal bindersnatch thing
20:36:48 <prometheanfire> anyway, once it's fixed we can update the uc on networkx and the two google libs
20:37:02 <tonyb> is there a 30sec summary of the problem?
20:37:13 <prometheanfire> they ran out of space
20:37:29 <tonyb> Oh okay  that's well less that 30secs
20:37:30 <prometheanfire> said it'll be fixed in 15 minutes or so
20:37:30 <smcginnis> Taskflow uploads ~500m of packages nightly. Mirroring locally got too big.
20:37:42 <smcginnis> Darn storage.
20:37:43 <dhellmann> I guess we need a bandersnatch->swift converter
20:37:52 <smcginnis> I saw some discussion of that.
20:37:54 <prometheanfire> 500M for taskflow?
20:37:58 <smcginnis> prometheanfire: Yep
20:38:01 <tonyb> smcginnis: Really? what are they uploading?
20:38:01 <prometheanfire> wow
20:38:08 <smcginnis> The world?
20:38:10 <dhellmann> what on earth
20:38:22 <smcginnis> I guess some of the ML stuff is pretty big for some reason.
20:38:25 <tonyb> dhellmann: I dont' think so I think that the mirrors are staycing on AFS it's only logs that are moving
20:38:35 <dhellmann> tonyb : yeah, I meant for the future
20:38:52 <prometheanfire> anyway, moving on then
20:38:59 <prometheanfire> #topic freeze
20:39:23 <dhellmann> smcginnis : tensorflow?
20:39:37 <prometheanfire> I'll need to find more mr freeze puns, maybe rewatch batman & robin
20:39:38 <smcginnis> Oh, sorry. Yes, tensorflow.
20:39:48 <prometheanfire> the things I suffer for you
20:39:50 <dhellmann> ok, whew.
20:39:56 <tonyb> smcginnis, dhellmann: Oh that makes away mmore sense!
20:39:56 <smcginnis> prometheanfire: :D
20:40:06 <dhellmann> prometheanfire : I hear they're rebooting the iceman comics
20:40:08 <tonyb> the earth stops shaking ;P
20:40:12 <prometheanfire> dhellmann: :D
20:40:24 <smcginnis> Heh, sorry. That would be really bad if taskflow was that big.
20:40:31 <prometheanfire> this freeze should be easier because of the per-project reqs
20:40:33 <dhellmann> well, we could at least do something about that :-)
20:40:37 <tonyb> With the freeze and pre-project requirements we need to think a little about minimum bumping
20:41:37 <prometheanfire> this is still the process https://etherpad.openstack.org/p/requirements-cycle-process
20:41:41 <tonyb> in that say (bad example) oslo.privsep adds a new API/constant today and it merges and is released next week at what point can python-swiftclient bump the minimum to require that (for them)?
20:42:10 <tonyb> what about say $clientlibrary that is pulled into python-openstackclient?
20:42:38 <dhellmann> as long as the u-c list isn't changed, does it matter?
20:42:41 <prometheanfire> they could bump it immediately, as long as their min is not over our UC
20:43:03 <prometheanfire> packagers would need to look at swifts lower-constraints to find swifts min
20:44:11 <tonyb> dhellmann: I'm not sure, which is why I'm asking ;P  I feel like if python-openstackclient took bumped a minimum that kinda of implies that they'd like a release and uc bump
20:44:22 <tonyb> which I guess will all come out in the exception proecess
20:44:36 <tonyb> I just wanted to discuss it ahead of time
20:44:39 <dhellmann> sure
20:44:44 <prometheanfire> so you want to make sure that their reqs are in any given release?
20:44:46 <dhellmann> I'm also making sure I'm not missing something :-)
20:45:04 <tonyb> prometheanfire: Yup and we have dhellmann build tool to make life slightly easier for packagers to work that out
20:45:17 <dhellmann> I do think if we end up needing a release of a library for a bug fix, we need to coordinate the u-c update ahead of time. But I think we generally do that in this period of the cycle anyway, right?
20:45:38 <prometheanfire> tonyb: ya, you talking about the global-lc.txt builder?
20:45:50 <dhellmann> we don't want to have a case where we release a library and then don't allow the u-c update, because then we have a thing out in the wild that we're not using in our own tests
20:45:54 <tonyb> prometheanfire: Yup
20:46:34 <prometheanfire> ok, anything else for now about the freeze?
20:46:52 <tonyb> dhellmann: Yup I agree and I guess without the bot updates a release is less impactful in that only projects that are actually impacted by the bug need to update
20:47:09 <tonyb> and if a project works that out after the fact the fixed version shoudl be available
20:47:14 <dhellmann> tonyb : ideally we wouldn't update minimums just for bug fixes, but yeah
20:48:00 <tonyb> dhellmann: true, but it does happen esp when a bug fix changes an API (by say adding a new kwarg)
20:48:19 <prometheanfire> yarp
20:48:29 <dhellmann> tonyb : sure
20:48:53 <tonyb> it may be I'm over thinking ia and worrying too much due to not enough coffee
20:49:33 <dhellmann> nah, it's good to review this stuff
20:49:43 <prometheanfire> ya, just in case
20:49:46 <dhellmann> maybe we should write the reasoning down so we don't have to think so hard next time :-)
20:50:03 <prometheanfire> dhellmann: formalizing the release process for reqs would be nice
20:50:05 <tonyb> dhellmann: good plan
20:50:12 <prometheanfire> turning https://etherpad.openstack.org/p/requirements-cycle-process more official
20:50:19 <dhellmann> yeah, that made a big difference in our confidence on the release team
20:50:20 <tonyb> prometheanfire: want to add something to the end of that etherpad
20:50:33 <prometheanfire> tonyb: go ahead?
20:50:41 <dhellmann> we have a rst file in the releases repo now; that etherpad could move over to the requirements repo eventually
20:50:46 <tonyb> prometheanfire: +1 having some process docs in repo (like releases) would be good
20:50:56 <tonyb> dhellmann: snap :_
20:51:00 <smcginnis> That might be a good place for that.
20:51:00 <dhellmann> maybe that's a project for stein
20:51:44 <tonyb> prometheanfire: I mean *can you* add something to the end of that etherpad (or PROCESS.rst)
20:52:10 <prometheanfire> tonyb: of course, it's just a review away
20:52:11 <tonyb> dhellmann: I feel like we can start now.  it has no material impact on the release
20:52:33 <tonyb> prometheanfire: cool
20:52:43 <dhellmann> oh, sure, I just meant that we could be deliberate about updating it throughout the cycle at different points where the rules change or whatever
20:52:47 <prometheanfire> #topic open floor
20:52:55 <tonyb> so we have 8mins left do we want to discuss post releases in general?
20:53:00 <tonyb> and pyldap specifically?
20:53:01 <dhellmann> do we have any topics lined up to discuss at the ptg?
20:53:11 <dhellmann> oh, sure, that's more important, let's do that
20:53:12 <prometheanfire> tonyb: sure
20:53:33 <prometheanfire> dhellmann: I think I'll start thinking about ptg next week (after freeze)
20:53:40 <dhellmann> makes sense
20:53:45 * tonyb wont be at the PTG but if y'all want to hangouts me into anything I can be available pening TZs
20:53:55 <dhellmann> aw!
20:54:03 <prometheanfire> tonyb: but you'll miss the choo choo puting you to sleep
20:54:06 <smcginnis> That's unfortunate.
20:54:09 <tonyb> dhellmann, smcginnis: goes for rleases to
20:54:22 <prometheanfire> tonyb: so, post release?
20:54:23 <dhellmann> yeah, we'll keep that in mind
20:54:30 <tonyb> so post releases
20:54:31 <smcginnis> tonyb: Thanks. If we have some topics we'll try to find a convenient time.
20:54:34 <dhellmann> so, what's the deal with post releases? why are we worried about them?
20:54:54 <tonyb> I think they got lumped with pre releases and theer was some unreasonable FUD about accepting them
20:55:05 <dhellmann> ah, ok
20:55:10 <tonyb> I think we shoudl at a minimum evealuate each, if not just accept them
20:55:20 <dhellmann> iirc, the only reason we don't like pre-releases is that pip needs special instructions to install them
20:55:29 <dhellmann> does that apply to post releases?
20:55:30 <prometheanfire> oh, those .post releases
20:55:39 <tonyb> and pyldap is a trivial one to accept as it's litterally the same code as 3.0.0
20:55:53 <prometheanfire> I think we should just accept them
20:56:15 <prometheanfire> we don't need to disable pre-releases because they are already disabled by default
20:56:28 <prometheanfire> that would make my review not needed
20:56:31 <tonyb> dhellmann: I thought there was more it it than that in that there is a perception that they're more risky
20:56:44 <dhellmann> I guess I don't understand that perception
20:57:01 <tonyb> put perhaps that's linked to pre releases of $major bumps
20:57:02 <dhellmann> if the same package was tagged with a new version number we wouldn't even behaving the conversation, right?
20:57:03 <prometheanfire> doesn't debian have something like that?
20:57:09 <prometheanfire> 1.2.3-2
20:57:15 <prometheanfire> gentoo has 1.2.3-r2 for instance
20:57:22 <prometheanfire> it's the same concept
20:57:40 <tonyb> dhellmann: that's true
20:57:46 <prometheanfire> this is a question of if we allow revision bumps, I can't see a reason why
20:57:53 <prometheanfire> s/why/why not
20:57:53 <dhellmann> perhaps we should treat them like any other update, and let the test jobs evaluate them on a case-by-case basis
20:58:09 <tonyb> I guess some of it came from before we had the cross-project gateing we have now
20:58:31 <dhellmann> hmm
20:58:36 <prometheanfire> so, everyone fine with just allowing them? if so I'll drop my review
20:58:43 <dhellmann> +1 for allowing them
20:58:54 <tonyb> prometheanfire: for varying values of fine ;P
20:59:00 <dhellmann> at least until things blow up in our face, I guess :-)
20:59:06 <tonyb> but yeah I think I'm good with it
20:59:14 <prometheanfire> sgtm
20:59:27 <tonyb> dhellmann: at least a u-c (partial) revert is easy
20:59:33 <dhellmann> yeah
21:00:03 <tonyb> .pre/.post releases for EVERYBODY ;p
21:00:03 <prometheanfire> tonyb: anything else?
21:00:14 <tonyb> prometheanfire: I don't think so
21:00:28 <prometheanfire> tonyb: I don't think pre-releases are good, but those have to be manually proposed anyway
21:00:43 <tonyb> prometheanfire: not all of them
21:00:44 <prometheanfire> ok, going to endmeeting
21:00:54 <prometheanfire> #endmeeting