07:01:06 #startmeeting requirements 07:01:07 Meeting started Wed Feb 14 07:01:06 2018 UTC and is due to finish in 60 minutes. The chair is prometheanfire. Information about MeetBot at http://wiki.debian.org/MeetBot. 07:01:08 looks pretty empty now :) 07:01:09 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 07:01:11 The meeting name has been set to 'requirements' 07:01:28 #topic Roll-call 07:01:35 tonyb, prometheanfire, number80, dirk, coolsvap, toabctl, smcginnis 07:01:44 tonyb: yep, it's awesome 07:01:58 o/ 07:02:00 also 07:02:01 o/ 07:02:16 o/ 07:03:47 #topic Any controversies in the Queue? 07:03:59 #link https://review.openstack.org/543934 07:04:05 the networkx thing for pike 07:05:10 prometheanfire: Yeah we need a reply from watcher 07:05:16 ya 07:05:34 o/ 07:05:41 networkx was always a bit of a pain there's a review to raise the cap 07:05:48 dirk: good vacation? 07:05:50 They're using constraints so I don't get how they're getting 2.x 07:07:02 how did we merge the cap to master in the first place? that looks odd 07:07:09 I don't think they were using constraints 07:07:18 prometheanfire: yes, too short as always, thanks ;-) 07:07:18 the link to the irc log makes me think that 07:07:39 dirk: merge the cap? 07:08:11 well, I see in a tools/grep-all output that we already have the cap in master and are backporting it to pike now 07:08:15 prometheanfire: http://git.openstack.org/cgit/openstack/watcher/tree/tox.ini?h=stable/pike#n10 07:08:51 ah 07:09:13 tonyb: I don't think they used tox 07:09:14 https://review.openstack.org/#/c/505905/ this is the master change 07:09:27 dirk: that's a godo point, we should investiagte why we added it and if that makes sense on pike 07:09:38 dirk: oh ya, I remember now 07:09:42 but I'd still really like to know how they're hitting the issue 07:10:09 hi 07:10:23 prometheanfire: que? Oh you mean they did the install wrong and bad things happend? 07:10:27 tonyb: ya, still need a response 07:10:34 tonyb: imagine that, yes 07:10:51 I think it's probably valid 07:11:17 I think it was earlier today that a lib that depends on networkx just added 2.x support 07:11:34 prometheanfire: in theory the backport to stable/pike would be a cherry-pick of the master change, with the same change id.. 07:11:42 not sure if we want to reject the current change based on that reason 07:11:50 pypowervm was it 07:12:20 dirk: we do, we shoudl follow the correct process so when we need to undo it we have easy tools 07:12:21 dirk: good point, I think, if we decide to approve it, that it should require that method 07:12:49 for now that's all I have for controversies 07:13:09 I cherry-picked th e master change now in a separate review 07:13:50 ok 07:14:02 moving on 07:14:06 #topic unfreeze 07:14:26 basically we can merge gr bumps but not uc bumps 07:14:39 we could merge https://review.openstack.org/#/c/518416/ then ;-) 07:14:49 tripleo stuff might be an exception if only used by them (cycle trailing and all) 07:15:10 dirk: two of us voted :P 07:15:43 tonyb: about dirks review, we are sort of chasing the tail on it, getting it half way there (or more) would help move things along 07:15:51 dirk: Sure I'm fine with it 07:16:04 +W 07:16:15 cool 07:16:45 anything else for the unfreeze? 07:17:05 write an announcement? 07:17:07 Oh dirk and I raced on the upload of that cherry pick but 'cause they have the same chang-id mine is a new PS to his ;P 07:17:12 dirk: email was sent monday 07:17:23 prometheanfire: excellent 07:17:29 tonyb: nice, sounds like a bug 07:17:42 no feature! 07:17:56 :D 07:18:07 #topic ptg 07:18:16 #link https://etherpad.openstack.org/p/rocky-PTG-requirements 07:19:17 I have no more than that, may (should) have more details next week 07:19:33 when do people get to town? 07:20:06 checking 07:20:23 sunday here, probably going to the distillery and brewery to stay awake, was invited 07:20:52 ~7am Sunday 07:21:09 also leave on sunday, doing the 'game of thrones tour' that saturday 07:21:16 my plan is to walk many miles 07:21:26 tonyb: I think dave kevin and I will land at about the same time 07:21:55 if you keep walking you could pick me up from the airport monday afternoon ;-) 07:22:03 okay we'll to the hangouts thing to say hello and maybe share a cab 07:22:07 10:05AM landing 07:22:16 so maybe not 07:22:21 oh that's not quite like 7am ;P 07:22:29 ya, used to brussels :P 07:22:40 dirk: I don't think I could piggy back you that far ;P 07:22:50 tonyb: you switch half way 07:22:59 brussels where 7am is like 1045am :D 07:23:24 where the extra travel time plus the timezone change make it close enough :P 07:23:24 dirk: you are only coming from Germany so what's that like 15mins? 07:23:41 anything else for the ptg section? 07:23:53 tonyb: flight time? or time zone switch? 07:24:10 haven't checked, but flight time should be like 1-2 hours I guess 07:24:14 dirk: flight time ... I'm being silly 07:24:28 yeah, I comfortable leave my home at 10am in the morning 07:24:29 Mines a quick 34hours 07:24:31 takes longer getting to chicago from here lol 07:24:40 (I can't travel on sunday due to family event) 07:24:44 tonyb: sounds about right :| 07:25:02 tonyb: wow ;/ 07:25:04 By 2100 Sunday night I will be soooo ready for bed ;P 07:25:04 #topic Open Discussion 07:25:16 tonyb: you get a good plane at least? 07:25:27 * tonyb thinks we kinda wandered off into open discussion already ;P 07:25:33 no plane is good enough for 34 hours 07:25:43 prometheanfire: nope 747, Economy #joy 07:26:02 :( 07:26:18 ya, going to end the meeting in a min 07:26:35 tonyb: I hope you invested well into bose noise cancelling technologies ;-) 07:26:35 At least I skip LHR CBR->SYD->ABU->DUB 07:26:58 ya... that's a mess 07:27:02 dirk: I will before I get on the plane ;P 07:27:40 so while we're on the open discussion 07:27:52 we've recently hit an issue with pycrypto vs pycryptodome 07:28:07 the latter seems to be a fork of the former, but most of openstack seems to use the former 07:28:16 both install files in the same place, so its "fun" to package 07:28:25 dirk: Yeah we've been trying to force the switch 07:28:40 I think we shoudl give up and many it an 'S' goal 07:28:40 does anyone know more about that? its not a new issue, but slowly dependencies are becoming incompatible 07:28:40 dirk: ya, it's complicated 07:29:17 here's gentoo's bug to transition https://bugs.gentoo.org/611568 07:29:33 I think it is not found in the gate because pip just happily overwrites files 07:29:42 but it sounds like a recipe for hard to trace bugs 07:29:51 are we tracking both? 07:30:05 prometheanfire: we must be right 07:30:10 we are 07:30:29 pycryptodomex===3.4.7 and pycrypto===2.6.1 07:30:30 prometheanfire: no, we only track pycrypto 07:30:35 in g-r at least 07:30:48 but some of the deps are indirectly switching 07:30:51 so it clashes 07:31:01 the case I found was pysnmp vs tooz iirc 07:31:20 prometheanfire: so perhaps we should have an adhoc session at the PTG to get buy-in from the main users 07:31:37 we need to force a move to either pycryptodomex or cryptography 07:31:45 tonyb: ya, 07:31:51 Oh yeah I thought that was the path 07:31:56 well, there are 3 libs it seems: pycrypto, pycrytodome and pycryptodomex 07:32:11 pycrypto users are supposed to got straight to cryptography wthout passing go 07:32:20 pycryptodome seems to be a drop in for pycrypto. so that switch seems easy 07:32:30 but I'm not sure if thats where we want to end up with 07:32:35 added to etherpad 07:33:01 it's not the optimal solution 07:33:05 prometheanfire: We need to ping the PTLs of directly affected projects 07:33:13 agreed 07:33:30 added todo item to figure it out when I wake up 07:33:31 we potententially might want to start the discussion beforehand 07:33:37 gonna start a thread 07:33:39 so that people can think about the impact 07:34:19 yep 07:34:21 https://github.com/Legrandin/pycryptodome/blob/master/README.rst 07:34:22 prometheanfire: sounds good I hope it gets more traction than the last 2-3 times we tried 07:34:34 tonyb: a little bit every time... 07:34:37 so it looks like pycryptodome is a drop in (so switch should be easy) and pycryptodomex is a new api 07:34:42 so the latter will take more time 07:35:00 tonyb: this cycle I want to start on the uncapping/moving a bit more vigorously 07:35:01 "an almost drop-in replacement for the old PyCrypto library" I bet we hit the cases where it isn't :D 07:35:10 so a path to success might be to switch pycrypto-pycrytodome and then figure out who wants to move to pycryptodomex vs cryptography 07:35:10 that and publish the constraints better 07:35:41 dirk: because it's a dropin we can't track both in g-r IIRC 07:35:43 dirk: I think that's just pushing the ball 07:35:58 tonyb: yep 07:36:04 anyway we can save if for the thread and PTG 07:36:04 tonyb: well, pip doesn't care :) 07:36:09 yp 07:36:10 yup 07:36:15 dirk: True 07:37:04 I think we're really at the point where we can do the publishing of constraints now but I don't have the time to learn how zuul v3 does publishing 07:37:13 can eitehr of you do that? 07:37:20 tonyb: ya, I was going to take it 07:37:21 no need to wait for anything IMO 07:37:24 when I have time 07:37:28 so you know, never... 07:37:37 but seriously, I'll do it 07:37:46 I might potentially have to learn it, as I wanted to create the wheels-builder job for opensuse 07:38:00 although that might be a different publishing path 07:38:00 I need more zuulv3 learnings anyway 07:38:27 prometheanfire: we might be able to get some steps on that with the infra people in one room at the ptg 07:38:46 in the worst case they tell us to rtfm 07:39:03 prometheanfire: we eitehr need to backport tools/publish_constraints.sh or limt the job to master while we rolel things out 07:39:13 Yeah if we haven 07:39:14 dirk: yep, that's the plan, more or less 07:39:36 't done it by then We can grab ianw and ask for help 07:39:37 this weekend is gentoo meeting hell time 07:39:58 anyway, endmeeting? 07:40:12 #endmeeting