| *** andreas_s has joined #openstack-requirements | 00:37 | |
| *** andreas_s has quit IRC | 00:41 | |
| rm_work | Can someone refresh my memory -- do projects need to match EXACTLY the line from global-requirements for every package, or does it just have to be mostly matching? | 00:42 |
|---|---|---|
| prometheanfire | rm_work: what do you mean? | 00:46 |
| prometheanfire | we are working on per project requirements https://bugs.launchpad.net/openstack-requirements/+bug/1719009 | 00:47 |
| openstack | Launchpad bug 1719009 in OpenStack Global Requirements "per project requirements" [High,New] | 00:47 |
| rm_work | prometheanfire: like gr here https://github.com/openstack/requirements/blob/master/global-requirements.txt#L84 | 00:48 |
| rm_work | can we have a slightly different line? | 00:48 |
| rm_work | as long as it doesn't violate u-c | 00:48 |
| prometheanfire | what do you want it to be? | 00:48 |
| rm_work | >=2.10 | 00:48 |
| prometheanfire | (and why) | 00:48 |
| rm_work | because that's what we require | 00:48 |
| rm_work | our project literally breaks on lower | 00:48 |
| prometheanfire | if that's what your project requires then update GR | 00:50 |
| rm_work | k | 00:50 |
| rm_work | can we backport g-r? | 00:50 |
| prometheanfire | (and lower) | 00:50 |
| rm_work | to queens | 00:50 |
| prometheanfire | no, the biggest thing we allow for stable branches is disallowing something | 00:50 |
| rm_work | :/ | 00:51 |
| rm_work | k | 00:51 |
| prometheanfire | ya | 00:51 |
| *** oanson has quit IRC | 00:56 | |
| rm_work | it SHOULD be fine but apparently RDO doesn't use upper-constraints and like ... just grabbed the lowest version of jinja2 that is valid (2.8.1) which doesn't work | 00:56 |
| rm_work | and i can't argue that technically our requirements are wrong since we imply 2.8.1 would work | 00:57 |
| *** oanson has joined #openstack-requirements | 01:03 | |
| prometheanfire | so you are not using upper-constraints? | 01:05 |
| prometheanfire | that seems like asking for trouble | 01:05 |
| openstackgerrit | Adam Harwell proposed openstack/requirements master: Bump jinja2 minimum to 2.10 (required by Octavia) https://review.openstack.org/549913 | 01:40 |
| rm_work | prometheanfire: it's not me, it's the RDO packagers | 01:41 |
| rm_work | and I told them as much :P | 01:41 |
| rm_work | prometheanfire: ^^ feel free to ask why they aren't using u-c on that patch, lol | 01:47 |
| prometheanfire | lol | 01:47 |
| prometheanfire | our UC is still good, we can accept a gr block on something if it breaks something, but so far it looks like this is all RDO packagers' fault | 01:48 |
| *** ying_zuo_ is now known as ying_zuo | 03:51 | |
| *** andreas_s has joined #openstack-requirements | 04:06 | |
| *** andreas_s has quit IRC | 04:11 | |
| *** udesale has joined #openstack-requirements | 04:44 | |
| *** andreas_s has joined #openstack-requirements | 06:48 | |
| *** andreas_s has quit IRC | 06:53 | |
| *** andreas_s has joined #openstack-requirements | 07:25 | |
| *** florianf has joined #openstack-requirements | 08:31 | |
| *** oanson has quit IRC | 08:41 | |
| *** ralonsoh has joined #openstack-requirements | 08:49 | |
| *** oanson has joined #openstack-requirements | 08:50 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/requirements master: update constraint for openstacksdk to new release 0.12.0 https://review.openstack.org/549983 | 09:11 |
| *** dims has quit IRC | 09:45 | |
| *** dims has joined #openstack-requirements | 09:49 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/requirements master: update constraint for openstackdocstheme to new release 1.19.0 https://review.openstack.org/549998 | 10:11 |
| *** ralonsoh_ has joined #openstack-requirements | 10:28 | |
| *** ralonsoh has quit IRC | 10:32 | |
| *** ralonsoh_ is now known as ralonsoh | 10:33 | |
| openstackgerrit | Merged openstack/requirements master: Bump jinja2 minimum to 2.10 (required by Octavia) https://review.openstack.org/549913 | 12:30 |
| *** edmondsw has joined #openstack-requirements | 13:27 | |
| *** udesale has quit IRC | 13:29 | |
| *** udesale_ has joined #openstack-requirements | 13:29 | |
| *** florianf has quit IRC | 14:38 | |
| *** florianf has joined #openstack-requirements | 14:39 | |
| *** andreas_s has quit IRC | 16:23 | |
| *** andreas_s has joined #openstack-requirements | 16:28 | |
| prometheanfire | dhellmann: when would you like to go over the per project requirements? | 16:30 |
| dhellmann | prometheanfire : I'm free now. Do you want to schedule a time when dirk and/or tonyb can be included? | 16:31 |
| prometheanfire | dhellmann: yes, so probably ~5 hours I think | 16:40 |
| dhellmann | that seems late for dirk, but I don't know his work schedule | 16:40 |
| prometheanfire | dhellmann: our normal meeting time is 1AM Central | 16:41 |
| dhellmann | ah | 16:41 |
| prometheanfire | which is tonight | 16:41 |
| prometheanfire | actually | 16:41 |
| *** andreas_s has quit IRC | 16:42 | |
| dhellmann | ok, I can be available to discuss it then | 16:44 |
| prometheanfire | dhellmann: ok, thanks, depending on when people are around we may be able to do it beforehand] | 16:47 |
| dhellmann | ok | 16:47 |
| *** udesale_ has quit IRC | 16:53 | |
| *** ralonsoh has quit IRC | 18:09 | |
| *** florianf has quit IRC | 18:30 | |
| *** florianf has joined #openstack-requirements | 18:45 | |
| *** florianf has quit IRC | 20:00 | |
| openstackgerrit | Paul Belanger proposed openstack/requirements master: Begin testing wheel builds with ubuntu-bionic https://review.openstack.org/550253 | 20:49 |
| openstackgerrit | Paul Belanger proposed openstack/requirements master: [WIP] test wheel builds on bionic https://review.openstack.org/550254 | 20:49 |
| *** prometheanfire has quit IRC | 21:10 | |
| openstackgerrit | Brianna Poulos proposed openstack/requirements master: Adds castellan-ui https://review.openstack.org/550262 | 21:18 |
| -openstackstatus- NOTICE: The infrastructure team is aware of replication issues between review.openstack.org and github.com repositories. We're planning a maintenance to try and address the issue. We recommend using our official supported mirrors instead located at https://git.openstack.org. | 21:18 | |
| *** prometheanfire has joined #openstack-requirements | 21:28 | |
| *** vpickard is now known as vpickard_ | 21:54 | |
| dhellmann | prometheanfire , tonyb, dirk : I have added some notes about an implementation plan to the bottom of https://etherpad.openstack.org/p/ocata-requirements-notes | 22:09 |
| dhellmann | I would appreciate it if you could confirm that I'm interpreting the test behaviors correctly and that I have thought through the sequence of steps correctly | 22:10 |
| dhellmann | once we agree in general, I can write up a more detailed plan to send to the -dev list | 22:10 |
| prometheanfire | dhellmann: https://bugs.launchpad.net/openstack-requirements/+bug/1719009 is our rough draft | 22:23 |
| openstack | Launchpad bug 1719009 in OpenStack Global Requirements "per project requirements" [High,New] | 22:23 |
| dhellmann | prometheanfire : I'm still confused by all of the stuff about global lower requirements in that bug :-( | 22:30 |
| prometheanfire | dhellmann: I think that's lower-constraints now | 22:31 |
| dhellmann | it feels like some completely different project; like we're working on different things | 22:31 |
| prometheanfire | the end goal is to allow projects to disable their reqs sync | 22:31 |
| dhellmann | prometheanfire : I'm also not personally interested in setting up gate jobs for projects to test their lower bounds, so I've left that part out. | 22:31 |
| prometheanfire | in order to do so they need to test their minimums | 22:31 |
| dhellmann | ah, no, see, I don't consider that testing to be a prerequisite | 22:32 |
| dhellmann | because we're not testing it now | 22:32 |
| prometheanfire | true, but does that mean that the projects can just stop requirements updates? | 22:32 |
| dhellmann | adding testing for lower bounds is orthogonal to what I want, which is to stop having a bot keep the requirements.txt entries in order | 22:32 |
| dhellmann | yes, after we adjust the test that verifies that their requirements files have "valid" entries we can just delete the job that does the sync | 22:33 |
| dhellmann | at least I think so, based on the notes I dropped into that etherpad | 22:33 |
| dhellmann | and fwiw, I'm trying to come up with steps to let us just turn it off completely, not turn it off one project at a time | 22:34 |
| prometheanfire | I'm not sure I think projects will actually update their requirements files | 22:34 |
| prometheanfire | what happens if everyone just opts out? | 22:34 |
| dhellmann | what ensures that anyone updates the minimums in g-r today? | 22:34 |
| prometheanfire | how is this diferent than projects just not syncing at all? | 22:34 |
| prometheanfire | nothing | 22:35 |
| dhellmann | so we're no worse off | 22:35 |
| dhellmann | this is different from not syncing at all because projects still ensure that all of their deps are in g-r and that they work with the u-c version | 22:35 |
| prometheanfire | true | 22:35 |
| dhellmann | so we still maintain the co-installability feature | 22:35 |
| prometheanfire | just seems like a step back | 22:35 |
| dhellmann | which was one of the main purposes of the g-r list in the first place | 22:36 |
| prometheanfire | sure | 22:36 |
| dhellmann | how is it a "step back" | 22:36 |
| dhellmann | ? | 22:36 |
| dhellmann | that feels like you see a negative outcome that I may not be seeing | 22:37 |
| prometheanfire | I'm not sure I actually do | 22:37 |
| prometheanfire | s/seems/feels | 22:37 |
| dhellmann | it definitely loosens the reins | 22:38 |
| dhellmann | but we have the same benefit because we still have that constraints list | 22:38 |
| dhellmann | syncing made sense before we had constraints, but now that pip supports constraints we don't need to be so tightly in sync | 22:38 |
| prometheanfire | one of the things I wanted before marking a project as 'self managed' is that they have something to ensure that they are looking at their requirements and updating them as needed | 22:39 |
| prometheanfire | I get that we don't really lose anything | 22:39 |
| prometheanfire | sure | 22:39 |
| dhellmann | I agree having something like that would be nice. | 22:39 |
| prometheanfire | that's what the lower-constraints testing was meant to do | 22:40 |
| dhellmann | sure, that was per-project though | 22:40 |
| prometheanfire | yep | 22:40 |
| *** andreas_s has joined #openstack-requirements | 22:41 | |
| prometheanfire | they only use of the global lower constraints is to fill in the blanks of what gets generated per project (since deps of requirements are not tracked) | 22:41 |
| prometheanfire | it could be fully self managed as well though, as dirk is doing/did for the requirements project | 22:42 |
| dhellmann | ah, so the global list will be used to produce the per-project lists? | 22:42 |
| dhellmann | at least the initial versions? | 22:43 |
| prometheanfire | fill in the blanks | 22:43 |
| prometheanfire | projects would have their first pass be is defined in their requirements file | 22:43 |
| dhellmann | I guess I'm lazy, I was just going to populate it with the lower bounds listed in their requirements files and let people add older versions of other things as they felt the need | 22:43 |
| prometheanfire | that's probably better, allows projects to self manage/override as well | 22:44 |
| prometheanfire | and self document | 22:44 |
| dhellmann | but again, I'm way less interested in that lower bounds testing | 22:44 |
| dhellmann | not all projects need to do it | 22:44 |
| prometheanfire | yep, I was only wanting it for projects self managing reqs | 22:44 |
| dhellmann | because we already tell distributors to look at the u-c list | 22:44 |
| dhellmann | so only projects that want to be installable via pip on their own will actually benefit from it | 22:44 |
| dhellmann | or projects that are being packaged on their own somehow outside of the major os vendors | 22:45 |
| dhellmann | and that list of projects is pretty small | 22:45 |
| prometheanfire | aparently rdo doesn't follow uc, not sure lc would help though | 22:45 |
| prometheanfire | true | 22:45 |
| dhellmann | yeah, if they're not following it they're going to have to do their own testing anyway | 22:45 |
| prometheanfire | yep | 22:45 |
| dhellmann | by "projects self managing reqs" do you mean that are not validating against g-r at all? | 22:46 |
| dirk | dhellmann: not sure it is actually true that vendors follow uc. they use it as inspiration, but it isn't always followed 100$ | 22:46 |
| dirk | 100% | 22:46 |
| *** andreas_s has quit IRC | 22:46 | |
| prometheanfire | no, I mean they still validate | 22:46 |
| * prometheanfire needs to find a term for this inbetween | 22:46 | |
| dhellmann | dirk : true, but we *tell* them to look at that list. We give them 1 set of versions we know works. If they don't like those versions, testing is up to them. | 22:46 |
| dirk | dhellmann: I think upstream could be more collaborative than "either this way or the high way" | 22:47 |
| dhellmann | prometheanfire : my goal is to have projects that validate against g-r and u-c, and projects that ignore the requirements repo entirely | 22:47 |
| dirk | dhellmann: did you intentionally add the rocky plan to the ocata etherpad? feels weird | 22:47 |
| *** jrist has quit IRC | 22:47 | |
| * prometheanfire wasn't going to say anything | 22:48 | |
| dhellmann | dirk : sure. I'm not saying we won't. I'm just saying I'm not personally interested in implementing all of that because I think the gain is very small. So if someone else wants to implement it, I support them doing that work. | 22:48 |
| *** jrist has joined #openstack-requirements | 22:48 | |
| prometheanfire | dhellmann: yes, that is the min neede to get this working | 22:48 |
| dhellmann | yeah, I was trying to keep all of the notes in one place | 22:48 |
| dirk | prometheanfire: sorry, I'm not good at knowing when to shut up ;-) | 22:48 |
| prometheanfire | dhellmann: I wanted to tie it to lower bounds testing because I just don't trust projects to self manage I guess :P | 22:49 |
| dhellmann | I shouldn't have named that etherpad with "ocata" in the title to begin with | 22:49 |
| prometheanfire | dhellmann: we had a separate etherpad for this, but that was moved into the bug | 22:49 |
| dhellmann | prometheanfire : like I said, they have nothing in place to cause them to do it now. | 22:49 |
| prometheanfire | true | 22:49 |
| dirk | dhellmann: I would actually say that the uc-working-for-all-projects-at-any-point-in-time is actually a myth more than reality to some extend | 22:49 |
| prometheanfire | it's prefrence to want to move to something tighter than to the same | 22:50 |
| dhellmann | dirk : why? because our test coverage isn't high enough? | 22:50 |
| prometheanfire | it's a weaker prefrence than an hour ago though | 22:50 |
| dirk | dhellmann: its only really tested by the "devstack integration" job, and that one uses pip for many deps (aka releases aka snapshots of a project from the past) | 22:50 |
| prometheanfire | dirk: that's what the freeze period is for though | 22:50 |
| dirk | at any point in time there are always some projects gating job broken by some uc chnage samewhere, and some things already more ahead and some more behind | 22:50 |
| dhellmann | isn't that pip call using the constraints list, though? isn't that exactly the point of having that list? | 22:50 |
| dhellmann | ah | 22:51 |
| dirk | dhellmann: sure, but the factor is time: releases happen in the past, uc is independent of that | 22:51 |
| dirk | it is a bit synced via a freeze period as prometheanfire mentioned, and it certainly is not a big issue, but to some extend there is always a slight mismatch | 22:51 |
| dhellmann | yes, I see what you mean, there is some slack in the system | 22:52 |
| prometheanfire | technically constraints/reqs support defining git urls, but we shouldn't do that | 22:52 |
| dirk | dhellmann: it certainly would become better if the "coinstallable set" is more towards the lower bounds than the upper bounds like it is now | 22:53 |
| dirk | especially if there is some testing on the project-minimums being "correct" | 22:54 |
| dhellmann | dirk : I can't imagine us saying that we need to maintain coinstallability for anything other than the upper-bounds of our own releases, though | 22:54 |
| dhellmann | we're not going to say that projects need to work with old versions of oslo libraries that do not include the features they need | 22:54 |
| dhellmann | we could technically have a different constraint list for every distro if we wanted to | 22:55 |
| dhellmann | where the distros specified the versions of the non-openstack packages they ship and we keep it up to date with the versions of the things we're releasing | 22:55 |
| dhellmann | the problem is we can't gate on that, because we try to use the requirements list to push distros to adopt the new versions of non-openstack things that we require for new features | 22:56 |
| prometheanfire | boil the oceans | 22:56 |
| dhellmann | and as soon as we gate on something they won't change, we stop being able to advance | 22:56 |
| dhellmann | I'm just trying to fill a tea kettle over here | 22:57 |
| dirk | dhellmann: not saying that projects should do anything other than now. if they use a feature of an oslo library, they should jus trequire it | 22:58 |
| prometheanfire | I think I'll have some tea tomorrow for my fix | 22:58 |
| * prometheanfire likes rusts dep model | 22:58 | |
| dirk | dhellmann: I certainly don't want to encourage to remain on old stuff, I just want to have lower bounds be correct in the sense of reflecting what the project actually requires (and yes that can be newer and older than what it could get via g-r) | 22:59 |
| * dirk needs to get some sleep before the meeting tomorrow though :) | 22:59 | |
| dhellmann | dirk : I do support having testing for lower bounds, just not enough to implement it myself. :-) | 22:59 |
| dirk | I'll try to check the etherpad tomorrow morning, but I'll likely not be able before the meeting starts | 22:59 |
| dirk | dhellmann: fine ;) | 22:59 |
| dirk | I actually carea bout that part quite a bit, so we can collaborate | 23:00 |
| dirk | anyway, not today anymore. | 23:00 |
| dirk | afk | 23:00 |
| prometheanfire | cya | 23:00 |
| prometheanfire | nn | 23:00 |
| * dhellmann should sign off for the evening, too | 23:00 | |
| prometheanfire | dhellmann: not at meeting tonight (just want to know if we should wait if you are not around) | 23:01 |
| dhellmann | what time was that agin? | 23:01 |
| prometheanfire | 1am | 23:01 |
| prometheanfire | central | 23:01 |
| dhellmann | what tz? | 23:01 |
| prometheanfire | -6 | 23:01 |
| dhellmann | ok, that's 2am here in EST? I think I'm very unlikely to be coherent at that time tonight given my day. :-( | 23:02 |
| prometheanfire | k | 23:02 |
| prometheanfire | np, catch you later | 23:02 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!