| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-11-19 | |||
| 14:25:42 | mriedem | as i've said earlier, i would put the bare minimum effort into making nova work with py27, i.e. let it install but after that all bets are off | |
| 14:25:54 | mriedem | if it doesn't work with py27 it's a bug until we can really say we no longer care | |
| 14:26:01 | stephenfin | mriedem: I also like that approach | |
| 14:26:07 | sean-k-mooney | ya same | |
| 14:26:19 | sean-k-mooney | which means remove 1 line in setup.cfg | |
| 14:26:20 | stephenfin | tosky: yeah, if you want to just revert the setup.cfg changes, I'll be happy with that | |
| 14:26:28 | mriedem | stephenfin: and likely the requirements stuff | |
| 14:26:34 | stephenfin | also that, yes | |
| 14:26:46 | sean-k-mooney | the requirement chagne for contextlib2 is only half needed | |
| 14:26:53 | mriedem | there should probably be a TODO in the setup.cfg line saying to remove python-requires once devstack uses py3 by default | |
| 14:27:05 | sean-k-mooney | the way devstack works it will get installed by other projects but sure | |
| 14:27:15 | stephenfin | and that, yup yup yup | |
| 14:27:22 | mriedem | sean-k-mooney: but kolla! | |
| 14:27:26 | mriedem | osa, etc | |
| 14:27:31 | mriedem | anything running nova in a container or venv | |
| 14:27:34 | tosky | stephenfin: but don't you think that also reverting requirements.txt is needed? There is at least one condition which affects py2, and that may break as well | |
| 14:27:44 | mriedem | tosky: we just said yes | |
| 14:27:52 | tosky | mriedem: sorry, missed in the stream | |
| 14:27:56 | stephenfin | tosky: it is. I missed that. Fortunately mriedem caught it (y) | |
| 14:28:03 | sean-k-mooney | kolla is python3 by default alredy but yes fair point | |
| 14:30:32 | mriedem | stephenfin: technically tosky brought up the requirements thing first so credit goes to him for that | |
| 14:30:32 | KeithMnemonic | lyarwood do you think you may have time to review this patch this week? https://review.opendev.org/683008 | |
| 14:32:31 | efried | so is someone going to modify ianw's patch or propose a new one? Probably the latter is better. | |
| 14:33:26 | sean-k-mooney | maybe we should send an update too the list too whith what our plan is | |
| 14:34:18 | openstackgerrit | Luigi Toscano proposed openstack/nova master: (Temporarily) readd bare support for py27 https://review.opendev.org/695007 | |
| 14:34:28 | tosky | let's see if it works ^^ | |
| 14:34:52 | corvus | tosky: test results will not be helpful here -- there is no testing of py27 | |
| 14:35:08 | tosky | corvus: one monent | |
| 14:35:27 | sean-k-mooney | i dont think we should add the py 27 calssifier back | |
| 14:35:35 | sean-k-mooney | but otherwise it looks fine | |
| 14:35:52 | stephenfin | sean-k-mooney: just left a comment to that effect | |
| 14:36:09 | sean-k-mooney | so did i :) | |
| 14:36:42 | openstackgerrit | Luigi Toscano proposed openstack/nova master: (Temporarily) readd bare support for py27 https://review.opendev.org/695007 | |
| 14:37:15 | tosky | I'm using this: https://review.opendev.org/694949 | |
| 14:37:17 | tosky | to test it | |
| 14:37:57 | stephenfin | tosky: one more comment, sorry /o\ (the Python 3 Only classifier still makes sense) | |
| 14:38:07 | tosky | stephenfin: oh | |
| 14:38:16 | stephenfin | it's just metadata so you can keep it | |
| 14:38:19 | sean-k-mooney | as does the cpython on | |
| 14:38:22 | stephenfin | yup | |
| 14:38:40 | stephenfin | after that, I'm +2 pending https://review.opendev.org/694949 | |
| 14:38:46 | openstackgerrit | Luigi Toscano proposed openstack/nova master: (Temporarily) readd bare support for py27 https://review.opendev.org/695007 | |
| 14:40:25 | corvus | i'd really love for there to be a job which tests py2.7. can we not just add one until the default switches? | |
| 14:41:13 | corvus | testing with depends-on changes seems error-prone | |
| 14:41:37 | tosky | a non-voting one? A voting one? | |
| 14:41:44 | tosky | I'd defer that to nova cores | |
| 14:42:38 | sean-k-mooney | the whole point of the orginal change was nova wanted to drop all python 2 support and therefor the testing | |
| 14:42:45 | sean-k-mooney | but yes we could | |
| 14:43:13 | sean-k-mooney | gibi: im reviewing your comments on the composable flavors spec | |
| 14:43:27 | sean-k-mooney | gibi: so far i more or less agree | |
| 14:43:38 | stephenfin | if we were to add one, it'd have to be voting for it to be of any use but per above, I don't think we should add one | |
| 14:43:55 | corvus | sean-k-mooney: right, i'm just suggesting that given the importance of devstack, that we should endeavor to keep it working | |
| 14:44:08 | dansmith | I'm coming into this late, but what is the controversy? | |
| 14:44:27 | dansmith | whether or not we continue to work on py2 at all via a test? | |
| 14:44:36 | stephenfin | dansmith: we've marked nova as python3 only, but devstack defaults to installing stuff in python2 mode | |
| 14:44:43 | corvus | stephenfin: i'm really puzzled -- i don't understand why there is so little concern about whether devstack works | |
| 14:44:46 | stephenfin | so anyone that doesn't override the defaults is broken | |
| 14:44:57 | dansmith | so how is this even a question? | |
| 14:45:19 | stephenfin | dansmith: It's not, we're reverting the part that broke everyone | |
| 14:45:43 | dansmith | reverting what? the dropping of a py2 job? | |
| 14:45:49 | corvus | but there are no current plans to have tests to ensure that it doesn't happen again (and even the revert is only tested indirectly with a depends-on patch) | |
| 14:45:54 | stephenfin | dansmith: https://review.opendev.org/#/c/695007/ | |
| 14:46:24 | dansmith | corvus: yeah, then you and I agree. | |
| 14:46:34 | sean-k-mooney | dansmith: we had a require-pyton >= 3.6 line in setup config that prevented nova installing on less then our new supported version | |
| 14:46:37 | dansmith | what's the argument for not having a py2 test until the devstack default switches? principle? | |
| 14:47:00 | stephenfin | mostly, yeah | |
| 14:47:04 | corvus | dansmith: https://review.opendev.org/694891 is the full revert which seems to be losing mindshare | |
| 14:47:19 | dansmith | stephenfin: I'm not on board with that | |
| 14:47:51 | sean-k-mooney | dansmith: that beeing testing py2 out of principal? | |
| 14:47:58 | sean-k-mooney | or breaking the devstack default | |
| 14:48:07 | sean-k-mooney | because im really hoping we change that today | |
| 14:48:45 | dansmith | sean-k-mooney: I'm not on board with us not testing nova in the default devstack config, and not okay with *not* doing that because of the principle that py2 is (soon) out of support | |
| 14:48:46 | dansmith | if we change the devstack default, then fine not testing py2 | |
| 14:48:51 | dansmith | if we don't, then we *have* to, IMHO | |
| 14:49:22 | sean-k-mooney | ok that is a clear statement | |
| 14:50:48 | dansmith | what is the blocker for the devstack py3 default? | |
| 14:50:50 | sean-k-mooney | the devstack defautl in that case need to be change in the next 23 days or we cant drop py2 testing by m1 in nova per the tc plan | |
| 14:51:03 | stephenfin | I figured someone needs to lead the way on this and it didn't seem anyone else was doing it. I also don't get why DevStack can't just change their default | |
| 14:51:27 | dansmith | sean-k-mooney: I totes don't care about the tc's schedule as much as I care that devstack works out of the box | |
| 14:51:44 | dansmith | (I don't care about the TC's schedule at all actually, but.. :P) | |
| 14:51:58 | dansmith | I know mriedem was on top of the devstack py3 default, | |
| 14:52:19 | dansmith | but he approved the switch prior to that being ready, which surprises me | |
| 14:53:06 | mriedem | i didn't know that the nova change meant you couldn't install nova on py2 anymore | |
| 14:53:23 | mriedem | i thought it was (a) what we tested against and (b) classifiers / support statement | |
| 14:53:27 | dansmith | ack | |
| 14:54:05 | gmann | mriedem: stephenfin dansmith tosky removing condition from setup.cfg should be ok as and add that once all projects moved to py3-only. | |
| 14:54:14 | dansmith | mriedem: from the two patches and the votes, it looks like you want to re-allow install on py2, but not test it, yeah? problem is "it" is the "default devstack arrangement" right? | |
| 14:54:15 | stephenfin | and I didn't know that this would break other people. I thought we were safe since we were consumers (a service), rather than the consumed (libs, clients). Didn't think of the DevStack implications | |
| 14:54:23 | gmann | if nova is breaking on py2 by code then we do not need to fix and ask consumer to move to py3-only | |
| 14:54:53 | dansmith | gmann: it's a matter of devstack working out of the box.. nothing to do with python versions, IMHO | |
| 14:55:07 | mriedem | dansmith: correct | |
| 14:55:08 | dansmith | if we're not testing devstack in remotely default form, that's a problem, IMHO | |
| 14:55:11 | gmann | my argument for devstack move to py3 by default can migrate all py2 jobs to py3 which is planned before schedule, like library etc | |
| 14:55:23 | mriedem | dansmith: i want more people to give a shit | |
| 14:55:29 | mriedem | and pain is usually how that happens | |
| 14:55:42 | stephenfin | ^ that, sadly | |
| 14:55:49 | dansmith | mriedem: so you want it to break to prove a point? :) | |
| 14:56:12 | mriedem | that's assuming we're going to land nova changes that break on py2 in the next what, 3-4 weeks? | |
| 14:56:15 | efried | Does anybody object to | |
| 14:56:15 | efried | - Re-enable py2 (revert setup.cfg and requirements changes) | |