Thursday, 2022-12-15

opendevreviewMerged openstack/project-config master: Update post-review zuul pipeline definiton
opendevreviewIan Wienand proposed openstack/diskimage-builder master: tox jobs: pin to correct nodesets; use host networking for containerfile
opendevreviewIan Wienand proposed openstack/diskimage-builder master: tox jobs: pin to correct nodesets; use host networking for containerfile
fungilooks like the mm3 list mods let my questions through:
*** rlandy|bbl is now known as rlandy|out02:24
*** ysandeep is now known as ysandeep|afk04:41
*** yadnesh|away is now known as yadnesh05:08
*** soniya29 is now known as soniya29|pto05:57
*** ysandeep|afk is now known as ysandeep06:08
*** jpena|off is now known as jpena07:54
*** ysandeep is now known as ysandeep|lunch09:00
opendevreviewAde Lee proposed openstack/project-config master: Add FIPS job for ubuntu
*** ysandeep|lunch is now known as ysandeep11:01
*** yadnesh is now known as yadnesh|afk11:05
*** dviroel|out is now known as dviroel|rover11:12
*** rlandy|out is now known as rlandy11:14
*** yadnesh|afk is now known as yadnesh12:02
*** gthiemon1e is now known as gthiemonge12:41
opendevreviewThierry Carrez proposed openstack/project-config master: Bring back the PTL+1 column in release dashboard
*** ysandeep is now known as ysandeep|dinner13:29
*** pojadhav is now known as pojadhav|dr_appt13:50
*** ysandeep|dinner is now known as ysandeep13:52
*** dasm|off is now known as dasm14:30
*** frenzy_friday is now known as frenzy_friday|food14:45
*** yadnesh is now known as yadnesh|away15:01
*** ysandeep is now known as ysandeep|afk15:23
clarkb!/story/2010472 I think the issue there is the calling environment's locale is used15:27
clarkbI hesitate to hard code utf8 in git-review since you could be operating on other encodings and using the calling environment locale as it should already do might be necessary for other users?15:28
clarkbfungi: ^ do you have ideas on that?15:28
fungilooks like the reporter did not mention what their locale is, nor does the traceback/exception say what locale it's assuming during the decode15:30
fungithat said, the plan in pep 686 (accepted by the sc) is to make utf-8 the default for everyone in python 3.1515:31
clarkbfungi: the traceback does show windows which is almost certainly not utf8 by default iirc15:32
fungiand in python 3.11 you can set PYTHONUTF8=1 or pass -X utf8=1 to get that behavior early15:32
clarkbI don't know what windows users expect but I expect if you run in a utf16 environment that you should be writing utf16 by default...15:32
*** frenzy_friday|food is now known as frenzy_friday15:33
fungithere's some rationale in that it's safe for windows15:34
fungi"Visual Studio Code and Windows Notepad use UTF-8 by default"15:34
fungi"this change mostly affects Windows users"15:34
opendevreviewMerged openstack/project-config master: Bring back the PTL+1 column in release dashboard
fungi"Ruby changed the default external_encoding to UTF-8 on Windows in Ruby 3.0 (2020)."15:35
clarkbfungi: I'm mostly concerned that if this person uses notepad to write a utf16 file that they will break in the other direction15:35
clarkbI don't know how realistic that is. But I'm kind of thinking we should ask the user to set the locale properly instead of hardcode utf8?15:35
fungiyes, or turn on PYTHONUTF8 in their environment (however windows users do that)15:36
clarkbthe underlying problem here being that the python process is interacting with files generated by git and $EDITOR which are likely to respect locale?15:36
fungioh, actually "utf-8 mode" is available as far back as 3.7 according to pep 540, so they wouldn't have to upgrade the interpreter to test it15:39
fungiand yes, that's git's stdout which git-review is choking on, i think15:40
fungiclarkb: i've added a comment with some suggestions, but also i don't think this is a bug for git-review to solve15:47
clarkbfungi: I guess, we could accept a git-review flag to set locale expectations, but python will already respect the calling environment for tha15:48
fungiright. "fixing" this on the git or python side is the right solution in my opinion, and since the python community is already planning to force utf-8 by default rather than doing locale detection for 3.15 it seems like the sort of thing which users should be trying out to make sure it does solve problems like this15:50
fungibasically the python devs' collective opinion is that if your locale isn't utf-8 then that's probably a mistake. if you really know for sure that you don't want python to ignore your locale settings then you can still flip a switch to turn on the old locale detection/guessing behaviors15:52
*** ysandeep|afk is now known as ysandeep16:06
clarkbinfra-root: I'm thinking today is a good day to start winding down iweb? That way its done before holidays begin in earnest?16:30
clarkbIf I hear no objections I'll approve after breakfast16:30
fungiclarkb: sgtm, i didn't know if those weren't wip for a reason so was hesitant to approve yet16:32
clarkbmostly just wanting to be sure we landed them when we are ready. I think we are there now considering how things slow down for holidays. I've also got enough work to do when things slow down with annual report stuff and so on16:35
fungiyeah, sounds fine. things have been pretty slow from a ci volume perspective for weeks already16:40
clarkbI've approved it and will go find breakfast now16:43
opendevreviewMerged openstack/project-config master: Set iweb max-servers to 0
*** jpena is now known as jpena|off17:16
*** ysandeep is now known as ysandeep|out17:41
*** pojadhav|dr_appt is now known as pojadhav18:01
clarkbiweb utilization has dropped to zero
clarkbI'm going to double check that we don't have any held nodes there since the graph doesn't seem to show held nodes?20:27
clarkbcorvus: ^ there is one node left and it is for a hold you created. Looks like zuul py38 stuff. Should I clean that up or do you want to take a look?20:28
corvusclarkb: i marked it deleting, should be gone soon20:55
clarkbyup its gone, thanks!20:56
clarkbI'm going to approve now to remove iweb diskimages and labels20:56
corvusfungi: i've not been paying attention so i don't know if there are changes, but now shows all the lists, not just zuul lists21:06
opendevreviewMerged openstack/project-config master: Set iweb to empty labels and diskimages
clarkbcorvus that was a whole thing yesterday and the day before21:08
clarkbcorvus: it turns out that when you login as admin toe django that side effects the creation of site records that then impact the vhosting of mailman21:08
corvussorry! i'll try to catch up sometime21:08
clarkbcorvus: on a held test node fungi wasable to reproduce this and fix it for both under postorius and hyperkitty. But we were only able to fix it for under hyperkitty.21:08
fungithat hopefully catches you up21:08
corvusoh okay so this is related to the "link in the top left" thing we talked about at mtg21:09
clarkband ya that email upstream is meant to try and figure out how to fix postorius under then we can make a plan to fix it21:09
fungiright, i pulled on that thread and a lot of stuff unraveled21:09
fungiclarkb: i feel like we have a bit of a blind spot on pbr-related packaging discussions too. this also seems to have cropped up in the wake of tox v4:
fungi"Use of 'pbr' results in "Multiple top-level packages" error and "unkillable" process"21:12
fungii think this is what you were observing21:12
fungistephenfin: ^ we should try to coordinate21:12
corvusfungi: thanks for the link, i feel much more caught up now21:12
clarkbfungi: stephenfin adding the find: directive fixed it. So I don't think thats much of a problem21:12
clarkbfungi: stephenfin: the real issue is that tox v4 appears to be invoking via the pyproject.toml invocation method which is not valid21:13
clarkbwhen you use pbr with pyproject.toml you need to invokce it via the pbr entrypoint that will then invoke the setuptools entrypoint. And importantly build sorts all of this out for you21:13
clarkb is what tox is using21:15
clarkbwhich calls into
clarkbfungi: stephenfin my takeaway from this is that you can only use tox v4 reliably if you are a prproject.toml project21:16
*** dviroel|rover is now known as dviroel|rover|afk21:16
clarkb thats the source of the issue21:17
fungiyes, or at least gabor doesn't understand that the project in question doesn't use pyproject.toml at all21:17
fungii guess build doesn't call directly into setuptools.build_meta:__legacy__21:18
clarkbI think weshould pick a brave soul to write an issue that explains tox doesn't properly support legacy projects anymore and explain our situation as the example21:19
clarkbwe can suggest build as the fallback and have tox continue to do its pyproject.toml things for modern projects21:20
fungii wonder if build doesn't use _DEFAULT_BACKEND under some circumstances21:20
clarkbfungi: I think it may only do that when pyproject.toml exists but doesn't declare the entrypoint21:20
clarkbtox is using it when pyproject.toml does not exist at all which is fundamentally wrong21:21
fungithe docstring here seems to imply they want to return that when there's no pyproject.toml:
* clarkb hacks up the local install of build21:22
opendevreviewJames E. Blair proposed zuul/zuul-jobs master: Add ensure-nox role
fungii need to take a break to cook/eat dinner but shall return21:23
clarkbya I haven't eaten lunch yet, but this too interesting21:24
JayFWhere does "Organizations" information come from in
clarkbJayF: bitergia maintains an affiliations db21:28
clarkbI believe foundation staff can edit it for the openstack instance, but they also use some central content they've collected over time or something21:28
JayFThat sounds a bit ominous21:28
JayFwhere can I look and see how Sauronergia has me categorized? lol21:28
JayFYeah; it has me categorized as "OpenStack Foundation", so I should fix that. Is there a documented process for fixing that?21:31
JayFand also I know several Ironic contributors are misattributed as well21:32
clarkbI'm not sure. I know I can do it for zuul but that is a separate instance. Let me ask21:33
clarkbfungi: stephenfin: I did a lot more digging and both tox and build seem to call the setuptools build meta backend in the same way. At first this had me very stumpted but then I noticed that tox does use a slgiht difference. They run a separate process that processes potnatially many commands to setuptools legacy but build uses a fresh process22:13
clarkbI think that very likely explains why the process isn't exiting (its getting stuck in the loop for that system) and may explain the version issues22:13
ianwsorry running a bit late.  assuming tox has broken everything :)22:15
fungiJayF: clarkb: there's work in progress to collect affiliation information from the openinfra member profiles22:16
fungithat should ideally become the official source of truth in time22:17
JayFIt makes me a little uncomfortable that my work is being misattributed, and that our project is now relying on information presumedly mined from somewhere on the internet.22:17
fungithe idea is that contributors should be able to directly adjust their affiliation info22:18
fungithere is a thread on the foundation mailing list where that work in progress is being discussed (note that the dashboard is incomplete, still being worked on)22:18
JayFI think my concerns are probably unique enough that such participation would be roadblocking without a better solution. I try not to NACK something unless I have a positive alternative to suggest, and in this case I have nothing :)22:19
fungiyeah, it's also not done yet, as i mentioned. work in progress22:27
*** cloudnull1 is now known as cloudnull22:29
fungithat dashboard is mainly only publicly reachable because they don't have a good way to share it with us otherwise, but it's not ready for real use yet, just a preview22:30
JayFHonestly, my concern was more with "it's not very open to use data that was mined/came from ??????" than the current status22:32
JayFI don't love the ability of using data mined using "vendor magic" as part of our open statistics 22:32
JayFbut there are lots of things I don't love and the world keeps spinning :P 22:33
clarkbianw: long story short is in trying to make tox v4 work with zuul I think we may have discovered a more problematic issue with tox v4. sfinucan discovered it independently (and now I realized i used the wrong nick earlier)22:33
clarkbianw: I've updated the sfinucan issue upstream with what I've found:
fungiJayF: that "magic" is mostly guessing based on people's e-mail address domains, as i understand it. inaccurate i agree, which is why we want it to use foundation profiles which are under the direct control of the contributors to adjust as they desire22:37
*** rlandy is now known as rlandy|out22:49
JayFinteresting that it tagged my email as openstack; given I commit from an email domain only used by my immediate family :) maybe a default? 22:51
JayFthat weirds me out a *lot* less though, knowing it's by domain22:51
JayFI had ideas of like, scraping github and linkedin profiles and the like :) 22:51
opendevreviewIan Wienand proposed openstack/diskimage-builder master: tox jobs: pin to correct nodesets; use host networking for containerfile
ianwi forgot to update the gate jobs in ^22:53
*** sfinucan is now known as stephenfin22:55
fungiJayF: i think they normally expect project leaders to go through and assign affiliations to all the contributors, but openstack is just too big for that to be a viable approach, which is why we want them to base it on something self-service from the contributor standpoint22:57
ianwclarkb: thanks, watching the issue with interest, and quite a bit of relief that you seem to have it under control :)23:19
clarkbI dunno if I would call this control...23:20
clarkbI've tried to naively use the existing "SubprocessFrontend" which is supposed to create a new backend per command invocation and the whole thing just hangs running the first command23:20
clarkbit is interesting that they've written a system that can do what I describe though but don't seem to use it...23:20
clarkbI see now that the pep517 frontend is a bit specialized which probably explains why my naive version didn't work23:26
*** dasm is now known as dasm|off23:59

Generated by 2.17.3 by Marius Gedminas - find it at!