| *** rlandy is now known as rlandy|out | 00:28 | |
| *** dviroel|rover|out is now known as dviroel|rover | 01:12 | |
| *** dviroel|rover is now known as dviroel|rover|out | 01:36 | |
| *** blarnath is now known as d34dh0r53 | 03:37 | |
| *** blarnath is now known as d34dh0r53 | 03:52 | |
| *** yadnesh|away is now known as yadnesh | 04:47 | |
| *** akekane is now known as abhishekk | 05:09 | |
| *** jpena|off is now known as jpena | 08:27 | |
| opendevreview | Merged openstack/project-config master: Flip jeepyb over to building Gerrit 3.6 https://review.opendev.org/c/openstack/project-config/+/867314 | 09:41 |
|---|---|---|
| *** rlandy|out is now known as rlandy | 11:17 | |
| *** dviroel|rover|out is now known as dviroel|rover | 11:18 | |
| *** frenzy_friday is now known as frenzy_friday|doc | 11:34 | |
| *** sfinucan is now known as stephenfin | 12:04 | |
| *** yadnesh is now known as yadnesh|away | 13:34 | |
| *** dasm|off is now known as dasm | 14:00 | |
| *** dulek_ is now known as dulek | 14:06 | |
| *** akekane is now known as abhishekk | 14:10 | |
| *** frenzy_friday|doc is now known as frenzy_friday | 14:57 | |
| sean-k-mooney | i assume a new verion fo gerrit rolled out recently | 15:03 |
| fungi | sean-k-mooney: yes, it was announced maintenance for 20:00 utc yesterday | 15:04 |
| fungi | upgraded from gerrit 3.5.x to latest 3.6.x | 15:05 |
| sean-k-mooney | that explaine the ugly +/- votes | 15:05 |
| sean-k-mooney | the grouping of the +1becide codereview and addign it to the reviews names at the top is ux regression | 15:07 |
| sean-k-mooney | i assume like the comments propagation that this is not configurable | 15:07 |
| clarkb | it isn't as far as I know | 15:09 |
| sean-k-mooney | ok that kind of sucks it makes it less clear at a glance if people i trust have reviewd a patch | 15:11 |
| clarkb | there was upstream discussion about it briefly not too long ago but I didn't follow it too closely. As I've mentioned in #opendev yesterday we can't really drive customization of Gerrit like we did in the past. It was the major reason for us being stuck on ancient Gerrit for so long. If we want to improve things we need to do that upstream first. The good news is upstream is | 15:11 |
| clarkb | receptive to the feedbacka nd if people write changes (of which I've written a few since we got to 3.2) they are often landed pretty quickly | 15:11 |
| clarkb | sean-k-mooney: you can hover the votes to see the details | 15:11 |
| clarkb | and the comments list at the bottom of the page still lists them all when they were set | 15:11 |
| sean-k-mooney | yep i know thats still a regression in ux | 15:11 |
| sean-k-mooney | ill adapt but its just kind of annoying | 15:12 |
| clarkb | sure, I'm just making sure people are aware that the info is still a vailable as that wasn't clear to some people yesterday | 15:12 |
| clarkb | it is a regression in useability not functionality | 15:12 |
| clarkb | if that makes sense | 15:12 |
| sean-k-mooney | its also in the reviers section in the chagne info | 15:12 |
| sean-k-mooney | but that just more noisy | 15:12 |
| sean-k-mooney | oh ya the info is still there its just presented in a less clear way and form a usablity point of view hover based info is generally not good for acciblity in general | 15:13 |
| fungi | more generally, i do wish developers responsible for user interfaces realized that every minor "improvement" to a ui comes as significant cost to users whose workflows have vitrified around the extant state of that ui | 15:14 |
| sean-k-mooney | they did add a tickbox to disable all keyboard shortcuts | 15:15 |
| clarkb | part of the struggle here I think is that the majority of UX work is driven by google. This is beacuse they have two large gerrit installs (chrome and android) informing their needs and they are willing to put devs on it | 15:15 |
| sean-k-mooney | thats an improvement | 15:15 |
| fungi | and so "minor improvements" to a ui are generally a bad idea. making a noticeable change to a ui really needs substantial justification in terms of benefits that outweigh the substantial cost to users | 15:15 |
| clarkb | And until anyone else provides feedback thats the direction it goes | 15:15 |
| clarkb | sean-k-mooney: just 10 minutes ago thunderbird did things I didn't understand because I hit some keys when I thought focus was elsewhere. I didn't realize gerrit had that option now thats a good one | 15:16 |
| fungi | well, ui developers really don't like to hear "please stop changing things" but for the most part users like software interfaces to remain stable and not constantly change | 15:16 |
| clarkb | also gerrit highlights the current line in the diff view a bit better which makes it easier to see what your context is if using keyboard shortcuts to control stuff | 15:16 |
| sean-k-mooney | ctrl-a ctrl-d in screen is detach | 15:16 |
| sean-k-mooney | in outlook its select all email and delete them | 15:16 |
| sean-k-mooney | clarkb: actully i was looking for how to turn that off becasue it also kep the selection box when you change files | 15:17 |
| fungi | i also really wish "web applications" didn't implement keybindings, or that browsers made it easy to completely disable them | 15:17 |
| fungi | i have my own keybindings already, and don't want to have to deal with every single application developer's opinions on what my keypresses should mean | 15:18 |
| sean-k-mooney | i dont actuly mind some of them but i do not use vim and i hate when vim keybinding are turned on on such apps and cant be trunned off | 15:18 |
| clarkb | sean-k-mooney: I don't think it is a selection box its just showing the current line. In the past there was a tiny cursor and that was often overlooked | 15:18 |
| sean-k-mooney | its a box around the current line ya | 15:18 |
| fungi | i use a browser extension which implements vim-like keybindings for navigation and control | 15:18 |
| sean-k-mooney | i was looking to see if i could revert it back | 15:18 |
| fungi | because i hate using a pointer any more than i can help | 15:19 |
| sean-k-mooney | ya so we both like non configurable keyboard shortcuts for different reasons | 15:20 |
| sean-k-mooney | i dont object to them but i generally want to turn most of them off | 15:20 |
| clarkb | anyway upstream gerrit is receptive to ui feedback. They don't necessarily make every possible change based on the feedback. They are also receptive to changes. Now that we are running an almost completely up to date Gerrit we have the ability to push patches upstream first and deploy them when they land rathe than forking and getting stuck forever | 15:20 |
| sean-k-mooney | do you know how i would file a bug or provide that feedback | 15:21 |
| sean-k-mooney | im happy they added "diable all keyboard shortcuts" and "diable token highlighting on hover" | 15:21 |
| clarkb | yes, they use chrome's bug tracker, https://bugs.chromium.org/p/gerrit/issues/list?can=2, anyone with a google/gmail/whatever account can file there aiui. | 15:22 |
| clarkb | I'm happy to act as proxy for some of those, usually when they are less opinion based and objectively a problem | 15:22 |
| clarkb | its better for things with more nuance and opinion to have actual reporters involved to help capture those details | 15:23 |
| clarkb | the other option is to send email to their mailing list to generate discussion/consensus around general direction | 15:23 |
| clarkb | that might be good for higher level things like "don't make seemingly small changes every release as it impacts user workflows" and "Think about accessability with things like hovering for critical info" | 15:23 |
| sean-k-mooney | what ill proably do is reflect on it over the break and see how i feel in the new year | 15:24 |
| clarkb | in thunderbird 'p' jumps to the first email in a folder and 't' jumps to the next unread message? | 15:25 |
| sean-k-mooney | i would have expecte p to be previous if anything | 15:25 |
| sean-k-mooney | i used to use tunderbird as my primary emial client for years but have been using evolution since i goined redhat since we are usign google mail accounts | 15:26 |
| fungi | example of my growing hate for keybindings in webapps. i have / configured to search for text in page but sites like github overload that and jump the cursor to their search input form. firefox also understands ctrl-f for search in page, but e.g. gitlab has mapped that to some scrolling feature and also overloads /, so my only option there is to click the drop-down menu in the browser | 15:29 |
| fungi | and choose search in page that way | 15:29 |
| fungi | something as simple as searching for a pattern in page content or scrolling around a page shouldn't be up to the whims of every single site to decide for themselves | 15:30 |
| fungi | which is why i still just use console-based browsers a lot of the time, when i can get away with it | 15:31 |
| sean-k-mooney | ctrl-f is the default in most things for seach | 15:31 |
| fungi | well, in windows and in software which copied windows norms | 15:31 |
| fungi | like f5 for refresh | 15:31 |
| fungi | or ctrl-c/ctrl-v for cut and paste | 15:31 |
| fungi | i didn't have those for years in x11, so am still not accustomed to them even today | 15:32 |
| fungi | but it seems lots of application developers have emulated microsoft's keybindings in order to ease transition for former windows users | 15:33 |
| sean-k-mooney | i started using ctrl-insert and shift-insert for copy paste when i found out that worka across windows linux and mac pretty much everywhere | 15:33 |
| sean-k-mooney | then i found out there was no insert key on a mac keyboard... | 15:33 |
| fungi | i still miss the sun workstation keyboards i used for many years | 15:34 |
| fungi | where's my stop key? | 15:34 |
| clarkb | sun type 7 unix keyboards had their super fans when I was at university | 15:35 |
| sean-k-mooney | i found out year later that the handme down pc my uncel gave me when i was 6/7 was a second or third generation ibm pc | 15:35 |
| sean-k-mooney | i was sad that i only knew that after it had been dumpted | 15:36 |
| clarkb | I'm typing on a model m that is older than some of my siblings :) | 15:36 |
| fungi | my early computers (not counting the altair and heathkit assemblies) were tandy/radio shack | 15:36 |
| sean-k-mooney | i almost had an operturnity to learn about non windows pc when i was young but we only had that for 6 months before i got a windows 3.1 systme also being an handoff form a university | 15:37 |
| fungi | the first computer i had with a full text display with an original model trs-80 (before they started giving them model designations) | 15:37 |
| sean-k-mooney | for what its worth other then the ux thing the new version is working fine for me | 15:47 |
| clarkb | ya 3.5 -> 3.6 is a fairly small delta I think. Especially when you consider that 3.5 eventually got its jgit and sshd versions updated | 15:50 |
| clarkb | 3.7 is a much bigger gap and we'll likely need to spend a bit of time working out that upgrade | 15:50 |
| *** JasonF is now known as JayF | 15:53 | |
| paladox | Underhood there are big changes - more components in the UI use Lit (migrating from polymer). | 15:54 |
| *** dviroel|rover is now known as dviroel|rover|lunch | 15:57 | |
| opendevreview | Merged openstack/project-config master: Scale iweb back to 25% of possible quota https://review.opendev.org/c/openstack/project-config/+/867261 | 16:13 |
| *** dviroel|rover|lunch is now known as dviroel|rover | 16:50 | |
| opendevreview | Ade Lee proposed openstack/project-config master: Add FIPS job for ubuntu https://review.opendev.org/c/openstack/project-config/+/867112 | 16:57 |
| opendevreview | Ade Lee proposed openstack/project-config master: Add FIPS job for ubuntu https://review.opendev.org/c/openstack/project-config/+/867112 | 17:15 |
| ade_lee__ | clarkb, fungi please take a look at https://review.opendev.org/c/openstack/project-config/+/867112 and https://review.opendev.org/c/zuul/zuul-jobs/+/866881 | 17:17 |
| ade_lee__ | of course we'll the first to be merged before the other will pass - but you'll get a sense of the direction | 17:17 |
| fungi | ralonsoh: on the tox v4 failures, it looks like neutron made their own tox plugin? those have to be rewritten for v4 since the plugin architecture has completely changed | 17:32 |
| fungi | ralonsoh: also even if the default changes in ensure-tox, you can still override it in your jobs to tox<4 in order to stick with the last 3.x release | 17:32 |
| ralonsoh | fungi, sorry, I need help on this: where in the Neutron CI are we implementing our own tox plugin? | 17:33 |
| ralonsoh | can you help me on this? | 17:33 |
| fungi | ralonsoh: oh, actually it's a flake8 plugin sorry | 17:35 |
| fungi | not a tox plugin | 17:35 |
| fungi | so looks like the error is about neutron not being installed into the pep8 testenv | 17:36 |
| ralonsoh | seems that yes, because stevedore is not finding the modules | 17:36 |
| *** jpena is now known as jpena|off | 17:36 | |
| fungi | looks like usedevelop is true in the base testenv, so in theory it should be getting installed into all the ones that inherit from it (i don't see it overridden in any of them) | 17:40 |
| ralonsoh | yes, we set it to True in the default env | 17:41 |
| fungi | ralonsoh: i wonder if it needs editable-legacy | 17:41 |
| fungi | https://tox.wiki/en/latest/faq.html#tox-4-packaging-changes | 17:41 |
| fungi | "The --develop CLI flag or the use_develop settings now enables editable installations via the PEP 660 mechanism rather than the legacy pip install -e behaviour. The old functionality can still be forced by setting the package setting for the run environment to editable-legacy." | 17:42 |
| fungi | can't tell if that means that usedevelop will now only work for projects with a pyproject.toml by default | 17:43 |
| ralonsoh | fungi, where the project (for example in py38) is being installed? | 17:44 |
| ralonsoh | I'll check other projects | 17:44 |
| fungi | i'm trying to work that out from the build logs myself, yeah | 17:47 |
| fungi | ralonsoh: so here it is happening in tox v3: https://zuul.opendev.org/t/openstack/build/ea7e9b2cd36749a3848b3fd37fd8678a/log/job-output.txt#1236-1239 | 17:52 |
| fungi | i don't see anything similar in the failing https://zuul.opendev.org/t/openstack/build/192d67d4367a46da828acb22de91f9a0/log/job-output.txt | 17:53 |
| fungi | like, no indication that it did a develop (editable) install at all | 17:53 |
| ralonsoh | no, I see this | 17:54 |
| ralonsoh | 2022-12-13 17:12:52.805033 | ubuntu-focal | Installing collected packages: packaging, pluggy, chardet, platformdirs, cachetools, tomli, pyproject-api, distlib, filelock, virtualenv, colorama, tox | 17:54 |
| ralonsoh | but nothing related to Neutron | 17:54 |
| fungi | if it happened like with v3 i'd expect to see something about it right after https://zuul.opendev.org/t/openstack/build/192d67d4367a46da828acb22de91f9a0/log/job-output.txt#1371 | 17:55 |
| fungi | clarkb: ^ have you observed that on any other tox v4 builds (editable installs simply not happening at all) | 17:55 |
| clarkb | fungi: usedevelop isn't a thing anymore I think? | 17:55 |
| clarkb | if you do a showconfig it will show things it is ignoring commented out | 17:55 |
| clarkb | I would start there and correct any of those items, and then debug any behavioral issues | 17:56 |
| fungi | clarkb: the breaking changes faq says it is but uses pep 660 methods (which might mean it only happens if there's a pyproject.toml) and suggests a editable-legacy option if you don't want that | 17:56 |
| fungi | oh, or maybe i'm misreading that faq entry | 17:57 |
| fungi | it talks about "the use_develop settings" | 17:58 |
| clarkb | what target is being run? | 17:58 |
| fungi | oh! there's an underscore now? | 17:58 |
| clarkb | # !!! unused: usedevelop is what showconfig shows by default | 17:58 |
| fungi | maybe usedevelop became use_develop? | 17:58 |
| clarkb | so it is definitely ignoring that | 17:58 |
| fungi | neat! | 17:58 |
| fungi | ralonsoh: ^ smoking gun | 17:58 |
| clarkb | but also you need to show config with the exact env you want to run against | 17:59 |
| ralonsoh | ok I'll try in this patch "use_develop" | 17:59 |
| clarkb | ralonsoh: you should run it locally | 17:59 |
| ralonsoh | ok | 17:59 |
| clarkb | I found that really helped in understanding this stuff | 17:59 |
| fungi | the example in this case was the pep8 target for openstack/neutron but it seems likely that not having neutron installed in any testenvs is going to also be at least part of the cause for unit test failures too | 18:00 |
| fungi | and now i have another meeting to get to | 18:00 |
| ralonsoh | clarkb, fungi thanks! I'll check it locally | 18:00 |
| clarkb | ya so if you've disabled regular install and use develop isn't set then it ma just not install your package | 18:00 |
| fungi | and yes, i have a feeling there are *tons* of projects relying on usedevelop=true in order to install their projects in at least some testenvs | 18:03 |
| fungi | the current tox docs still claim usedevelop is valid? https://tox.wiki/en/latest/config.html#usedevelop | 18:18 |
| fungi | how odd | 18:18 |
| clarkb | may be a bug | 18:18 |
| clarkb | showconfig definitely complains about it | 18:26 |
| clarkb | # !!! unused: use_develop it complains about this one too | 18:27 |
| fungi | how odd | 18:27 |
| fungi | tox itself has usedevelop true in its own tox.ini file, and v4 is able to parse it there for some reason | 18:36 |
| fungi | $ tox --showconfig -e dev|grep develop | 18:36 |
| fungi | use_develop = True | 18:36 |
| fungi | so maybe something else is conflicting with it in neutron | 18:37 |
| fungi | though tox also has a pyproject.toml | 18:37 |
| clarkb | the docs talk about not using it with install command | 18:43 |
| clarkb | maybe that is the problem? | 18:43 |
| fungi | yeah, i was wondering about that possibility myself | 18:45 |
| fungi | though it doesn't seem like neutron is doing that in their tox.ini file | 18:52 |
| fungi | looks like editable-legacy is only used with pyproject.toml using projects, or at least it's only checked in src/tox/tox_env/python/virtual_env/package/pyproject.py | 18:56 |
| *** dasm is now known as dasm|off | 19:26 | |
| clarkb | ralonsoh isn't here anymore... | 21:35 |
| clarkb | anyway I did more digging because nova actually appears to be tox v4 compat but does use usedevelop | 21:36 |
| clarkb | in comparing the difference between nova's tox.ini and neutron's I've discovered that nova does not set skipsdist = True. COmmenting that out of neutron's tox.ini fixes the showconfig comment complaint about usedevelop | 21:36 |
| JayF | oh, that makes sense | 21:52 |
| clarkb | I pushed a new ps to https://review.opendev.org/867048 in order to test this more fully | 21:57 |
| *** dviroel|rover is now known as dviroel|out | 21:58 | |
| opendevreview | Merged openstack/project-config master: Ensure NetworkManager doesn't override /etc/resolv.conf https://review.opendev.org/c/openstack/project-config/+/866475 | 21:59 |
| clarkb | I think it is actually running now so some sort of package install expectations around usedevelop and skipsdist (I don't understand why they would interfere like this as usedevelop should imply skipsdist?) | 21:59 |
| clarkb | yup pep8 is successful now after removing skipsdist | 22:03 |
| fungi | not sure if JayF was being ironic, but it makes zero sensions to me that turning on skipsdist would make usedevelop invalid now | 22:26 |
| JayF | I viewed them as like, mutually exclusive options | 22:28 |
| JayF | either "don't build an sdist" or "install in develop mode" | 22:28 |
| JayF | you're right in that skipsdist=true is the behavior you get in usedevelop=true | 22:28 |
| fungi | they weren't in 3.x but maybe that's how it's different now | 22:28 |
| JayF | I'm saying that's how it's different now based on the observations; they're just considering it syntactically invalid to provide the extra "and no sdist" | 22:29 |
| JayF | because skipsdist=false in that config would be really, really weird, right? | 22:29 |
| JayF | skipsdist=false with usedevelop doesn't even make sense | 22:30 |
| JayF | so why should they allow it to be set to true? | 22:30 |
| * JayF is not saying that's a good decision, or one he would've made, but at least I can reason out why they woulda done it | 22:30 | |
| clarkb | I think skipsdist=True may be a proxy for "never install the project we just want a convenient virtualenv" | 22:30 |
| clarkb | then separately usedevelop is pip install -e instead of pip install an sdist | 22:31 |
| JayF | that's how I've always used it in the past | 22:31 |
| clarkb | skipsdist might be better named skipinstall | 22:31 |
| JayF | clarkb more tersely restated my understanding (me? verbose? never!) | 22:31 |
| fungi | yeah, i had previously treated skipsdist as "don't go to the additional step of building the sdist for no reason" | 22:33 |
| *** rlandy is now known as rlandy|out | 22:43 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!