| -@gerrit:opendev.org- Tristan Cacqueray https://matrix.to/#/@tristanc_:matrix.org proposed: [zuul/zuul] 926803: Add support for log_url bigger than 255 characters https://review.opendev.org/c/zuul/zuul/+/926803 | 00:16 | |
| -@gerrit:opendev.org- Zuul merged on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul] 926862: Remove docs python version pin https://review.opendev.org/c/zuul/zuul/+/926862 | 03:32 | |
| -@gerrit:opendev.org- Zuul merged on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul] 926865: Fix regex syntax warning in docs/conf.py https://review.opendev.org/c/zuul/zuul/+/926865 | 03:37 | |
| -@gerrit:opendev.org- Simon Westphahl proposed: [zuul/zuul] 926891: Exclude empty change queues from status JSON https://review.opendev.org/c/zuul/zuul/+/926891 | 10:12 | |
| -@gerrit:opendev.org- Damian Fajfer proposed: [zuul/zuul-operator] 919808: Fix schema for env variables https://review.opendev.org/c/zuul/zuul-operator/+/919808 | 12:23 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 924807: Add configure-projects https://review.opendev.org/c/zuul/zuul/+/924807 | 14:52 | |
| -@gerrit:opendev.org- Zuul merged on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul-jobs] 925539: Update test-prepare-workspace-git to use a module https://review.opendev.org/c/zuul/zuul-jobs/+/925539 | 15:06 | |
| @tristanc_:matrix.org | Hello folks, not sure what else is needed for 926803, could you please have a look, it's a fix for an obscure bug. Thanks! | 15:07 |
|---|---|---|
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 923084: WIP: Try docker compose with podman https://review.opendev.org/c/zuul/zuul/+/923084 | 15:21 | |
| -@gerrit:opendev.org- Tristan Cacqueray https://matrix.to/#/@tristanc_:matrix.org proposed: [zuul/zuul-jobs] 926931: ensure-docker: extract the restart procedure in a task file https://review.opendev.org/c/zuul/zuul-jobs/+/926931 | 15:40 | |
| -@gerrit:opendev.org- Tristan Cacqueray https://matrix.to/#/@tristanc_:matrix.org proposed: [zuul/zuul-operator] 926932: DNM: testing zuul-jobs https://review.opendev.org/c/zuul/zuul-operator/+/926932 | 15:45 | |
| -@gerrit:opendev.org- Tristan Cacqueray https://matrix.to/#/@tristanc_:matrix.org proposed: [zuul/zuul-jobs] 926931: ensure-docker: extract the restart procedure in a task file https://review.opendev.org/c/zuul/zuul-jobs/+/926931 | 16:01 | |
| -@gerrit:opendev.org- Zuul merged on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul] 924807: Add configure-projects https://review.opendev.org/c/zuul/zuul/+/924807 | 16:09 | |
| -@gerrit:opendev.org- Tristan Cacqueray https://matrix.to/#/@tristanc_:matrix.org proposed: [zuul/zuul-jobs] 926931: ensure-docker: extract the restart procedure in a task file https://review.opendev.org/c/zuul/zuul-jobs/+/926931 | 16:36 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 923084: WIP: Try docker compose with podman https://review.opendev.org/c/zuul/zuul/+/923084 | 17:27 | |
| @jim:acmegating.com | Clark fungi tristanC mhu in the quickstart we tell people to globally pip install git-review; that's a no-go on noble... should we tell them to apt-get/yum install it? https://zuul.opendev.org/t/zuul/build/3157f479353946cdb9b5a3190a6e55e7/console | 18:10 |
| @jim:acmegating.com | basically, i think i'm asking "on modern ubuntu/fedora is the package-managed git-review the best way to go?" | 18:10 |
| @jim:acmegating.com | (or should we update the instructions to install it in a venv?) | 18:11 |
| @clarkb:matrix.org | corvus: I think using a venv is probably ideal | 18:12 |
| @clarkb:matrix.org | I think you can do a pip user install to accomplish that? | 18:13 |
| @jim:acmegating.com | why are git-review packages bad? | 18:13 |
| @clarkb:matrix.org | they have in the past gotten stale. I haven't checked yet but I suspect noble's packaged version still tries to fetch the commit msg hook via scp by default but the latest release will install a vendored version which allows us to avoid the deprecated scp toolchain. That said if scp works on noble without extra flags to make it work then it is probably fine (some newer distros force you to allow that to happen iirc) | 18:14 |
| @clarkb:matrix.org | to be clear the issue is as openssh updates they want people to stop using scp and instead use sftp but gerrit's SSH server doesn't have sftp implemented so you have to set a flag or something to force openssh to go ahead wtih the deprecated appraoch | 18:15 |
| @jim:acmegating.com | 2.3.1 from 2022 | 18:15 |
| @clarkb:matrix.org | if Noble's openssh doesn't force you through that dance then it may be sufficient (since it is an LTS they are unlikely to make major behavior changes) | 18:15 |
| @jim:acmegating.com | either way, it does sound like continuing to recommend pip is the most universal thing; we probably still want this to work on jammy for a while | 18:16 |
| @jim:acmegating.com | unfortunately, `python3 -m pip install --user podman-compose` throws the same error | 18:17 |
| @tristanc_:matrix.org | We could also look into leveraging `uv`, it looks like a promising toolchain, and it features a `uvx git-review` usage that supposedly takes care of the venv stuff. (see: https://astral.sh/blog/uv-unified-python-packaging) | 18:17 |
| @clarkb:matrix.org | well I think you can do `pip install --user $thing` oh maybe not | 18:17 |
| @jim:acmegating.com | so apparently there's no "pip install --user" on noble... :/ | 18:18 |
| @jim:acmegating.com | the debian folks are recommending pipx | 18:18 |
| @jim:acmegating.com | i wonder if pipx or uv is universal enough to work on current and recent releases of ubuntu/fedora? | 18:19 |
| @clarkb:matrix.org | tristanC: I hesitate to recommend people possibly new to python and our tools use tools like uv until they have had more time to bake. In particular I worry about consistent behavor over time (the major reason pip is in the situation it has today is maintaining compatibility and uv hasn't proven they will do that) | 18:19 |
| @clarkb:matrix.org | the uv toolchain is also heavily dependent on pyproject.toml. Not sure what they do for projects that don't have that (they must have a compatibility shim but I don't know what the expectations of that are) | 18:20 |
| @clarkb:matrix.org | corvus: you mean there is no python3-pip package on ubuntu noble? | 18:20 |
| @jim:acmegating.com | no i just meant it throws that error | 18:21 |
| @jim:acmegating.com | https://zuul.opendev.org/t/zuul/build/3157f479353946cdb9b5a3190a6e55e7/console#3/0/1/ubuntu-noble | 18:21 |
| @jim:acmegating.com | that happens for both system and user installs | 18:21 |
| @clarkb:matrix.org | corvus: that example didn't use --user though | 18:22 |
| @fungicide:matrix.org | Clark: pip install --user is also disallowed, i think | 18:22 |
| @clarkb:matrix.org | --user should put it in a virtualenv for you | 18:22 |
| @clarkb:matrix.org | wut | 18:22 |
| @jim:acmegating.com | yeah that's what i'm saying | 18:23 |
| @jim:acmegating.com | i tried --user locally | 18:23 |
| @jim:acmegating.com | same output | 18:23 |
| @clarkb:matrix.org | I thought the whole point of --user was to solve this exact problem of not wanting things to go in a system wide install but also not wanting to manage a million venvs | 18:23 |
| @fungicide:matrix.org | unless --user changed how it works, it doesn't use a venv | 18:23 |
| @clarkb:matrix.org | fungi: it installs to ~/.local or similar which is effectively a venv | 18:23 |
| @fungicide:matrix.org | pip install --user does some sort of weird hybrid of installing the packages into a fixed path in the homedir and adjusting the python import path to include that | 18:23 |
| @clarkb:matrix.org | anyway if that doesn't work then ya I think we either use distro git-review or git-review in a venv | 18:23 |
| @fungicide:matrix.org | or pipx install or whatever | 18:24 |
| @fungicide:matrix.org | i'm not opposed to distro packages of git-review though | 18:24 |
| @jim:acmegating.com | pipx is available in jammy and noble; but they want a user PATH update, so that's a little tricky | 18:25 |
| @jim:acmegating.com | newer pipx (newer than noble) has a --global option, but that's not available to us | 18:25 |
| @jim:acmegating.com | given the pipx complexity, maybe just a venv along with an "activate" line is the best approach for the tutorial | 18:25 |
| @fungicide:matrix.org | just checked, latest git-review is 2.4.0 from march of this year. ubuntu noble and debian bookworm have 2.3.1 (the previous version) from 2022, so probably still fine | 18:27 |
| @clarkb:matrix.org | ya the main issue with the older version of git-review is it tries to use scp by default which may not work on new platforms. If it works for noble's version of openssh without config edits its fine | 18:29 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 923084: WIP: Try docker compose with podman https://review.opendev.org/c/zuul/zuul/+/923084 | 18:32 | |
| @jim:acmegating.com | well, there's the question of how new of a system do we want to require for the quickstart? if we want to support (slightly) older versions like jammy, we may want to test a few more boxes in the matrix. if we just want to bump everything up to latest then we may be fine. | 18:33 |
| @jim:acmegating.com | (i have not tested the actual use of git-review on noble yet, but that patchset does, so if it gets to that point, we'll know) | 18:34 |
| @clarkb:matrix.org | I think picking one platform and testing that well is what I would do. People can (and in fact I am) adapt to other platforms if they wish | 18:34 |
| @clarkb:matrix.org | I've been running the quickstart on tumbleweed | 18:34 |
| @jim:acmegating.com | the quickstart does have instructions for ubuntu, debian, centos, fedora, and opensuse... | 18:35 |
| @jim:acmegating.com | i think doing the pip install of git-review there let us presume that would just work. | 18:35 |
| @clarkb:matrix.org | thats a good point | 18:36 |
| @clarkb:matrix.org | just thinking out loud here: using a venv might be an extra annoying step but it may also be a good one to introduce to zuul users and devs? | 18:38 |
| @clarkb:matrix.org | they are super valuable for doing development and its cool that `uv tool` will hide all of that for you but maybe making it explicit is actually more helpful to the audience | 18:38 |
| @clarkb:matrix.org | (I personally do what uv tool does for you and have for years just making venvs for tools then symlinking the commands into my path) | 18:39 |
| @jim:acmegating.com | well, the audience here is zuul users (not just devs), and i personally think it's tragic that it's become harder to use git-review, which was supposed to be more or less the easiest zero-configuration way to use gerrit | 18:39 |
| @jim:acmegating.com | if we go through with the virtualenv, it will actually legitimately be easier for users to just "git push gerrit HEAD:refs/for/master" | 18:40 |
| @clarkb:matrix.org | ya I'm honestly surprised that pip install --user is not expected to work anymore. I was under the impression this was pips solution to breaking global installs | 18:40 |
| @clarkb:matrix.org | fungi: do you know why python3-pip exists at all if you can't use it anywhere but in a virtualenv which will include pip? Is it to provide the code that would normally get vendored by virtualenv/venv in a standalone package managed by the distro? | 18:46 |
| @jim:acmegating.com | we could recommend `pip install --user --break-system-packages git-review` (which sounds like nonsense, but does work) | 18:48 |
| @clarkb:matrix.org | I think I would just use the distro package if that was the alternative | 18:49 |
| @clarkb:matrix.org | because --break-system-packages is something that might turn away prospective users | 18:49 |
| @jim:acmegating.com | because it looks ridiculous, or because of a concern it might actually be bad? | 18:49 |
| @jim:acmegating.com | yeah, makes sense | 18:49 |
| @clarkb:matrix.org | like when `curl foo | sh` is suggested to me I'm more likely than not to not move on and not use the software | 18:49 |
| @jim:acmegating.com | (it shouldn't be our place to tell users "--break-system-packages" doesn't mean "break system packages" really trust us) | 18:50 |
| @clarkb:matrix.org | ya beacuse it looks bad and if that is one of the first things people see they get a bad impression | 18:50 |
| @jim:acmegating.com | not suggesting this, but jeff geerling feels the same way, but somehow decided this solution would be better: https://www.jeffgeerling.com/blog/2023/how-solve-error-externally-managed-environment-when-installing-pip3 | 18:51 |
| @jim:acmegating.com | (i disagree with that, just sharing to commiserate) | 18:52 |
| @clarkb:matrix.org | If we did use pipx how do we get that installed? | 18:53 |
| @clarkb:matrix.org | I think uv is a curl | sh situation | 18:53 |
| @jim:acmegating.com | apt in noble | 18:56 |
| @jim:acmegating.com | but then we have to update the user's path and restart the shell | 18:56 |
| @jim:acmegating.com | (not sure if we can tell pipx to use ~/.local/bin) | 18:56 |
| @clarkb:matrix.org | Ack | 18:57 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 923084: WIP: Try docker compose with podman https://review.opendev.org/c/zuul/zuul/+/923084 | 19:22 | |
| -@gerrit:opendev.org- Jan Gutter proposed: [zuul/zuul-jobs] 924970: Update ensure-kubernetes with podman support https://review.opendev.org/c/zuul/zuul-jobs/+/924970 | 19:42 | |
| -@gerrit:opendev.org- Jan Gutter proposed: | 19:49 | |
| - [zuul/zuul-jobs] 926013: Fix k8s-crio buildset registry test https://review.opendev.org/c/zuul/zuul-jobs/+/926013 | ||
| - [zuul/zuul-jobs] 924970: Update ensure-kubernetes with podman support https://review.opendev.org/c/zuul/zuul-jobs/+/924970 | ||
| -@gerrit:opendev.org- Jan Gutter proposed: | 19:51 | |
| - [zuul/zuul-jobs] 926013: Fix k8s-crio buildset registry test https://review.opendev.org/c/zuul/zuul-jobs/+/926013 | ||
| - [zuul/zuul-jobs] 924970: Update ensure-kubernetes with podman support https://review.opendev.org/c/zuul/zuul-jobs/+/924970 | ||
| -@gerrit:opendev.org- Tristan Cacqueray https://matrix.to/#/@tristanc_:matrix.org proposed: [zuul/zuul-jobs] 926931: ensure-docker: extract the restart procedure in a task file https://review.opendev.org/c/zuul/zuul-jobs/+/926931 | 19:51 | |
| @fungicide:matrix.org | Clark: pip3 exists as a holdover from back when it was common to have both python and python3 installed and to install things into the system environments for each of them | 20:27 |
| @fungicide:matrix.org | but since pip3 may still reside in some scripts and so on, it's been retained for compatibility reasons | 20:28 |
| @fungicide:matrix.org | also keep in mind that not all distros tell pip not to allow installing into the system context, and pip does have a cli option to override it as well as other ways to get around it with configuration (it's really just a flagfile which turns on that behavior to begin with anyway) | 20:29 |
| @fungicide:matrix.org | further, these days you can't assume pip is installed in all venvs, since pip can also be run from outside a venv and told what venv you want it to operate on | 20:30 |
| @clarkb:matrix.org | fungi: ya I'm more confusing about making --user not work. Since that seems like a reasonable reason to have the package | 20:30 |
| @clarkb:matrix.org | basically if --user worked then having a system package that can install to specific users with a single flag makes a lot of sense. but having a pcakge that won't install that way without a ton of workarounds in an effort to get you to use system packages makes it seem less valuable | 20:31 |
| @fungicide:matrix.org | the reason --user got blocked the same as system python is because it doesn't have an on/off switch like activating a venv does, so if you pip install --user some lib which is also a dependency of a python-based tool in your system context, you can break things | 20:31 |
| @clarkb:matrix.org | maybe the idea is that people can do the hack that jeff geerling used | 20:31 |
| @clarkb:matrix.org | but you should only break things for that user not for the system | 20:31 |
| @clarkb:matrix.org | since a different user won't see the packages installed to ~foo/.local | 20:32 |
| @fungicide:matrix.org | right, but the trend is to protect users from themselves, apparently | 20:32 |
| @fungicide:matrix.org | each time this comes up, people trot out the hordes of scientific researchers who apparently can't figure out which is the sharp end of their keyboard, who like to cut and paste commands they find taped to the underside of wastebaskets and in train station bathrooms | 20:33 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 923084: WIP: Try docker compose with podman https://review.opendev.org/c/zuul/zuul/+/923084 | 20:34 | |
| @clarkb:matrix.org | ya I mean I can understand not wanting pip interferring with system level packaging. That seems totally reasonable because then the distro gets complaints about cloud-init not working or whatever. But if I'm installing to my local user package space thats on me | 20:34 |
| @fungicide:matrix.org | i think a lot of it is a defeatist attitude toward trying to find compromise/middle ground between maintainers of language ecosystem package tools and maintainers of distribution package tools, and the "solution" was here's a switch you can turn on to break users' ability to install stuff so you can be the one to get yelled at not us | 20:35 |
| @fungicide:matrix.org | as for whether pip install --user is less of a break-the-system thing than sudo pip install, the assumption is that the majority of posix systems these days where either of those might be run are effectively single-user, so "only broken for my account" when mine is the only account on the system is still ~synonymous with "broken for everyone on the system" | 20:39 |
| @fungicide:matrix.org | sure it won't overwrite/delete files under /usr, but recovering from it still requires the user to figure out what to blow away in their homedir | 20:40 |
| @clarkb:matrix.org | ya I guess in my mind there is a difference between breaking the ability to boot properly or install packages and a specific user (either for desktop use or dedicated to service running atop the base platform) failing. | 20:41 |
| @clarkb:matrix.org | the thing that we should be careful about is the base system functionality. If users pollute their homedir that isn't a distro problem | 20:41 |
| @clarkb:matrix.org | anyway this isn't really a venue for that argument. And its unlikely we'll sway any minds at the distros so just have to accept they've done this | 20:42 |
| @clarkb:matrix.org | I'm just still really confused why they would make ap ackage for a tool that they explicitly break | 20:42 |
| @fungicide:matrix.org | it's technically not broken, just produces a giant hard-to-miss warning that tells you in no uncertain terms that the distro is not responsible if you go ahead and add the extra scary command-line option named along the lines of how your distro maintainers view the action you're about to take | 20:44 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!