| coderanger | Hey there, what's the best way to get in contact with the Python maintainer team for Ubuntu? Because the Python 3 packages for 18.04 are all kinds of broken right now. | 01:07 |
|---|---|---|
| sarnold | coderanger: best is to file a bug report with ubuntu-bug <packagename> | 01:08 |
| coderanger | No, I need to talk to them directly. | 01:08 |
| coderanger | This isn't a bug, it's a structural failing somehow | 01:08 |
| coderanger | After Barry left, I'm not sure who took over | 01:08 |
| coderanger | It looks like someone decided to move distutils and venv into their own packages, which is fine, but then they never reworked things so python3.6 or one of the related stdlib packages depends on them. | 01:10 |
| coderanger | doko: Looks like the package files say you're it :) | 01:11 |
| coderanger | doko: Hopefully this was just an oversight and we can get it cleaned up easily. | 01:14 |
| coderanger | (3AM there, rats) | 01:15 |
| nacc | coderanger: we fixed most of them, pre-release, iirc | 05:17 |
| coderanger | nacc: Yeah, looks like distutils was fixed upstream right around feature freeze, so might have just been missed :) | 05:17 |
| tarzeau | we use ubuntu 18.04 with d-i, and 5 out of 10 installations hanged at the "update-grub" screen 66%, and alt-f4 showed os-prober. kill -9 `ps | 06:31 |
| tarzeau | |grep mounted` made it go further. any ideas? | 06:31 |
| tarzeau | i have images i can post by imessage (or email) if anyone cares | 06:32 |
| === stgraber_ is now known as stgraber | ||
| tkamppeter | mdeslaur, I do not know anything wrong with QPDF 8.0.2, please go ahead and do the update. | 12:26 |
| sergiusens | doko: where can I get a chroot/lxd image to test the adt failure? | 12:40 |
| doko | sergiusens: I don't think we already have cosmic images | 12:41 |
| doko | infinity, Odd_Bloke: ^^^ | 12:41 |
| Odd_Bloke | We do not yet have cosmic images. | 12:41 |
| infinity | sergiusens: Boot bionic image, 'sed -i -e "s/bionic/cosmic/g" /etc/apt/sources.list && apt-get update && apt-get dist-upgrade', congrats, it's cosmic. | 12:42 |
| infinity | sergiusens: But yes, no lxd images yet, AFAIK (not my department). | 12:42 |
| sergiusens | wfm | 12:45 |
| Laney | sergiusens: you can pass RELEASE=cosmic to autopkgtest-build-lxd to build a bionic image and then dist-upgrade it | 12:50 |
| Laney | that's what we did to get the actual ones that are running | 12:50 |
| infinity | Or that. | 12:50 |
| infinity | Laney: Your suggestion offended him right off IRC. | 12:51 |
| Laney | ENOBRAINSPC | 12:52 |
| sergiusens | thanks Laney, will look at that next | 13:19 |
| nacc | coderanger: ah could be | 15:34 |
| infinity | coderanger: distutils not being installed by default is a feature, not a bug. If there are still things in the archive that (incorectly) depend on it at runtime, please file bugs. If you have things locally that do, either fix them or install it, IMO. | 16:34 |
| infinity | tarzeau: Filing a bug on debian-installer with reproduction info would be helpful. Assuming you're using our d-i build and not a custom one. | 16:36 |
| coderanger | infinity: That 1) breaks how Python is supposed to work and 2) implies that people mostly install Python apps via apt rather than the python-native system. Distutils is an important part of the stdlib and I'm not really sure why it would be specifically carved out. There are exceptions in place for stuff like Tk where it's in the stdlib but installing it requires a lot of support libs, but distutils is pure python soooooo | 16:36 |
| cjwatson | fg | 16:52 |
| cjwatson | oops | 16:52 |
| coderanger | infinity: There is also the pythonX-minimal packages for people that want to opt-in to a reduced view of the world :) | 16:53 |
| infinity | coderanger: By the "python-native system", do you mean setuptools (which depends on distutils) or pip (which recommends setuptools), or something else? | 16:56 |
| coderanger | pip, which is not always installed via apt | 16:56 |
| coderanger | Basically it's weird that `apt install python3` doesn't result in a fully-functional python environment (modulo that "fully-functional" is a very loaded term and as mentioned there have been good reasons in the past to punt some small bits of the stdlib out of the default package set) | 16:57 |
| jbicha | coderanger: totally off topic, but did you see https://xkcd.com/1987/ yet? | 16:57 |
| coderanger | Especially since this is a change from the debian version of the same packages | 16:57 |
| infinity | coderanger: Anyhow, the change seemed sensible to me, but feel free to file a bug and/or yell at doko. Same changes landed in Debian first, so perhaps best to argue the case with him in the Debian BTS. | 16:57 |
| coderanger | jbicha: I can neither confirm nor deny ... O:-) | 16:57 |
| coderanger | infinity: Yep, put some in last night :) | 16:58 |
| infinity | coderanger: It's not a change from Debian, Debian's got the same setup in unstable. | 16:58 |
| coderanger | Ahh okay, I wasn't sure how to read that part of the changelog. | 16:59 |
| infinity | jbicha: See, I'm more inclined to say "I prefer to use a non-packaged package manager, but have it magically work with the packaged language runtime" is https://xkcd.com/1172/ | 17:00 |
| infinity | But I also despise all things python, especially the way its ecosystem is managed, so I'll bow out from the conversation. :P | 17:01 |
| coderanger | infinity: It's the continual tension between distro packages as the most available option and the conflict of distro vs. upstream bundling policies :) | 17:01 |
| infinity | coderanger: Sure, I get that language ecosystems and distro packaging will always be at odds, but it seems daft to use a non-packaged pip when a packaged one is designed to work with the packaged runtime, IMO. | 17:02 |
| coderanger | infinity: My specific case is I maintain the Chef ecosystem support for Python so I want to install pip from source to get the same behavior across all OSes. | 17:03 |
| coderanger | Tooling dislikes special cases :D | 17:03 |
| infinity | (Though I also understand that, in the past, the packaged pip and venv didn't really work right on Debian, and maybe that's led to all sorts of curious practices) | 17:03 |
| coderanger | There also appears to be an odd mismatch where python3-dev depends on distutils, but python3.6-dev does not | 17:04 |
| infinity | coderanger: That's definitely by design. pythonX.X-dev is meant for C-linking (ie: building modules), while python3-dev/python-dev are meant for building extensions, hence the deps on all supported python versions and distutils and dh-python. | 17:06 |
| infinity | coderanger: Not voicing an opinion on if that was "right", but that's how it's been pretty much forever. | 17:07 |
| coderanger | Fair :) | 17:07 |
| coderanger | The whole design of the debian python layout is definitely showing its age | 17:07 |
| coderanger | Some day Python 2 will die and we can all have the nice things, right? | 17:07 |
| * coderanger looks around nervously | 17:07 | |
| infinity | 2020. | 17:07 |
| infinity | Not sure that'll suddenly make python3 nice, but it can't make it worse. | 17:08 |
| coderanger | 2025 because LTS :) | 17:08 |
| coderanger | Though I'm sure RHEL 6 will find a way to make everything terrible forever, as its job. | 17:09 |
| infinity | RHEL5 is still making things terrible forever, nevermind 6. | 17:09 |
| tarzeau | infinity: we're using yours, however i have no idea how to reproduce, we'll try to figure out (maybe the old 16.04 install on the disk causes it), however it only happened on about 50% of the machines | 18:47 |
| tarzeau | i'll try to find the launchpad bugs page for d-i ubuntu 18.04 and post the console screenshots there | 18:47 |
| doko | infinity, just fyi, there are no differences between Debian and Ubuntu | 19:15 |
| tarzeau | mhm, we use the preseeded d-i with stretch and xenial, no problems. but since 18.04 half of my installs hang as described | 19:21 |
| tarzeau | we do have lvm and btrfs/xfs/ext4 (and i'm not sure if the problem is with update-grub or os-prober) | 19:22 |
| tarzeau | if it helps anyone i can post the photos more early and try to debug later | 19:22 |
| tarzeau | but it seems, we're the only ones with that problem, otherwise i would have found some other reports on the interweb | 19:22 |
| infinity | tarzeau: https://bugs.launchpad.net/ubuntu/+source/debian-installer/+filebug | 19:40 |
| infinity | tarzeau: If you're using a preseed, attaching that (with any private info scrubbed) would be super helpful. | 19:41 |
| infinity | tarzeau: Also, /var/log/syslog from one of the systems at the point where it stalls. | 19:41 |
| infinity | tarzeau: (Or grab it later from /var/log/installer once you kill and finish) | 19:41 |
| shiftplusone | Would anyone (chrisccoulson?) happen to know how where the orig tarballs for chromium come from? | 19:44 |
| shiftplusone | ah, https://commondatastorage.googleapis.com/chromium-browser-official/chromium-$(ORIG_VERSION).tar.xz actually works | 19:46 |
| cyphermox | tarzeau: any chance it's on a really really really huge disk? | 20:36 |
| cyphermox | tarzeau: I'd be very interested in seeing a bug report with the details | 20:36 |
| === nOOb__ is now known as LargePrime | ||
| sarnold | bdmurray: 1768866 is the second bug I've seen with "PythonDetails: /root/Error: command ['which', 'python'] failed" | 22:32 |
| sarnold | bdmurray: is /root/Error really being constructed? it feels a bity funny. I can't say it's wrong, but it doesn't feel right :) | 22:33 |
| sarnold | bdmurray: is this working as intended? or is something else at play? | 22:33 |
| bdmurray | sarnold: They probably don't have python installed so the python details bit is poorly constructed | 22:35 |
| sarnold | bdmurray: thanks, that's probably fine then :) | 22:38 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!