IRC channel logs

<mirai>jmes: maybe (provision '(xorg-server display-manager)) ?

<mirai>my bad, not provision

<mirai>put requirement on 'display-manager

<mirai>I've noticed that gdm only provisions 'xorg-server but not 'display-manager, that might be worth a look

<jmes>mirai: Thanks! Let me have a look at what the display-manager shepherd service does

<mirai>its something provisioned by sddm/lightdm I think (along with 'xorg-server)

<jmes>Hm, would a home shepherd service aware of system shepherd services? Also what about the case where the user doesn't use a display manager (as I currently do not)?

<jmes>be aware*

<mirai>hmm… I think it wouldn't

<mirai>if no one comes up with an answer try mailing to help-guix

<jmes>mirai: the approach that comes to my mind is making a home service to manage the wayland compositor. Those services can provision 'wayland and then the notification daemon service can require 'wayland

<jmes>Thanks a bunch for your input :)

<cow_2001>how do i find out what phases each build system has?

<cow_2001>there is the documentation, but from within guile?

<ireallyhateirc>you can find some hints in ./guix/build/ in the source code

<ireallyhateirc>%standard-phases are for example defined in ./guix/build/gnu-build-system.scm

<cow_2001>hmm!

<cow_2001>no way of getting it from inside the repl?

<ireallyhateirc>I'm not that advanced, you need to ask someone else for that. But I suppose you could just do use-modules and the name of the module that defines it?

<cow_2001>wait a minute.. hmm.. it is %standard-phases i guess O_O

<ireallyhateirc>Try (use-modules (guix build gnu-build-system)) or something like that idk

<cow_2001>oh, woops it is a different module

<cow_2001>i was looking at (guix build-system cmake)

<cow_2001>not (guix build cmake-build-system)

<cow_2001>this is a bit of a confusing naming scheme

<cow_2001>ireallyhateirc: thank you!

<ireallyhateirc>so it worked huh? That's great since I'm mostly trying to hack together basic packages.

<cow_2001>no, but now i know where to look

<ireallyhateirc>that's something

<cow_2001>i knew it before but forgot

<cow_2001>need to write it down

<meaty>Is there a way to configure zsh via the zsh home service without it changing your default/login shell?

<old>Is someone here used to package kernel modules?

<old>I've contributed lttng-module a long time ago. However, it keeps breaking whenever a new kernel version is available. I need to update the module to a newer version

<old>However, by the time my patch gets merged, there can be weeks

<old>So I have to either not update until my patch is merged, or remove the module from my operating-system definition

<old>Currently, I keep my own private definition of the module that I manually update whenever I see that the newer kernel is not compatible with the module

<podiki>perhaps cc the kernel team on patches if you haven't already? maybe that will help with a quicker review from relevant people

<podiki>or even join the kernel team :)

<unmush>looks like the python-requests hash has changed, I get 1qfidaynsrci4wymrw3srz8v1zy7xxpcna8sxpm91mwqixsmlb4l

<unmush>I don't have the original, so I can't see what's changed

<jadzi>Hello all! I'm trying to define a tree-sitter-phpdoc package for my project, it requires a higher tree-sitter version than guix has. I've built that package's version, and now i'm curious if there is a way to override a build system's build-inputs/host-inputs in general (or for tree sitter build system in particular) without copy/modify the whole build-system procedures?

<x8dcc>Well this is just great. Now my machine can't even boot because grub can't find some module after 'guix system reconfigure'

<andreas-e>You should always be able to boot into a previous generation.

<x8dcc>How would I do that if I can't reach the bootloader

<x8dcc>Grub can't find a module, so it doesn't even start. It just goes into rescue mode

<andreas-e>Strange. I have never seen this, and thought the module was missing during the start of the kernel.

<andreas-e>One option would be to start on a USB stick and to repair things from there, but I do not know how the details would work.

<x8dcc>I wouldn't even know what to do from a USB stick, I don't even know what caused the issue

<x8dcc>This was after reconfiguring the system to a previous configuration, by the way. So much for reproducible environments, or whatever

<df>I'd search for the specific error, if it's missing a module I'd guess it's failing to mount its second stage because it needs that module for the hardware or filesystem in question

<csantosb>Ey Guix ! Funny, I have a v1 patch succeeding, https://qa.guix.gnu.org/issue/68153, except it succeeds in something unrelated called mqueue, is that an error ?

<futurile>Morning all

<civodul>o/

<cbaines>csantosb, the mqueue related changes are an existing issue with the asdf build system, which manifests as packages changing unexpectedly, you can read more about it here https://issues.guix.gnu.org/76582

<cbaines>it's not something to do with your patch

<cbaines>csantosb, it is odd though that the new package isn't showing up, unfortunately you have to go in to the data service logs to work out why

<cbaines>error: while processing python-cocotb ignoring error: unbound-variable: (#f Unbound variable: ~S (python-find-libpython) #f)

<cbaines>csantosb, I'd maybe send one patch series including both python-cocotb and python-find-libpython

<nmeum>does anybody have the time to give me a review on https://issues.guix.gnu.org/75934 ? I think this would be nice to have in order to replace isc-dhcp (which has been EOL since 2022)

<nmeum>comes with tests and documentation and I have been using in on my Guix system for a while now without running into issues

<cbaines>futurile, how's your key situation now? I did get your email but I haven't got around to replying yet

<csantosb>cbaines: ok, thanks, I'll do that

<futurile>cbaines: I think that Maxim fixed it last night, so I'm going to create a branch today and push it (that way if it's messed up I won't break anything).

<cbaines>futurile, +1

<andreas-e>Hello cbaines!

<cbaines>andreas-e, o/

<andreas-e>I have just activated marsiling for the bordeaux build farm again. It has a slightly shaky connection to its charger and ran out of battery. Since there were no builds in the x86 queues, I left it turned off. But now I see that the plans have filled up.

<andreas-e>Excellent!

<andreas-e>But in the builds, I see lots of "path `/gnu/store/79b71gb175wqlifz96ipvx9jqv5c6nkj-charls-2.3.4.drv' does not exist and cannot be created"

<andreas-e>I am sure you already explained to me what this is. Maybe related to the fact that only old builds are sent to the machines (the priority is 0), of which the inputs have disappeared?

<andreas-e>On the activity page, strangely the bold face "origins" of the packages are missing for marsiling. What does this mean?

<cbaines>I'm not sure what you mean by "origins", do you mean some of the tags for builds?

<mange>vhns: A little while ago we talked about using Wireguard with a random ListenPort. I went to make a change to handle #f, but it turns out the current code in Guix already does that! If you set (port #f) then it will omit the ListenPort line in the wireguard config.

<andreas-e>I mean on the page https://bordeaux.guix.gnu.org/activity the bold face "branch: gnome-team category: package revision: 9f..." under each package. On marsiling (and hydra-guix-128) they are absent. As if these are derivations of which it has been forgotten where they come from.

<andreas-e>So I wonder if this explains why they do not build.

<andreas-e>Probably so. Now the page has changed, there are packages of higher priority with their bold face information, and something builds.

<cbaines>andreas-e, those are the build tags, and if builds aren't tagged then it's probably me (or someone else) submitting them manually

<cbaines>and yeah, there's still a bunch of builds that need cleaning up

<andreas-e>Ah, interesting!

<andreas-e>Well, they seem to clean themselves by not building :)

<andreas-e>Maybe we could decrease the number of trials from 20 to 10 or 5? Are there cases where downloads succeed on the 20th trial?

<cbaines>andreas-e, the retries are set based on the narinfo negative and transient error TTLs, they did go down recently so it might be possible to reduce the retry time

<cbaines>so it's less of 20th time lucky, but more by the 20th retry the cache for the 404 or 504 response will have expired and the substitution will actually be retried and maybe succeed

<andreas-e>Alternatively, reduce the number of trials from 20 to 5 and increase the delay between two trials by a factor of 4? That should have a similar effect with less work by the server on the other side.

<andreas-e>Or the cache is locally so that a retry costs nothing?

<cbaines>the cache is local

<andreas-e>Okay, then it amounts to the same. Less cluttering of the log file, but that is not really important.

<andreas-e>It is great to see QA running again! For instance it is working on https://issues.guix.gnu.org/76626 which was submitted just yesterday.

<cbaines>yeah, it's still not reliable, but when it works it does seem to do OK

<cbaines>you can see at the bottom of https://data.qa.guix.gnu.org/ it's managed to process a whole bunch of patches

<andreas-e>Including some old ones, which is nice. Also on qa.guix.gnu.org one sees that reviewers and committers pick them up.

<futurile>oh https://data.qa.guix.gnu.org/ seems to show it's really moving along - that's great

<futurile>did anything particular change cbaines?

<cbaines>futurile, it's hard to point at a single thing

<cbaines>after resisting for a while, I got the data service just to ignore a load of systems and systems+targets https://git.savannah.gnu.org/cgit/guix/maintenance.git/commit/hydra/deploy-node-130.scm?id=473a5b55be49052d520a67207315a251bbf09f20

<cbaines>that's an example of just doing less which can speed things up

<cbaines>then there's changes like https://git.cbaines.net/guile/knots/commit/?id=a11cd24e57ad334c26700691dca0164f410cc392 where I just had some inefficient code, and fixing the design improved the performance

<futurile>Maybe when we do the next release we can figure out some official status on each architecture - we have the data from the survey - a bit like Debian does - not holding ourselves up on certain architectures does make sense

<futurile>does my idea of "double the size of the build farm" still have any sense to it? Now that we have the potential for funding VM's out, we could have a "trial" of adding more build machines

<futurile>what do you think?

<futurile>cbaines: -^

<Rutherther>anyone familiar with packaging node? trying to package math-preview package (https://gitlab.com/matsievskiysv/math-preview), it dpeends on mathjax-full and I have trouble packaging it since it seems to look for tsc (there are some ENOTCACHED errors and then it tries to execute tsc) - failing command is "sh -c npx rimraf js; npx rimraf dist; ./node_modules/.bin/tsc; ./node_modules/.bin/webpack"

<cbaines>futurile, in a sense we're already trialing that, hydra-guix-124 is now building things for the bordeaux build farm, and if we could get hydra-guix-128 online as well that would be a second new machine

<cbaines>futurile, I'm also working on getting a couple more arm machines online, I think it's something to circle back to in maybe a few weeks, in particular it would be good to have some concensus about the direction we want to take longer term

<futurile>cbaines: understood

<df>is there a way (in both guix system and package) to list/delete all but the most recent N generations?

<futurile>df: guix package --list-generations or uh guix system list-generations (I think)

<futurile>df: then guix package --delete-generations=1m - or 1w for 1 month or 1 week

<futurile>df: you can be more specific with: guix package --delete-generations=5..10

<futurile>df: and then running guix gc will remove anything in the database that's not needed - it's aggressive so only use it if you need the space

<df>I'm looking to write this as a cron job so it won't know the existing generation numbers without a bit of hackery

<df>was wondering if there was something like --list-generations=-5

<df>hmm, although I guess you'd probably expect that to include *only* the most recent 5 rather than the other way around

<andreas-e>I have what should be a simple guile question. I define the following macro:

<andreas-e>(define-syntax N

<andreas-e> (syntax-rules ()

<andreas-e> ((N) 0)

<andreas-e> ((N x y ...) 2)))

<andreas-e>Now when I type (N 'a) I obtain 2.

<andreas-e>Why does this match the (N x y) list?

<andreas-e>Without the ... I obtain (as expected) a syntax error, that is, no match.

<civodul>andreas-e: (N x y ...) matches a list of two elements or more

<civodul>in total

<civodul>that is, “y ...” means zero or more

<andreas-e>civodul: The strange thing is that (N x y ...) also matches (N 'a), that is, 1 element.

<icepic1984>Hi guix! Is it possible to reference a git repo hosted on ssh in a channel file? If i do something like `(url "ssh://icepic@xxx.de:2222/bla.git")` in the channel config, i get the following error on pull: ` Git error: failed to start SSH session: Unable to exchange encryption keys` ? I use public key authentication for my ssh server

<andreas-e>I would have expected "..." to match 0 or more elements, and "y ..." to match 1 or more elements, and "x y ..." to match 2 or more elements.

<futurile>I kinda simultaneously hate debbugs/bts and think it's brilliant

<futurile>it does some really good things and it's kinda amazing using it with email, but the hodge-podge of confusing commands and rubbish error messages makes me want to scream

<futurile>plus the fact that the GNU and Debian installations don't match is just the worst part

<cow_2001>guix repl --listen="$PWD"/listen-socket => guix repl: error: /home/yuval/foo/guix-kakafarm-channel/listen-socket: invalid listen specification

<cow_2001>is there a way of running the guix repl with listening over a fifo file socket?

<cow_2001>oh, the guix shell --help did not help, but the info manual was informative

<cow_2001>err

<cow_2001>guix repl --help

<cow_2001>you just need to --listen=unix:$path

<futurile>cow_2001: https://guix.gnu.org/manual/en/html_node/Invoking-guix-repl.html also in the manual. Never used it though, did it work?

<cow_2001>futurile: worked! :D now i can geiser-connect-local :D

<cow_2001>futurile: why shudder?

<futurile>cow_2001: emacs :-))

<cbaines>any ideas why the Guix arm64 linux-libre config has CONFIG_VSOCKETS=m, but my running system has "CONFIG_VSOCKETS is not set" in /proc/config.gz ?

<ekaitz>efraim: what's the status of the riscv bootstrap? are we in time for the release?

IRC channel logs

2025-02-28.log