IRC channel logs

2024-08-05.log

back to list of logs

<mod0>hello
<mod0>I am new to irc, used it in early 2000s. I don't see any messages on this server and don't know if Im logged in correctly
<mod0>ah, I see my messages in the log
<nikolar>mod0: you are
<nikolar>we see yo
<mod0>thanks
<ryan77627>Hi, quick question that I couldn't really find a great answer to. When I am redefining a package to just change some build flags, is there a function that allows me to append/prepend/remove options from the #:configure-flags list like with inputs? Or do I just have to redefine the entire arguments parameter?
<FlaminWalrus>I'm trying to define a package for the Twelf logical framework. However, I can't seem to get the Standard ML of New Jersey compiler to work as an input: I've done =(#:use-module (gnu packages sml))=, in which case =guix build= tells me I forgot a =use-modules= form, and =(specification->package "smlnj")=, in which case apparently =(gnu packages sml)= does not contain the variable =smlnj= despite the fact I can see that it's introduced
<FlaminWalrus>with =define-public= via =guix edit=.
<tschilp>Hi! I'm receiving http://paste.debian.net/1325494 at system reconfigure on e482d04d0f3de934a617429b1c774513638aa985. I haven't changed my configuration except for adding ~(containderd-service-type)~ as needed. Things seem to get stuck at '9 drafts for texlivetexmf-20230313'. My memory does not seem to fill up and cpu usage is mediocre, however my pc seems to become quite laggy.
<tschilp>I could only see the mentioned 'error' after C-c C-c ing the first attempt to reconfigure. The situation is the same on attempt no 2 -- no moving vertical bar next to the graft information for texlivetexmf.
<jaft_r>tschilp: what file system are you using for where your /gnu/store is located? The reference to memory leaks is what it's worried about, not what's causing it to deter (I believe); the reference to blocks is likely blocks of the file system. Unfortunately, I don't know how to address that, though…
<tschilp>jaft_r: thanks, it's ext4. I've been hitting memory and filled up disk in the past when reconfiguring my system, but neither of these is the case at the moment. Mhm.
<tschilp>I just pulled to 3c309a7b7020a9645bfe775d9f53f43f05d076bf, situation seems to be the same there.
<jaft_r>tschilp: of course; is that the only error you're getting back or was there any other logs, etc.?
<tschilp>jaft_r: No, I can't see any errors in my terminal. Even the warning just shows a few times before the process starts 'as usual', just then things become very slow. It's somehow 'intervally' though, so from one moment to the next my keyboard becomes unreactive, even here in my irc buffer.
<tschilp>guix home reconfiguration goes fine though
<tschilp>maybe this teaches me getting rid of the full texlive :)
<futurile>Morning all
<theesm1>good morning
<nikolar>well building ffmpeg for aarch64 seems to have gotten stuck again
<nikolar>this time it's apparently rust
<nikolar>which is blocked on llvm
<nikolar>it seems that the build just timed out
<kaij>Does someone here know what guix time-machine -C .
<kaij>wups sorry. what that will build? I can't seem to figure it out from staring at the source
<mekeor>kaij: it just fails for me: guix time-machine: error: failed to load '.': Is a directory
<kaij>ah sorry, I have a channel file there that it's supposed to use. you can try with the default one (iirc ~/.config/guix/channels.scm). it starts building bootstrap packages (as it's a custom mirror of upstream guix) and then eventually fails. I want to manually build whatever it needs before calling it (with guix build) but no idea what it actually needs.
<mekeor>kaij: "guix time-machine -C ~/.config/guix/channels.scm" fails with "guix time-machine: warning: no command specified; nothing to do". sorry that i haven't been helpful at all so far!
<redacted>kaij: can you share the full command you're running?
<redacted>mekeor: guix time-machine runs a guix command under a specific version of guix. `guix time-machine --help` will show you the syntax. Note that "COMMAND" is a guix command like `build' or `package'.
<kaij>just doing `guix time-machine -C ./channels.scm -- describe -f channels`. it's kicking off some build. presumably the same that guix pull would (?) but not sure what that's trying to build either.
<futurile>kaij: it will always build the 'guix' cli/modules etc for the point you're specifying in channels.scm
<futurile>kaij: like you said, it's exactly the same process as 'guix pull'
<kaij>so it'll build (& install?) the guix package? so if I successfully build the guix package it'll also succeed?
<futurile>kaij: yep. Hopefully, you've found it works for you. Time machine runs just the one command, if you want to run multiple commands you can do -- shell
<cria>got a weird issue hoping someone can shed some light on. i'm starting keepassxc as a home shepherd service and it launches fine, but seems like it's only registering to dbus for things that were started at the same time as it via shepherd. apps i run afterward don't see it listed as a secret service. should i be extending a different service?
<cria>naturally if i disable the service and start it manually, everything sees it
<Oleand3r>Hi, I'm still experiencing the issue I reported above. I thought my shell was misconfigured but I was wrong. This issue reoccurs every time a new update is downloaded via guix pull. Then I reconfigure and the command hangs at "guix system: bootloader successfully installed on /(dev/sda)". The only way to shutdown my laptop after this, is by pressing the power button, loginctl poweroff and sudo poweroff hangs forever and do
<Oleand3r>nothing. Then I turn my laptop on, log in and run guix pull && sudo guix system reconfigure config.scm and if no update is available, this time the command succeeds.
<fnat>Do you think it might be worth to have a least-authority version of transmission-service-type?
<fnat>Ooops, it's actually 'transmission-daemon-shepherd-service'. :)
<fnat>How would one go about it? Rewriting the whole 'transmission-daemon-shepherd-service' definition? Is there any other (more lazy) way to use 'least-authority-wrapper' while reusing the service as it's written already? Hm... probably not.
<craigbro>I am trying to install guix on a qemu vm on a intel host, and it keeps hanging while builing a derivation, the moduled-import is I recall correctly.
<craigbro>this is using the 1.4.0 install iso
<craigbro>ran that with 1024m ram, will try increasing..
<PotentialUser-78>Hello. There is something strange going on. I got complains on guix ... is not a descendent of ... and some packages can not be found. guix search for emacs-consult-denote does not give a result
<futurile>PotentialUser-78: have you been messing around with time-machine or inferiors - maybe you switched your guix to a different commit/channel/branch
<dckc>ok... so... emacs + magit were working, and now they're not. I should be able to roll back to where they work working, yes?
<Rutherther>that will for sure depend on whether the reason they stopped working is related to something that was declared or not. what is the issue?
<PotentialUser-78>futurile: No. I am on a foreign distro Asahi Ubuntu. Everything else is as always
<dckc>Jul 21 17:23:07 <dckc> ooh! now I'm cookin' with gas... emacs, magit, direnv all talking nice to each other...
<dckc>^ that was the happy state
<Rutherther>since you are on foreign distro, do you "source ~/.config/guix/current" somewhere?
<dckc>Aug 03 16:09:23 <dckc> nooooo! magit is horked now! when I hit b, I get: transient-setup: Symbol’s function definition is void: transient-prefix-object
<dckc>current... checking...
<dckc>my bashrc has:
<dckc>GUIX_PROFILE="/home/connolly/.guix-profile"
<dckc>. "$GUIX_PROFILE/etc/profile"
<Rutherther>dckc: sorry that was for PotentialUser-78
<dckc>np
<PotentialUser-78>Rutherther: This is still confusing to me. But after sourcing it is there
<PotentialUser-78>Rutherther: Thank you
<dckc>ok... I see https://guix.gnu.org/manual/en/html_node/Invoking-guix-package.html#index-rolling-back ... now how to list profile generations with dates?
<Rutherther>PotentialUser-78: after you do guix pull the guix script is put to "~/.config/guix/current/bin/guix", you need to switch that one. I suppose you are using different one accidentally
<dckc>`guix package -l` seems to do it...
<dckc>$ guix package -S 10
<dckc>switched from generation 26 to 10
<dckc>ok! winning!
<dckc>I must admit that it shakes my confidence greatly that magit would get busted like this. Surely lots of people use magit?
<vats>Oh, did magit break?
<Rutherther>seems so, also not working for me when I get it out of guix
<vats>I can't see any outstanding bug reports at https://github.com/magit/magit/issues .
<Rutherther>ah. you meant like that. I suspect dependencies in guix are wrong, not magit itself
<vats>I see. I use it from straight.el.
<Rutherther>already investigated by someone, the byte compiled version doesn't contain the function for some reason https://lists.gnu.org/archive/html/help-guix/2024-08/msg00021.html
<kaij>is there a way to use package x instead of y on the whole system? in general something like meta packages from other packager manager systems where any package that provides x can be used?
<kaij>just to give a reasonably common example, wireguard. A tool that relies on wireguard needs there to be one these options. A kernel with wireguard compiled in, wireguard as a kernel module or the go userspace implementation. I'm not sure how to express something like this in guix. not sure how a user would make the selection either.
<mfg`>How can i skip tests when using guix time-machine?
<mfg`>--without-tests isn't available at that commit it seems
<mfg`>I just want to use an old toolchain. (built with an old glibc) But that doesn't seem to work. Something around 2018 would be nice
<Rutherther> [mfg"](https://matrix.to/#/@irc_libera_mfg=60:ditigal.xyz): have you tried what --without-tests does - modifying the arguments to have "(#:tests? #f)`
<Rutherther>hm, heisenbridge sent the mention wrong, sorry for that
<kaij>seems like things like cc are special cased
<Rutherther>kaij why do you want to replace it on the whole system? I am not sure that would be possible as you can use multiple channel definitions with stuff like -C or time-travel, each user has their channel etc.
<mekeor>redacted: i know. i was just following the instructions provided by the question.
<kaij>well for experimentation for example. specifically looking at the zig compiler, coreutils, that kind of thing.
<kaij>seems like cc being special cased makes some sense, but I can't find a general mechanism similar to cc-for-target
<mfg`>Rutherther: How do i modify the package definition, if i time-machine'd to that commit? can i just guix edit it? If so, i didn't know that
<mfg`>wouldn't really make sense tho
<kaij>like, I'd want to try out replacing gnu coreutils with the rust project. it may or may not work, but I'd be interested. of course I can make a complete duplicate of the mainline guix channel, but that seems overkill for something where really there's just a few choices between packages?
<Rutherther>mfg` what operation are you using? guix build? shell/environment? you should be able to use a scm file and specify the modification in it
<Rutherther>mfg` in case I wasn't clear, transformation without-tests actually uses this function https://termbin.com/46m1, so you can just use this, call the package you want with it and it will modify it to not include tests
<mfg`>Ah, indeed. Thanks for the tip :)
<mfg`>Rutherther: The package where i want to disable the tests is a dependency, so i also have to modify the parent package to overwrite the dependency with this new package?
<Rutherther>mfg"yeah, I think you will have to do that. One way is to just something like"(package/inherit your-package (inputs (modify-inputs (package-inputs your-package) (replace "dep-name" dep-name-no-tests)))`. Or you could do deep override of the dependency, the transformations can again work like an example. Even the without-tests is actually defined to be able to rewrite all dependencies of the packages, though I sent only the part that does it for one...
<Rutherther>... package
<mfg`>i see
<ieure>I am once again begging anyone with a commit bit to review this patch series which contains security fixes, which has been sitting unreviewed since June. https://issues.guix.gnu.org/71832
<ieure>Do I need to be cc'ing a security team here, or like... what
<vagrantc>seems reasonable if there is a long list of CVEs that it fixes ...
<ieure>vagrantc, 16 CVEs fixed: https://issues.guix.gnu.org/71832#16
<vagrantc>right
<ieure>Eeehhhh I need to update the upstream version again. This will be the fourth upstream release since sending the original patch.
<vagrantc>as a drive-by person glancing at it ... it unfortunately looks like a large diff, but i suspect that is due to some whitespace changes triggered by appropriately indenting something with a deeper () level
<vagrantc>e.g. the switch from (package (version ... to (let (version ... (package ...
<ieure>vagrantc, It's mostly refactoring. When I initially opened the issue, it was an upstream version bump without CVEs. But since it's been sitting for so long, now there are security issues as well.
<vagrantc>sure
<ieure>On a personal level, it's frustrating to me that "but who will maintain this" was a question for acceptance of that package, and I'm doing that work, but the review process is inhibiting it.
<vagrantc>understandable, yeah ...