IRC channel logs
2024-09-16.log
back to list of logs
<PotentialUser-84>I am getting an error upgrading package "powerstat". guix package --upgrade says that the build of a derivation failed. Are these events common? <podiki>PotentialUser-84: no, shouldn't be common but things break. seems that an update for powerstat will fix it, i will double check and push that later today <podiki>so possibly this broke as other things got updated and powerstat's version was too old <PotentialUser-84>podiki Oh. Thank you. I did not mean to trouble you. Is there a more proper way to notify or call attention when something of the kind happens? <PuercoPop>PotentialUser-84: I'm also a new user coming from nix. Two things: my impression is that guix home is newish and people configure much less of their system using guix compared to what is achievable using nix home-manager today. This is not a technical limitation though. <PuercoPop>PotentialUser-84: The other is that guix users are more likely to value of free-software and self-hosting more than nix users, so you'll find less of the configs in GitHub. More in codeberg, sourcehut or self-hosted forgejo. Unfortunately sourcehut's search is not as good as GitHub's. <PuercoPop>PotentialUser-84: Because it was made clear that the Nix leadership didn't share the same values as me nor the part of the nix community the left. <podiki>PotentialUser-84: we have a bug/issue submission (via email), but searching on the issues website and/or asking here is good too <PuercoPop>PotentialUser-84: No, that wasn't the issue, but I feel that discussing the issue would be off-topic here. Another thing, is that I asssumed that having a full language would make the configuration a lot more straightforward compared to nixlang. So far so good with Guix! <PuercoPop>I also likethat Guix recommends a consensus driven decision-making process <aswjrisp>I noticed that a difference in the error output of reconfigure if I add a package but do not add it's module. The error asks me if I forgot a use module. Sometimes it tells me the module for that package like with aria2. Other times it does no tell me the module like with agda. It is not a problem when it does not say the module as I can run a guix search. I just wanted to point out the inconsistent <efraim>visudo, other than not being setuid, tries to run /usr/bin/vi. should we set it to use nvi? <aldum>it should be running whatever is in your $EDITOR or $VISUAL, no? <efraim>oh, I was going to see about adding a check for the sudoers file to run visudo --check but I see it's already there <efraim>I think there might be reasons to hardcode the path, otherwise I would've suggested changing it to just 'vi' <efraim>there's also SUDO_EDITOR, so I guess it's not really needed in theory <apteryx>is libera.chat overloaded, or is my network connection unstable? I keep disconnecting. <nckx>ACTION is not aware of any interruptions at this time. <nckx>efraim: visudo setuid? You sure? <efraim>visudo isn't currently setuid, and we configure /etc/sudoers using the sudoers-file, so I don't think it's necessary <efraim>although someone looking to make a quick change could run SUDO_EDITOR=$EDITOR sudo visudo <nckx>It's just that I don't understand what making it setuid would do. <divya>Hello, everyone. Does anyone use Emacs' `ess` package fo data science/statistics? How can one use it in tandem with `guix shell`? I have an environment ready with the necessary R packages, but `ess-R` isn't able to find those and thus I can't use the R packages I downloaded using guix. Any help will be appreciated. <Kabouik>I use it for R programming, but I am afraid I can't help with the `guix shell` use you have in mind, I have no experience with that. <Kabouik>My R packages are installed with Guix (either from the main Guix channel or the Cran channel), and are immediately found by ESS. <divya>Okay, I just tested. Even if I install something using `guix install r-flextable` that doesn't get recognized by ESS. <Kabouik>Is that with `guix shell` in the equation, or just directly in your normal environment? Let me check if I needed some configuration for ESS. <divya>Do you have any specific configuration for setting up ESS ang guix Kabouik? <Kabouik>I can't find anything specific to R so far, except QOL things. <Kabouik>But I see nothing related to packages. <Kabouik>Maybe the way you start R differs? I just M-x R Return <divya>Nope, that's the same thing. Thanks for the config, I'll take a look. I think it's probably because of the guix profiles maybe. <aswjrisp>When I run `guix system reconfigure /etc/config.scm` there is about 2 minutes with no output. Then it goes to output about updating susbstitues. This makes it look like it might have frozen up. It would be nice if there was a visual indication that it is working at the beginning. Like the spinner (- \ | /) that I have seen in pull or reconfigure. <civodul>aswjrisp: 2 minutes is a very long time indeed <Rutherther>Oof, two minutes is a lot, are you using some computation heavy functions? <civodul>i think that can happen when the substitute server is slow (or connection is bad) <Rutherther>aswjrisp do you get this behavior with --no-substitutes or does it start to build faster? <aswjrisp>Rutherther: I do not think I am using any computation heavy functions. It is the config you saw yesterday when you were helping me with just more packages added. <aswjrisp>civodul: I am okay with a slow substitute server. I just think a spinner or some other output to indicate things are happening would be a nice ui improvement for that command line tool. <aswjrisp>Rutherther: It is still about 2 minutes with --no-substitutes. <Rutherther>Then it must be the evaluation and not slow substitute <aswjrisp>Maybe slow hardware. I am probably not the only person with slow hardware. Some visual output would be nice. <aswjrisp>The command works so this is more of a ui enchancement suggestion. <civodul>aswjrisp: yes, i agree, there should be some feedback <attila_lendvai>should guix deploy work with dropbear? i'm getting a backtrace, the error is: Throw to key `guile-ssh-error' with args `("channel-request-exec" "Socket error: disconnected" (#<unknown channel (freed) 7ff1b3a8d520> "guix repl -t machine") #f)' <Rutherther>any idea why relocatable guix pack with python + a python package makes python repl's readline not work? it's fine in a guix shell... <tachymelia>any way to resume a system boot after an exception? <tachymelia>there's like a single failed mknod I don't care about and it's interrupting boot <tachymelia>thanks but this is on a vps and they don't let me mount arbitrary images <Rutherther>I don't even have to set any env vars related to python, and python from a relocatable pack works if there are no other python packages in the pack, but as soon as I add a package, it stops working, that's curious <Rutherther>I am unable to reproduce with anything other than a pack <Rutherther>also it's only if I execute the python from the merged profile, although it's just a symlink to python-3.10.7R. If I open python directly from there, it's fine <Rutherther>okay, I found the difference that causes this, but not what causes this difference - https://paste.debian.net/1329549/. The python3 from merged profile is missing "/gnu/store" inside of the sys.path, so the stuff cannot be loaded <Rutherther>(also forgot cd /tmp/b as second command in the paste) <tachymelia>okay figured it out: you can just. (load "/run/current-system/boot") to re-run the boot sequence without rebooting <tachymelia>or call shepherd manualy like that file does to manually activate <pjals>What happened to that one patch (39728) that added parallel downloads? <attila_lendvai>if a test is flaky, then shall i limit test threads to 1, or disable the test that fails when not on the main thread? package is python-pycurl <fireking04>Hi guys, what's the canonical way of adding patches to packages in your own channel? I'm using: <ieure>fireking04, There was a thread about this recently on guix-devel. <aswjrisp>I have installed guix on a laptop and am not getting unexpected behavior. When I power off the computer, it turns off. When I press the power button to power it back on it starts but after about 2 seconds it hard powers off. This makes the spinning disk harddrive make a bad noice. If I press the power button again a second time to the computer starts normally. <aswjrisp>This only started happening after I installed guix. It was not happening with the previous operating system. <aswjrisp>Does anyone know how to prevent it from hard powering off during boot up? <aswjrisp>I am thinking the hard powering off on boot up will destroy the harddrive. <ieure>You're probably hearing the head snap back into the parked position, which is something that prevents damage. <ieure>I don't know how Guix would have any effect on any of that. Sounds like it's powering off before any OS is loaded during the boot process. <aswjrisp>ieure: So maybe an issue with the bios? Why would it only start now after I have guix installed? <aswjrisp>I should probably see if I can find an update for the bios software. <ieure>aswjrisp, I'm honestly not sure what the problem is. I'm just also not sure how *any* OS would cause the problem you're experiencing. <ieure>aswjrisp, When it powers off, how far does it get in the boot process. <aswjrisp>I just tried to power off and then power it on. I think my previous perception of what happened was wrong. <aswjrisp>This is what it looks like is happening. From poweroff. I press the power button. The light on the power button turns on and the screen backlight turns on. It is a black screen. <aswjrisp>Then after 8 seconds (not 2 seconds) the power button light turns off, the screen backlight turns off and the spinning harddrive makes the noise. <aswjrisp>Then after a couple more seconds (during which I had previously pressed the power button but did not this time) the power button light turns on again, the screen backlight turns on again, the harddrive makes normal noises. Then I get the prompt to decrypt the drive. Things work normally from that point on in previous boot ups. <aswjrisp>Nothing shows on the screen until the prompt to decryt the harddrive. <ieure>aswjrisp, Honestly don't know. Maybe you can find someone with the same laptop make/model to see if they can reproduce the problem? <aswjrisp>ieure: It is booting but in the wierdest way I have seen. I will be regularily back up the harddrive and will try to reduce the number of power off on cycles. Thanks for the reasurance that it probably will not hurt the drive. <ieure>aswjrisp, If you can, I'd strongly recommend replacing the HDD with a SSD. <aswjrisp>ieure: Why, to protect against data loss in case there is an issue with the drive? I know SSD are faster. I will setup an automated backup of the data on this harddrive to other drives. <ieure>aswjrisp, SSDs don't make scary noises. :) <aswjrisp>Right and then I am not talking about scary noises on here. <nckx>attila_lendvai: Why does this feel like you know the answers already… ;-) The Highly Korrekt answer is skip the test in the 'check phase and run it alone, in a separate phase. But if the test suite isn't that long you can get away with -j1. In either case, note the reason in a comment. <attila_lendvai>nckx, meanwhile i've decided to just skip the test. i needed to skip another one anyway (expired CA) <attila_lendvai>but the decision wasn't obvious... i'll certainly not introduce a new phase to run two tests in single threaded mode. i'll leave fixing their code to upstream. <Rutherther>what should I type in e-mail body so that mumi web interface lists it as a code section? <nckx>‘i'll certainly not introduce a new phase’ is what I expected. <nckx>I don't understand why, but it's fine to skip it. Plenty of other flaky tests are. <ArneBab>With that fix, I have a working build of unbescape. <attila_lendvai>nckx, wasted effort. i'll spend the same time on updating the package (doing that now), or reporting the issue (long-term fix) <attila_lendvai>meanwhile: the next version fails with address already in use. its test system is not thread safe, needs n=1 <attila_lendvai>heh, the entire package db seems to depend on curl... :) let me just throw away these pending updates, and deescalate this fix. <nckx>I'm a bit surprised that our everything-under-the-sun-depends-on curl is also the end-users-install-this-one curl. <nckx>Not disappointed, just surprised. <attila_lendvai>yeah, not ideal. but it wouldn't help me here, because the next python-pycurl needs a newer libcurl. we would probably need to have 1) curl-latest-to-install, 2) curl-for-packages, and 3) curl-for-packages-in-the-next-core-updates <attila_lendvai>Usage: guix gc [OPTION]... PATHS... but it should be [PATHS], and what's worse, it does not accept any paths: this fails with extraneous arguments: guix gc /gnu/store/[hash]-python-pycurl-7.45.2 <ieure>attila_lendvai, I think it should be `guix gc -D [PATHS]'. <ieure>The guix CLI could use a lot of work IMO. <nckx>Comparatively nitpicky, but shouldn't it be [PATH…] then? <ieure>What's a reasonable way to test a patch which alters `guix system reconfigure' behavior? I tried changing my ~/.config/guix/channels.scm file to point to a local repo, but `guix pull' complains about no branch tracking refs/heads/master (or something very similar). <ieure>Using `guix shell -D guix --pure' and a pre-inst-env doesn't work because sudo isn't available. Adding sudo to the pre-inst-env doesn't work, because it's not suid. <nckx>Does it really require ‘reconfigure’ and not, say, ‘build’? <nckx>I'd just put it in my local repository, which is what's already in my channels.scm. I have no idea what your tracking branch error is about. <ieure>nckx, Hmm, well, I'm changing the `guix-activation' procedure. Does `guix system build' activate the built system? <nckx>There's also no reason you can't use $(which sudo) with --pure, it's not a --container. <nckx>That answers my question. <nckx>(or sudo guix shell …? Or maybe there's some gotcha I'm missing.) <ieure>nckx, Pretty sure which(1) checks for its arg on $PATH, which doesn't include sudo if the --pure flag is given. I could omit --pure and use sudo, but I'm not sure if that would cause pre-inst-env issues. <nckx>At least if you mean ‘guix shell -D guix --pure -- $(which sudo) ./pre-inst-env guix system reconfigure’. <nckx>You're not running which in the pure shell, after all. <ieure>I can give that a try, I'm not used to hacking on Guix like that, though. <nckx>I'm not sure exactly which permutation of all these things will work, and which will have some hidden gotcha, but I'm positive there's one that will work fine. <nckx>ACTION away to leave you permutin' in peace. <aswjrisp>I am trying to setup openntpd-service-type. system search tells me it is from the service module networking. I have (use-service-modules networking ...) and (service openntpd-service-type) in my list of services. <aswjrisp>When I reconfigure I get "guix system: error: service 'ntpd' requires 'networking' which is not provided by any service". <mirai>aswjrisp: you need a service that provides networking <mirai>e.g. networkmanager, dhcp or static <aswjrisp>mirai: That is interesting currently I am manually running ip and dhclient to get network connectivity using an ethernet cable. <aswjrisp>I know I do not want networkmanager. I am fine manually running dhcp so I don't think I want that as a service. <aswjrisp>mirai: I will look into the static service. Thanks for the tip. <aswjrisp>static-networking also turn on networking automatically. <aswjrisp>I guess what I shoulp probably do is forget about the services and just install the package for openntpd and start it when I want to after I manually connect to the internet. <aswjrisp>Maybe a bash script to connect to the internet and turn on ntp and ssh. And another one to turn them all off. <aswjrisp>Is there a way to do this in a more quix way? <ieure>aswjrisp, The normal Guix way to do this is with services. <aswjrisp>Or maybe I can configure those services to be off. Then I start them with herd when I want them started. <ieure>aswjrisp, If you want to sync time without a daemon, ntpdate is the best way to do that. I'm not sure what package that's in. <ieure>I'm just confused why you're asking for the "Guix way" to do this when you've been told what that is and have decided not to use it. <aswjrisp>ieure: No I think I would like to have the services. I am just not sure how to have them default to off so I can manually start them with herd when I want. <aswjrisp>I did read the services section of the manual as well as the system configuration section. Maybe I missed the part on having a service default to off. <aswjrisp>Like how ssh-daemon is off by default on the install image. <ieure>aswjrisp, Looks like it's a special %auto-start? field in the openssh-configuration, and not a thing you can control for every service. <nckx>Apologies if ieure sent their answer before mine. But don't waste your time searching for ‘disabled by default’ services; they aren't a thing. <attila_lendvai>is it only me, or `make check-system` is broken for months? my work toward fixing it is ignored (e.g. it's a simple, non-controversial build fix: https://issues.guix.gnu.org/72207). what's the deal? no one is running the system tests? <aswjrisp>nckx: thanks for letting me know. A disable by default option for services would be nice to have. I think someone else must have already made this feature request. <aswjrisp>I guess I will go with a shell script to turn on daemons as that will give me the off by default behavior. <ieure>aswjrisp, Is this just for ntpd? <aswjrisp>ieure: turning on networking interface, wpa_supplicant, getting ip address, starting vpn, starting npt and starting ssh. <ieure>Why don't you want that to happen automatically? <aswjrisp>I only want them on if I am planning on using them. <aswjrisp>Download documentation, turn off internet and focus on a task. <ieure>I see that I'm unlikely to get anywhere productive, so, I hope you find a setup that agrees with you. <aswjrisp>ieure: thanks for your help and thanks for mentioning ntpdate.