IRC channel logs
2024-10-09.log
back to list of logs
<__cornflake___>also can I just download your file and install it? or should I add your repo as a channel? <PotentialUser-0>Rutherther I made it work with distrobox, than you. I just added the policy.json and also started it with root. After that I started installing a few packages in fedora it started to work. <another-fake-san>hi folks - I am trying to use guixsd on a laptop, and having trouble getting nmcli/nmtui available trying to add the network-manager-service-type to my operating-system config. When I try to add it I get "service 'networking' provided more than once.". this is on a fresh install from the 1.4.0 ISO using %base-services, any pointers? <__cornflake___>another-fake-san: could you please create a pastebin of your system config file? you can do it on paste.debian.net with the "upload file" button, then post the link to the pastebin here! <jaft_r>another-fake-san: ditto to what __cornflake___ mentioned though, just from your description of your problem, it sounds like you might be trying to include network-manager-service along with the normal default %desktop-services; those already include the network-manager-service which is probably why you're getting a warning about it already being provided. <another-fake-san>I'm building off %base-services and not %desktop-services; but I think I just figured out what was wrong; I had added dhcp-client-service-type which appears to mutually exclusive with using network-manager-service-type ? <Guest79>Hi is https://ci.guix.gnu.org still a valid site, sorry new to guix and trying to figure out an issue installing guile-emacs ( is that still maintained? ) <__cornflake___>another-fake-san: I'd love to help but it's pretty hard to figure out what's going on if we can't see your config file. We'd be able to give you more help if you created a pastebin of your config file and posted it, then we could look at the file too. <__cornflake___>You can create the pastebin at https://paste.debian.net by clicking the "Browse..." button at the bottom of the page, then selecting your file, or by pasting the contents of the file into the text box, then click "Send" and copy the link it generates & post it here <bigbookofbug>i'm running into an error with trying to compile the sample program from the guile reference manual. running 'gcc -o simple-guile simple-guile.c \ <bigbookofbug>in a shell, i recieve the following error: line 30 of scm.h: "libguile/scmconfig.h: No such file or directory" <bigbookofbug>i was advised to ask here in #guile, as i feel that guixes filesystem may affect the loadpaths. i have #include <guile/3.0/libguile.h> in my program currently. are there any guix-specific flags i need to pass to compile c programs in guix shell? <Rutherther>sneek, later tell *cornflake*: you don't have to use my repo as a channel. You can either copy the function where you need it or the whole file. For whole file, just add the folder you put it in to your load path (-L) and adapt name of the module <Rutherther>Matrix heisenbridge translated the name wrong :( <Rutherther>Gooberpatrol_66 you've put gexp inside of gexp without ungexping the second one #~#~, so the second is passed to the daemon and it doesn't know it (nor it should) <meaty>I know guix can make graphs to visualize sevice dependencies and has 'guix size', is there a way I can use that functionality to create a kind of 'store usage analyzer' that tells me which packages are responsible for the most disk usage <meaty>I guess I should actually browse the mailing list <simendsjo>Think I found a bug, but can someone else check to see if the following fails before I report it? guix shell -e '((@@ (gnu packages embedded) make-arm-none-eabi-toolchain-7-2018-q2-update))' <futurile>meaty: you should ask efraim when he's around - as he is the *entire* rust-team heh <efraim>I'm deep in bootstrap territory this week and likely next. As far as merging the rust-team branch there's a compatability issue between icecat and the version of rust, so I can either downgrade rust so it just works (not sure how far), work with gnuzilla to support the next ESR branch or create a patch to update the offending crates to newer versions <Rutherther>simendsjo I cant check now, but could you send the error? <efraim>I think once I unbreak the x86 bootstrap chain I'll spend the time to merge my many squash and fixup patches and rebase the branch <simendsjo>Rutherther: Throw to key `lzlib-error' with args `(lz-decompress-read 5)'. substitution of /gnu/store/mrbb807fzq3vvxbpj1lqbdxwlq596b99-gcc-arm-embedded-7-2018-q2-update-1.261907-checkout failed. guix shell: error: corrupt input while restoring archive from #<closed: file 7fd576335d90>. <efraim>simendsjo: try restarting guix-daemon <simendsjo>efraim: Thanks, restarting guix-daemon worked! <neox>Hi there, is guix.gnu.org down ? <sneek>Welcome back neox, you have 3 messages! <sneek>neox, mirai says: exactly what value/path is to be set for TEXMF for the wrapped dblatex? <sneek>neox, andreas-e says: texlive-team has been built successfully on QA, I let you the pleasure to merge and delete the branch. Congratulations! <sneek>neox, andreas-e says: This will break the compatibility of the monolithic texlive with packages such as texlive-biber, right? So I am not too fond of doing it for the time being. <nckhexen>neox: Not g.g.o itself, but services like ci.g.g.o are indeed down with the sickness. I try to note dowtime in the channel topic whenever I'm around. <nckhexen>Some g.g.o (download) links do point to CI, so those will be broken unfortunately. <nckhexen>Based only on the fact that I can't SSH into either berlin or its back-up rackmate, this seems like a datacentre problem. <simendsjo>Can I add udev rules directly from a file? /etc/udev/rules.d is read-only. I have some udev rules for qmk, but I don't want to add this for my entire system, just use it adhoc. How can I add udev rules when I cannot copy it to rules.d? <meaty>what's the proper order of operations for doing a system reconfigure just after a pull? just reconf? upgrade>reconf>upgrade? <Guest79>after a pull I did "sudo guix system reconfigure /etc/config.scm" meaty <meaty>Also, is it possible to upgrade/administer/etc. a guix system without rebooting? <Guest79>meaty - I think I read that a reboot is best, however I am new to this so ymmv. <mange>I believe a reconfigure will do most of the same things as a reboot, but it won't restart running services. You can do that with "herd restart X" for the services you're interested in restarting - Guix won't restart them for you. <mange>There were some patches towards automatically restarting services, in https://debbugs.gnu.org/cgi/bugreport.cgi?bug=33508 (issues.guix.gnu.org is down, so this was much harder to find than I thought it would be!). That work didn't really go anywhere, though, and I'm not aware of any further work since then. <abbe>with the recent outages to gnu/fsf infrastructure, what's the best way to use the mirror of guix repository ? just update $HOME/.config/guix/channels.scm with mirror URL ? or something more ? <rekado>nckhexen: I'm not working right now, so I'm out of the loop; I do see a reminder in my work mailbox about planned maintenance of the medium and low voltage supply to the data centre. "Dienstag, den 08.10.2024 ab 22.00 Uhr bis Mittwoch, den 09.10.2024, 06.00 Uhr" <civodul>abbe: either change the URL in your channels.scm, or just ‘guix pull --url=…’ with a mirror URL <sneek>civodul, you have 1 message! <sneek>civodul, dariqq says: I think i know what the cause of my privileged-programs troubles is. The first time we try to get the gid of a new group is failing because group does not exist yet. And after reboot groups are not recreated yet so it is using the info from the previous /etc/group . (This "works" for groups and therefore opensmtpd because "group not found" is a 'system-error being caught by the exception-handler and "user not fou <rekado>nckhexen: best get in touch with Madalin for a visit to the data centre. <abbe>thanks, so i'll propagate this channels.scm change to all my VMs, and hopefully that'll be all. thanks! <efraim>don't forget a straw to breathe! <abbe>nope, was just curious about this <meaty>tests fail fore building the latest lsof, anyone else have this issue? <nckhexen>meaty: I've heard talk of it. Clearly non-deterministic, although it might be consistent per machine. <futurile>all - don't forget that we're doing an online meet-up later today - so come along if you just want to hangout in a 'hacking room' - or do some patch reviews! <futurile>(might be less hacking and more chatting if our services are down heh) <mehrad>I'm trying to have dev env for an R project and I want to use guix.scm, but the issue is that some of the dependencies are not in guix proper repo. Is there any way to add channels to the `guix shell`? <mehrad>some of my dependencies are for example in guix-cran channel <jlicht>mehrad: you could use guix time-machine with a --channels=your-channels.scm <mehrad>jlicht: thanks, but doesn't that mean that I have to also tie the channels to specific commits? <mehrad>I already have a time-machine for the project, but I though the dev env should be kinda based on the HEAD and not a time in history <Rutherther>mehrad if you don't specify commits of the channels, newest will be fetched every time. Which might not be what you are after as well <Rutherther>Could be scripted with something like Makefile to create guix instance under the project based on the channels.scm. So you could have sort of a project specific "guix pull" <mehrad2>Rutherther: thanks, I didn't know that. Then only advantage of guix.scm would be using `guix shell`, where as here I need to have a larger started command. <boo_>Hello! I haven't been able to reach ci.guix.gnu.org for a while. I just wanted to know if anyone knows what the situation is? Is there any sort of ETA? Thanks :) <boo_>futurile: Alright, the issue is that bordeaux doesn't have all the substitutes, but oh well. I'll have to let it build in the meantime. <cbaines>boo_, what substitutes is bordeaux missing? <boo_>cbaines: Actually, now that I try rebuilding it looks to have more substitutes. I remember ibus and some python packages from yesterday, but I didn't watch it too closely, more just saw that there were many packages and let it run. <boo_>or, more accurately, it seems to have substitutes for the packages I tried to install today. <galois>The error comes from including <iostream> or <string> or similar <ieure>I thought mumi was a client program that interacted with the mailing list. <weary-traveler>ieure: it also provides the web interface accessible at issues.guix.gnu.org unless i'm mistaken <ieure>Guix infrastructure seems to break way more often than it should. <dariqq>omg, the order of services is important for the order of the activation scripts .... I guess it makes sense, but is not really something i am thinking about <civodul>hey dariqq! saw you found the reason for the setgid failure <dariqq>would it be possible to move the accounts-service type further down / privileged-programs further up in operating-system-default-essential-services? have not tested it yet but the resulting folded activation-service has a different ordering? <dariqq>yay, the warning from privileging is gone, Test still fails though <dariqq>This seems like a bug in the test not using the setgid smtpdctl though <dariqq>cant tell you this. I guess nobody tried to set something setuid to a nonroot user (that currently causes a bad crash). The setgid stuff "works" after a reboot, it would only be visible as a warning during the first reconfigure <nckhexen>boo_: There was planned power supply maintenance at the MDC data centre today and the machines haven't responded since then. Someone who works there should be able to investigate but I've not heard back yet. Probably not for today (Eurotime). <nckhexen>futurile: Good luck! At least Savannah is up for now. <weary-traveler>ieure: yeah, my normative expectations when it comes to services like mumi are "almost never". reality is more like "every so often" <boo_>nckhexen: Makes sense. I have a system which can build substitutes for my weaker x86 machines, so that's fine for the time being. <dariqq>Not sure why the opensmtpd test is still failing . Is there a way I can get more output from the test-runner? <boo_>By the way, I do have a question regarding substitutes; I have a weaker ARM system which I want to make sure has substitutes available for heavier packages like a kernel with a custom configuration later on. I would like to host substitutes but the hashes don't match for the same package even when built for aarch64-linux-gnu, so it doesn't get used. Where would I start looking to find out why? <boo_>To be clear, my build server is x86-64. <Rutherther>boo_: so are you cross compiling? the hash won't match between cross compiled packages and native ones, because the dependencies are different (cross compilation toolchain as opposed to native one etc.) and thus the hash is different <Rutherther>probably will be easiest to use guix deploy instead of substituting <dariqq>aah, opensmtpd seems to prever /var/mail instead of /var/spool/mail and the test looks in the wrong directory <Rutherther>boo_: other than that you could build with emulation, but that's quite slow. Or ensure that you are getting the same cross compiled version on the other system (so you would basically be building some of the packages like on x86, but cross compile to aarch64 - since they are going to be substituted you might not even have to setup qemu emulation, but I am not sure about that.) <dariqq>Something is overriding the mbox directory: checking system mail directory... /var/mail from _PATH_MAILDIR <x8dcc>Hello, I am considering switching to Guix as my primary package manager. I was wondering if there is anything similar to Gentoo's USE flags. <ekaitz>x8dcc: what are gentoo's USE flags? <x8dcc>They are used by Gentoo's package manager to essentially enable or disable features of a package, usually at compile time. For example, you might want to compile X package without bluetooth support, or decide to use some interface instead of some other (e.g. gtk/qt/ncurses) <x8dcc>In Gentoo this makes a lot of sense since essentially everything is compiled from source. I am not sure if it makes so much sense in Guix, though. <sepeth>x8dcc: not something general but if you need a specific flag for a specific package, it is easy to modify one, you can inherit a package and modify it as you like, assuming you have your own channel (or contribute back to the original package if it is useful to everybody). <x8dcc>sepeth: That sounds very interesting. Sometimes it's more about "preference" that "usefulness", but I already assumed I would need my own channel at some point, at least from reading the Guix docs. <ieure>x8dcc, Running your own channel is pretty straightforward, I set one up within a few weeks of using Guix. The thing that's harder is setting up your own substitutes, if you have multiple machines that need to build big projects. <ieure>Substitutes may not be needed for your usecase. I packaged LibreWolf, because that's my daily browser, and it takes ~1hr to compile. Not much fun on a laptop. <sepeth>x8dcc: portage may have guix already perhaps. You can try it out as normal user as well. <luca>1hr sounds quite quick for what is essentially a firefox fork <ieure>luca, Doesn't feel quick when you have >1 machine that needs the package, one of which is scorching your lap while it compiles. <x8dcc>1h of compilation is very standard in the Gentoo life. Sometimes it's easier to just run a crypto miner instead, and buy a computer that has firefox installed :) <x8dcc>sepeth: I started using Guix on my main machine, running Arch. Since my daily packages were already installed, I haven't used it too much yet. <sepeth>guix shell still could be useful for projects you working on <x8dcc>ieure: About "substitutes may not be needed for your usecase", do you say that because there might be other alternatives to compilation? <x8dcc>sepeth: Yes, I haven't read about it in-depth, but I did read about nix shells back in the day, and they looked very, very interesting. <x8dcc>However, I hate nix's syntax with all my life, and also love Lisp with (almost) all my life. <ieure>x8dcc, No, I say it because if your channel only has stuff that builds in a couple minutes, you probably don't have to care about rebuilding it. <ieure>But when you have two machines that both need a package which takes 1hr to compile, it's a real drag. <sepeth>x8dcc: nix packages have a lot of bash inside it as well :( <x8dcc>Right. Same problem happens in Gentoo. I like the whole USE flag deal, but having to recompile a big ass package whenever you decide to change some detail is a pain. Not even talking about updating the whole system. <x8dcc>sepeth: Not sure what you mean by that, sorry <x8dcc>And as a side note about Gentoo, I am a bit tired about their package manager since it's written in python and it can be really, really slow. That's why I considered getting the hang of Guix as a package manager in my Arch machine, and switching to the Guix distro in my other Gentoo machines. <x8dcc>sepeth: Oh my god... I understand now. <Rutherther>x8dcc: in guix source you know the whole dependency tree of packages and you can map the whole tree with a function called "package-mapping". So if you wanted you can add some optimization flags, swap library for another etc., of course it will be up to you to specify the exact transformations for each package, but there are standard build systems so usually you could be able to just declare something per build system. <luca>Any tips to getting `./pre-inst-env guix build <rustpkg-here>` to NOT build rust itself? <Rutherther>luca: that's not possible, rustc is needed to build the rust package <luca>Can I not ask it to use an existing substitution? Since I don't want to build rust itself, but something that is built using rust. Such as a crate or app <Rutherther>luca: if that specific rustc you have in that revision is available from a substituter you have configured, it will be substituted <luca>Hmm ok, running `./pre-inst-env guix weather rust` does tell me no substitutions are available <luca>If I am on the newest checkout of the rust-team branch, should I rebase against master? Or what? <Rutherther>luca: rebasing against mater cannot help you with rust package, if anything it could make it worse as dependencies could change and the builds happen on rust-team branch itself. Normally you can find what branch revision was built on https://ci.guix.gnu.org/, but it's currently down. Maybe it even has the current rust-team revision built as well. <luca>Actually I am getting `guix weather: warning: ci.guix.gnu.org: connection failed: No route to host`, so maybe it's just bad timing on my part <luca>Hmm, `./pre-inst-env guix weather rust --substitute-urls=https://bordeaux.guix.gnu.org` says 0.0% substitutes available and `(continuous integration information unavailable)` <luca>but it works from master, sooooo I am lost <nckhexen>bordeaux. doesn't share job specs with ci. <x8dcc>Rutherther: I don't know about this mapping thing. I will try to remember it in the future. <nckhexen>luca: If this is only about your machine struggling to build Rust in a reasonable time, I could build it for you. <luca>I mean I guess I could build rust myself, but it seems like I shouldn't and that there should be something wrong I am doing <luca>Hmmm ok. Then building rust it is <nckhexen>There's nothing wrong with what you're doing. <Rutherther>even if ci. was up, you can even get substitute misses on master, and on team branches it is worse, they are not built that often <nckhexen>You're building a non-upstream(/master) Rust, and the only machine that builds the same Rust on a schedule is currently down. <nckhexen>I can manually ask a server to build it, but building it at home is technically safer. <luca>I see, so a combination of bad timing and non-master branch <nckhexen>And the tech team at the MDC data centre being a bit of a klutz, TBH, going by their recent record. <luca>No need for a builder, I think I have a powerful enough of a machine to build everything myself. Thanks for the offer though <nckhexen>The Max Delbruck Centre in Berlin, where CI is hosted (hence its hostname ‘berlin’ you might see mentioned here). <vdtoorn>Hi guix, is this maintenance on the services? Or is it unknown when everything will be available? Thanks! <ekaitz>hi! i'm having issues with ruby dependencies... is it normal? <vdtoorn>ekaitz that's why I'm here, but apparently services are down <vdtoorn>I have it in a message above the chat <luca>Yes ci.guix.gnu.org is down, bordeaux.guix.gnu.org is still alive and kicking <ekaitz>each time i want to upgrade something happens <vdtoorn>yes, hopefully it will be fixed soon <nckhexen>Don't count on it until the sun rises over Berlin, at the least. <nckhexen>Christ that was painful. Weechat + mosh + emojos = pain. <dariqq>my service-type that I was originally trying to write that started this quest is now working with setuid/setgid. Crazy that the solution is as simple as changing the order in a list. But learned (maybe a bit more than i wanted to know) about the service extension mechansm. Definitly a very unique and elegant system <futurile>it's got to be the worst sort of 'broken' when you're stuck and can't login <shudder> <luca>huh, emojis in topics are messing up my weechat a bit. Neat <nckhexen>I want it to be mildly obnoxious and there's still no support for <blink> yet, so… <vdtoorn>allright, I'll be here waiting until the sun comes up over Berlin ;) <nckhexen>Not to be a killjoy, but that's a lower limit :-) We're at the mercy of remote hands. <boo_>Rutherther: I didn't actually know about Guix deploy, will look into it. It looks really promising for how I want to manage and push configuration updates. Thank you :) <luca>nckhexen: Nah not too badly, just channel numbers on the left side are missing, so it goes "22." then "|3.", and it only happens when I am focused on this channel. I'd say it's acceptable for me <luca>Oh the line comes from the previous line in the terminal <luca>So I guess it more so breaks my terminal emulator (foot)? eh, small enough for me not to bother to report as a bug <luca>Anyway after a couple hours of stage0 rust builds I ended up with 'failed to select a version for the requirement `once_cell = "^1.20"', so that's fun <luca>so idk if it's appropriate of me to update the existing rust-once-cell-1 version or not, but might as well give it a try <luca>and rust-regex and rust-sysinfo and boy does the rabbit hole go deep <ieure>Rust deps are as bad as node_modules. <nikolar>if not worse, because rust deps go much deeper in the tree <nikolar>take a look at why mpv isn't building on aarch64 <[>nckhexen: some irc clients support blinking text, but solanum strips colours from topics <[>scrolling past the topic change messes up my terminal so I need to /redraw <nckhexen>You're talking to the bore who set +c on the channel to begin with. <dthompson>I keep hoping the "rewrite it in rust" meme will die soon... <dthompson>it feels like the programming community at large is starting to become a little more open to criticism of rust <dthompson>but criticizing language package managers like cargo and npm is just as poorly received as it was 10 years when I got involved with guix <nckhexen>Pity, since they are far worse than their respective languages. <dthompson>maybe the mpv issue would be a good case study for a blog post <dthompson>and using some guix graph output to show off some absurd dependency graphs <nikolar>dthompson: "rewrite it in rust" isn't slowing down yet <[>linux takes long enough to compile as it is <nikolar>but don't you say that in the presence of a rust user <[>I hate how so much rust software requires nightly rust, which is basically unpackagable (and will probably be outdated by the time you even finish building it) <[>Rutherther: because it updates nightly <[>you can package it but it will be out of date almost instantly