IRC channel logs
2024-02-29.log
back to list of logs
<darkexior>Hello guys, i installed a game called widelands from terminal <darkexior>but i cant open it from terminal by typing widelands <darkexior>it has happenede before with a couple other games <mange>When you say "can't open", what do you mean? Are you getting an error message? <darkexior>no, there's just no command for opening it that i know of <darkexior>like, i can open minetest with the word minetest <darkexior>the error is that command is unknown for the bash <old>lol what: guix build: package 'widelands' has been superseded by 'widelands' <mange>I don't know the answer, but if you run "ls $(guix build widelands)/bin" that might help you find the right command to run. <darkexior>okay thank you that is useful but its not doing anything at the moment, probably its on widelands side? ill have to check later, thank you my friends <old>darkexior: widelans id under /gnu/store/.../bing/games <old>I believe it is not correctly packaged <darkexior>im doing this and ls $(guix build widelands)/bin/games/widelands <old>so if you can to play it: run /gnu/store/.../bin/games/widelands, where .. is the sha256 value from guix build widelands <old>what about ls -R $(guix build widelands) <old>ls -R $(guix build widelands)/bin ** <old>you will get lots of noise otherwise <old>I get: /gnu/store/fvjlvns4qff4i83ylh6jksj79n29wqmb-widelands-1.1/bin/games/widelands <old>do you get the same sha256? <darkexior>my sha256 is gnu/store/2154pjqmmg0k5b2g524nplv5rfvyzc54-widelands-1.1/bin/games/widelands <old>okay so maybe the package has been updated since <old>I'm at bf7991a8c767abd32cfb2c92e3d57665a7cabf00 on guix main channel <mange>It doesn't look like the package definition has been updated in the last year. <mange>At any rate, if you ran "guix install widelands" you should be able to run ~/.guix-profile/bin/games/widelands <old>ahh the change in sha256 could be explained by its inputs changed I guess <hapst3r>Hi guix o/ can someone please reply so I know I successfully connected? Thanks :-* <hapst3r>Is someone here experienced with installing guix on ancient Hardware and has got some tips or knows good resources on the subject? After failing to install on a MacBook pro 2010, I am at my wits' End when it comes to installing on an ancient <efraim>I used to have an x120e until it fried itself in my backback <efraim>I would assume it should just work on those devices, although wifi is always iffy <hapst3r>efraim: OK but was there anything specific about the Installation process on the x120e? Because using the guix-provided image, some parts would not work like gpu and wifi (expected), but the Installation failed saying not enough space available (I suspect either the root partiton was too small or I chose EFI over grub by mistake). <efraim>I think on that machine I used the whole drive for /, although I don't remember if it was an EFI machine or not. IIRC the installer is supposed to "get it right" between EFI and plain grub, although formatting the drive as GPT would probably help <adanska>hmm.. mutter failing a test on master <adanska>will rebase gnome team onto master and see if theres problems there too <hapst3r>efraim: thanks, will Experiment with that :) <efraim>mutter built for me on x86_64 on master on my machine, also having it build on berlin <adanska>efraim: i feel like im going bananas! just an hour ago, the build was consistently failing and now its working perfectly. didnt pull or anything <efraim>more tests than I would like fail under high load <adanska>i guess it was in the midst of a reconfigure, but when i built from my checkout that was fairly standard load <fnat>During a fresh install, I run `guix system init /mnt/etc/config.scm /mnt', which seems to terminate successfully. But then, after rebooting, I get stuck at a grub prompt, with error `disk lvmid/... not found'. Any ideas on what I should be checking? <fnat>In config.scm, I've added cryptsetup-static and lvm2-static to my package list. <fnat>I've also added dm-crypt to the initrd-modules. <fnat>My drive has a GPT table, with a /boot/efi partition plus a LVM-on-LUKS partition with root and swap. <fnat>Unfortunately that doesn't seem to have been resolved, at least by the way the ML thread looks. <jpoiret>it doesn't work for now until we update grub <jpoiret>I've got a patch for grub2.12 locally if you want to try <jpoiret>it's just not ready for upstream unfortunately (also it'll need loads of tests) <fnat>Yeah, I'd be glad to try it, thanks! Alternatively, should I switch to LUKS1? <jpoiret>switching to luks1 would work as well, yes <jpoiret>and once Grub is fixed, you can convert to luks2 in place <fnat>(Do you know if I can also retro-convert LUKS2->LUKS1 in place?) <fnat>`cryptsetup convert --type luks1 /dev/my-dev' seemed to work just fine. Now going to see if it works. <fnat>Hm... it still fails to boot... grub reports the same error re `lvmid/... not found'. <fnat>ACTION scratches their head. <fnat>jpoiret: The manual seems to mention that LUKS2 + pbkdf2 is ok? I've tried converting to LUKS1 and I get the same error. Anything else I could try/look at? <ayatsfer>in nix it's possible to take a package, and change some of its attributes, for example to change the src: "mypkg.overrideAttrs (oldAttrs: { src = ....; })" Is this easily doable in guix? <lysgaard>I would like to modify the `gcc-toolchain` package to add a `cc` binary which is just a symlink to `gcc`. I have tried a lot of things, but not succeded. This is the current snippet I have, it compiles, but here is no `cc` available in it: <lysgaard>Anyone have tips on on how to proceed? It seems as my extra phase simply is not executed. <jpoiret>fnat: ah, just converting the FS won't work, you need to also reinstall the bootloader <jpoiret>so a chrooted reconfigure would probably work <jpoiret>lysgaard: is there any reason for that? <jpoiret>most build systems respect the CC environment variable <fnat>jpoiret: Ah, I see! Thanks, going to try that then. <lysgaard>jpoiret yeah, I have some rust packages I want to compile, and they do not respect the CC variable :( <jpoiret>lysgaard: and you can't substitute cc with gcc in the source? <jpoiret>that's probably the easiest solution <jpoiret>rather, substitute "cc" by the result of cc-for-target, see other packages using that procedure <lysgaard>I would like to build the rust package not using guix, but using rusts cargo. I am using guix to create a reproducible build environment, but would like to use the standard rust cargo build system for the actual source I am working on and its rust dependencies. <lysgaard>So I am using guix to get a reproducible system (native library) situation, and then rust cargo to build inside this environment. <lysgaard>jpoiret does that make sense, or am I doing something stupid? <jpoiret>yeah, taht does make sense. Are you using the fhs-container thing? <jpoiret>you should be able to symlink gcc to cc using it without resorting to a new definition for gcc <jpoiret>and in any case you could also add a trivial package with a single symlink if that doesn't work, instead of rebuilding all of gcc just for that <lysgaard>I am not using a fhs-container thing. Wouldn't only overriding the install step use the cached build outputs from substitutes? <lysgaard>jpoiret, if substitutes don't work with that, I guess my best guess would be to define a package. Any clues how to do that? Do I need the symlink to point to the path of the `gcc` binary in the `gcc-toolchain` package? If so, how do I query the `gcc-toolchain` package for the path of one of its outputs? <jpoiret>lysgaard: no, phases aren't independently built and substituted <lysgaard>Then I guess my best option is to have this `cc-appendix` package that somehow symlinks to the `gcc` package form `gcc-toolchain`? <jpoiret>lysgaard: something like (symlink (search-input-file inputs (string-append "/bin/" #$(cc-for-target)))) in a phase <jpoiret>uh, forgot the second argument to symlink <jpoiret>that's the base gist of it, then you probably want to use trivial-build-system, have that as the builder but use (mkdir-p (string-append #$output "/bin")) before <janneke>lysgaard: why not patch the packages to respect $CC and send the patch upstream? <lysgaard>I do want to do that, but I also want to get on with my development <lysgaard>jpoiret I am making the trivial package, but it requires a source, which this package doesn't really have? What does one do then? <lysgaard>janneke I will try to submit a patch, the problem seems to lie in the libc bindings of rust, which in some way is the most fundational package in the whole rust ecosystem <graywolf>Is it expected that guix system reconfigure respects umask? <jpoiret>graywolf: well, respecting umask is the default behavior unfortunately <jpoiret>but yeah, I also think we should reset umask <graywolf>Ah ok. I will adjust my script to reset it. <jpoiret>same goes for all other guix commands though <graywolf>For me it meant the system was not useable without reboot after each reconfigure <ekaitz>i'm finding many issues related with this <lysgaard>Is it possible for `guix environment` to set environment variables in a declarative way, eg. declare them in a manifest? <lysgaard>Eg. `guix environment -m my-manifest.scm` and when I am inside this shell, I would like some environment variables set to specific values that are a function of some scheme code in my `my-manifest.scm` file. <old>lysgaard: AFAIK, guix environment is deprecated. Use `guix shell'. For the environment variable themself, they can only be passed through the --preserve option I thik <jpoiret>lysgaard: for env variables I usually combine `guix shell` with a shell script that does `export XXX=VVVV; export YYYYY=WWWW; exec $@`, just like guix's pre-inst-env <lysgaard>jpoiret that makes sense, but what if the environment variables directly relate to paths that are computed during execution of my scheme code, eg. store paths for different packages? <old>lysgaard: also one way would be to do something like `guix shell --preserve=COMMON_PREFIX.+', then in your manifest you can do `(setnv "COMMON_PREFIX_..." "foo")' <graywolf>I wonder if you could create a dummy package and (miss)use search-paths? <old>ofc that works only if you control the environment name that you're targeting <jpoiret>lysgaard: you can refer to the profile of the environment with $GUIX_ENVIRONMENT, and use eg. $GUIX_ENVIRONMENT/lib/pkgconfig/ <graywolf>If I do, `guix shell -C coreutils gtk+ -- env', for examble XDG_DATA_DIRS is filled. Cannot the same mechanism be used for arbitrary environment variables? <jpoiret>graywolf: you could, but it's often less trouble than a shell-script <jpoiret>and also the scope of what you can do like that without modifying existing packages or rebuilding them is quite narrow <graywolf>I mean I could just have dummy package doing just the environment directly in the manifest file <graywolf>(I am not saying this is better then shell script, I was just curious whether it is possible) <reedm>Is it possible to inspect the data for current system services using "guix repl"? <reedm>I want to write some custom services, but I'm having trouble understanding the way "extensions" work, and figuring out the service-types for the services listed by "herd status" <rekado>reedm: exactly how extensions work depends on the service implementing them. <reedm>rekado: If I just want to do a "simple-service", is there an easy way to determine the service-type of something like "file-system-/data" for a file system configured in /etc/config.scm? <reedm>If I wanted to make a service that does a chmod/chown on all files in /data, would I do something like (simple-service 'change-folder-permissions <file-system-/data-service-type???> #~(<code goes here to actually change the ownerships, etc...>))? <freakingpenguin>reedm: Part of the confusion is that shepherd-service != guix service (or service-type). A guix service may create a shepherd service, but not all do. <freakingpenguin>file-system-service-type or fstab-service-type is probably what you'd want to look at, unfortunately I don't think the manual documents them. This would be for mounting/unmounting a file system though, not running a recursive chmod. <reedm>freakingpenguin: got it. I think I need to dig through the code a bit more. I've read through the services portion of the manual a few times and still don't understand how to actually hack on any of it. <freakingpenguin>When writing services I usually just throw some debug code in the file to create + fold services & extensions and see if my services compose+extend properly with the repl. <reedm>freakingpenguin: Can you point me to a minimal example I can start with? <freakingpenguin>TL;DR is I wanted to adjust the xorg configuration from multiple services, but guix login managers (gdm, sddm) only use the first extension and discard the rest. Wrote a custom service that extends login manager and can be extended multiple times. <Altadil>Hi, I don’t know if it’s been reported already, but the download of the guix latest USB/DVD ISO installer seems down. It times out. The one for guix 1.4 is fine. <pkill9>davidl: dont think you can,you can use flatpak tho <davidl>pkill9: thanks for the suggestion <vagrantc>huh. i notice that i am sometimes only CCed on patches in a multi-patch series for areas i am presumably in etc/teams.scm ... in one sense, this makes sense, but it actually makes it harder to review the full series :/ <vagrantc>i can of course re-download the bug mbox and extract the patches that way, or grab from the guix-patches git (although i've had mixed results with that actually being up to date in the past) <ekaitz>i got a weird question to ask... if i use a branch name in a `version` of a package and I push --force to the repo, guix continues to build it even without changing the hash <ekaitz>it only realizes the hash changed if i manipulate the hash <ekaitz>this happens also after a guix gc <ekaitz>what triggers this? looks like some kind of weird caching but it happened to me in machines that never downloaded the previous branch <ieure>ekaitz, The source of packages ends up in /gnu/store, and Guix won't redownload if it's in the store. It should be impossible for you to force-push to a repo, then build that package on a machine which has never built it before, since there'll be a mismatch between the hash in the package and what got downloaded. <ekaitz>ieure: guix gc doesn't clean those? <ieure>ekaitz, I'm not sure; did you check? <ekaitz>something weird happened in my build machine... maybe i built them some time ago? hmmm <ieure>ekaitz, If you have instructions to reproduce the latter behavior, please file a bug which includes them. But I strongly suspect there's a copy of the old source in the store somehow. <ekaitz>okay, that's what I suspected too. thanks for clarifying <beastwic1>Is there an official guix testing or unstable channel? <ieure>beastwic1, Not really. There are branches for long-lived update work, but not a distro-wide thing like Debian's experimental or unstable repos. <beastwic1>okay, just noticing that some packages are relatively out of date <ieure>beastwic1, Yes, I've also noticed that more stuff than I'd expect is stale for a rolling distro. <ieure>beastwic1, If it's something smallish, it's good way to start contributing! Most package updates are very straightfoward. <beastwic1>indeed, I have some experience, atm just messing around with different package agnostic managers after realizing I don't want containers if I am just poking holes in them just to run some newer apps <vagrantc>there is always more room for more people helping, surely. but you could say that for much larger projects too. <vagrantc>seems small compared to debian, at least, my main other reference point. :) <beastwic1>something along the lines of ebuild/pkgbuild <vagrantc>they vary widely, of course, but overall, i find many packages are pretty straightforward. <vagrantc>depends also on how quickly you can get familiar with guile/scheme/lisp and whatnot ... i've been stumbling through it for many years now and manage ok. :) <ieure>beastwic1, I've never used Gentoo, so, can't really compare. Version bumps are easy in Guix, but there are some restrictions vs. traditional packaging systems which take a minute to get your head around. <ieure>beastwic1, Stuff like, everything must build from source, and builds can't use the network at all. <vagrantc>biggest challenge tends to be assumptions about test environments ... many things want to download random stuff off the internet to run tests, and that is not possible in guix. <vagrantc>beastwic1: just pick a package you like and run "guix edit PACKAGE" and see what it looks like ... or check out the git repository and take a look at gnu/packages/ <vagrantc>i love that all the packages are all defined in a single repository ... makes it easy to cargo-cult, er, borrow, from other package definitions <vagrantc>although "guix edit" is usually a bit misleading, in that it typically brings you to a copy of a read-only file ... :/ <ekaitz>beastwic1: hey but you can rollback easily (just in case) <beastwic1>I am confused, I did a foreign install, but it tells be to set GUIX_PROFILE twice, to different locations, which one is it? <beastwic1>GUIX_PROFILE="$HOME/.guix-profile" or GUIX_PROFILE="$HOME/.config/guix/current" <oriansj>well depends if you are doing something like guix reconfigure or using your newly built profile. reconfigure needs GUIX_PROFILE="$HOME/.config/guix/current" <oriansj>otherwise GUIX_PROFILE="$HOME/.guix-profile" will point to your default folder holding your installed packages <beastwic1>okay, so it depends on what I am doing, I need the latter now, but any operation to guix itself I should switch it to the former? <beastwic1>btw really love the build --with-source command, so very cool <beastwic1>I am trying to build current obs, but I think it needs qtbase-6, how can I specify this? <beastwic1>guix build obs --with-source=obs-30.0.2.tar.gz --no-substitutes --no-grafts -c 14 --with-input=qtbase-6 <hapst3r>Has anyone here successfully installed Guix System on an ancient MacBook 2010 and has got some tips for a desparate chap :)? <hapst3r>Has anyone already encountered a message such as this during Installation: "not enough space on /boot/efi to install system"? It happens as the last step, after /mnt has been populated etc. <hapst3r>It is on an ancient Lenovo x121e Laptop, with a /boot/efi partition size of 512MB, which I have come to understand is comfortably sufficient for installing an OS bootloader. I restarted Installation with a larger partition (2GB), but I dont understand What is going on. One thing I could imagine -- even though it doesn't make sense -- is that the sdd overall is too small (240GB). Any ideas? <PotentialUser-3>Hello, pardon my language as I'm not native english, I have installed the latest guix fresh, not on top of other systems, I might be missing or misunderstanding something in the documentation, where can I see the list of use-modules and use-service-modules available? And the list for bundles like %base-packages? Thank you in advance <ekaitz>PotentialUser-3: what's your native language maybe someone can help you in that <ekaitz>hm! almost! I speak spanish, but I can try help you in english (or basque LOL) <ekaitz>PotentialUser-3: %base-packages is defined in gnu/system.scm <ekaitz>wait /gnu/store is where your running software is stored and your packages and stuff <ekaitz>but i'm talking about the guix codebase <ekaitz>the codebase i mean guix's source code <PotentialUser-3>for example I see there is no "file" utility, or htop, or screen or tmux for that matter, is there a bundle that has like a base system or I have define my own and then append it in the main config? <ekaitz>in the end Guix is just guile code so you can search the codebase and find stuff <ekaitz>there i'm extending %base-packages with other stuff <ekaitz>i think that's exactly what you need <qecep>something else I don't understand, auth.log from sshd is not where is supposed to be, is it with shepperd like with systemd made into a binary? and for example if I would use something like fail2ban the jail is already "compatible/configured" or do I need to configure the sshd service in shepperd to write into auth.log, hopefully is not too <ekaitz>qecep: i'm not sure about that because I don't have that set up, but let me see if i can find an answer for you <adamnr>Hi the IRC log archive for Hurd and Guix appears to have stopped logging <qecep>yes but there is no auth.log, sshd is running, i connected to localhost just to be sure <yelninei>qecep: on my system sshd writes logs into /var/log/secure. There is also a fail2ban-service-type but I have not looked into how to configure it with guix <qecep>oh my bad, how could have i missed that, thank you <ekaitz>qecep: fail2ban has good documentation <qecep>f2b, auditd and firewalld are the self defense tools i can't live without, but i've already seen on the wiki in wishlist that firewalld isn't packaged yet <bam94>Hi all :) I have a newby question: I had hooked up an external monitor to my laptop running guix and it did work. Now after a reboot, with or without the external monitor, the graphical environment is for some reason broken. How can I "reset" to a standard configuration (that worked previously), eg undo what I did in the graphical user interface to <bam94>configure the external monitor? I naively thought that guix reconfigure would solve the issue, but it is unaffected. <qecep>xrandr --auto does anything for you bam94? <qecep>or do in a terminal xrandr --current to view the situation <bam94>actually, the command xandr is not found, neither as user nor as root.. <qecep>and then xrandr --output nameoftheoutput --off <qecep>do you use a desktop environment? isn't there a screen configuration application? in xfce Settings->Screens <bam94>The desktop environment is broken (no visible display), which is exactly what I am trying to debug. So for now I can only interfere from a remote shell login. <qecep>oh i see, what about the xorg.conf file? <qecep>might want to back it up and regenerate it fresh? <bam94>Sure, but - again newby warning - where do I find it? Either as user or as root, if I look eg at /var/guix/profiles/per-user/root/current-guix/etc/ there is no xorg.conf there.Same for /var/guix/profiles/system/etc/. I guess I am overlooking something very trivial .. <qecep>oh yes, I see that myself now, pardon me I also just installed guix, this interests me too, gonna browse the docs <qecep>bam94 (++) Using config file: "/gnu/store/89f81415ss6ry46sb8dmkw8wpmfb1iyv-xserver.conf" <bam94>Hey we are at the same level.. also just found that.. but disappointment, because that file is so simple it cannot be the cause of the error, and there is nothing referring to the external display, which I very strongly suspect to be the issue. <ekaitz>in order to see the one running I normally htop and search <qecep>probably something in the user folder, either in .cache .config or .local, good luck <bam94>Thanks! Must be off now, but will check this out later. Indeed I started now with .cache/gdm/session.log <qecep>but it seems odd to me that xrandr is not installed <ekaitz>bam94: i don't know how to fix your issue but taking a look to the config might help <ekaitz>qecep: i had to install it by hand <ekaitz>so it's not installed by default <ekaitz>you can install it `guix install xrandr` <qecep>yes, i just installed it too <ekaitz>xrandr is just a utility, so it doesn't come by default <qecep>so, the kernel definitions are linux linux-libre and linux-lts? am I understanding right? this 3 names keeps popping out when I search for customizing/recompiling the kernel or using binaries