IRC channel logs
2025-09-02.log
back to list of logs
<podiki>unfortunately the web frontend has been really bad recently (more than a few months) <podiki>perhaps because of the merge/build load/gc going on today <FuncProgLinux>I haven't watched the guix infra, I'm afraid my scheme skills are still very green to know how to contribute there :( <euouae>Hello, what is the latest and greatest on encrypted partitions? <euouae>for now I was going to go with encrypted /home and place swapfile in /home/swapfile <euouae>but /tmp is left hanging like that <euouae>Hm, still no argon in Grub. sigh. <euouae>looking into file systems, can I bind-mount /tmp into /home/tmp? <apteryx>someone reported that facebook videos can't play audio in icecat, and I can reproduce. Has someone investigated that problem yet? <euouae->Hello I am trying to define / as a dependency of /home like so: (define root-fs (file-system ...)) (define home-fs (file-system ... (dependencies (list root-fs)))) <euouae->But I get `guix system: service file-system-/home requires file-system-/ but not provided by any service`. What am I doing wrong? <euouae->I have both appended to %base-file-systems in the operating-system record <Rutherther>apteryx: do you see an error in the output? If so, what is it? <apteryx>Rutherther: I couldn't see an error that seemed related to audio not being decoded or something <Rutherther>apteryx: yes, what I am looking for specifically is if something is logged when you try to play it, it would probably be undefined symbol or missing library. <Rutherther>euouae: I am afraid the root file system is handled in a special way, so you cannot really mark it as dependency. It is required to be mounted as first though, so there is an implicit dependency on it <euouae>Oh so that explains a couple of things. Thank you. <apteryx>it's reproducible in current icecat and also icecat 140.2 that I'm trying to finish <apteryx>Rutherther: no error in the browser console <apteryx>happens also when in troubleshooting mode (no plugins) <apteryx>so seems something at the system interaction/library level that is broken <Rutherther>apteryx: I meant in the stdout/stderr, not browser console. Anyway it doesnt seem to be what I had in mind then. Firefox and librewolf had similar issue with videos and it was because of an inaccessible library due to sandboxing <apteryx>my previous reply (the paste.centos.org thing) was about the output seen in a terminal <euouae>Rutherther: thank you, that worked. I might send a patch for docs to mention root fs cannot be dependency & is implicit dependency in the documentation for the dependencies field for the file-system record <euouae>perhaps even more ideally, I should also send a patch that fixes the error message to something more legible <sneek>civodul, you have 1 message! <sneek>civodul, rlb says: -- hadn't understood them well enough yet to know if I knew :) Happy if we don't need them. <mfg>Hi, i found https://issues.guix.gnu.org/69828 and it seems i have the same problem on one of my machines. The bug report mentions that logs have been found for the greeter, but it is not shown where/how. Does anyone know where/how i can get the logs? <efraim>I thought there was network access during gexp->derivation <Rutherther>efraim: only fixed output derivations have network access <efraim>lilyp: the patch to add hare didn't register the patches <lilyp>efraim ugh, thanks for the catch <lilyp>did you already push a fix or should I do so? <gabber>can i pass a manifest to `guix shell -e` or do i have to extract the packages from that manifest in the guile expression? <efraim>wouldn't it be `guix shell -m <manifest-file>` <untrusem>what would be the easiest way to backup my home folder, I am using btrfs, I am not using subvolumes <DefaultUser>I am trying to set up Cloudflare-Warp as a service using wireguard with guix home <DefaultUser>Though getting an error saying it depends on networking, which is provided in the system definition though <DefaultUser>guix home: error: service 'warp' requires 'networking', which is not provided by any service <charlesroelli>I had a similar error when I forgot to run (service dhcpcd-service-type) in a guix system <DefaultUser>Also will the service in guix home be able to see that service? <identity>it is included in ‘%desktop-services’, not in ‘%base-services’ <identity>the user shepherd that runs the home services is a different shepherd from the one that runs system services and provides 'networking <DefaultUser>Ok so what should I do then to provide that to home services? <DefaultUser>Attach warp-service to guix system rather than home? <identity>DefaultUser: why? just drop the 'network requirement <DefaultUser>build of /gnu/store/v7x3spcvavd3l7p8zwk3s7zfz8y82zbn-wgcf-2.2.20.drv failed <DefaultUser>View build log at '/var/log/guix/drvs/v7/x3spcvavd3l7p8zwk3s7zfz8y82zbn-wgcf-2.2.20.drv.gz'. <DefaultUser>%exception #<&invoke-error program: "go" arguments: ("install" "-ldflags=-s -w" "-trimpath" "github.com/ViRb3/wgcf") exit-status: 1 term-signal: #f stop-signal: #f> <DefaultUser>command "go" "install" "-ldflags=-s -w" "-trimpath" "github.com/ViRb3/wgcf" failed with status 1 <DefaultUser>I don't know enough about either guix or go to figure it out <identity>DefaultUser: how did you end up with the package definition? <DefaultUser>I would like to learn it, but I am under a time constraint right now and getting warp to run is quite necessary as soon as possible <identity>so learning how to do it, and then doing it, is not an option for some reason? <DefaultUser>Yeah, I have exams coming up in a week and a conference to attend just before that <identity>DefaultUser: read the manual section on ‘go-build-system’, and look at go packages defined in the guix channel <gnucode>hey guix! Ya'll rock...also I just realized that as much as I love and use guix... <gnucode>Wasn't there a collective group that was just getting started to advance guix development somehow ? <ieure>gnucode, It's a consultancy firm, you can donate to them, but it's fairly indirect support. <ieure>gnucode, I don't know if there's a way to directly financially support Guix specifically. It's a GNU project, so you could donate to GNU. <gnucode>guix pull fails on power9 at the moment. <gnucode>unsupported hare target: "powerpc64le-linux" <apteryx>ACTION wonders if something could be done to avoid ci.guix.gnu.org returning 504s. That's a serious usability problem. <apteryx>is it because the load is too high? it seems berlin should be beefy enough to be able to serve a website even when the load is high, so I'd assume something problematic in the design. <gnucode>apteryx: I feel like guix weather should be re-worked. ci.guix.gnu.org should run guix weather once a day for all CPU platforms. then client machines running "guix weather" should just report the cached result. <apteryx>(you usually give it some packages at the end, not the whole collection) <ieure>I do not believe it is, doesn't `guix weather' effectively issue a HEAD request to substitute servers? <apteryx>I think the observation is that by default, 'guix weather' will compute *all* derivations <apteryx>e.g., 'calcul de 29 898 dérivations de paquets pour x86_64-linux…' <ieure>That happens on the client though, yeah? <ieure>Okay so.... yeah, not going to help. <apteryx>it's like doing 'guix refresh' without a package; it'll try to update the whole package collection <cbaines>gnucode, the bottleneck with guix weather is generally populating the local narinfo cache <ieure>Oh, yeah. I've never even run `guix weather' without a specific package. <cbaines>if you hack the code not to cache the narinfo responses, it's very quick to run <apteryx>ACTION wanted to see which job was set for gnuzilla... I think it's 'gnuzilla-updates' <apteryx>if someone wants to test icecat 140.2 and icedove 140.2 ESR, feel free to checkout the 'gnuzilla-updates' branch and build them. <apteryx>I still have more cleanups I want to do there, but it at least builds and runs <civodul>meanwhile, ‘guix gc’ has been running for 36h or so on bayfront <apteryx>ugh, icecat 140.2 parameters page appears broken. <cbaines>civodul, maybe just kill it, we're not short of space on bayfront at the moment <gnucode>cbaines: then I humbly suggest that guix weather be eventually (not ASAP) reworked so that it just shows what's at qa.guix.gnu.org/branch/master <cbaines>gnucode, running guix weather gives you data for the revision of guix you're using, including the specific substitute servers and other channels I think <apteryx>hm, 2025-09-01 04:34:18 guix gc: error: executing SQLite statement: FOREIGN KEY constraint failed <cbaines>that's different from what QA is showing, so you're not comparing like for like <gnucode>my guix weather is saying that ci.guix.gnu.org has 80% substitutes available. <ieure>gnucode, `guix weather' is really not going to move the needle here. <ieure>gnucode, The state of things is that everyone knows *what* is happening (CI serves 50x errors much too often), nobody knows *why* it is happening. <ieure>So proposing a fix without diagnosing the problem can only improve things on accident. <gnucode>ieuru you are saying that "guix weather" is not currently a bottleneck. Gotcha. <ieure>gnucode, I do not believe it to be. <apteryx>in /var/log/gc.log on berlin there are lots of "skipping invalid root from `/var/guix/profiles/per-user/cuirass/cuirass/..." lines <gnucode>cbaines: I feel like what I'm learning from this conversation is that "reproducible builds as done by guix" is better than other options, but much more complicated than say the way Debian currently does it. (I'm speaking as someone who really doesn't get all the technical stuff). <cbaines>I thought we were talking about guix weather? <gnucode>we were. I'm sorry if I'm coming across as critical. :( <gnucode>I'll go ahead and make that bug report about 'guix pull" failing on power9 unless someone already knows it exists ? <gnucode>I am really happy to see hare packaged! <civodul>old: re LWN, let’s see how we can fix the reputational damage now… <civodul>apteryx: “skipping invalid root”, weird <m4xxed>Hey, is there any way I could check the subsitute build status of a specific package? I know of `guix weather` but it only shows me that there are no substitutes for the current qtwebengine, not what the state of the build is.. because locally, my build fails <ieure>m4xxed, You have to go rummage around on ci. <identity>m4xxed: if the build fails localy, it is likely that the build fails on the build farms <m4xxed>identity yes I figured, that is why I wanted to check :) <m4xxed>ieure could you elaborate a little bit? Is there any frontend of the CI that I could search through? <euouae>Hello if I have some management scripts that I want to run after /etc/config.scm how do I do it? In particular I want to open some luks devices. My attempts with mapped-devices are floundering and I suspect it's because I'm using Argon2. <efraim>I think I have a fix for the target->hare-arch breakage <Rutherther>euouae: what does 'run after /etc/config.scm' mean? config.scm does not 'run' <Rutherther>if you want to run something at startup of your system, yes, shepherd service is appropriate <euouae>I have this huge pet peeve about people picking up on words <Rutherther>this is not word picking, saying config.scm runs shows a deep misunderstanding and is confusing, the most probable meaning I would guess is that you mean when you reconfigure. But you didn't mean that according to what you're trying to do <euouae>I did not say that which you accuse me of <ieure>"if I have some management scripts that I want to run after /etc/config.scm" <ieure>After /etc/config.scm ... what? <euouae>Exactly -- there's something implied there. <euouae>It does not necessarily show a "deep misunderstanding". <ieure>euouae, The implication of "run after /etc/config.scm" is that both config.scm and the management scripts "run." Which is what Rutherther is saying.. <ieure>It's not clear if this is misunderstanding or imprecise language. Many Guix users are non-native English speakers, which can sometimes get in the way of the precise technical wording needed to help troubleshoot. <euouae>That is only because you're assuming the person communicating has a misunderstanding -- there are other guesses for what the filled in part is. <ieure>I try to be understanding and ask for clarification. <ieure>euouae, Would you please clearly state your usecase? <ieure>I think we can stop haggling over word meanings if you just state the thing you want to accomplish. <Rutherther>(and that is what I already asked in beginning of my first reply to the message) <euouae>Let me ask this question, can LUKS encrypted partitions with Argon2 be dev-mapped from config.scm? <dariqq>finally reconfigured the i686 machine again, why do I keep doing this. guix pull OOMing, rebuilding qt because of OOM in make-dynamic-linker-cache, rebuilding both webkitgtk <ieure>euouae, I strongly suspect yes, but don't know 100%. LUKS encrypted partitions can be mapped in config.scm, all my machines use it. But none of them use Argon2. I believe the specific LUKS algo doesn't matter for Guix configuration, as long as the cryptsetup the system uses supports them. <ieure>euouae, I will also tell you: this question has nothing to do with management scripts or "after config.scm" though, which makes me wonder which of those two things you said is the thing you want to do. <euouae>How can I debug system reconfigure's where it's getting stuck at splash screen? I tried adding nosplash but I still saw the penguins and gnus <Rutherther>what splash screen? I don't know of a splash screen on reconfigure <euouae>perhaps I'm talking of something else then <euouae>Anyway I think I need to look at the generated crypttab and so on in the generation files <ieure>euouae, Does Guix even use crypttab? <euouae>I'm not sure how it does whatever it does <euouae>but in any case I need to look at the generation files and see what it generates for the mounts. Maybe I should try to use pbkdf2 and see if that solves it too, that's an alternative route <Rutherther>euouae: yes, person saying X perhaps means something else than X, but I would rather expect they mean X than something else <euouae>Rutherther: alternatively you can expect they mean something else when X is not right <identity>so when i try to login with GDM, it gives me a black screen for a moment and then takes me back to the login screen; prior to this, i was mucking around with display managers, but i have reverted the changes, really, triple checked, triple reconfigured and triple rebooted, so what the hell? <identity>i should try switch-generation, but i would expect reconfigure to do the trick <Rutherther>No, I wont expect person saying X means something else, I will ask them to clarify. But if they wont, there is nothing else I can do. <identity>well apparently i was wrong and booting into an older generation works. not really eager to discover the cause <euouae>I tried PBKDF2 and it didn't work -- does mapped-devices not work with passwords (only keyfiles?) <Rutherther>euouae: mapped devices work with passwords. Have you tried typing your password and RET even without seeing the prompt, when you are stuck? <euouae>Yeah I tried that, it didn't work <euouae>How can I get rid of the splash screen? I want to see what's happening at boot <FuncProgLinux>About that splash screen Isn't it "plymouth"? I didn't know guix could be configured with it :o <Rutherther>there is no splashscreen on Guix by default, so you will have to clarify what you mean <euouae>The 100% libre splash with gnu and penguin <Rutherther>never seen such splash screen on Guix System, it is definitely not there by default <euouae>maybe youre' using the nonfree kernel <identity>i do not thing i ever saw the linux-libre splash screen on guix, i guess it is not there on linux-libre version that is the default on current guix <FuncProgLinux>luca: You could use (local-file "./path-to-your-file") relative to the package module you are working on. I don't know if that's a best practice though, I copied that from the "aux-files" directory the main repository has. <luca>Sure, but I can't point biboumi to "a file". I need to point it to "a directory" <Rutherther>euouae: I have just downloaded the Guix System iso and ran it in a VM, no splash screen at all <FuncProgLinux>luca: (local-file "./aux-files/biboumi/" #:recursive? #t) I think :) <euouae>Rutherther: you have to run 1.4.0 <Rutherther>luca: you can put whatever to the store. But are you sure you want to do that? The store is immutable, by pointing the software to a given location in the store everyone has to use the same policies, or recompile the software themselves. While by pointing to somewhere like /etc, users can configure the policies during runtime. This all then comes down to the question if users are expected to be changing this or if everyone is supposed to use the same... <FuncProgLinux>Isn't that "splash" the same you can see when booting into sysvinit/openrc systems with the tux and the fancy colours on screen? <euouae>Rutherther: you installed or ran it? <cancername>hi #guix! I'd like to build Linux incrementally for development, like usual, then package the result as a guix package so I can boot into it. does something for that use case exist already, or do I need to homegrow it? <luca>Rutherther: I believe the user expectation here should be that the application is configured at "runtime" the same way prosody is configured for example. (not sure what the right word here is. "Deploy time"?) Certainly not compile time. Ideally still declarative <euouae>Rutherther: I think it might not show on VM <cancername>euouae: not quite, but thanks! I'm running Guix System and need to recompile my own kernel often, and because Guix doesn't support ccache and builds are isolated, I need to wait around half a day for Linux to finish building, so I'd like to build Linux and then put the artifacts in a package. I was asking whether something like that already existed, or if I had to do it myself with copy-build-system <Rutherther>luca: right, good, you can definitely use local-file inside a config you create by a service for example then <luca>Can I use local-file from the repl to see how it would end up in the store? <dariqq>luca: ,build (local-file ...) . Don't remeber what module this is part of <identity>cancername: you are probably better off with getting a guix shell --development linux-libre and compiling it from a local checkout, at least for development <luca>Actually I'm confused. I ran `,build (local-file "./dir" #:recursive? #t)` and got `$4 = #<<local-file> file: "./dir" absolute: #<promise #<procedure 34ba2810 at <unknown port>:7:7 ()>> name: "dir" recursive?: #t select?: #<procedure true (file stat)>>` <identity>then you could do some hacks to put the linux&co. in a package and make an image out of that <cancername>fyi, I'm getting "Unknown meta command: build" even after (use-modules (gnu)) <cancername>identity: I see, yeah. I know the current structure, so I should be able to do it <Rutherther>gnu is the correct module, but use-modules won't work for meta commands <Rutherther>oh sorry, it's actually (guix). Interesting, I typed guix from memory in the repl to test this and then I claim here it's gnu lol <euouae>I wonder if there's a way to disable boot-logo-patch <FuncProgLinux>luca: I haven't learned to read scheme correctly :( I test my modules/packages in the repl and (usually) I think that means your statement was evaluated successfully <luca>FuncProgLinux: Sure, it being evaluated successfully is great and fine. I'd just like a path to the store it made <Rutherther>euouae: I have just booted the iso on a real computer, no splash <euouae>Rutherther: You have to install it. <euouae>There's no splash when you boot the ISO <Tadhgmister>k how do I get my system to stop using emacs.desktop and instead use emacsclient.desktop? I have the home-xdg-mime-applications-service setup to remove a bunch of associations with emacs.desktop and see that in ~/.config/mimeapps.list show the removed associations but `xdg-mime query default text/x-c` still reports emacs.desktop <Tadhgmister>I've even relogged, so it isn't a cache thing... I am at a loss <podiki>cancername: kernel compile will be much faster if you limit which modules are built, and customize-linux procedure just got support for using a file from e.g. modprobed-db <FuncProgLinux>luca: I have to think of a script to see if G-Expressions can help with your problem <luca>I get "Unknown meta command: build" <luca>oooooo that's what you meant. Nope, but I'll do that now <luca>Yep, that works perfectly <luca>I assume I can use ,build outside of repl as well? <Tadhgmister>how do you run `info` to a specific page you know the name of? would love to share the command to open info to the page instead of a link to the web view <identity>Tadhgmister: you mean like (info "(guile) REPL Commands") <FuncProgLinux>Anybody else experiencing issues building qtwebengine-6.8.2 after world-rebuild merge? :( <Tadhgmister>yes exactly that, the info page is in parenthasis thank you much <Rutherther>luca: depends what the file is... if you want to refer to files of packages, you can use file-append, ie. (file-append sway "/bin/sway") <luca>Thing is that errors out with `Wrong type (expecting string)`, so it sounds like it wants a path to the store <Rutherther>can you send a stack trace for that? it should go to mixed-text-file call that should support file-append <luca>(ignore the messiness and prosody and what not. I'll clean up some time) <Rutherther>"so it sounds like it wants a path to the store" you cannot get paths to the store at the evaluation side as strings... or like you can, but not in a pretty way, you would make it a dynamic dependency, it's not a good way to approach a goal, everything should support gexps producing file-like objects nowadays for this use case <Rutherther>oh I see it now! I missed it, someone decided to use string-append inside the mixed-text-file <Rutherther>the default-session-command should be fine, but the initial-session-command currently won't work :( I would call this a bug definitely <luca>I see. So the "right" fix here would be for initial- to follow default- and work the same? Would this be a breaking change? <luca>but mixed-text-file in mixed-text-file is certainly a bad idea 😅 <euouae>luca: looks fine to me but you could do (define (initial-session user command) ...) and then after "command = " default-session command "\n you can insert (initial-session ...) <euouae>it would still be the same code essentially but there wouldn't be any if/else logic in the mixed-text-file <euouae>(the initial-session would be defined inside make-greetd-terminal-configuration-file as a helper function) <luca>What would initial-session return? <euouae>the entire if block is inside it -- either the formated text or "" <euouae>another option is that it returns #f and you do (or (initial-session user cmd) "") if you want the logic to be visible inside mixed-text-file <luca>I'll give it a shot, thanks! <Rutherther>euouae: I have successfully installed guix system on my laptop, version 1.4.0, linux-libre, no splash screen <Rutherther>euouae: I have successfully installed guix system on my laptop, version 1.4.0, linux-libre, no splash screen <euouae>Well then, you don't see it and I see it. <Rutherther>since I can't say where it would be coming fom when booting Guix System, I can't say how to get rid of it <euouae>I know instlling nonfree kernels gets rid of it so that's what I'll do <euouae>later I might try to figure out how to send a patch that shows the kernel text beneath the boot logo so that people can debug their issues <FuncProgLinux>qtwebengine is failing on latest guix with: cd /tmp/guix-build-qtwebengine-6.8.2.drv-0/build/src/core && /gnu/store/31h3cyip6fa0x9g49ijp64mx7hc8c74y-ninja-1.13.1/bin/ninja -k1 -j28 -C /tmp/guix-build-qtwebengine-6.8.2.drv-0/build/src/core/RelWithDebInfo/x86_64 QtWebEngineCore <FuncProgLinux>ninja: build stopped: subcommand failed. The log doesn't provide much more information other than "subcommand failed" <FuncProgLinux>Fixed by uninstalling electrum wallet, although someone using it might encounter the same issue <luca>Can I test changes in the guix repo on a system config that uses third party channels? <luca>Any tips on how? Using pre-inst-env doesn't seem to use my channels.scm, and -L at a first look doesn't work. But that may be my bad code edit <luca>all right, back to the drawing board <FuncProgLinux>luca: If it's a clone of the Guix repository it should work without setting a load path. Or is it a personal channel? <luca>It's a clone of the guix repo with my own modifications <mgd>FuncProgLinux: It's crashing for me as well but I don't have electrum wallet. I'll have to see what's using it <Rutherther>you also need to put #$ before initial-session-user <FuncProgLinux>mgd: kde software like kde-games, neochat, discover, etc depend on this package. I'm not a KDE user but I think guix graph could show you what you're searching for <luca>Rutherther: Sure, but doing that and -L doesn't change the error at all. Same error as if I hadn't done anything at all <Rutherther>luca: so you're doing ./pre-inst-env guix system build in a pure shell? <jfred>luca: Would `guix time-machine -C channels.scm -- system reconfigure <your config>` work for you? <luca>In a normal non-pure-non-container-non-nothing shell `guix system build -L $HOME/src/guix system-turret-config.scm` <Rutherther>jfred: that would definitely work as long as the url to the guix channel is set to the local path, can be passed through --url <jfred>(assuming `channels.scm` is a channel file pointing at both your custom channel + your modified clone of guix) <Rutherther>luca: when you mentioned -L I assumed you meant you are putting the other channels to -L, not guix. Guix can't be put to -L, that is always going to be shadowed by the original guix <luca>Oh sorry for not being clear. So I should pre-inst-env + time-machine + channels.scm? <Rutherther>if you're going to use time-machine, there is no reason to use pre inst env <Rutherther>you can either use time machine, or use pre inst env, both is unnecessary <luca>Ok, `git fetch origin keyring:keyring` and `guix time-machine -C $HOME/.config/guix/channels.scm --url=$HOME/src/guix -- system build system-turret-config.scm` seems to be the way to go <Rutherther>you might have to commit it and disable authentication <Rutherther>channels definitely should not specify guix as a depenency, guix is an implicit dependency of every channel <Rutherther>removing guix from channels.scm should also not be done <luca>Removing guix as a dependency and adding it back to channels.scm did not give me a new error, just the same old one. Any way to get a bigger stack trace or something? <Rutherther>did you do what I said - commit and disable authentications? <luca>I didn't commit to master, do I need to do that? <Rutherther>no, you don't need to do that, but you need to point time-machine to the branch you made the commit in, I presume you're currently pointing it to master <luca>I see. I was expecting it to use HEAD but it used master. Setting a branch got me on the right path now <luca>aaaaand after all that it didn't even work ._. but at least that's a greetd thing and not guix. So I'll look more into it <unmush>apteryx: I happened to be reading the logs and saw that you ran into the same icecat problem as I did. If you look in security/sandbox/linux/broker/SandboxBrokerPolicyFactory.cpp and look for "/nix/store", you'll see that they are apparently relying on hardcoding some things. <unmush>we have a 'build-sandbox-whitelist phase for this, but it turns out that the security.sandbox.content.read_path_whitelist preference is only honored by GetContentPolicy, not GetRDDPolicy or GetSocketProcessPolicy <unmush>it can be worked around by adding the denied library to LD_LIBRARY_PATH, since all directories in there are added to the whitelist at runtime IIRC <unmush>personally I've locally just added policy->AddDir(rdonly, "/gnu/store") after the /nix/store ones <ieure>LibreWolf has fixes for this. 