IRC channel logs
back to list of logs
<midgardian[m]><bricewge> "Tcache: You better register your..." <- What is wrong with [m]? <midgardian[m]><unmatched-paren> "(not sure if they're even..." <- If it is Arc GPUs it requires nonlibre firmware <midgardian[m]><rekado> "the problem with that server..." <- Is there some problem with Guix using btrfs? <nckx>Just with booting degraded ‘raid’ file systems that don't get assembled in a timely manner by shoddy firmware. <jackhill>Ideas as to why I would get "guix lint: warning: while connecting to Software Heritage: TLS certificate error: X.509 certificate of 'archive.softwareheritage.org' could not be verified:" from `guix lint`. I have nss-certs installed in my profile <apteryx>jgibbons[m] nckx our current mono already relies on blobs, If I recally someone passing by here mentioning it. Perhaps we should propose to remove it from our collection, if it cannot be fixed. It's unfair in the face of our other efforts such as a properly bootstrapped rust. <jackhill>apteryx: mono's not the only one. See also ghc and chicken scheme (at least). I support having some for of policy guidance. I suspect a lot of those got in before we had full source bootstrap for all the rest :) <midgardian[m]><jackhill> "apteryx: mono's not the only one..." <- What about MIT/GNU Scheme? <jackhill>chez scheme has the problem too I believe, but racket has a way to bootstrap thier chez fork, and I think there's some hope of making using of that for the non-forked version as well <hnhx[m]>Hey! Could someone help me? I can't really get emojis to display on Guix. Which font do I have to install? <nckx>hnhx[m]: For example, font-google-noto (warning: huge boi). I would not recommend font-gnu-unifont for this. Some applications (like foot) need explicit configuration to render emojos at all. <nckx>apteryx: As tempting as it would be to (re)move all yoghurts, I don't think we can afford to lose GHC... <nckx>hnhx[m]: Yes. Maybe refresh your cache (fc-cache -rv). <nckx>apteryx: ...what triggered me about mono was my reading that they keep adding whole repos of blobs. We should say enough at some point. But jgibbons[m]: which repo was this? <hnhx[m]>oh okay it works now, I already installed that font in the past so it seems like I just had to do fc-cache -rv to make it work <nckx>It's like 'hash guix': Guix can't do that as currently designed. I can think of some hacks, none of which I'd want on my machine. <hnhx[m]>does anyone here use llvmpipe for xorg? <hnhx[m]>I tried to get picom to work with blur but its extremely laggy : / <hnhx[m]>I wish amdgpu would work without non free firmware zzz <unmatched-paren>midgardian[m]: fortunately it's not an Arc GPU, i think it's the new Xe integrated one, which was introduced in gen 11. this laptop is gen 10, so it missed out :( <unmatched-paren>although it's slightly irritating that even Intel lock up their discrete GPUs <sneek>raghavgururajan, you have 1 message! <sneek>raghavgururajan, nckx says: Did somebody say botsnacks? <nckx> I forgot sneek was cross-channel. That was from #wie. <unmatched-paren>maybe it should say where the message came from if it's not the current channel? <yarl>Right permissions for /etc/guix/acl are root 400, right? <yarl>I don't see where this is defined in the code. <unmatched-paren>it's not exactly a particularly important site; it's just a prettier UI for debbugs.gnu.org <bost>unmatched-paren: aah. I just wanted to ask where else can I browse the bugs. <yarl>Can someone confirm please what /etc/guix/acl mode should be? Thanks! <lilili>Hello everyone, I made my Guix LiveCD using `guix system image -t iso9660` for portable use and I found that the optical drive would stall when idle, does keeping the optical drive spinning improve the read speed? <nckx>unmatched-paren: No no, I wish. It's somebody in #guile (no secret who, but I'm not going to ping them here). <nckx> yarl: It can also be a regular file (e.g. if you use 'guix archive --authorize' imperatively). All that matters is that only root can write to it. World readable is fine. <nckx>bost: No, we can't do anything about it *right* now, but it will be fixed in at most a very few days. <bricewge>We are just missing a service and optionally rtkit with it's own service <nckx>lilili: Yes, although I'm not sure where you're going with that. <jpoiret>bricewge: pipewire doesn't really need a service <jpoiret>it should be started on demand by dbus <unmatched-paren>jpoiret: does that mean i could get pipewire working in my sway desktop right now? <bricewge>Yes a service which at least expend `dbus-service-type` <bricewge>unmatched-paren: No, we are missing some bits and pieces right now <jpoiret>although i launched `pipewire-pulse`, `pipewire` and `wireplumber` manually <jpoiret>although yes we're missing the optional rtkit <bricewge>jpoiret: I'll work on it this afternoon. Do you have any code to share about it? <nckx>Audio throughput, not latency, enjoyer. <nckx>Best-effort 48bit is the future. <nckx>(And yet part of me wants to try. Is this collection of dbus hacks public anywhere, bricewge?) <yarl>nckx bricewge I ran `chmod 444 /etc/guix/acl` then `sudo guix archive --authorize < /bla/bla/bordeaux.guix.gnu.org.pub`. <yarl>mode of /etc/guix/acl is back to 600. <yarl>Can you confirm this behaviour nckx bricewge? <nckx>I guess I can confirm if you really want. Sec. <yarl>I am not judging, the problem is that then you can't run "guix substitute --help" or "guix substitute --version" as a regular user. <nckx>I got distracted by Satan. <nckx>guix substitute isn't intended for use anyway. It is a cosmetic issue, but a very tiny one. <nckx>I'm sure you could swap the order of operations to fix --help if you want to submit a patch. <oriansj>nckx: stupid sexy satan on your brain huh... <yarl>nckx 1/ I know. but it is good picks for me too become familiar with guix (and guile) without hurting to much. 2/ Yes I began to do that but I needed to be sure about the mode of /etc/guix/acl as if it is others readable, guix substitute does not crash. <yarl>...does not crash as regular user with --help <yarl>Checking the readability is probably better. <nckx>yarl: I'm ( :( ) on a telephone right now, I don't have a comfortable way to read Guix code in context. I would have wrongly assumed that --help and --version were centrally handled. Guess not. <nckx>What do you mean by ‘crash’? If it's ‘just’ a backtrace with ‘/etc/guix/acl: permission denied’ somewhere in it, I think that's OK TBH. <nckx>Backtraces aren't crashes. But if it's not, yes, we report a better error at read time (so no separate check beforehand please). <nckx>*aren't always crashes that need to be hidden at any cost. <yarl>nckx Yes alright this is a backtrace not a crash. <nckx>Yeah, we don't need the (likely flawed) ‘am I root’ check. But the rest is great. <yarl>Anyway, `guix substitute --help` is not supposed to try to open /etc/guix/acl, right? <bricewge>nckx: I'll try to publish a patch about it before the end of the day <nckx>I feel like that image is biased in some way but can't quite put my finger on it… <nckx>Plus, I only wear mine to sleep in. And at FOSDEM. <yarl>unmatched-paren alright, thanks, why?
***GNUtoo_ is now known as GNUtoo
<akonai>it depends on what the expected return value from the function is <unmatched-paren>so if you want the match to display some kind of failure value if it falls through to `_`, then you should use #f <yarl>I just want the function to go one if _ matches. <yarl>Well, strange. I thought I could run `sudo guix substitute --query /gnu/store/HASH-hello.2.12` <yarl>It's not how it is supposed to be ran? <yarl>Anhydrous Pacific Ocean? <yarl>unfortunately, I got other things to do. I'll be back. <unmatched-paren>"take my advice with the amount of salt necessary to fill a black hole" <oriansj>I prefer, everything I say may be wrong or intensionally evil designed to destroy you and everything you love. So don't trust me but perhaps you may find value in what I share either dirrectly or indirectly. <unmatched-paren>oriansj: "you may find value in what I share" only applies to smart people like you :P <oriansj>unmatched-paren: The best level 1 tech I know has an IQ of 62 but he spends every waking minute of his day mastering his craft. I find he has a great deal of value to share when people need help with their computers. Everyone can provide great value to others or choose not to. <oriansj>some people do it by collecting, some by creating, some by sharing and others by pointing out the problems that exist <oriansj>a person screaming in your face is trying to express something that matters (perhaps only to them but that isn't the point) and people acting in good faith will defuse quickly when you actively listen and create a shared understanding. Those acting in bad faith will enable you to spot another area where you can improve things. So even bad people can contribute meaningfully to you if you think of them with consideration and care. <oriansj>for there are no "stupid users" only people who you haven't taken the time to understand or time to teach in a way that connects with them. (the connects with them is the really important part) <oriansj>no one is born knowing how computers work nor how your specific program/language layers on top of that tree of abstractions. So the things that trip up that person probably trips up others as well, so use it as a chance to make things better for everyone or accept it can't be made better (because sometimes that is the reality of unsolvable problems) but provide a way for them to know why and create a greater shared understanding of the <oriansj>nothing repels trolls better than kindness, consideration and loving supportive community; which either convert to productive members of the community or get bored by lack of reaction. <yarl>Ok well, I don't see any difference between #f, #t or even '() or anything else since the result of the match expression is not used. the match is here only for side effect. <yarl>Is it possible to call `guix substitute --query` directly? How? <civodul>yarl: the easiest option is to use "guix weather", which calls the same code <yarl>civodul, hi. The thing is, I found what might be call a bug in `guix substitute`, I don't know if you read through the irc log. <yarl>It seem to be working via `guix weather`. <maximed>Does someone know any software for computing all linearisations of a graph? <sneek>Welcome back maximed, you have 1 message! <sneek>maximed, unmatched-paren says: will #:skip-builds be unnecessary in antioxidant-build-system? <civodul>yarl: ah no sorry, i missed the earlier discussions <maximed>sneek: later tell unmatched-paren: Yes, it doesn't have #:skip-build?, it doesn't need source code of dependencies and never copies source code. <civodul>in that case, you can try something along the lines of "echo info ITEM1 ITEM2 | ./pre-inst-env guix substitute --query" <civodul>where ITEM1 and ITEM2 are store file names <unmatched-paren>maximed: awesome! :D so we'll be able to remove any crates specific to windoze, smack, or redox. that'll probably be a non-trivial reduction in crates-io.scm <sneek>Welcome back unmatched-paren, you have 1 message! <sneek>unmatched-paren, maximed says: Yes, it doesn't have #:skip-build?, it doesn't need source code of dependencies and never copies source code. <maximed> unmatched-paren: Yes, they are already being automatically removed <maximed>in the "guix.scm" in the repo of antioxidant-build-system <yarl>civodul: echo info /gnu/store/13xz4nghg39wpymivlwghy08yzj97hlj-hello-2.12/ | ./pre-inst-env guix substitute --query <yarl>guix substitute: warning: ACL for archive imports seems to be uninitialized, substitutes may be unavailable <yarl>guix substitute: error: '/gnu/store/13xz4nghg39wpymivlwghy08yzj97hlj-hello-2.12/' does not name a store item <maximed>It's not part of the store item name IIRC <yarl>Still the warning but not the error. <yarl>What is still bothering me is that I can't spot where in the code the mode of /etc/guix/acl is set to 600 by `guix archive authorize` <yarl>I don't want to touch that but I'd like to see where it happends. <civodul>this must come from 'authorized-key' in (guix scripts archive) <yarl>civodul yes, I'm reading it but did'nt spot yet <civodul>i think with-atomic-file-output has that side effect <civodul>yarl: hmm yes, though "guix substitute" is not meant to be invoked directly :-) <yarl>civodul ooh right. with-atomic-file-output uses mkstemp! (note the !, I can't find it in guile reference, only mkstemp without the !) and that uses stdlib mkstemp and from the man, "the file is created with permissions 0600". <yarl>civodul. Yes, I thought this was a nice and harmless thing to do to learn guix and guile. <civodul>on both mkstemp and learning/contributing i mean
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | use https://bugs.gnu.org/guix & https://bugs.gnu.org/guix-patches for now | https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | paste: https://paste.debian.net/ | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: https://logs.guix.gnu.org'
***ChanServ sets mode: -o nckx
<maximed>Could perhaps be turned in a proper Guile library or such <cbaines>I've also noticed that guix weather is unhappy for the latest commit (08756c831cda0dc8cc8a58221378bcf76a4ceff4) <cbaines>guix weather: error: #<<platform> target: "x86_64-linux-gnu" system: "x86_64-linux" linux-architecture: "x86_64" glibc-dynamic-linker: "/lib/ld-linux-x86-64.so.2">: invalid G-expression input <bricewge>Does some one has an exemple of a package with multiple source? <bricewge>ie. a package which adds code from an additional repo which isn't upstream *unmatched-paren just remembered that they still haven't figured out how to get valgrind to run under guix <unmatched-paren>ok, so: how do i run valgrind on a program straight out of the -oven- compiler without it exploding? <nckx>bricewge: Anything matching ‘grep '(".*" ,(origin' gnu/packages/*scm’, for one, although many packages probably split that over 2 lines. The point is to add it as an input, then explicitly unpack it somewhere into your existing source tree, assuming you keep the default 'unpack phase to unpack (source …). <nckx>You should still have exactly 1 source field, so you need to decide what the ‘primary source’ is. ‘guix build P --source’ will return only that, not any of the extra input origins. <bricewge>nckx: Thanks, I ended up going in an other way. I'm writing a `pipewire-launcher` package with the trivial-build-system. It has an hard coded `pipewire.desktop` file and `pipewire-launcher` shell script. <nckx>Meh, it might be preferable over pointing to an external one (there are plenty of packages that do both of those things, and for the better — but note that there's a make-desktop-entry-file helper if you didn't know/use that yet). Each package is different. <nckx>Even my own t-b-s packages often feel ugly to me… <bricewge>Ok. I didn't knew about `make-desktop-entry-file`, I'm looking into it. <nckx>It's kind of… inextensible. <nckx>But should cover most cases. <bricewge>I'm missing some specific KDE and Gnome directives and can't manage to use the `#:rest all-args` escape hatch <kaelyn>Hello #guix! Does anyone here use the bitlbee system service? <minikN>Hello, when writing a package definition, how can I symlink a file to <output>/bin? I'm currently using `install-file`, but that just copies the file. <silicius>I think stumpwm's info is not building right. There's a bunch lines with "@" that look like they miss something <pashencija[m]>How to correctly handle problem described in above bugs if I face them writing my own kernel package? <bricewge>minikN: Did you find the procedure `symlink`? <minikN>bricewge: Not yet. I'm scouting guix/build/utils.scm <maximed>So you'll have to look in the source code of Guile <minikN>maximed: It actually looks quite straight forward, thank you, I'll try it. <maximed>Does someone know if #sr.ht requires registration to post messages? <maximed>maximed: I mean, the #sr.ht IRC channel <maximed>The tutorial says that the e-mail appears in the ‘monentarily’, do you know roughly how long this is? <bricewge>For me its was instantly, like most other emails <maximed>(It's not appearing there, and a mail I sent to another one of the lists hasn't been received yet even though it has been ~10 hours) <maximed>Looks like if I use the graphical e-mail program, it is received, but not when using "git send-email" <akonai>Are activation-service-type gexps ran as root? I'm getting a "mkdir: permission denied" error with guix system reconfigure for some reason <maximed>However, possibly it's actually being run outside the gexp. <maximed>And activation gexp can switch to non-root. <maximed>E.g., root cannot write to /gnu/store/ (though there are some work-arounds) <maximed>could I have a look at the gexp code you wrote? <akonai>though I'm not sure if it is the gexp because I can't see the format message in the build log <akonai>but I don't know why etc-service-type would error out because of mkdir and I don't see any other place where I'm creating directories <maximed>The %rspamd-activation looks ok to me, but maybe check if you have imported (guix gexp) in that modulee <maximed>akonai: Is it the gexp that is erroring out or the etc-service-type? <maximed>Also, is there a backtrace or such (for context) <maximed>(FWIW, you can use mkdir-p/perms to create a directory owned by a user+group+permission-bits) <akonai>also, thanks, didn't know that existed <maximed>akonai: I don't know what's going on, but you could try minimising the %rspamd-activation a bit (e.g., is it caused by /var/log/rspamd, or /var/lib/rspamd?) <maximed>(or maybe it isn't %rspamd-activation related, which you could test by removing %rspamd-activation) <akonai>yeah, it seems to be unrelated since if I comment it out, it still errors out <akonai>okay, it was in fact the etc-service-type service. rewriting it to use nested file-unions seems to have fixed it <PotentialUser-57>Where does the libvert config file live in guix? Trying to add my user to the libvert group in "/etc/libvirt/qemu.conf" so I don't need root for qemu but it seems to be elsewhere <lilyp>PotentialUser-57: all system configuration is derived from your config.scm for libvirt, the libvirt-service-type is part of %desktop-services, I believe <lilyp>you can (modify-services ...) that to change the config to your liking (within the constraints of what's implemented in Guix; if something's missing, do raise a bug) <PotentialUser-57>lilyp: I haven't gotten to that part of the manual yet. I got to like halfway past chapter 5 but I saw I think twice it refer to things in chapter 2 which I skipped since I installed the guix system rather than just guix ontop of another linux distro, so I figured I'd start a virtual machine and go through chapter 2 with it before proceeding <the_tubular>Also, PotentialUser-57, the answer to "Where does the X config file live in guix?" Should always be config.scm :) <the_tubular>Well, maybe not config.scm, I meant to say a scheme file <PotentialUser-57>maybe it would be simpler to just add my user to the group that has access to the KVM, since that is what QEMU is complaining about, anyone know what group KVM usage is tied to, or how to find that out? <PotentialUser-57>i looked in my terminal, i wrote it lower case. do I need to refresh it somehow before it takes effect?
***Tracnac_ is now known as Tracnac
<the_tubular>Is the name PotentialUser-57 given by a IRC client I'm unaware of, or did you chose that nick ? <the_tubular>I though it refered to guix user, guess it refers to libera user <PotentialUser-57>anyways those are the error messages I get after trying to run QEMU, after declaring that im part of the kvm group in my config.scm and running reconfigure and verifying that im in the group with the "groups" command. Maybe it's QEMU that needs to be in the kvm group, not me? <kaelyn>rekado_: Thanks for the confirmation. I hadn't seen anything here (also checked the channel message) or on guix-devel. <rekado_>the problem is an unresponsive server with nobody around to lower the risk of rebooting it.
***jesopo is now known as jess
<nckx>kaelyn: It's in the topic… but not the channel message. I'm just glad you read either. <PotentialUser-57>"sudo chmod +666 /dev/kvm" fixes it but seems maybe a bit unsafe to give everyone access to kvm? *nckx wishes they could help, but uses KVM through QEMU only (and adding themselves to kvm sufficed). <nckx>PotentialUser-57: It's also not permanent, but you probably knew that. <bjc>is there a way to use kvm without qemu? <nckx>Well, I meant ‘QEMU directly’. <nckx>Silly question, but did you reboot after reconfiguring, PotentialUser-57? It's generally not needed, but it can fix weird state. <PotentialUser-57>i haven't rebooted, only ran the system reconfigure. but good idea, i'll try rebooting <kaelyn>PotentialUser-57: If you added "kvm" as a supplimentary group for your user account in your system's config.scm, then you'd need to at least log out and back in for the group change to take effect. <nckx>(They are away whilst rebooting.) <nckx>(Which will almost certainly solve the issue but ssh, they don't know that yet.) *kaelyn checks if part/join messages are being suppressed in hexchat <nckx>22 May 22:43:30* PotentialUser-57 has quit (Quit: Client closed) <bjc>fwiw, if you don't want to log out/in, there's a ‘newgrp’ command that can also do the job, but only for the current shell *kaelyn unchecks the option to hide those messages <nckx>bjc: Would that fix the whole libvirt stack though, since it's a (or more) service? I don't know. <bjc>if the issue is that you can't run libvirt commands because of your current group, it'll fix that <bjc>libvirt will run as its own user/group, but that should be happening as soon as the service is activated <nckx>I know it's a cop-out, and might give some flashbacks to other OSes, but I've learnt over the years that ‘reboot’ is just a good first step to cover 5 bases at once. <nckx>If that costs me 1337pointz, so be it. <bjc>i reboot after a system reconfigure pretty much every time. there are some kinds of state that unix systems have a hard time cleaning up while staying up <PotentialUser-57>sometimes the simple customer service reply of just turn it on and off again is all it takes to solve a stubborn problem <bjc>i'm sure i'm overdoing it, but eh <nckx>I play with my kernel so much it's usually justified :o) <PotentialUser-57>by the way, i added myself to libvert group as well (long before trying rebooting). Do i just need to be in kvm group or should i be in libvert group too? <bjc>libvirt should be enough, is my recollection <PotentialUser-57>what im saying is i dont think i need to be in libvert so i should be able to remove that part? <bjc>i /think/ libvirtd handles all the interaction with /dev/kvm, so as long as you're using virt-manager, virsh, etc al, you should be fine <bjc>in that case you'll need to be able to read/write /dev/kvm <bjc>you don't need libvirt if you're not using its tools, no <silicius>I found out why stumpwm's documentation is compiling wrong. Last time it was a problem with guix package definition <silicius>this time it's an upstream issue with a super simple fix <bjc>running into problems with pulling the perl5 substitute, so i'm using --no-substitutes, and looking at how much it has to do, i think i may have made a mistake <nckx>bjc: If the problem truly is one bad substitute, you can use something like ‘guix shell -D perl5 -- true && guix build perl5 --no-substitutes’. <bjc>that's a great idea. i'm not looking forward to compiling gcc and linux <bjc>for some reason, it's still trying to grab perl-5.34.0 <nckx>OK, erm… I can't reproduce that (although I'm not using perl@5). <nckx>This is the same perl you're trying to build? <bjc>default perl is perl5, and it's part of the base packages <nckx>Maybe (this is really just a guess) try with --no-grafts. <bjc>my guess is that perl5 is needed for part of the bootstrap or something, so it's always pulled in <bjc>well, how long can it /possibly/ take to build gcc and glibc and tcc and all the mes and boot0 stuff? <nckx>Eh, if you're near me you were going to bed anyway 😉 <nckx>Yes, perl is *very* low in the graph, but I'm missing why the above hack wouldn't work for it. <nckx>Surely something obvious I'm missing because I'm → 😴💤 <lechner>Hi, do Guix packages declare installation prerequisites? For example, 'password-store' requires 'pinentry', and <lechner>byobu requires screen or tmux, but in both cases i had to install them manually <bjc>guix packages should specify everything necessary to not only run, but build the package <bjc>that said, things like pinentry tend not to be strict dependencies, as there are lots of implementations and they're loosely coupled (in this case, over dbus, iirc)
***Bryan is now known as Bryan-H
<bjc>it's sort of like how you need to be running x to use xterm, but you could be using gnome or plasma or whatever. so xterm doesn't say it needs anything other than the x11 libraries in order to build, but leaves the runtime support up to you <Bryan-H>Hi all, where is the clangd package located? It used to be in clang:extras but that doesn't exist anymore