IRC channel logs
back to list of logs
<pkill9>:why is it not common practise to make a launcher that keeps a log of the output of applications that are launched <lechner>fnstudio: xclip -selection clipboard ?
***ChanServ sets mode: +o litharge
***litharge sets mode: +b *!*@2001:470:69fc:105::2:8118
***wes234[m] was kicked by litharge (You are banned from this channel)
***litharge sets mode: -o litharge
*apteryx answering self from earlier question: to configure the amount of memory for a marionette system, you set it via 'virtual-machine', a record <apteryx>e.g., (define vm (virtual-machine os (memory-size "1G"))) <apteryx>err, (virtual-machine (operating-system os) (memory-size "1024")) <andi->while guix home isn't entirely "done"/stable yet how do people manage their emacs init files? I'd love to manage it via my system configuration just like most other things on the box. *vagrantc eyeballs new activity on the version-1.4.0 branch <apteryx>is there a way to have the QEMU invoked through a 'make check-system' test graphical? <apteryx>even defining the virtual-machine os with graphic? #f and commenting out '-nographic' from make-marionette, it still runs in terminal mode <shareni>Hi! Can someone help me out? I've been trying to install guix as an OS in virtualbox, and followed the SystemCrafters video about it. The first time it worked, but when i started updating the system it hung while compiling the new kernel (nonfree). After a few hours I C-c'd and restarted it a few times, but to the same result. After that I've rebooted the VM, and it wouldn't work past grub. Following that I've deleted the vm, and tried <shareni>reinstalling guix multiple times, following the same steps, but during login it shows the following error: "Sep 20 02:20:59 localhost elogind: /gnu/store/wi0wljc58pf1h2dsgrkw1wksxmz34ljx-logind.conf:11: Failed to parse handle action setting, ignoring:" and it becomes unresponsive. <ulfvonbelow>graphic? #f? I'm no expert, but that sounds like "off" to me <zamfofex>shareni: Does it work well with Linux Libre as the kernel? Because if so, then maybe you should ask for help in the forbidden channel. (Proprietary software is not appropriate here.) <ulfvonbelow>the "Failed to parse handle action setting, ignoring" line is something I've seen before on startup, I think it's caused by an option being set of the form "SomeOption=" - that is, no value set. I think it's fairly benign. <shareni>zamfofex: i haven't tried it. Is it allowed to say the name of that channel? <shareni>ulfvonbelow: I've tried googling it but couldn't find any results with that error message. Got any tips where I should look? <ulfvonbelow>you could look at the file in question and see what's on line 11 <ulfvonbelow>assuming you have some way to access the guest file system from the host <GNUtoo>shareni: why do you need a nonfree kenrel in a VM? It's probably way easier with the official installer <GNUtoo>shareni: Also we have interesting alternative to virtualbox, like virt-manager for instance. It's not a drop in replacement though <GNUtoo>Though it probably have better support for some of the hardware (like USB). The downsides are things like 3D acceleration only working under GNU/Linux. <shareni>ulfvonbelow: thank you, i'll try to chroot into it <shareni>GNUtoo: Yeah, but I'm trying it out as my new main distro, and need proprietary drivers for wifi on my laptop. I just picked virtualbox since it first came to mind. <shareni>GNUtoo: but thank you for the suggestion, i'll have to check it out at some point. Virtualbox really doesn't like i3wm for some reason <GNUtoo>OK, in the case of virt-manager there are the choice to emulate a wide variety of GPUs (things like "VGA" "cirrus" "virtio", etc) so there is often a way around issues like that <GNUtoo>Usually the virtio virtual hardware is more efficient though <shareni>Good to know, but in this case the UI is really buggy. Like when I try to fullscreen it it goes black and unresponsive until I spam i3 fullscreen keybinding and it goes back to the original size. Scaled sort of works, but stretches everything out. It's a mess, but hopefully good enough to figure out installing guix. Def trying out virt-manager next time i need a vm <shareni>ulfvonbelow: it was "HandleLidSwitchExternalPower=". I've added suspend and ignore, now it's freezing without an error message. <shareni>same results with the libre kernel. time to try out virt-manager i guess
***Xenguy_ is now known as Xenguy
<shareni>it was virtualbox all along... virt-manager shows the same error, but follows it up with the login screen. Thanks for the help <apteryx>is 'make check-system TESTS=lightdm' failing for anybody else? <apteryx>passes on one machine, fails on another <lechner>for some reason that last version fails to build with Throw to key `match-error' with args `("match" "no matching pattern" #<procedure inputs (a)>)'. <lilyp>lechner: you still need %build-inputs in make-flags <lilyp>also you shouldn't refer to (guix build utils) inside (gnu packages) – the (search-input-directory ) call needs to be evaluated at build side IIUC <ulfvonbelow>lechner: the difference between this-package-input and search-input-directory is, basically, when they are run. this-package-input searches the input labels of the package's inputs and propagated-inputs fields, and returns a corresponding <package>. search-input-directory searches an association list of (label . store-path) pairs for a store-path with a given directory, and returns that directory. in https://www.pastery.net/rydpgy/ whe <ulfvonbelow>you invoke (search-input-directory inputs ...), the 'inputs' there refers to the <package> field of the same name. If you dropped the #$ in front of (search-input-directory ...), it would instead be evaluated at build-time, in a context where no variable named 'inputs' is bound. %build-inputs, however, is globally-visible at build-time, so you can use that instead. <ulfvonbelow>the first version you sent (https://www.pastery.net/mxhzae/) would have the same problem we ran into before, since 'this-package-input' still uses input labels to find the package in question, and the input label being searched for is still hard-coded while the fallback input label is still taken from the package name. Note that in that version, you could replace #$(this-package-input "xkeyboard-config") with just #$xkeyboard-config an <tschilptschilp23>Hi guix! I just noticed a detail -- how can it be that vlc from android sees an (well, not accessible) sftp share on my laptop, while I don't have an ssh-server configured, while nmapping it from another device shows all ports closed? Also, 'netstat -tulpn | grep -i listen' from inside does not report anything?! <da77a>Hi. Guix seems a good idea. Ive been trying to install guixsd from the 1.3.0 ISO as a virtualbox vm hosted on windows 10. I get errors during the download/install of packages - "substitute .. TLS ... Error in the push function" . This aborts the install but I can resume from the config selections (last step) and it seems to make some progress / get <da77a>further each time. Any idea what this means/how I can avoid this? <jpoiret>da77a: I'd suggest using the latest installer instead of the 1.3, that version is now pretty old and we've got a couple of bugfixes on master <jpoiret>no guarantee that this will resolve your specific problem though <jpoiret>also, virtualbox is less supported than say QEMU <jpoiret>anyone know if we have the libc manual in guix? <jpoiret>ah, i'm stupid, i don't have the glibc package installed, that's the reason <da77a>jpoiret: I installed the version I found at https://guix.gnu.org/en/download/ - you are saying using latest development build is a "good idea"? As to the platform - for (corporate) reasons I'm running on a windows host and unless things have improved radically QEMU is just too slow on that host. My interest is in running on metal and windows vms as <da77a>the majority of targets - but its a place to start and a way to have a "dev" host that is usable. <jpoiret>right, i was just pointing out that you might have issues with that kind of host <jpoiret>and yes, i am saying that the latest installer should work better than the point release <jpoiret>point release are just snapshots in time that help with marketing :p <mothacehe>Hey Guix! 10 years anniversary seemed awesome, i was able to catch some really nice presentations using the live stream :) <civodul>mothacehe: hey, good that you were able to see some of it! *abrenon has to watch all the recordings of sunday afternoon <jpoiret>what are the "gotchas" i need to look out for when primitive-forking with fibers? <jpoiret>i know that fibers installs non-blocking ports or something similar, so i'm worried about what happens in forked off processes <civodul>the whole crowd of actual people is back in digital space :-) <bost>I have two versions of a package in two different channels. How do I specify which channel I want to use when running 'guix install ...'? I.e. does 'guix package' have any '--channel' switch? <abrenon>bost: doesn't the usual @version work ? it works for different versions within one channel <bost>abrenon: basically I have following situation: <bost>$ guix package -A maven\$ <bost>maven 3.8.6 out gnu/packages/maven.scm:2134:2 <bost>maven 3.8.6 out bost/packages/maven.scm:2143:2 <bost>And I'd like to use maven from that bost/packages/maven.scm <rekado>bost: “guix package” can take an expression with “-e” <rekado>so you could do “guix package -e '(@ (bost packages maven) maven)'” <bost>rekado: Ah interesting. Thx. Now I have to figure out how to pass such a parameter specification to the 'home-environment -> packages' from guix home. Ohje. <bost>rekado: and later probably to the guix shell. <rekado>does the home environment *require* package specifications? Or can you use package variables? <rekado>guix shell takes a manifest, so you can explicitly select a package from a Guile module in the manifest <rekado>generally I recommend the use of manifests, because they make transformations and special cases so much more convenient due to the fact that they are just Guile code. <efraim>I have the same content in my manifest file as I have in the packages field of my home config <bost>rekado: I have commented out the (package ...) in the home environment definition and "guix home ... container ..." works, so apparently the package specification is optional. <florhizome[m]>well guix weather says it has 1/2 substitutes but it has to build it <ArneBab>I just discovered --tune for guix package. Nice! <ArneBab>(I’ve been wishing for something like that) <GNUtoo>seer: later tell gnucode: I've managed to try gnome-clocks on Guix on my X200 (working to fix the i686 breakaga also made it possible for me to try gnome-clocks) and I've the same error than you ("Couldn't open libGLESv2.so.2: libGLESv2.so.2: cannot open shared object file: No such file or directory") <GNUtoo>seer: later tell gnucode: maybe it doesn't work because that library is dlopened somehow
***wielaard is now known as mjw
<florhizome[m]><florhizome[m]> "what does that mean? I don’t..." <- And everything that depends on it 😭 <dgcampea>how does guix manage its network connections by default? I'd like to configure some DHCPv6 DUIDs and IPv6 tokens (this is due to dynamic prefixes from ISP).
***Server sets mode: +Ccntz
<pkill9>somethign we don't see is live demos on web pages for software, using an embedded vnc instance <jpoiret>sneek, later tell dgcampea: Guix doesn't have network connection defaults, but the %desktop-services that you might be using contains networkmanager <apteryx>civodul: hi! I'm free for a bit if you'd like to look at the overdrive1 file system <apteryx>trying '# btrfs balance start -dusage=5 /' to sanity check things
***jesopo is now known as jess
<jpoiret>has anyone used eshell with guix profiles? <dgcampea>jpoiret: I do have desktop-services already but haven't directly used networkmanager yet although guix can autoconnect to a wired connection. Is this autoconnect due to networkmanager? <sneek>dgcampea, you have 1 message! <sneek>dgcampea, jpoiret says: Guix doesn't have network connection defaults, but the %desktop-services that you might be using contains networkmanager <dgcampea>if I didn't have desktopservices, would guix just not autoconnect to the wired network? <apteryx>civodul: I don't think the balance succeeded, it doesn't have enough space when doing "btrfs balance start -dusage=15 -musage=15 /" <efraim>apteryx: does 'btrfs balance status /' show anything useful? <apteryx>there's not enough free space to run a balance, so I'm running 'guix gc'
***Dynom_ is now known as Guest8894
<apteryx>qemu 7.1.0 seems unstable a bit for me; it sometimes appear to hang <jpoiret>you would have to set the interfaces manually and also run a dhcp client if applicable <civodul>apteryx: it didn't say that it failed though :-) *civodul tries again with more "free" space <apteryx>civodul: you ran it with what usage value when it didn't fail? for me, above 15 it fails with clearly with an out of space message <apteryx>below 15 it worked, but it didn't reclaim any unallocated space <efraim>you're checking with 'btrfs filesystem df /' and not just 'df /' right? <civodul>apteryx: i think i pasted that to guix-sysadmin, that was 50 because that's what efraim had suggested *civodul has no clue what they're doing <civodul>efraim: "btrfs ... df" is consistent with "df": 114G "free" <civodul>yet, "btrfs balance start -dusage=50 -musage=50 /" fails with enospc <civodul>does one of you want to connect to overdrive1.guix and give it a try? <efraim>I assume I'd have to connect through berlin? <civodul>i just don't get the "free space isn't free" thing <civodul>efraim: you can also connect directly on the right port <apteryx>civodul: I'm already connected to it, running 'guix gc' <apteryx>civodul: free space is a bit of a tricky thing to account for with btrfs, it's best to never run out of it ;-) but basically when using btrfs its 'unallocated' space is what matters <abrenon>civodul: it's a matter of unbalanced trees <abrenon>I had the same "what ?!" moment when our server ran out of space with 90G "free" <abrenon>there's space there somewhere alright, but the way btrfs adds things means that these free slots are somehow unreachable
***maximed is now known as antipode
<civodul>abrenon, apteryx: alright, but as a user, what should i do? i normally trust "df", and if deleting 10s of GiBs isn't enough, what else can i do? <civodul>apteryx: do you want to connect to overdrive1 to give it a spin? <apteryx>civodul: a balance job running daily should keep you out from thinking about it as a user <cbaines>monokuma (the Overdrive I have) gc's so that it has 200G free every day, rather than just 50G <cbaines>that might be something to tweak first <cbaines>the balance is probably less important if you have more free space, and I don't think there's a reason why the disk needs to be so full <apteryx>trick to debug marionette-control actions or a WM in a system test: take screenshots with (marionette-control "screendump before.ppm" marionette); they get saved to the failed build output <abrenon>apteryx: doesn't it slow things too much ? on our server we run it manually when we think we can and it won't affect too critical services <abrenon>maybe running daily is faster because each time it doesn't have much to do ? <abrenon>the first time we had to free the 100G imbalance, it took hours to complete (then maybe it's just the slow VPS) <apteryx>you can run it at night using low values for usage, e.g. -dusage=5 <apteryx>on node 129 (build farm), it runs every 3 days with -dusage=5 <apteryx>see hydra/deploy-node-129.scm in guix-maintenance <apteryx>looks like I haven't put that job yet on berlin itself, I should do that <abrenon>I dunno where that is but I was mainly fishing for config tips to improve our setup anyway <apteryx>it's one of the Guix project repos on Savannah <lechner>ulfvonbelow: thanks for the explanation! i am inclined to go with #$xkeyboard-config, but would that go against lilyp's position that all input requirements should be "searched" for using relative paths, for an effective decoupling? <ulfvonbelow>¯\_(ツ)_/¯ I'm not familiar with packaging guidelines, I mostly just package stuff for myself. Using search-input-directory with %build-inputs makes sense, I guess, as long as you know there won't be another input with the same directory that hides what you're actually looking for. Makes it work better with transformations, too (I assume package transformations don't apply to gexp-inputs?) <lechner>Hi, are there contribution guidelines for core-updates? Which style should be used? Also, guix seem to recompile every time i run 'guix build' <Reventlov>Hello. Is there something to make guix pull a little bit faster ? I mean, on an up-to-date channel, it takes 2 minutes doing "nothing" (for some definition of nothing) <rekado>lechner: what does Guix recompile? <stfnbms>After updating my Guix installation, M-x guix p r now gives me: guix-start-repl-maybe: Symbol’s value as variable is void: geiser-repl-company-p. What is up with that? <rekado>Reventlov: this happens in build-aux/build-self.scm <stfnbms>Mmh. Now I get: Starting Guix REPL ... <stfnbms>geiser-repl--wait-for-prompt: No prompt found! <jorge[m]1>When my electricity fails my Guix system fails to reset the correct time and my browser crashes. <ulfvonbelow>ah, I remember that happening to me before I replaced my motherboard's battery. Had to manually run ntpd to get it back to how it was. *apteryx has a special xvnc test running a remote desktop from inside the test VM <lechner>ulfvonbelow lilyp: just replacing inputs with %build-inputs, i get error: %build-inputs: unbound variable <ulfvonbelow>right, that's because the ungexp #$ is causing it to be evaluated at the same time as the gexp is evaluated instead of being quoted now and evaluated later during the build <lechner>rekado: this occurs every time i re-enter the enviroment (but without any git checkouts in between). please ignore the errors in the log. i also know, per lilyp, not to include (guix build tools) <lechner>ulfvonbelow: should i remove the #$ ? <ulfvonbelow>remove the #$ (that is, go from #$(search-input-directory ...) to (search-input-directory ...)) and it should work <lechner>ulfvonbelow: thanks, that worked great! is that an acceptable patch? <ulfvonbelow>I would think the patch replacing the assoc-ref with search-input-directory would be acceptable, but again, I'm not familiar with the guidelines. Regarding the test failure, I have no clue; only way to know for sure is to see whether it happens without the changes in question. <fnstudio>hey, i just tried 'guix shell --container guile guile-readline coreutils' and i was surprised that 'ls' shows all files in my home folder... anything i'm missing? shouldn't i be given a clean, isolated environment? <cbaines>florhizome[m], it's possible, email isn't always instant <podiki[m]>fnstudio: depends on where you ran the container, was it in your home folder? i believe by default you will just have access to current directory <fnstudio>podiki[m]: oh i see, interesting, let me try in a diff folder <cbaines>fnstudio, and see the --user and --no-cwd options of guix shell <podiki[m]>otherwise see the options for setting a different home or what to share <podiki[m]>I like to have a separate directory as my "shell home" that I use often <fnstudio>lol, stupid me! "The container [...] shares no files *other than the current working directory* with the surrounding environment." <fnstudio>if only i had parsed the *whole* doc line before asking here :) <apteryx>efraim: still running 'guix gc' on that overdrive1 box. it's collected ~900 GiB for deletion so far <civodul>apteryx: isn't it a bit... too much? <civodul>i would normally arrange to collect just enough to let the machine run for another 24h <civodul>900G is wiping most of the store i guess? <apteryx>yeah, I'm just letting 'guix gc' complete <apteryx>I agree that in general we want to keep things there to avoid re-fetching things all the time, but in the full drive situation we are, I think starting from a clean slate is a good thing to do <apteryx>then we can recover our unallocated blocks, we verify that zstd compression is enabled, put that balance mcron job and forget about it
***mark_ is now known as mjw
***Noisytoot_ is now known as Noisytoot
<fnstudio>is there anything similar to 'home-xdg-configuration-files-service-type' but for saving files to '.local/bin/' instead? i suppose not, but i thought of asking <fnstudio>i suppose i can simply use 'home-files-service-type' for that? <apteryx>so you need to call it (polkit-for-system) <apteryx>perhaps there's a mistake somewhere its used <ikappaki>I have been discussing this with Julien, but the process went cold
***warren_ is now known as warren
<apteryx>when I try to restart an inetd style shepherd service, I get: "In procedure bind: Address already in use". This seems related to using `make-socket-address' in the start procedure. <apteryx>TopExpert: thanks for sharing. you may want to contribute it as a patch <apteryx>OK, my "bind address already in use" earlier was my fault, I had some forgotten SSH tunnel <lilyp>well, at least it wasn't a log file or something <podiki[m]>ooo fancy weather icons in the guix weather output :) <TopExpert>apteryx: thank I've tried before but my Scheme is not top notch so things don't get accepted, so I just made a channel since many people need this <apteryx>ugh, since switching from slim to gdm, my remote machine goes to sleep after a while. how can I prevent that? <apteryx>TopExpert: OK. Thanks for putting it out there.
***maximed is now known as antipode
<jackhill>would be cool to get dconf setting in the guix service config, but I haven't look at it. <apteryx>haha, and now the UEFI is buggy, following resume from hiberation or something <apteryx>error: efibootmgr failed to register the boot entry: Block device required <apteryx>eh, our openbox has a 'firefox' menu <the_tubular>Everything I read online doesn't seem to work for me ... <TopExpert>apteryx: @ slim, this should be configurable with elogind for when it goes to sleep <TopExpert>ideally, you'd want to try Greetd, but it that's just the login daemon, it needs frontends like tuigreet or gdkgreet, very modern <TopExpert>the_tubular: is that the windows subsystem? what's the problem <TopExpert>apteryx: I've added most of the rust packages to build tuigreet if you want that <the_tubular>I'm encoutering a lot of problem, well first I can't even get a shell inside WSL when I import busybox like instructed on various github pages I found online. <TopExpert>the_tubular: is this GuixSD or Guix on foreign distros? <TopExpert>the_tubular: I have no idea about busybox, but if you can boot the installation media you have everything you need <TopExpert>you should try installing guix in a virtual machine, then simply copy the root filesystem into WSL <TopExpert>and you need an operating-system configuration that doesn't install a bootloader <TopExpert>I believe the docs around creating a virtual machine with Guix shows how to install such a bootloader <the_tubular>Yeah, problem is that it's a corporate computer. Guess I could try ... <TopExpert>what are the error messages, what's going wrong <TopExpert>(I don't know anything about WSL), but maybe there is some log that shows what's going on during the boot <the_tubular>Exacttly as he was doing and I fail at that line : Does --exec work? <TopExpert>the_tubular: try wsl -d guix --exec /bin/bash <TopExpert>it's possible /busybox doesn't exist as a file on the guix filesystem <TopExpert>"sh" command on guix is basically somewhere in /gnu/store/xyz12312312-shell/bin/sh so it's not on the path to execute just "sh" <the_tubular>But yeah, guix isn't installed yet in my WSL machien so "wsl -d guix --exec /busybox sh" should work ... <TopExpert>does anybody knows how "outputs" work in package declarations? <TopExpert>I have (outputs (cons "clippy" (package-outputs base-rust)) in the package declaration, but the resulting directory in the store is virtually empty <TopExpert>where does this declared "clippy" output come from? It is built in the package <TopExpert>is the "out" output, just a directory? Do I need to build the "clippy" binary into a "clippy" directory <pkill9>someone is porting microsoft's clippy to guix? nice *TopExpert needs his assistant