IRC channel logs
2022-12-30.log
back to list of logs
<the_tubular>I changed directory and reran the command and it worked ... <nckx>Most commands don't like running in a deleted current working directory. <nckx>(Is it still a cwd if it doesn't exist…? Most commands don't like philosophy.) <the_tubular>Umm, the directory does exist, but i was messing with docker mounts, so that might be it <nckx>Well, it didn't exist to that process at that point, is all I can say. <nckx>Nope. I'm a bit surprised you've never encountered this before (it's not that hard to produce — here it happens mostly when I'm messing around in a --keep-failed /tmp/guix-build-… in one shell, then re-run the build [so a new but identically-named directory is created] in another, then forget to ‘cd $PWD’ in the first shell.) <the_tubular>I see, well that's a good thing I encounter it rarely <nckx>Yes. I meant ‘it's nothing to worry about, not as arcane an error as your situation might have led you to believe’, not ‘omg you noob’. <mfiano>is there a message leaving bot in here? <the_tubular>My guix learning as hit a cap, and can't find good ressources to watch/read to keep learning :( <nckx>mfiano: ‘sneek: later tell NICK: blah’. <nckx>There's also a very useless feature if you forget the ‘later’: <nckx>sneek: tell mfiano: Good night! <sneek>mfiano:, nckx says: Good night! <mfiano>sneek: later tell luis-felipe: I'll try again. I may have just tried it on a random broken commit. <mbakke>the_tubular: what are some topics you would like to learn more about? <the_tubular>Guix / Scheme, but I learn a lot with videos and watched pretty much all I could find. I've read the manual and pretty much all of the little schemer and still feel like a noob :( <mfiano>Ah, well are you new to programming in general, or just Scheme? <the_tubular>Aka i just know a bit of python to get me through boring tasks. <mfiano>You have a couple of options for a good learning path, depending on how you feel. <mfiano>If you want to learn programming very well, using Scheme as a vehicle, just read SICP. <the_tubular>I'm open to all ideas, but learn better in video format. Never heard of SICP <mfiano>Scheme though, is a very small language, and you could also start by reading TSPL3 or TSPL4 (I recommend the latter if you will be using Guile). <mfiano>SICP is actually a course, in both text and video form. <mfiano>Structure and Interpretation of Computer Programming <mfiano>It's where to start if you are new to programming, and want to start with an easy, small, language <the_tubular>I'll definitely give it a shot, any other recommendations ? <mfiano>Otherwise, The Scheme Programming Language 3 (for R5RS) or 4 (for R6R6) is a more throw you in the deep end kind of start, but with good exercises along the way, and you can also ask questions on IRC. <mfiano>A revision of the original IEEE Scheme standard <jman>hello quick question about contributing patches to guix: since it's under the GNU org. umbrella I would need to assign copyright to the FSF through the usual procedure, correct? <mfiano>There are many variants of Scheme, and those are efforts to have them all more or less follow the same paths <the_tubular>Got it, thanks a lot mfiano, and if you ever think of something else, feel free to ping me <mfiano>the_tubular: Check out #guile for newbie help. I'm new to guile, but not other Lisps. <the_tubular>But yeah, learning guix is a lot different than learning other distro like gentoo, well that helps that I could read portage as it's written in python. <mfiano>the_tubular: All you need to really know is Guile follows all of R5, R6, and R7 (with some slight variations that it mentions in like 1 paragraph of the Guile manual) <mfiano>So read either TSPL3 or 4. Doesn't matter too much. 6 is larger. <jgart[m]>mfiano: Do you happen to know if scheme is a smaller lang than lua? I have a feeling core lua is smaller but I haven't counted the primitives or whatnot <the_tubular>Also, any podcast ? I know of FOSS & crafts, but it's not really "guix" focused <mfiano>I can't answer that because Scheme isn't an implementation, and afaik, Lua is. But I know very little about Lua. From what I know, I would say base Scheme is much much smaller. <haugh><mfiano> "Structure and Interpretation..." <- SICP is a general computer science textbook that happens to be written in Scheme. It is an iconic book, but it does not paint an accurate picture of modern Scheme in practice. <mfiano>Lisp is full of historical treasures. I can't think of any good CL book that didn't age well. <the_tubular>What does that mean "Scheme isn't an implementation" ? <nckx>jman: No. There is no copyright assignment at all. <mfiano>the_tubular: I probably won't ping you here, but feel free to ping me in #guile about any guile/scheme questions. If I can't answer, someone can! <jgart[m]>the_tubular: I think he means that guile is an implementation but scheme is a collection of standards for how to implement a scheme <the_tubular>Also is there some "easier" stuff I could read for just reading lower level stuff like writing bindings to configure a program ? <the_tubular>Guix's manual kind of touch on it, but it's not enough ... <jgart[m]>But Guile doesn't conform to any RNRS although it supports some of them <mfiano>The thing is Guile is a kind of Scheme which is a kind of Lisp. There are many kinds of Lisps, Scheme is one family of them. There are also many languages (implementations like Guile) in the Scheme family. <the_tubular>I understand that, but I don't understand which "level" do what mfiano <mfiano>You download a language and use it <the_tubular>What does Guile brings to Scheme that brings to Lisp ? <mfiano>I can't understand that question? <jgart[m]>the_tubular: re ffi You have existing code, blog posts, irc <jgart[m]>But not sure if someone has written a good blog post on guile ffi <the_tubular>I mean, what is a kind of scheme and what is a kind of lisp ? <haugh>Guix is a program written and configured in Guile, which is a dialect of Scheme, which is a standard for how to write a programming language in the Lisp family of languages, which includes Common Lisp and Clojure <jgart[m]>Guile is a kind of scheme which is a kind of lisp <mfiano>a kind of lisp is anything that derives from the original McCarthy 1958 invention called LISP <panadestein[m]>How to get Plymouth to work in Guix System? For instance, do you simply add it to your programs in the configuration file, or you create a service? <jgart[m]>the_tubular: TLDR Guile has one namespace whereas CL has two namespaces, one for functions and one for variables <mfiano>Lisp-2 is just a label Scheme people throw at CL to make distinguish that it has more than 1, but it has like 10 or something. <mfiano>If you read PAIP I think it is mentioned in that. <mfiano>I don't recall. I remember answering that in one of my reddit comments a couple years ago if you really wanted to dig. <jgart[m]>I just meant the books that are deemed important <jgart[m]>although that piece is in "the canon" in addition to being a musical canon <haugh>The cannon Graham keeps in his garage <Guest30>From my quick look it doesn't seem like the cuirass manual or tarball paths are ever populated <trevdev>Has anyone got a .dir-env.el method for activating a guix shell for a project? <trevdev>I'm about to muddle my way through writing an eval cons and try to muddle my way through it <trevdev>lechner: That's a start! Ideally I'd like to write something that will make emacs behave like it started from a guix-shell env, but I realize that might be tricky. Maybe it's simpler to do just that: activate guix shell then run emacs from there. <lechner>trevdev / doesn't buffer-env provide exactly the desired environment on a per-buffer basis? <trevdev>lechner: I'm nearly done the article, I'm a slow reader. I think it might be all I need <Guest30>it might be good at some point in the future to add a note about `buffer-env` in the guix docs, since it seems like it has some pretty common use-cases in guix (for those of us using emacs) <trevdev>lechner: I think this needs some work if it's going to see a `manifest.scm` up the directory tree. I have not read the source yet, maybe the author has accounted for this with project.el <lechner>trevdev / okay, i was just trying to help. i got buffer-env to work one time but then it stopped. you are probably right <trevdev>lechner: Hey, I appreciate the help! It was helpful :) Part of a learning process <trevdev>lechner: I think I found a better lead in exec-path-from-shell. I may be able to adapt the code to at least easily grab the PATH or any other specified environment variable <efraim>zacchae[m]: I make sure to not reconfigure from a tty. as for qutebrowser, we still have the qtwebengine bug which is why I have ("QTWEBENGINE_CHROMIUM_FLAGS" . "--disable-seccomp-filter-sandbox") in my environment-variables. <Parnikkapore_m>I'm giving the plover package some TLC; the app starts with the following error message (visible with --log-level debug, does not block usage) <Parnikkapore_m>OSError: libdbus-1.so.3: cannot open shared object file: No such file or directory <netamuffin[m]>Is there a way to make a shepherd service that runs stuff in a container that has access to only one of my network interfaces? I am pretty new to guix and guile, so reading the source of make-forkexec-constructor/container didn't seem to have that option. <sughosha>Hi all, is it possible to modify ~/.profile of from Guix Home? <sughosha>I want to source a file from ~/.profile, so I want to know if I can customize ~/.profile. <sughosha>Or if there is any service like home-environment-variables-service-type where I can source a script along with "export"ing environment variables. <paul_j>sughosha: First of all, a disclaimer - I am not very familiar with the details of guix. As I understand it, you would edit the .profile in a different location, then add it to your home configuration. When you run "guix home reconfigure", it will load your edited .profile into the store, and create a desktop link to it from ~/.profile. Since the store is not writable, you cannot just edit the .profile via the link - you need to re-edit <paul_j>your original file and run "guix home reconfigure" again. <paul_j>Actually - apologies - seems like I have mixed up my understanding. I have configured home-bash-service-type to add environment variables, but I haven't directly created .profile as I explained above. I do follow this approach for other config files (such as mpd.conf). <PurpleSym1>munksgaard: In light of #54729 (sorry, still not merged ☹️) I’m looking at wip-haskell again. I’ll probably add GHC 9.2 first, then try to update packages to Stackage LTS 20. <munksgaard>PurpleSym: No problem, I've not had time to work on guix stuff anyway. But if you do that upgrade, I'll take a look at itds patch. <PurpleSym1>munksgaard: Yeah, maybe we have to adjust your initial (or rather, simon’s updated) patch to fix attoparsec 0.14.4, which is part of Stackage 20.2. <sughosha>paul_j: Thanks for taking your time. `home-shell-profile-service-type` is what I was looking for. I had a look at shells.scm, but I didn't pay attention to this service other than the services for bash,zsh,fish,etc. <Parnikkapore_m>re Plover error: I'll try adding #@dbus to LD_LIBRARY_PATH and see what happens <axelbdt>Hello, I wish to deploy Guix System on a cloud instance supporting custom images (iso, qemu...). How can I find or generate an image that would allow basic ssh and subsequent `guix deploy` to the machine? I am a bit overwhelmed buy the manual and cookbooks. Thanks! <jgart[m]>They provide Guix System images in the cloud ready to go <jgart[m]>Or wait 5~ years for me to code up a guix deploy service... <axelbdt>Haha, thanks for the reply. So I should use the Iso from Guix download page and go through the graphic installer on the VPS instance? Won't I end up with a "desktop setup" with XFCE and other things? <axelbdt>Nevermind, not choosing a DE was actually an option ^^ Installing right now <mirai>netamuffin[m]: what you want to do is pass a namespace <mirai>you setup a new namespace which has its own routing table and add the desired interface to this namespace <rekado>cbaines: I was going to reconfigure bayfront to apply some minor changes to goggles, but since yesterday the big garbage collector lock is held. Is this expected? <netamuffin[m]>mirai: thanks. I will look into that, once i am back at my pc. <jas4711[m]>hi guix! i was just bitten by a GnuPG bug that was already fixed, and noticed Guix uses really old GnuPG. Any reason for this, or should I just try to build recently released stable 2.4 and submit patches for it? Should that be done against core-updates? <rekado>hmm, “guix refresh” appears to be broken <rekado>In procedure update-spec-package: Wrong type argument: #<package node-openzwave-shared@1.7.2 gnu/packages/zwave.scm:94 7fb352f918f0> <nckx>jas4711[m]: Go ahead, and yes (see ‘guix refresh -l gnupg’). <jas4711[m]><nckx> "jas4711: Go ahead, and yes (see..." <- thanks for mentioning that command! i suppose that if 'guix refresh -l FOO' is large enough, contributions go onto the core-updates branch? Is this documented anywhere? Learning that the core-updates branch existed was key for my understanding how grafts would work, and maybe the 'guix refresh -l FOO' command workflow teaches me something more about the Guix development workflow <nckx>jas4711[m]: Yes, see (guix)Submitting Patches (although it's the first match for ‘core-updates’, just search for that 😛). <jas4711[m]>D'oh! Thanks gain for the pointer! I read that long time ago now that I see it, but it never sunk in. Will try again <nckx>I re-read it a while back and realised I'd forgot (or didn't process) a few things… <nckx>The naive rebuild-count approach should disappear one day, anyway. <nckx>Thanks for updating gnupg! <lechner>nckx / should we instead look at the aggregate rebuilding time a change is expected to incur? <Parnikkapore_m>is the main concern here "number of things that might break" or "length of time this will take on the build farms"? <lechner>i think the point is to control the impact of rebuilding <nckx>s/ on the build farms// — this is also to be mindful of those building at home, not using substitutes. <nckx>But yes, it also touches on breakage: I expect people to rebuild all dependents of their master patches locally; for staging/c-u this is generally not done because it's prohibitive. <cbaines>rekado, regarding bayfront GC, yes, I've got a GC running to clear some space on bayfront. Feel free to kill it though if you want to reconfigure the machine. <jas4711[m]>yeah, i realized rebuilding core-updates locally took some time, the rebuild of the existing gnupg in core-updates just finished on my laptop... i'm hoping building a refreshed version of it will be fast now though <netamuffin[m]>(mirai) I've somewhat figured out how network namespaces work now. Do you know of a proper way to define the setup of a namespace as part of a service? I would like to have a containerized service that can only use one of my interfaces, ideally even without the shared interface disappearing from the default namespace. <jas4711[m]>are there substitutions available for core-updates? changing even tiny things in gnupg.scm causes a lot of things to be rebuilt (curl, python, ...) <nckx>Then you didn't really rebuild the existing core-updates gnupg earlier. You would have needed those then. Python is a (transitive) input to pcsc-lite, which is an input to gnupg. <nckx><are there substitutions available for core-updates> If you didn't get any, then… no. <nckx>Assuming you have substitutes configured already. <jas4711[m]>ok i'll just let it build instead of understanding the dependency tree. i have updated libksba, libgpg-error so far which maybe triggered more rebuilding to happen <lechner>jas4711[m] which bug did you encounter? <nckx>Oof, yes, libgpg-error is a major rebuilder too. <jas4711[m]>the bug is fixed in gnupg 2.2.41 too, but I reckon I should add GnuPG 2.4 too while I'm at it <jas4711[m]>my idea was to rename the current gnupg to gnupg-2.2 and bump that to 2.2.41 in case we need the gnupg 2.2 branch somewhere <nckx>jas4711[m]: You could just add gnupg 2.4 to master without touching anything else. <jas4711[m]>let's see if rebuilding works or not :-) i realize it may be too early to move to 2.4.0 but we won't know until we try and fail <nckx>That would fix users of ‘guix install gnupg’ at least. If you rename the variable (more work, but also won't cause rebuilds), it will address all user-side users of gnupg. <nckx>jas4711[m]: Same process. <nckx>(define-public libgpg-error-x.y (package (inherit libgpg-error) …)) <jas4711[m]>nckx: ah, you mean introducing libgpg-error-1.46? interesting <tsvallender>I'm installing packages with Guix Home via (home-environment (packages (list perl ruby node))), but how can I specify a version? I thought it would be @v but that doesn't work. For Ruby it seems to work to specify ruby-3.2, but the same syntax doesn't work for Node <sneek>tsvallender, you have 1 message! <sneek>tsvallender, nckhexen says: Import (gnu system setuid). <nckx>tsvallender: It's not a ‘syntax’, it's just a different variable name. Guix defines a ‘ruby’ and a ‘ruby-3.2’ variable, this does not apply to every thunkable name-version combo. <lechner>tsvallender / you are using variables or specification strings? <nckx>It seems Guix does not have any other versions of node, only ‘node’ (10.24.1). <nckx>700 lines further down 😛 <tsvallender>Yeah it's the v14 I wanted, that's node-lts? Where did you find that? <nckx>The A is for ‘available’, the argument is a regex (so just ‘node’ would work but return a lot of noise). <nckx>I actually first looked at gnu/packages/node.scm, which has node near the top, which is how I missed the node-lts definition near the bottom. So don't do that. <nckx>Instead, use the ‘location:’ field that ‘guix package’ returns. <nckx>Yes, this is a bit tedious now that I've typed it out. <nckx>IWBN for the CLI to print the variable, if that could be done reliably. <tsvallender>So to see the name, look at the line given by the location? That's how I should have found "node-lts"? <nckx>Yeah, wasn't it obvious? /s <jas4711[m]><lechner> "jas4711 / did you see this..." <- Yes -- although -oPubkeyAuthentication=unbound is a better workaround to still use PQ crypto. I can get into my systems with the workaround, but I think we should have more recent GnuPG since it is an important package <nckx>jas4711[m]: There's no reference from openssh to gnupg I'm missing, right? That is, ssh uses gnupg from PATH (or some other environmental place) if it's installed? <lechner>jas4711[m] / i did not mean to advocate against the update; the purpose was merely to provide immediate relief <nckx><i can't SSH into modern boxes any more> So does this imply that Guix's OpenSSHd isn't modern? (I didn't check.) <jas4711[m]>nckx: ssh talks to gpg-agent and all this is working, I get the error from scdaemon which is running under gpg-agent. so the call path works but there is a bug in scdaemon <nckx>So just updating the gnupg (agent) in the environment should fix this? <jas4711[m]>nckx: It doesn't happen for everyone -- only if you have a Ed25519 key on an OpenPGP card and want to use it for authentication <nckx>Here I was trying to reproduce it :) <nckx>I use EC keys but not on a card. <jas4711[m]>updating GnuPG in Guix seems to be a perfect learning experience for me. Right now I'm stuck adding a new ./configure parameter that the updated libgpg-error requires, but this helps me get along with all the paranthesis. will unfortunately run out of time for today, but will keep working on it <rekado>what’s the code and command leading to this output? <rekado>lechner: uhm, you’re compiling guix.scm there <rekado>is this with the guile-build-system? <rekado>unless you really mean to compile it <Guest30>i assume the guix.scm file contains the guix definition for the guile package you're building? <rekado>lechner: use not-compiled-file-regexp <rekado>see the “build” procedure in guix/build/guile-build-system.scm <lechner>yeah, it was an innocent addition to the repo. rekado hit the nail on the head <rekado>normally, for a Guile package you’d put your modules in a sub-directory <rekado>the guile-build-system can be told to compile everything in that directory then <nckx>lechner: I see that the Huh?-bot doesn't match debbugs.gnu.org/cgi/bugreport.cgi. <ss2>hello, I reinstalled a Guix on a lvm partition scheme. Did a reconfigure, but now grub fails to find/load the kernel. The previous generations too. I did no gc so far. <ss2>entering insmod lvm doesn't help either. <ss2>Ther kernel image in the store is there. I just checked. <lechner>nckx / i added XXX@debbugs.gnu.org a few days ago but not the two URL forms /XXXX and /cgi/bugreport.cgi?bug=XXX should I do so? <Guest30>ss2: there's a lot that could be going wrong there, is this the first time you've run reconfigure (i.e., is it a fresh install)? <nckx>That's up to you, but I don't see any value in the inconsistency. If the bot always outputs issues.guix.gnu.org/XXX, it's a nice way to advertise our tracker. <ss2>It was my first reboot after reconfiguring the third time. <ss2>So I don't know if it wen't wrong ealry. But I could reboot after the fresh install yesterday. Something must have gone wrong later. <nckx>ACTION missed the paste. So it does. <rekado>ss2: can you boot the first generation? <ss2> The error is: error: no such device: /gnu...bzImage; error: file `...' not found; error: you need to load the kernel first. <ss2>I noticed that I need to load lvm manually first, but it would boot just fine first. <lechner>ss2 / you can either boot into an earlier generation or work in the grub shell <ss2>Don't wanna reinstall again. It just took me the whole day. <ss2>nope, all generations are broken. <ss2>well, not bootable so far. <ss2>(lvm/device-NAMES) (proc) (hd0) etc. <ss2>sorry, have to type it. <ss2>the devices are there apparantly. <ss2> /boot is in its own partition. <rekado>ss2: can you access /boot from grub? <ss2>it is in its own partition outside of lvm, the rest is inside. <rekado>did you arrange for the kernel and initrd to be on /boot? <rekado>ss2: when you append a slash after the device name you might be able to see what’s on it <rekado>all you need for booting is for grub to be able to access the initrd and kernel image <lechner>ss2 / in the grub menu you can hit 'e' on any entry (for edit). what are the starting paths for linux and initrd ? <rekado>lechner: Guix only generates /gnu/store locations <rekado>gotta make Grub find the disk that holds /gnu/store, so that these locations can be resolved <ss2>yes, they go to the store, and root=/dev/mapper/device. <ss2>okay, how do I ls a directory in grub? I'll just read some docs for a while. <rekado>ss2: I suggest going to the grub console and ls’ing the devices <rekado>append a slash if nothing happens <nckx>Yeah. You can also use variables like $root and $prefix to figure out what GRUB thinks should be there. <nckx>rekado: It will print file system info without the /, nothing won't happen. <ss2>ah, now I got it. Yes, I can list /boot# <rekado>okay, but /boot doesn’t contain your kernel and initrd <rekado>what about the partition that has /gnu? <lechner>i think /dev/mapper/device is wrong. it should read /dev/mapper/${volumegroup}-${volume} <rekado>gotta identify it and then set the root variable, so that grub picks that disk to read from <nckx>You can list all detected partitions with just ‘ls’. <ss2>I've got all partitions in front of me. All lvm volumes to, which is system-rootfs. <lechner>i agree, however, that grub has to find it first, via "root" <nckx>lechner: That doesn't sound right. <nckx>Maybe some other distro does it like that. <ss2>and the kernel is sitting in there. <nckx>/dev/mapper/device is correct for any device named, er, ‘device’. <nckx>ss2: On which (GRUB) device, and what is $root? <nckx>I'm going to stop, you're getting too many instructions. <ss2>okay, I've got (lvm/system-{home,rootfs,swap}) and (hd0) (hd0,msdos{1,2,3}) and (proc) <nckx>lechner: Hm? You were correct, we were just asking too many questions. <ss2>great, I've bein playing with lvm the last days. Something had to fall apart eventually. ^^ <ss2>and I didn't place my kernel out to say /boot. It is left in system-rootfs, which wasn't smart maybe. So far my test runs had been working. <lechner>nckx / not sure what you are talking about <rekado>ss2: it’s okay not to have the kernel + initrd in /boot <nckx>ss2: If you ‘set root=(the-device/that-had-gnu-store)’ on the command line, then hit ESC (IIRC) to return to the menu, it should boot. Once. <rekado>it would only be a problem if grub wasn’t able to read from your LVM volumes <lechner>or hit 'e' and change to your liking <nckx>ss2: That's why I asked what $root is *now*, i.e., what GRUB was trying and failing to read. <lechner>then you can use the store path in the menu <ss2>after setting root, I've got: error: unknown filesystem. <nckx>That's true, but then you're just inserting the implicit $root by hand, twice. Setting $root is equivalent but less work. <nckx>ss2: You said you could ‘ls (something)’ that showed /gnu/store/stuff. But setting $root to that same ‘(something)’, then running ‘ls /’, doesn't work? <nckx>Smells like a typo or unclear instructions on my part. <nckx>* ‘ls (something)/’ that showed /gnu/store/stuff <ss2>ls / doesn't work now. It is error: unknown filesystem. <ss2>ufff, maybe the fs is having problems. Dmesg was throwing errors early on. Will fsck it for now. <nckx>So what does ‘echo $root’ say now? <nckx>Or did you never change it? <nckx>The silence makes me fear we'll never know. <sughosha>Could someone take a look at https://issues.guix.gnu.org/60201? This is first time I created a new .scm file, don't know if I have to add anything else. Also, I did some mistakes which I fixed in the next patches. <ss2>If I do ‘set root=(lvm/system-rootfs)’, then ‘ls /’ I get ‘error: unknown filesystem’. <nckx>sughosha: You need to register new files with gnu/local.scm under GNU_SYSTEM_MODULES (sic, that includes packages). <nckx>sughosha: Also, just squash (=‘merge’) the two commits when sending a v2. Revisions should stand on their own. <ss2>I just rebooted and run an fsck against the lv, which is fine. So maybe I borked it somewhere else. <ss2>ah, but i can do a ls / if I mount my lv home partition. Something is broken somewhere else. <sughosha>nckx: I don't know about squash and merge, I will look for that. Thanks. <ss2>okay, whatevers. Reinstalling will be it. <ss2>but then I was still wondering why grub (then when it was still working) would have me do an insmod lvm. <apteryx>mbakke: what was the install-guix.sh domain again, and how can it be usedÉ <nckx>ss2: For argument's sake, why would Guix generate a different grub.cfg if you reinstall? <nckx>apteryx: guix-install.sh <nckx>(Name of the script in etc/.) <ss2>lechner: I created the ext4 partition and it had no big directories features enabled and was constantly throwing errors until I did a tune2fs to make it quiet again. But I don't know how much had gone wrong actually, and it is a new system anyway. <sner>happy-shoes-1979: what output are you specifically seeing from straight when you attempt install? <nckx>Pretty sure that's on your end. <ss2>nckx: I don't suspect that guix did anything wrong. This is my running system config that has been working for ages. It has never failed on me. :) <ss2>but I was sloppy with my partitions early on. <unwox>apteryx: it works for me too <nckx>Both at home and in Munich. <nckx>I had a Canadian VPS until last week :) <apteryx>I'm using my internet provider's nameservers <nckx>apteryx: Maybe invalidate nscd cache? <nckx>you're still misspelling it. <sner>happens to the best of us <ss2>that's such a cool idea with the domain. :) <nckx>In apteryx's defence, I didn't notice it was still wrong in the wget example. <nckx>But then I always need coffee so maybe not much of a defence. <nckx>ACTION goes to get ☕ for everyone. <apteryx>ACTION blames the object-oriented-like naming of the script <ss2>okay, I'm gonna reset this machine again. :/ Thanks for your help still. <mroh>ss2: guix's grub (and generated grub.cfg) does not have "insmod lvm" (and "set root=(lvm/...), so out of the box its not possible to boot from lvm (w/o editing grub.cfg). Im working on it... but its time consuming (and not much fun) to test... <nckx>I'm a bit weary, but good luck ss2. <nckx>mroh: Do you have patches? <zacchae[m]>sneek tell efraim: Oh wow, and just like that my qutebrowser is fixed (as far as I'm concerned)! Thanks a ton <sneek>efraim:, zacchae[m] says: Oh wow, and just like that my qutebrowser is fixed (as far as I'm concerned)! Thanks a ton <mirai>netamuffin[m]: I don't yet though I've been planning on doing it for a while <ss2>oh, I'd be happy for patches too. I've got a good use case to test on. <mirai>just need to find some time for it <mroh>nckx: very much wip, yes... <nckx>apteryx: If you want to register install-guix, go ahead 😉 <zacchae[m]>sneek later tell efraim: Oh wow, and just like that my qutebrowser is fixed (as far as I'm concerned)! Thanks a ton <nckx>We could also add install.guix.gnu.org to make it look more trustworthy, but mbakke's unofficial hack is nicer, damn them. <apteryx>nckx: eh. I still haven't uploaded an apteryx photo to apteryx.ca, so I don't think I should manage extra domains ;-) <KarlJoad>Does guix deploy respect verbosity and debug flags? I cannot make figure out why the deploy fails SSH authentication when I can SSH there myself. The necessary information is in .ssh/config too. <sneek>KarlJoad, you have 1 message! <happy-shoes-1979>sner: I think this is a minimal reproducible error, but i'm still learning how guix shell works: <happy-shoes-1979>guix shell --network emacs emacs-straight-el git findutils coreutils -- env HOME=$HOME/emacs-test emacs --eval "(progn (add-to-list 'load-path \"$HOME/.guix-profile/share/emacs/site-lisp\") (guix-emacs-autoload-packages) (straight-check-all) )" <sner>nckx: the maintainer team still has `guix.info` registered, right? any thoughts on `install.guix.info`? <florhizome[m]>archwiki says, to enable hibernation I need “resume” in the initramfs hook <nckx>We don't have initramfs ‘hooks’. <nckx>The initrd code already handles hibernation. All you need is ‘resume=/dev/foo’ in your kernel arguments. <sner>happy-shoes-1979: let me take a look <nckx>florhizome[m]: Some more complex set-ups are supported (encryption), some aren't (files, I think?). <florhizome[m]>right now ‘sudo loginctl hibernate’ just seems to shut down the device <sner>happy-shoes-1979: also, try to keep your message lengths relatively short, the matrix->irc bridge truncates your messages to a link otherwise <nckx>sner: Just for length, you mean? I'm not sure we have a clear vision on what that .info domain is for, apart from its use by the HPC project. <nckx>I'm not familiar with loginctl. <nckx>florhizome[m]: Did you set up a swap partition and add it to resume=? <nckx>Those two things are all that's needed. <sner>nckx: i didn't realize the hpc site was sitting under that domain (thought it was going unused), disregard <nckx>My understanding is that it's for metaguix stuff (e.g., http://foundation.guix.info/, HPC, …) but I don't know if that was ever formalised. Maybe that opinion's not shared. And the choice to mirror the front page makes it less than clear. <nckx>(UUID?) Hm, could you try with the /dev file name instead? <nckx>Yeah, I'm not sure where you got UUID= from? I don't think that was ever supported. <nckx>I'm not sure where you getting this from. <nckx>Using Arch-specific docs for Guix will eventually end poorly. <nckx>I'll add ‘UUID=’ prefix support eventually, but for now, if UUIDs work at all, it's without that prefix: resume=d49b9fa5-a55d-4b98-a4fa-c3cb645480d4. <nckx>I tested that when I added the feature, but never since, so I hope it still works. <nckx>Until then, /dev/foo is what I use and safest. <nckx>The lack of ′UUID=’, ‘LABEL=’ prefix is because it re-uses existing Guix functions for that (in se, good), but Guix never followed the blkid-standard FOO= notation used by many other distroes (probably bad, because it's fundamentally ambiguous). <sughosha>If I should sum up all the patches and make into a single commit and resend, I am also fine with that. <nckx>sughosha: Yep. The v5 should be entirely self-contained, not assume that someone has the previous version applied. So one single patch that adds the entire file and adds it to gnu/local.mk at the same time. https://paste.debian.net/plainh/06c23579 <sughosha>nckx: Ok. I will create a new seperate branch and make single commit. Then send it as reply to the same issue. <nckx>Sounds good. The new branch isn't needed, but if that helps you keep things clear in your mind whilst you learn, go right ahead. <sner>happy-shoes-1979: glad to see you resolved it, sorry for the delay, had to go afk for a few mins there <netamuffin[m]>mirai: nice. tell me, in case you have you develop a solution. thanks again! <lechner>hi can guix boot of a luks-encrypted root drive? how does grub find the modules? thanks! <nckx>Yes. GRUB builds in the modules it (thinks it!) will need to load the others; this can include cryptodisk. <klat>Also, how to enable KMS in general <lechner>nckx / our grub includes cryptodisk, or i have to tell it to? <nckx>klat: There's no support for silent boot through plymouth yet. I think rekado was working on that. I use a screen-filling Linux boot logo for that, which works, but it's a gross hack. <nckx>lechner: It's automated. <klat>nckx: thanks, that saves me some time debugging <nckx>klat: I'm not sure what you mean by ‘enable KMS’. It's not opt-in, or at least not on any sane hardware (can't speak for Nvidia :). <klat>I am interested, how do you implement your hack :) <lechner>nckx / automated via mapped-devices? <klat>Yeah unfortunately I have an Nvidia laptop <sughosha>nckx: I sent v7 patch with one single commit. <nckx>It also assumes you build your own kernel, but who doesn't do that. <nckx>lechner: It could be, but I think we just set a CRYPTODISK_PLEASE variable unconditionally. <nckx>(Name obviously not accurate :) <KarlJoad>Does guix deploy not read SSH's config file (somehow)? Deploy says it cannot reach the machine, but everything is correct; I can manually SSH there. <nckx>It could be cleverer, but it's probably not worth the risk. GRUB still fits in any BIOS boot partition I've ever had. <florhizome[m]>I get an “/gnu/store/… initrd” not found by grub, for the last configuration too <nckx>The configuration before which exact change? And what's the error? <nckx>(I'd recommend a fsck, but hey, btrfs is magic and never breaks so doesn't have that.) <lechner>nckx / does the LUKS module thing also work with EFI boots? <nckx>It should but I can't say from experience. It used to in… 2018 or –19. <nckx>Haven't used UEFI+LUKS since. <nckx>I'd expect noise if it broke, though. <lechner>it's much easier when you are around <lechner>are there any hoops after adding a new channel? <KarlJoad>I figured out my issue. Turns out you have to specify the identity file. <KarlJoad>lechner: After adding a channel, usually no? I have never had an issue unless it was my own channel that I broke myself. <lechner>how may i check that a channel was pulled <KarlJoad>If you already pulled, you can use "guix describe" to find out the channels you are using. <KarlJoad>I just tried to use the guix deploy -x to reboot a computer, and got an error, despite the computer doing what I want. Is that expected behavior? <KarlJoad>lechner: "Throw to key `guile-ssh-error` with args ... ("channel-get-exit-status" "Parent session is not connected #<unknown channel (freed) ..." Ellipses indicate omitted parts. <lechner>KarlJoad / yeah, we have been getting those on deploy recently <KarlJoad>I'm pretty sure this one is a non-issue because I sent the command to reboot, so the SSH server and connection just died. <KarlJoad>I could send a pastebin to somewhere so others could view it. I could also submit a bug report, if that is the better move. <lechner>Hi, should the guix from the system profile ever be preferred over a personal user profile? <nckx>rapa: In future, paste the actual ‘View build log at’ build log. In this case, it reports ‘Component ASDF/USER::CL-GOPHER not found, required by #<NASDF-SYSTEM "nyxt">’. You need to add cl-gopher to the inputs. You also need to add cl-tld, because the CL god is a bloated beast that demands sacrifice. Then you'll get ‘Component ASDF/USER::DISSECT not found’, which is where I stopped. <rapa>nckx: my bad, I realised that my paste was not useful <rapa>One last question if I may <florhizome[m]><nckx> "The configuration before which..." <- before the last reconfigure which added the resume= line <nckx>Try booting with ‘noresume’ (by editing the GRUB ‘linux’ command line with ‘e’ at the menu). You previous ‘shutdown’ using loginctl probably wasn't a shutdown at all, but a hibernation, which was now resumed (which is very bad). At least that's my guess. <florhizome[m]>nckx well the btrfs error could have multiple issues but i guess it's because i resized the Partition to add swap <florhizome[m]>i actually rebooted into an older session, then shut that down again and then i arrived there <nckx>Well, it's possible that my guess above is wrong, and to be frank I hope it is. <nckx>rapa: If you're waiting for permission, no need. <mirai>Is there anything remaining to do for #60131? There's not many ways to handle AVIF images in Guix at the moment <rapa>nckx: thanks! I was actually thinking how to proprly ask, it is quite trivial though <florhizome[m]>I am downloading a guix image to reconfigure from a chroot anyways <rapa>(append (list cl-gopher cl-tld)) this is not working and I don't know why <rapa> (inputs (modify-inputs (package-inputs nyxt) <rapa> (append (list cl-gopher cl-tld)))) <nckx>It might be because nyxt uses old-style inputs. `(("label" ,package) …)) <nckx>I'm using (inputs `(,@(package-inputs nyxt) ("foo" ,foo) ("bar" ,bar)) myself. <rapa>So you don't use modify-inputs at all, ok <nckx>No, mainly because I was in a hurry. <nckx>florhizome[m]: If possible, throw away that btrfs, don't keep using it :-/ <nckx>At least until you can scrub it or otherwise confirm it's healthy. <nckx>(Does scrub verify all metadata? I don't know. I'd guess so.) <nckx>That builds, rapa, but I didn't run it. <rapa>I can feel your love for CL <lechner>Hi, anybody having problems getting M.2 SSDs recognized in UEFI mode (but they are present in CSM mode)? <nckx>rapa: Nah, I was just being silly. It's not CL's fault. <nckx>lechner: Is this a old Thinkpad by chance? <nckx>I ask because then I can at least try to reproduce it, but this particular one doesn't have UEFI. <lechner>didn't rekado also have problems booting off of M.2 on the embedded thing? <mirai>lechner: I'm running 2 M.2 ssds with btrfs-1 raid and no CSM <mirai>with a 2022 motherboard as well <mirai>if the issue is already present at the UEFI level maybe try updating the fw for it? <mirai>some nvme drives also may require a fw update <lechner>ACTION is the only one running ten-year old rotational media with sapphire bearings <mirai>if both ssd fw and mobo fw fail to solve the issue I'd report it to the vendor(s) <mirai>(you're paying not just for the hardware but for their support as well) <lechner>unfortunately, nobody is paying for my support <lechner>anything to win another soul over to Guix <lechner>i am so embarrassed because of my antique equipment; no SSDs or DDR4 memory here <florhizome[m]><nckx> "(Does scrub verify all metadata?..." <- I can check it first <leg7[m]><lechner> "Hi, anybody having problems..." <- Recognized as boot drives? <bjc>zfs scrub verifies every block that's accessible, including meta-data <nckx>florhizome[m]: I mant reformat it if possible, since you mentioned a guix image. <nckx>bjc: I presume the same is true of btrfs? (If merely by virtue of having to, well, walk the entire tree & checksums in order to perform the scrub.) <bjc>i would assume it's true for btrfs, since a scrub that doesn't check metadata is pretty broken <bjc>but then again, btrfs still has a write hole, so… <florhizome[m]>nckx the Image is the guix Installer that i'm downloading as a rescue tool <phodina[m]1>What is the best way to refer to the root directory of the channel in the code? E.g. when I want to use some file I add to the repository like defconfig (not a patch). <leg7[m]>lechner: Maybe because you installed your bootloader in legacy mode <lechner>but we will blow the bios. the board is coreboot approved <nckx>ACTION just uses search-patches… <rapa>nckx: nyxt is working like a charm, thanks again <rekado>lechner: no, wasn’t me. I have no M.2, nor do I have 2022 boards. <rapa>I want to bother you good people once last time today. I have something like this (service home-files-service-type `("init.el" ,(local-file "init.org"))) <rapa>This is an org-mode file that I usualy tangl manually <rapa>Can I do the tangling using scheme, preferably in the above code block? <KarlJoad>You could define a package that runs emacs in the build phase to tangle the file. That may overcomplicate though, I am not sure. <phodina[m]1><phodina[m]1> "What is the best way to refer to..." <- Solve by using `(search-path %load-path FILE)` <florhizome[m]>nckx i add 'noresume' at the end of the file after i Press e in grün? <omlet[m]>How long can the distrobox be added to the guix latest? <nckx>When the patch is revised and approved.