IRC channel logs
2025-01-20.log
back to list of logs
<ekaitz>civodul: i have a question though! the lib has to be pointed to by the browser, is there a proper way to make it easy to find in guix? <ekaitz>i don't want to paste the whole store path <ekaitz>.guix_profile/lib is a proper path I can use? <civodul>i guess you could make a symlink somewhere? <stochastic>Is there a way to do a `system init` with guix deploy? <civodul>stochastic: depends on the backend; the Digital Ocean backend effectively does that, but not the SSH backend <lockbox____>Fun problem rn, savannah cgit is getting ddosd, and guix bootstrap requires access to the cgit for the bins since non of the other mirrors have the current required bins <lockbox____>rwp on #savannah also mentioned any automated usage should be using the smart-git interface - which kind of poses a chicken and egg problem wrt bootstrap <lockbox____>Possible solution is to add a guile dumb-git walker to grab the necessary stuff for bootstrap <lockbox____>(Also smart-git still being up is a plus as opposed to being 429'd from the ddos) <lfam>lockbox____: Which bins are you looking for? <lockbox____>Oh yeah that should get added to bootstrap.scm as a mirror <lfam>A whole lot of mirrors that don't have the file too <lockbox____>I also have aarch64 and riscv64 bootstrap fails for the same rn <jaft_r>I dunno if anyone else is using XFCE but is it expected that the power manager will overwrite the default elogind service settings or does one still need to set those oneself rather than relying solely on the XFCE power manager? <paul_j`>Good morning! I am doing some local development requiring cross compilation in rust, and have set up a directory with my code. I need to have probe-rs installed for cargo. I am happy to do this locally, as currently there is no guix package for it. Build fails due to missing libudev. I am using guix shell to create my environment, so I searched for the libudev with `guix search`, and found rust-udev. This is located in <paul_j`>gnu/packages/crates-io.scm, so in manifest.scm I added the module (gnu packages crates-io) to my use modules declaration and rust-udev in my packages->manifest. I am still getting "error: rust-udev: unbound variable hint: Did you forget a `use-modules' form?" The very simple manifest.scm is here: https://pastebin.com/27irrzL8#aKnHds0B - `guix shell rust-udev` without the manifest.scm installs the package ok - any hints as to what I <efraim>paul_j`: you probably want eudev if you need libudev <nckx>The gnu.org Mailman UI seems to be down so some messages may be delayed. <paul_j`>efraim: Ah - so rust-udev won't pull that in automatically? <paul_j`>I have to go to the doctor - I'll pick up comments on my return... <efraim>paul_j`: no such luck, the rust packages generally don't pull in their non-rust dependants currently. Without seeing the error message I'd guess you'll also need pkg-config <paul_j`>efraim: (back from the doctor!) This is the error message: https://pastebin.com/i5ncCeFe It says "rust-udev: unbound variable", so I would be surprised if that was a result of not pulling in a dependency. I will add pkg-config and see what happens... <paul_j`>No - still the same error message. Seems like my `(use-modules (gnu packages crates-io))' is not being seen... <Rutherther>paul_j: there is no "rust-udev" symbol in crates-io <Rutherther>paul_j: yes, that means package with the name rust-udev is in there. Not that it's name of the symbol <paul_j`>I am using the rustup channel from declantsien - perhaps it's coming from there...? <Rutherther>only: rust-udev-0.8, rust-udev-0.7 symbols are there, no rust-udev <paul_j`>Ok - my misunderstanding. So if I call rust-udev-0.8, it should work <paul_j`>How am I supposed to understand the symbol name from the output of guix search? I thought the name: field was the correct one to use? <Rutherther>paul_j: guix search doesn't contain the symbol name <paul_j`>I guess the other source of confusion for me: a simple "guix install rust-udev" worked... <Rutherther>again, that is different thing. guix install rust-udev is looking up package by specifications, so it searches for package with the name rust-udev. It doesn't have direct relation to the guile symbol <paul_j`>Is there a syntax I can use for manifest.scm which selects the packages I want to include in my dev shell which avoids these issues? <Rutherther>no syntax, but you can just use specifications yourself, by using specifications->manifest to obtain a manifest out of list of specifications strings <paul_j`>Thanks - I will go and read up on this. <jaadu>I have an issue where one fonts glyphs are taller than the other ones and its interfering with the rendering in emacs. <jaadu>The unruly font is -GOOG-Noto Sans Symbols <jaadu>I believed its in the same package as the font I want to use <nero>what is the name of the installer "graphial" UI ? <nero>i have trouble installing guix because im in a corporate network behind a HTTP proxy, and the installer GUI restricts the textbox for the proxy URL to 12 chars or such <efraim>it uses guile-newt to interface with newt, which I believe was previously called whiptail <gabber>after rebuilding a VPS my renew-certbot-certificates service fails to renew the certificates - i guess i have forgotten to backup my .well-known directory with the challenges. now the service is unable to set the challenges up as they were? is this a thing? is it possible to re-initiate the service or am i thinking in a completely wrong direction? <gabber>ACTION needs to restart sway hoping the cursor will finally reappear <Kurayami>can Guix switch to something like forgejo/codeberg? <Kurayami>I mean, the contributor survey does indicate that not a whole lot of people like the email based workflow <rekado>codeberg is currently used by the guix-science group <futurile>I think "switch" might be the wrong word, "add" is probably something that realistically people who like the current flow can't argue with: otherwise, people who like the email flow face "loss" <rekado>it wouldn't be something that happens soon as more work would be needed to satisfy requirements for integration with other services and enable automation. <rekado>this is what happens in the context of guix-science at the moment <nero>guix system init fails with error: fport_read: Input/output error, i confirmed that both disk space and RAM are not exhausted, is there something i can do to debug? <JoeTriden994363>Hey guys... Joe Biden here. I've decided to step down from the White House to focus on other projects. Billionaires are a threat to democracy, so check out https://BidenCash.st to put them in the bullseye. Keep an eye on the CNN inauguration for a promo code! <nero>Deltafire: zisofs_inflate returned -3, i guess my iso is corrupted <efraim>janneke: Remember what I said about my C code earlier? I just found a load-bearing comment in my awk code. <FrenchNewbie>hello i was sent here to ask a question about nonguix-system image release. Why does it not include the installation and setup of the nonguix channel in a more easy way than having to switch TTY, create the channels.scm file and modifying the config.scm ? <FrenchNewbie>apparently newer version of nonguix include that but not the nonguix image because it was released 2 years ago and not updated since <rekado>FrenchNewbie: better ask on the nonguix channel, they would know. <FrenchNewbie>i asked and the response was "nonguix stuff should be talked about in nonguix channel. This is not nonguix, it's guix releases, latest guix release is two years old. After newer one is made, I presume nonguix one will be made as well" <Rutherther>FrenchNewbie: no, you were sent here to ask about guix release, not about nonguix system image release <Altadil>FrenchNewbie: I think there’s been a little confusion. Releases of nonguix are tied to those of guix. And this is the right place to ask about guix :) <Rutherther>FrenchNewbie: I said from the start nonguix image release MAPS to guix release <FrenchNewbie>all i want to do is learn and use guix, why do i have problems at the installer that i feel plural users might have encountered in 2 years and no actions have been made from the devs to fix those issues <rekado>the lack of releases has been a source of frustration to others as well, but it's not something that's easily explained here without a lot of ultimately pointless context. <rekado>on the plus side: you can build a new installer image from any system that has Guix as a package manager. <Altadil>FrenchNewbie: it’s a known issue, but the "devs" are all volunteers, so we can’t force them to work on one thing rather than another. ^^ <FrenchNewbie>i'm genuinly interested in that context because i don't feel like "build your own installer" is the right answer <rekado>Guix as a rolling release doesn't really *need* a blessed installer, but it would certainly be convenient. <rekado>I feel that the discussion of context is only fruitful when it involves people who are in a position to change the conditions. <rekado>this discussion is already taking place elsewhere. <FrenchNewbie>do i have to make a gitlab account to push a change request ? <Altadil>FrenchNewbie: nope, you only need an email account. <Altadil>Wait, I think you might be thinking of nonguix, which does use gitlab indeed. ^.^ <getstate>Does anyone configure nushell with guix home? <getstate>did* and if so, would you mind sharing your configuration? <User42>Hello i am trying to package a rust project, one sub-dependency is rust-llvm-sys that is not packaged yet. When packaging it, it builds fine (without tests),however something curious happens in the main project build. Guix complains about the hash mismatch, saying expected A,found B,the usual. However when i switch A for B, the error is now the <User42>opposite (expected B, found A) what could be going on here? <futurile>User42: sound weird. when you build the sub-dependency, does it's hash change when you build it? so using --check ? <User42>.. never mind i think i actually just changed the wrong hash once (parent project) and never noticed. that confusing A B thing was just a random accident from that. <ekaitz_>hi! anyone knows which locale should I use in guix if I want it to be in english but be reasonable when it comes to weekdays, 24h format in times and all that kind of things? <ekaitz_>Rutherther: but which one should I use for that? and how do I configure that on guix? <Rutherther>I don't understand, I've already said LC_TIME, or what are you asking now? <ekaitz_>i mean: should I set LC_TIME=C.UTF-8, that's what you suggest? <ekaitz_>and also, how can I make that config in my config.scm? <ekaitz_>i have the locales configured, but there's just one locale name that I suspect affects to LC_ALL, do we have any entry to be more specific? or should I use a service that sets a global variable? <Rutherther>ekaitz_: yes, I am suggesting to set lc time to something that will give you 24h format. You can do what you are saying, env vars. In the end the global locale just ends up in the environment service, setting LANG envvar <ekaitz_>if i just set C locale for everything, would that affect many other things? <ekaitz_>i'm not a computer, but neither I am an english speaker from the US <ekaitz_>Rutherther: but thank you for the help <ekaitz_>Rutherther: and not only for the help you gave me today. For all of it. You are doing a very good community service here. <acrow>The times here appear to be in the future! or am I behind the times? :) <acrow>I guess that's a Libera.chat problems, not guix. <Rutherther>acrow: probably the client itself, not really libera.chat, maybe your browser is in utc or other timezone? (or the client has no access to timezone) <acrow>Rutherther: maybe, but my system time is correct to nist and guix system is running ntpd. My client is emacs... erc... <acrow>Rutherther: I'm double checking... <acrow>Rutherther: Gah, you are right. The guix system is not syncing time correctly. <ekaitz>okay so now I just edited the locale-definitions part of the os definition and i get the GUIX_LOCPATH issue, what's the clearest way to fix? why should I set it myself? shouldn't the system definition set it up for me? <ekaitz>sorry for the noise, but as I was saying I don't understand very well the need of setting GUIX_LOCPATH manually when setting `locale-definitions` shouldn't that just work? <acrow>So, what is the smart-git interface that allows guix system reconfigure to complete for a trivial conf file change? I'm getting persistent git-ssl failures while trying to reconfigure... <Rutherther>ekaitz: are you sure you have the main locale in your locales? <acrow>If guix system reconfigure doesn't fail outright it stalls while collecting objects at about 20% completion. I presume this is a consequence of the DDOS attack on Savannah? <Rutherther>by main locale I mean the one in operating-sytem locale definition (going to LANG env var) - also don't forget that your LANG env var is not going to be changed by reconfigure <Rutherther>acrow: what package is being built that causes those failures? <ekaitz>but it's going to be changed in a reboot, right? <acrow>Rutherther: Acutally none. I've only adjusted my hosts; so, the git pull seems redundant bc there are no new substitutes needed (I think). <ekaitz>in any case, i can't find any proper locale that makes icedove show the Mondays as the first day of the week (even though i can manually configure that) and show the time in english but 24h format <acrow>Rutherther: eg... paste.debian.net/1346046 <acrow>Rutherther: consistent with savannah issues? <Rutherther>acrow: I don't know, I don't really understand this backlog. The thing is your system shouldn't usually contact savannah, unless you update or have guix-for-channels (or just channels specified in guix-configuration). Since you said you did just a small update, I don't know why your system would try contacting savannah <acrow>Is there a way to suppress the git pull? I thought that the git repo was cached? <acrow>Rutherther: I do have channels specified. They are local so I didn't expect... <Rutherther>acrow: if you have the channels pinned, it shouldn't need to pull new version as long as you didn't gc it. If you don't have it pinned, it will download every time you reconfigure to check for new version <Rutherther>ekaitz: en_GB has 24 hour format and should also give you monday first as far as I understand <acrow>k, I'll add a -C arg back in. <rekado>ACTION is about to merge python-team into master <ekaitz>there are supposed to be modifiers like @euro in locales <ekaitz>i'm trying to configure en_IE now but i cannot :( <mirai>laptop guix users, is there any secret to avoiding the LUKS (slooow) double-password torture? <mirai>don't feel that walking around with an unencrypted laptop to be a good idea <Rutherther>mirai: yeah, you have like three options: a. have guix store on unencrypted partition, b. put your key to extra initrd (and thus also to store, so in that case your store has to be encrypted), c. the best imo, making a bootloader that copies the kernel and initrd to /boot partition that is unencrypted, that way your grub doesn't have to unlock it, only the initrd llinux system (which does much faster job than grub, for 800 GB ssd I had to wait like a... <acrow>Rutherther: How do I pin a channel? -- it turns out, for this machine, the only channel in the describe (guix and guix system though different) is guix with a commit on master. Is that not enough? I'm puzzled. <Rutherther>acrow: yes, that should be enough. But the source still might be redownloaded if you gc it <Rutherther>as far as I know guix-for-channels doesn't know the target derivation, you need to compute it from the source, so for that you need the source, compute it and then, if it's already built, it will be used. But I might be wrong here. <mirai>Dracut on Fedora can handle encryption just fine, wonder what kind of tricks are they doing there <mirai>Rutherther: re option b, isn't the key-file left unencrypted thus the laptop is still trivially accessible by any random passerby? <Rutherther>mirai: the key file is in the gnu store, when your disk is already unlocked, yes, then it is unencrypted, but it's on the encrypted partition <Rutherther>you can't password-protect it as far as I know, at least not normally, you would have to craft something yourself <Rutherther>(but after that you would have to still enter two passwords) <mirai>hmmm… so the unencrypted key in initrd is used by grub first and after it boots the kernel the /gnu/store keyfile is used <mirai>kinda like how you have to enter the password a second time for the kernel to use it? <Rutherther>there is no keyfile itself in the gnu store, it's in the initrd filesystem <Rutherther>and yes, this keyfile is used to unlock the filesystem within the initrd. First grub unlocks your system, loads the initrd, then the pid 1 process in initrd will try to unlock your disk, again, and it will be without manual password, as it can use the keyfile <mirai>how do you achieve the c. method? <mirai>in the c. method, does it prompt you for a password (once)? <mirai>I haven't read the code line-by-line yet <Rutherther>grub doesn't need it, as it doesn't need to unlock anything, initrd and kernel are unencrypted <mirai>neat, I assume this works in a similar way that Dracut does? at least when it comes to decryption speed and prompt times <mirai>its whatever Fedora uses for its initramfs <mirai>silly question, "how do I use this" <Rutherther>I doubt that it's similar, but I don't know what Fedora does <mirai>didn't find a system config in your repo to steal from <mirai>… completely glanced over it <Rutherther>note that it has some drawbacks - currently it doesn't remove older kernels or initrds, so you might have to delete them after some time. /boot is assumed to be unencrypted, and ideally the efi partition itself (I don't know what will happen if /boot/esp is used as efi partition) <Rutherther>and of course from the name it's only efi bootloader, but it shouldn't be too hard to make it into non-efi version (as long as /boot is on separate, unencrypted partition) <Deltafire>i went with option a, it was probably the easiest to set up <Deltafire>but for some reason it doesn't recognise my password the first 3 times, so now i enter it 4 times instead of the 2 previous <mirai>Deltafire: is grub doing the decryption? <Deltafire>no, initrd & the rest of the store is unencrypted <mirai>do you not run into space troubles by keeping them as separate partitions <mirai>I find it easier to have everything in a flat partition <Deltafire>well, it was originally one partition shared with windows, then i ran out of space so shrank windows, but that left it as [windows] [free space] [windows recovery] [guix] <Deltafire>so i installed guix onto the [free space] and left /home on the encrypted [guix] <Deltafire>overwise i'd have to move partitions around, which meant booting from a recovery stick <Deltafire>tbh, i'd say leave it as default - even if it means entering the password twice.. how often do you reboot your laptop? mostly suspend/hibernate <Deltafire>although hibernate may need the password if you've got store and/or swap encrypted <mirai>it gets particularly agravating when doing multiple reconfigure's in a row <mirai>if you're writing some extra code <mirai>the prompt is infuriating as well <mirai>slow to detect keystrokes, slow to decrypt <Rutherther>multiple reconfigure's in a row? how does that relate to disk encryption? <mirai>for ones that require a system reboot to take effect <mirai>not often but also not fun when you have to