IRC channel logs
2025-03-20.log
back to list of logs
<RavenJoad>arebi: There are some limitations with the way haunt & commonmark work together, and that is one of them. I'm considering just writing my own extension to commonmark to do that kind of thing in a reasonable way. <arebi>thanks civodul: however, i understand that commonmark does not support html5 tags and attributes and especially <div> ., but i may be wrong., i have just started learning guile and haunt., so wanted to see how well a quick tag can be extended with attributes so that it can be created as a module for others to reuse., because i am sure the html5 attributes would be helping a wider userbase <arebi>civodul: do you have an example so that i can learn and try to code? i was thinking about the commonmark-reader post-process but with standard specification not supporting <div> not sure how to approach <arebi>mange: thanks for your suggestion. will it be in the site module? to learn about post-template and see how additional attributes could be generated based on the key value pair written in a post content file? md? <civodul>arebi: the example i gave above changes <img> tags to <video> tags where appropriate <arebi>RavenJoad: thanks for sharing your inputs, do you have an example or resource where this extension can be written? since i am a beginner trying to learn., i am looking for a sample so that I can start creating a module that others could also use if they find it a fit for the requirements that they have <mirai>RavenJoad: re: URW type font <mirai>there's an open issue in debbugs about it <RavenJoad>mirai: I figured it out. I added font-ghostscript as a native-input and it was fixed. <mirai>yeah, quick solution is that <arebi>civodul: thanks for the link #L325., let me access it now <mirai>though do note that font-ghostscript is not the recent if that matters <RavenJoad>arebi: I don't have any examples. I don't know how, but I am thinking of something like Racket's pollen. I want to be able to escape commonmark's markdown into something where I can put semi-arbitrary Guile and have that expand to SXML. <RavenJoad>mirai: The age of the font does not matter, no. This project has not received too many updates too recently. <RavenJoad>Yeah, that's pretty much it. I have some issues with the way Haunt ends up working too, but that is a discussion for the haunt channel, not here. <arebi>RavenJoad: thanks, the #haunt channel has two other members in it and total 3 including me. let me try to post this question there as well. appreciate your quick response! <RavenJoad>arebi: Sorry, there is no #haunt channel. dthompson's documentation says to use #guile. My bad. <arebi>RavenJoad: no worries. let me try and continue to see how the current functionality works., I was able to access this https://git.dthompson.us/haunt/tree/haunt/reader.scm#n82 where the key value pair written in the post content file is being returned as an alist (as per the comment)., so trying to see where valid key value pairs are being read and rendered to the respective html tags., let me post this in guile channel as you mentioned <muaddibb>how could I set up a full system installation with encryption and hibernation? <jA_cOp>kernel arguments can be added to your system configuration with the `kernel-arguments` field of `operating-system` <jA_cOp>full disk encryption with LUKS can be done with the `mapped-devices` field <jA_cOp>when I set it up I followed some examples I found online of other people's configuration, but I don't remember where <podiki>easier to read the markdown rendered there <lactose>How can I enable tapping and inverse scrolling for touchpads on Xorg? I've seen a snippet where they just add it to extra-config but I wanted to see if there's a better option <viaken>Where does X log to with gdm-service-type? <viaken>Ah, rootless X. ~/.local/share/xorg <untrusem`>anyway If I added a channel, How can I use its services? <untrusem`>for example small-guix channel have mullvad-vpn service <RavenJoad>Add the channel, guix pull to make all channels registered with the Guix on the command line, then just pull the service into your configuration the same way as any other Guix thing. <untrusem`>home and system services would be different as they are in main channels <RavenJoad>They actually behave the exact same way. The main guix channel is not truly special in any way. <untrusem`>So does anyone uses rde, I see there are many packages, what are the good ones you found use of? <untrusem`>so question, Emojis and many langauges font doesn't work <untrusem`>but still those languages doesn't show up, so I have to install those specific language font as well? <untrusem`>I don't any font that would work well for those languags <untrusem`>I installed google-noto-sans as suggesed in a reddit post as includes many languages, and ran `fc-cache -rv` but still, languages text does's show up <marmar22>untrusem`: for me I install 3 fonts, google-noto{,-sans-cjk,emoji} <marmar22>it might be those scripts that aren't rendering <divya>lilyp: Did you merge emacs-team recently? The builds are failing. <moksh>so used the guix rolling back thing <moksh>my sytem would just reboot after adding nix service and reconfiguring my system <moksh>I just logged into a older working generation <Kabouik>Hey #guix! I use Guix (system) as my daily machine and I work in academic research. So far, I have been allowed to use the distro I want and be admin, but I can feel this is changing and the IT department wants us to all use Windows for security reasons (for convenience actually because that would be a single OS for them to secure); and some Linux colleagues in another team have been forced to move to Windows already, with just "You're free to install <Kabouik> WSL if you want" as a consolation. We're starting a group among Linux users to defend a little better and show why forcing an OS onto scientists who use specific tools and takin away admin rights may be counterproductive. I'm looking into short arguments that I could use, both to defend allowing Linux, and ideally also allowing any distribution (because if only one distro is allowed, it will be Debian for sure). I already thought about personal <Kabouik>efficiency with habits and workflows that are sometimes optimized over years, freedom to use any DE which WSL doesn't allow, reproducibility with Guix and time-machine or manifests, etc. Any other key points I might be missing? <Kabouik>This may be best posted on Reddit but I don't really want my message to be indexed and found when the IT people search online what is Guix after hearing about it from me. <Kabouik>Yup I was including that in the reproducibility argument (but if I'm honest I don't use hpc yet so it's hard for me to show a track record of why this is key) <Kabouik>And also as far as I know most of these arguments remain true if using Guix package manager onto a foreign distro; which I guess wouldn't be too bad already, but would be a pain still. <Kabouik>I use it on a foreign distro but just on my Droidian phone. It's great really, and already allows me having almost the same environment on my phone using foreign distro and on my computer using Guix system, but this still adds the complexity of having to deal with two distinct package managers, and the difficult choice of what to install with what, which also hurts reproducibility. <Rutherther>so the argument is for the IT? are they going to care about such reasons? I personally like those reasons, I also like it's focused on freedom and that it's declarative - one person writes a service and you can just start using it by a line or a few lines in config. But if the IT guys are not going to care about this, I would be afraid they won't accept it as valid 'arguments' and that they would have to be convinced by something else <Kabouik>I'm fairly confident that they will step back on their "Windows for all" strategy because this is a local initiative and never was considered at the scale of the whole institute, it's a bit zealous from them. So if we can make this Linux group and they accept to join to discuss, they should allow us using Linux… But this will be Debian, and possibly not as admins. I guess it could be worse, but it still is a pain and I'm facing a honesty conflict <Kabouik>here because I want to use Science as an argument (and it is one), but the real reason it's annoying me is personal convenience and freedom, which I understand may clash with security from their perspective, of course. <Kabouik>I mean the last time we were attacked was from a printer they manage, not from a Linux user using whtie-listed packages from a package manager repository. <Kabouik>So I won't convince others to use Guix, but I guess I'm looking for ways to defend that they should not restrict the distributions they allow, because in a sense the security protections they may use should be distro-agnostic in the first place (e.g., firewalls), and distributions use package managers with curated packages, so more distributions don't really add risks. Would that be right too for Guix that may be lacking some packages available only <untrusem`>sorry for diverting topic, but I added nix-service-type as shown in the manual <untrusem`>but after reconfiguring the sytem in container <Rutherther>Kabouik: I would expect that from their perspective more package managers means more risks as there are more places to check for security fixes. It's unfortunate but that's how it is. But maybe I am just pessimistic from a corporate environment (where they for example don't allow any browser extensions except the ones "necessary for work" whatever that is supposed to mean) <Kabouik>Just curious, are you using a 0x0 script untrusem? I have been using one for years but it stopped working to upload text from my clipboard a few months ago. <Kabouik>Oh, I have this one too untrusem, thanks. But sometimes I need it out of emacs and that's where I was using the script. <Rutherther>untrusem: the relevant exception is higher up, the service load is somehow failing <Kabouik>Rutherther: yes, you might be right. I do understand the security aspects of not having too many different systems, let alone admins in the wild, but I also see all my colleagues using Windows being stuck for hours or even days sometimes because they can't install anything on their machine but need some R package, genomics-thingamajig, and so on, and still be more exposed to opening malicious email attachments in Thunderbird or Outlook compared to me <Kabouik>Thanks for pb untrusem, will give it a go. <untrusem`>so the log file i send also have path to script for running the container <untrusem`>for example can Rutherther run the container? <Rutherther>untrusem: re your error. It seems that either something has gone seriously wrong in the nix service type's extension of user accounts, or somehow the resulting .go file got corrupted. I am myself using the nix service and I don't see any changes in the last few days not even months to neither account service nor to the nix service. <Rutherther>I've already said the two possible causes I see. I don't see any other cause <Rutherther>and since I don't see any change to the services in few months I am going with corruption as most plausible at the current moment <untrusem`>I tried after removing extra configuration from nix-service-type <Rutherther>untrusem: your config (minus the nongnu and mullvad stuff which shouldn't affect this) is working for me in a container. So I am going with disk corruption as final guess <Rutherther>possibly, I can't say. What I can say is that I think that the .go files in your service log are corrupted <Rutherther>untrusem: how does your nix service type configuration look like? <Rutherther>untrusem: that is strange, again. Should work without it and works for me on newest guix <Rutherther>seeing eof there makes me think the file write operation has been aborted midway. But I am not sure completely. <Kabouik>Actually I see that multiple users/maintainers/founders of Guix work at INRIA. If any is here, I'm curious if you are researchers or part of the IT team, and if you are researchers, do you all have root permissions on your machines? If one French research institute can allow that (and also work on Guix as part of their research activities), I could use that as an argument in another French research institute. <civodul>Kabouik: yes, we’re root on our machines by default at Inria :-) <Kabouik>Nice. I am too, but I feel like I'm prey and my time is counted. <civodul>‘we’ being IT, engineers, and researchers <civodul>there’s occasionally a temptation to change that, but i don’t think it would fly in a computer science research institute <csantosb>Hi Guix ! Do we have a means to tell `guix import pypi -r ` to ignore non existing or wheel only packages ? <Kabouik>Thanks civodul. Coincidentally, I had lunch today with some people from the IT department of our institute at the regional scale, and took the opportunity to ask about that strategy to Windowize research units. They confirmed there is indeed sometimes incentive to encourage that, but nothing very firm, and anyway it shouldn't be forced in a research institute. So that confirms what might happen here is really local zealous initiative, and there's hope <Kabouik> it can be reasoned. I hope so. In the meantime, I'm feeding myself with talks from the Guix and Guix-HPC conferences, and will not miss the opportunity to state that Guix is something that partly comes from INRIA and is advocated for reproducible open science. I'm just not sure how to phrase it accurately because I don't know the exact history there except that some founders are INRIA people. <civodul>Kabouik: that Guix ‘partly comes from Inria’ is not wrong, but for a small part only: it’s a community project with 1,000+ people who contributed since its inception <civodul>(and they weren’t here when it began) <jcabieces>Hi everyone! How can I pack a previously build and installed (from file with -f option) package. It doesn't "find" it <acrow>jcabieces: Don't forget to include the directory of your file in the package path, usually with -L <dir> <jcabieces>acrow I used -f with an absolute file path, and it builds and install correctly. What does it change to add -L <dir> <acrow>jcabieces: probably just saving a few characters of typing. The full pathname is also a fine way to do it. <ieure>-L and -f are not equivalent <Rutherther>jcabieces: there isnt really a way to find it proper from installed profile. You could either pack the whole profile from store with path to it, or just use the file with instruction again with the pack. Packing just the store path with output of the package will mean loss of search paths, so is not really a supported use case. <Rutherther>As long as you are using the same guix instance, the same store path is used and no rebuild is needed. <xelxebar>debbugs should probably decode the RFC 2047 =?UTF-8?Q?...?= spam for us. <jcabieces>Rutherther my goal is to make a relocatable tarball to distribute among user who don't have guix. <jcabieces>Rutherther "just use the file with instruction again with the pack" -> I tried this way guix pack -c 14 --load-path=~/myconf/guix-examples/ --file=qgis-qt6.scm <Rutherther>Why are you putting a file to load path? Load path is for a folder with modules <jcabieces>Rutherther Unable to find file "guix/~/myconf/guix-examples/qgis-qt6.scm" in load path <untrusem`>the guix jupyter package have jupyter-notebook version 6, I want version7 but installing with pip would be a hassle <Rutherther>From that it seems that you input it in a way where your shell doesnt substitute the home path, so try substituting it manually <Rutherther>Yes it is, but shells arent going to substitute it if enfered as --option=~/... <Rutherther>Apart from that if your goal is to make it relocatable, you're going to have to use the relocatable flag, possibly twice if your goal is to make it work on systems without user linux namespaces <jcabieces>Rutherther I read that in the meantime and added the -RR option <Rutherther>guix-noob: Guix environment is deprecated command, better to use guix shell. Also you can just do guix shell without arguments in the checkout to get all you need thanks to the manifest.scm file <Rutherther>But I really have no idea why you encountered that error <ieure>You'd have to look at config.log, like the output tells you to. <lechner>xelxebar / debbugs will decode RFC2047 for you <Rutherther>Anyone knows why is guix updating to python 3.11 rather than 12 or 13? <rbrins>I'm running into a problem where my emacs packages installed by guix are not finding their way into the load-path for emacs. I see the package listed in .guix-profile/share/emacs/site-lisp/ but not sure if this is the right place to check installation <Rutherther>rbrins: do you have emacs installed in the same profile, and did you relog after installing the first package? <rbrins>I see more packages in that dir then I installed though when I do guix package --list-installed, which makes me think I overwrote the site-lisp directory with a backup <rbrins>Rutherther: unsure on the profile I haven't done anything explicitly with profiles, and yes to restarting (multiple times) <Rutherther>rbrins: when you guix install, you are installing to a profile. When you use guix home/system packages you are installing to a different profile. So did you use the same mechanism to get emacs and the packages, ie. Guix install both? <lechner>rbrins / i see fewer packages than I have installed via Guix Home <rbrins>Rutherther: tbh I thought emacs was preinstalled with guix (I'm using it as a system) but I've only used guix install package-name <lechner>My files are here but maybe that's because I'm using Guix Home ~/.guix-home/profile/share/emacs/site-lisp/ <Rutherther>rbrins: so you haven't installed emacs with guix at all? Then you don't have emacs it in the same profile and that is why you don't see your packages. The search path for EMACSLOADPATH is missing. It is contained with the guix emacs package <Rutherther>The guix emacs packages are partially incompatible with regular emacs, especially if they are native compiled, so it's better to use guix emacs with its patches <rbrins>Rutherther: I'll install now and check <Rutherther>youi will need to relog / source the profile file are you install it <rbrins>I'm going to lose this chat, but do you have a link for source the profile file? <Rutherther>and it's emacs that needs that env var, so you need to start emacs in the process that has the env vars from there <Rutherther>(unless if you were using something like exec-path-from-shell) <rbrins>Rutherther: thanks, I'll hopefully be back! <untrusem`>anyone else having problem with nix-service-type <ieure>untrusem`, You're going to need to paste more info, like the log of the failed build (using a pastebin). <arebi> section of the post markdown file but the generated html did not include them. is it something related to the serialized-artifact module? <rbrins>Rutherther: that did the trick and didn't even need to source anything after logging back in <Rutherther>yup, that was only for case you didn't relog. The source should be happening by your shell's profile file, that's how you setup guix initially <lechner>Rutherther / thank you for doing all this! you provide the best support here that I've seen since nckx stepped away <rbrins>Rutherther: agreed with lechner, thank you so much! <rbrins>to try to dive a bit deeper with guix and profiles does `guix install` go to a default user profile while the emacs I was originally using would have be akin to `guix --system install`? and are profiles used `guix shell package_names...` just in a temporary context? <Rutherther>untrusem: have you tried guix gc --verify=contents to rule out file corruption? <Rutherther>rbrins: there isn't really a guix --system install. Guix system uses a configuration file that you build and then just 'activate'. Yes, guix install goes to the default user profile - ~/.guix-profile pointing to /var/guix/profiles/per-user/$USER (where older generations are kept as well) <Rutherther>rbrins: guix shell uses profiles sort of temporarily, yes, but they will remain in your store unless you gc (garbage collect) them so they will be reused upon reruns of the same guix shell invocation <guix-noob>hi ...i build guix successfully...at least without a fail during make. tried executing 'guix system list-generations' but errors out. here the basic workflow: https://paste.debian.net/plain/1364223 .Doi have to specify something specific if i want to use guix from my built on my actual system? <guix-noob>Damn it...'guix shell -D guix -CPW' puts me in container...can i mount var inside? <Rutherther>yes, but why are you going to a container in the first place? <Rutherther>(definitely do not run reconfigure of your base system from a container though) <cancername>Rutherther: that couldn't break the base system though, right? <Rutherther>guix-noob: right, it shows you one way to do it, that doesn't mean it's the only one, just a guix shell without container is fine <guix-noob>not reconfigure...only wanted to see if i can list generations....to see if guix-command itself does something and in theory be usable <Rutherther>cancername: not really break, no, but it could cause inconsistencies if it finishes (it will fail probably, on bootloader change) <guix-noob>Rutherther: thnks. Tried 'guix shell -D guix --pure'. Seems to have worked. List the generations if used with './pre-inst-env'. Thx again... <untrusem`>Rutherther: running `gc -verify="contents,repair"` <guix-noob>Rutherther: since i am not in container...now i could use reconfigure on system? <lilyp>divya: I did not even submit it to merge yet <lilyp>if you have a bug to report please at least mention the package :) <Rutherther>guix-noob: definitely. But note that it will probably tell you it's unknown command. If that happens, first enter shell, ie. "./pre-inst-env bash" and only then do (sudo) guix system reconfigure. To verify the guix is actually the one from the repository, you can try "guix describe" <Rutherther>guix-noob: also since you did --pure, you won't have /run/privileged/bin in path where sudo is <Rutherther>yeah, so my guess with disk corruption seems to be right then <Rutherther>I suppose it tells you it won't be able to repair them <Rutherther>and in that case you will probably need to remove them (guix gc, either full or just guix gc -D for individual files, where because other paths reference them, you will need to get the referrers to them and remove them first) <guix-noob>Rutherther: to have bash available i need to add bsh to 'guix shell -D guix --pure', correct? <Rutherther>guix-noob: yes, or just don't use pure, it's not _that_ essential for this use case <guix-noob>Rutherther: if i don't use "--pure" i get error "no code for module (guix ui)" when using ./pre-ins-env guix system list-generations" <untrusem`>will the `guix gc --verify="contents,repair" keep on going <Rutherther>guix-noob: hmmm alright. Better to use pure then for sure. Did you run make btw? <Rutherther>untrusem: it will definitely exit after it's finished going through all the store paths, which can take a while <guix-noob>Rutherther: initially yes but not now when i left evn and re-entered without "--pure" <Rutherther>untrusem: it's probably better to leave it finish by itself to see all the corrupted paths if there are more. Also I would recommend you to run a disk check to make sure your disk is okay as long as power didn't go off recently / and you haven't forcefully powered down your computer <Rutherther>untrusem: btw are you sure the system build with nix service finished when you were powering down / rebooting your computer (even normally, not jut forcefully) <Rutherther>because some users reported recently that their fs did not get unmounted cleanly on poweroff / reboot <untrusem`>just reconfigured directly, and when I rebooted, it would just re-reboot <untrusem`>I am using btrfs, I should run a disk check on that <Rutherther>untrusem: and when, you didn't get fsck check of your disk by any chance? <untrusem`>it checks for disks, it something changes, It throws a warning <Rutherther>then you are definitely affected by the bug someone reported recently. And it could be the reason for the corruption here <untrusem`>I didn't manually enable it, all are defaults <untrusem`>should I add the service and build it in a container to see if it works <Rutherther>untrusem: it won't since the .go files aren't repaired, at least I think so, you haven't sent the output, but the only thing gc verify repair can repair are substituable paths and since the substitute server is not building those services, their output files aren't substitutable <ieure>guix-noob, Don't do that? The instructions I linked you specifically say to use --pure. <ieure>And you've just discovered the reason why. <Rutherther>untrusem: okay, good, then the container could work <untrusem`>what are extra configuration you recommend for that? <guix-noob>ieure: and i did...well...first i used "-CPW' ...then changed to "--pure" after Rutherther pointed it out... commands seem to work fine but for the user that build ...can sudeo <ieure>Perhaps you can use --pure and `/sbin/sudo ./pre-inst-env guix system ...'? <ieure>(Or whatever the path to your foreign-distro sudo is) <untrusem`>ok the container builds and it throws stuff in the container itself like about starting services, I haved logged into it, some error as well `shepherd[1]: could not create '/dev/autofs': Operation not permitted` <untrusem`>but when I log in into container and run `herd status`, it just keeps on running <untrusem`>when I exit the container it throws =guix container: error: process exited with status 130= <Rutherther>untrusem: that herd status hanging will happen if shepherd is busy with starting services. It can happen in a container that some of the service startup functions will somehow hang. Doesn't mean your system won't run properly on real hardware <Rutherther>yeah, shepherd is waiting for udev to start and it doesn't, it's unfortunate <Rutherther>you can run vm. Not really sure what the proper way to get udev service working in system container all the time <untrusem`>the last time I ran vm, I ran out of memory or idk, it hanged the system <Rutherther>the default amount of ram given to the vm is set to 512 MB <guix-noob>ieure: i'm on guixsd ... sudo is liked(???) in /run/ . Changed to root before entering env to avoid sudo...that worked...deleted a unused version... <Rutherther>not linked, copied from store to /run/privileged/bin so that the executable can get setuid <guix-noob>oki...i can execute while in [env] as user: "/run/privileged/bin/sudo ./pre-inst-env guix system list-generations" but get guix: system: command not found then... <Rutherther>yep, that's why I said to first enter bash with "./pre-inst-env bash" <untrusem>What is the default user and password for the VM? <guix-noob>Rutherther: oki...just did.To be able to use bash i had to add bash to shell command "guix shell -D guix bash --pure" . "pre-inst-env bash" works but "guix system list-generations" thwos again error "no code for module (guix ui)" ... <Rutherther>guix-noob: oh. I don't know how to use guix system properly from the pre inst env then. You can always pull the channels from the local checkout. Maybe someone else will <guix-noob>Rutherther: I will go with changin to root-user for now . Thank you so much <untrusem`>when I reconfigure the system it reconfigures but throws this at last line <untrusem`>guix system: error: error parsing derivation `/gnu/store/433fy0pwq7qjc4frddapmnsp41dxsihd-upgrade-shepherd-services.scm.drv': expected string `Derive([' <Rutherther>I don't know what you mean by that. The operation couldn't finish completely, that doesn't mean that part of it hasn't finished properly... <untrusem>What I mean is that the system reconfigures correctly even though it throws this warning. I am logged into the new generation. <untrusem>I will run the guix gc verify command later <Rutherther>untrusem: reconfigure is a multi step process which can succeed in some of the steps and fail in the last one (ie. the reconfigure of the running system). That's exactly what happened here. The upgrade of services failed. <platoxia>In the filesystems section for the system configuration, how do I refer to the previous entry (i.e. "/") as a dependency for the next entry? <divya>lilyp: Why exactly the build (specifically patches) for emacs failing then? <attila_lendvai>re syncthing: i move the *.pem files, but what about ~/.config/syncthing/index-v0.14.0.db/ ? that sure sounds like something for ~/.local/state/ <lilyp>divya: i don't even know what branch you're on <lilyp>successfully built /gnu/store/4c52xwbw9q4qdnwkx4h0fnlap53cijfs-emacs-29.4.drv <lilyp>/gnu/store/8l74c0za9s96brf2y342yqlil63ysmc8-emacs-29.4 <divya>Okay, I forgot. It's actually building from my channel not guix upstream <lilyp>looks like a bug in your channel then ;) <attila_lendvai>could someone with the commit bit please bump the version of gpaste to the lates (v45.2)? it fails to build currently, but this works: guix build gpaste --with-commit=gpaste=v45.2 <muaddibb>Rutherther: I'm getting, during boot, the message 'waiting for device <uuid-of-resume-kernel-param>' repeated 10-ish times for about 10 seconds, but booting normally otherwise <muaddibb>I tried changing the uuid of the kernel parameter to the uuid of the mapped deviced, but it froze my system after boot with the capslock light blinking <lechner>Rutherther / what makes you think Guix doesn't unmount cleanly? <ieure>muaddibb, I have the exact same bit in my config on a machine with FDE and it works fine. Did you create the swapfile and run `mkswap' on it? <muaddibb>ieure: I remembering running `mkswap', but I could be wrong. Could I just run it again on the file? Are you also passing the uuid of the mapped device to the kernel? <muaddibb>s/remembering/remember idk what got to me today <ieure>muaddibb, Yes, you can re-`mkswap' it. You can also `sudo swapon /swapfile' and see if that works, which will tell you if your issue is with the file itself or with Guix. And as I said, I have the exact same configuration for swap as you. <ieure>For the device containing the filesystem, yes, I'm using (source (uuid "...")) <ieure>Okay, not the exact same, my file is named "swap" instead of "swapfile." <muaddibb>weird, I got `swapon: /swapfile: software suspend data detected. Rewriting the swap signature.' I'm running `mkswap' again, just in case <ieure>ACTION re-ignores debbugs, but Forever this time <ieure>muaddibb, Yeah, hibernate saves your RAM contents to swap and restores it on boot. And since your machine didn't recover, the data's sitting there in your swap file. <ieure>muaddibb, Guessing there's an ordering issue and the swap isn't readable at the time the system is resuming from hibernation. <lechner>Rutherther / okay, i marked that bug as grage <attila_lendvai>ACTION once again tried to set up hibernate for his laptop a few months ago, but he failed <attila_lendvai>maybe debbugs messages should start with a visibly different pattern, like 3 spaces or something like that <muaddibb>ieure: just to see if I got it right, you also encrypt the partition of your swapfile? what do you use? ext4 and luks2? <ieure>muaddibb, I use full disk encryption, LUKS1 because it's an older install, and ext4. <ieure>I know the main LUKS2 blocker is fixed, but I don't know if the installer supports that yet. <muaddibb>would you mind sharing your config? or a portion of it <muaddibb>so, the uuid you pass to the kernel is `b437fd73-827b-49f8-9465-478116fa94ca'? <ieure>muaddibb, I replaced the real UUIDs with random ones. <ieure>I'm not sure what you mean by "pass to the kernel." That's the config. I don't do anything outside of it. <ieure>And there's no reference to those UUIDs other than the ones I pasted. <muaddibb>ieure: you didn't pass "resume=uuid=<uuid>" as an argument to the kernel? I'll try removing it from my config <ieure>muaddibb, I didn't, this is from a desktop machine which runs 24/7. <ieure>muaddibb, I can't help with hibernate issues. <ieure>Sure thing, good luck with the hibernate stuff. Be interested to hear if you get that working. <muaddibb>removing the kernel arguments from the config didn't work. I'll try again tomorrow <nikita_>does anyone know where i can grab an amd64 hurd/guix image? i understand its limitations, but im curious to try it out. or is the only supported way to get a linux binary and swap the kernel out like in the guide blog post? <muaddibb>got hibernation to work! I was passing `resume=UUID=<uuid>' to the kernel instead of the correct way `resume=<uuid>' <attila_lendvai>what can i use instead of file-append in a guix home config? it's not using gexps and file-append only works inside gexps... i'd like to refer to a /gnu/store/.../bin/foo directly. <attila_lendvai>maybe it's not even possible...? how would such a reference be registered as a gc root? 