IRC channel logs
2024-10-08.log
back to list of logs
<__cornflake__>The file also includes the output of guix package --install-from-file and the error message btw! <LesikEdelweiss>Hi! Does anyone know, how to configure PAM modules in Guix? I want to try oath-toolkit, which has pam_oath module, but I can't find a way to actually use it. <AwesomeAdam54321>__cornflake__: The xcursorgen package is undefined, so you need to use the right module that defines it <bigbookofbug>quick question: is pulling from different branches in the guix repository possible? i'd like to use the current rust version as defined in the rust-team branch for a project im working on, but initial tests of pulling from `rust-team` and `master` caused some issues in module definitions <bigbookofbug>or if theres a way to do it correctly in guix-shell, that would be the most desirable option ! <apteryx>on right click on links in librewolf, I see two checkboxes at the top, without any explanatory label (text). anyone else? <bavier>bigbookofbug, you can use `guix pull --branch=rust-team` <bigbookofbug>bavier: are you able to pull within the context of a shell ? i dont have any programming setups installed locally really <bavier>bigbookofbug: I guess you could give yourself a different guix in a new shell: `guix shell guix --with-branch=guix=rust-team` <bigbookofbug>apteryx: im currently using an inferior for librewolf from commit 33b4ac12340044b76b8ee8fe223fe50e86aff6df, if that helps any <bigbookofbug>np ! its a bit of a quick-fix but lets me use the browser still which is what i appreciate :) <apteryx>oof: librewolf-130.0.1-1.source.tar.gz 906.5MiB <apteryx>cbaines: bordeaux download speed seems to cap at 2.7 MiB/s here <apteryx>bigbookofbug, podiki I've pushed the fixup commits of ieure now <abbe>are there any guix modules for having a full-disk encryption based host with remote unlocking (preferably over ssh) ? <gabber>i noticed PYTHONPATH not being exported into guix shells, e.g. `guix shell python-django -- django-admin` throws an error. am i doing it wrong or is this an issue yet to be created? <Rutherther>gabber the shell itself should get the path only if you also include python in it. But binaries should be wrapped and not need that, so that looks like a bug in django package <gabber>Rutherther: thanks! i guess i'll craft a patch, then (: <mehrad>Is there a way in `guix shell` to create an environmental variable? I use `--pure` and `--preserve` but I also need to create some variable <mehrad>or should I use `setenv` function in the manifest? <mehrad>after a quick test, adding it to manifest didn't work. So any insight is highly appreciated <futurile>mehrad: Rutherther came up with this the other day: guix shell -- bash -c 'env TEST=$GUIX_ENVIRONMENT' <futurile>mehrad: if it's for your own development environment the other option I saw mentioned was to use direnv - but ofc that adds more 'stuff' to your stack <mehrad>Thanks. That of course is great, but it kinda force me to land in bash, where as ideally i would like to land in R :D of course that can be bypassed by `-c R` to the bash, but looks a bit weird <futurile>mehrad: guess it's going to be setenv then. It seems like a missing 'thing' that people want/need in 'guix shell' <mehrad>Thanks though, this for now get me out of deadlock. But would be cool if Guix can also add environmental variable in combination with --pure and --preserve <nckhexen>abbe: The Internet is big & all, but almost certainly not. <cbaines>(the mirrors seemed to not be handling that file, but I think I've fixed that now) <nckhexen>abbe: I think such a thing would imply upstream changes to the initrd code that we'd have seen. <nckhexen>mehrad: --preserve should suffice here. Why didn't it? <abbe>I was hoping it would just be a matter of creating a custom initrd/bootloader functions that one could import from external repository without making changes to the guix upstream <nckhexen>abbe: oh Absolutely, and that's the kind of patches I'd expect if someone were doing this: patches to make the initrd more 'modular'. Guix System's initrd isn't very flexible yet. <nckhexen>mehrad: I mean something like A=foo B=bar guix shell --preserve={A,B} -- R. Does that not work? If not, why not? There's no reason to use a shell. I don't think Guix needs yet another slightly-different option—I've heard people (OK, two, but that's people! :) say that they find the current number of CLI options intimidating on its own. <mehrad>nckhexen: that works, but that pollutes/overwrites the parent shell <dariqq>sneek later tell civodul: I think i know what the cause of my privileged-programs troubles is. The first time we try to get the gid of a new group is failing because group does not exist yet. And after reboot groups are not recreated yet so it is using the info from the previous /etc/group . (This "works" for groups and therefore opensmtpd because "group not found" is a 'system-error being caught by the exception-handler and "user not fou <dariqq>nd" is a 'misc-error instead which does not get handled causing the user not to be created). <PotentialUser-0>hey everyone,does i wanted to try an application and i have an appimage for that. I made it executeble but when i execute it by . ./mybinary, i get the answer from the prompt: cannot exectu binary file <PotentialUser-0>Does anyone has an idea, whether AppImage is working at all within guix <gabber>PotentialUser-0: i'm sure you can make it work. i think the easiest way to provide means to use "legacy" software/builds/executables is to invoke a shell/environment that emulates FHS (i've never done it, though) <Rutherther>PotentialUser-0 is that the whole error message? <gabber>PotentialUser-0: i think this stems from the executable not being able to find the c runtime library <PotentialUser-0>dave@host ~/Downloads$ guix shell --container --network --emulate-fhs \ <PotentialUser-0> --preserve='^DISPLAY$' --preserve='^XAUTHORITY$' --expose=$XAUTHORITY \ <Rutherther>PotentialUser-0 you did not send the error. Please use a paste site, IRC is a line based protocol. Your messages got trimmed because of how Fast they were being sent <PotentialUser-0>Rutherther The error is this: ./cursor-0.41.3x86_64.AppImage: error while loading shared libraries: libz.so.1: cannot open shared object file: No such file or directory <Rutherther>PotentialUser-0 see. Then add the zlib to your guix shell packages. That should suffice <dariqq>PotentialUser-0: sry i know nothing about appimages, just that there was a blog post. <PotentialUser-0>Rutherther okay thanks, but now i get: Failed to run /tmp/appimage_extracted_45f8c6f6e324ebd2f98efdceeb36ebb1/AppRun: No such file or directory <futurile>PotentialUser-0: that sounds like an appimage thing, it's trying to use /tmp/<blah>, so I guess you'll need to also --share=/tmp <futurile>PotentialUser-0: just going from the error - I have no appimage experience <nutcase>Hi. If I want to add a file into my system profile's etc (/run/current-system/profile/etc/), I can't use etc-service-type for that, right? It just allows me to add something to /etc. Do I need to create a package to have files in /run/current-system/profile/etc/somewhere/ or is there another service-type I can use? <Rutherther>nutcase yes, use a package and put that to your system profile. The service is called profile-service-type. So you can either extend it with other service or just put the package to your packages list <PotentialUser-0>futurile does not work unfortunately. Anybody another idea maybe? Failed to run /tmp/appimage_extracted_45f8c6f6e324ebd2f98efdceeb36ebb1/AppRun: No such file or directory <nutcase>Rutherther: thanks. And if I only want to place a file from another existing package into a specific directory of my system profile, is it possible to create a packate with (input) but without (source)? <PotentialUser-0>I feel a bit frustrated, how can I get to th root cause of this problem? <Rutherther>PotentialUser-0 to really get to it you would probably have to study how appimages work. I think the behavior of appimage is to extract the files to tmp. Not sure why it is not working <Rutherther>Alternatively you can try extracting the appimage yourself and just run the files it will produce <PotentialUser-0>Rutherther I think, you are right. I should first check out, how guix shell works and then work with appimage in order to get to the ground of that prolem, thank you. You guys helped me a lot so far, thank you for that. <futurile>cbaines: got to say, your patchworks instance is so useful sometimes <PotentialUser-0>I must say, I feel a bit frustrated though, but this is the time, where I need to put my frustration into something to learn, I guess. <futurile>PotentialUser-0: it's generally going to be more of a struggle to run a "compiled" application on Guix - because the libraries are in different places. If this is a really important app for you, I'd suggest you can try running it in a VM with a "standard" Linux distribution. And/Or see if there's a flatpak - as I know people run flatpak on Guix <futurile>PotentialUser-0: I'm saying run a VM with Debian/Arch/Ubuntu/Redhat - whatever you're comfortable with - I guess the app you're trying to run might have some supported environments, so whatever they recommend. <Rutherther>There is also distrobox that works on top of docker/podman and emulates other distros. Usually should be enough <futurile>yeah - do we have it packaged? Qemu and virt-manager definitely are <PotentialUser-0>thanks a lot, what do you recommend for ease of use, distrobox, virt-manager or qemue? <futurile>I would say if you have no experience with any of them virt-manager is pretty popular and you'll find lots of articles about it. <futurile>Distrobox is really nice but I don't think it's packaged for Guix <Rutherther>PotentialUser-0 virt-manager is a gui around qemu. So if you choose it you also have to have qemu. <futurile>Rutherther: brain fart - it literally shows up there <nigko>Hello Guix! I am building a package on a remote machine which is running guix-publish service. The package has 'debug' and 'doc' outputs in addition to standard 'out' output. When I install the package on my local machine with 'guix [package-name]' command I expect only 'out' output will be downloaded from the remote machine. However for some <nigko>reason 'debug' and 'doc' outputs are also dowloaded. Is it normal? <nigko>* 'guix install [package-name]' <Rutherther>nigko I would think so. While it is true that when thinking only about substitutes, it could make sense to download just one, when you are actually building the package instead you have to build all the outputs. Substitutes just download result of the build, being all outputs <nigko>Rutherther: Thanks for the answer! But when I install some package with multiple outputs, say 'guix install cmake', only cmake:out is downloaded, but not cmake:doc. Why? <nigko>*some package from official guix repo <nutcase>I get a "copy-file: ~A No such file or directory" in phase install <efraim>switch it from copy-file to install file and remove the `1eac` from the destination or create the destination directory first <efraim>although you'll need to change it so that you're installing to the output and not straight to /etc <efraim>install-file will install a file into the directory it's pointed at, so you'll want (string-append out "/etc/ModemManager/fcc-unlock.d/") <nutcase>yes, you're right. fixed. Now I try to get that into my operating-system's definition (via profile-service-type)? <efraim>There's something about an etc-service-type to install files into /etc, although I don't remember much about it off the top of my head <nutcase>In my case etc-service-type is not sufficient, I think. I tried that before, but it seems that ModemManager looks into /run/current-system/profile/etc. At least that is, what I want to try <nutcase>an etc-service-type would have been more easy <nutcase>I had an etc-service-type successfully defined and the file in /etc <nigko>Is it possible to build separate substitutes per each output of the package with guix-publish? So when installing an output only substitute for this output is downloaded from the remote machine? <futurile>nigko: `guix build blah:doc` or something? what's the package that you're struggling with, is it in the archive? <nigko>futurile: I am building a package on a remote machine which is running guix-publish service. The package has 'debug' and 'doc' outputs in addition to standard 'out' output. When I install the package on my local machine with 'guix install [package-name]' command I expect only 'out' output will be downloaded from the remote machine. However for some <nigko>reason 'debug' and 'doc' outputs are also dowloaded. <futurile>nigko: I think I'm struggling a bit to understand. You're doing "guix build <package>", that will always get the source and build all the outputs (out/debug/doc). That's expected right? <futurile>nigko: what's the package that you're trying to install on your machine? Maybe the "out" isn't specified correctly? <nigko>futurile: Yes, on remote mashine I ran "guix build <package>", and obtain all outputs (out/debug/doc) as expected. Then I run 'guix install <package>' on my local machine which uses remote machine as on of substitute server. <futurile>nigko: so if you do "guix install package:doc" on the remote machine it *only* installs the doc portion. BUT, if you do it on your local machine it's downloading *all* outputs (out/doc/debug/whatever). <nutcase>does anyone have an example of a profile-service-type within the (services) of an (operating-system)? <potatoespotatoes>hey all! long time nixos user here, trying to migrate over to guix. I'm running guix pull, but I'm hitting this known timebomb on openssl 1.1.1l during the test phase. How would I go about overriding this package to disable checks? <potatoespotatoes>I came across the "package variants" page, so I can probably hobble something together, but I don't know where to put a variant that will effect `guix pull`. For context, this is my first guix pull <weary-traveler>can mumi be used to apply patches corresponding to a specific reroll count? <weary-traveler>when i pass in a spec such as v<reroll-count> it seems to be ignored <futurile>potatoespotatoes: how did you install Guix? did you tell it to use substitutes? <futurile>nigko: I don't know why it's doing that - package definition makes sense. I guess you can't build it on the "local" machine to see if it is definitely something to do with "guix publish". It's very odd <futurile>nigko: maybe someone else will know later - maybe ask later on when the USA is around - if not I guess ask on the user mailing list <nigko>futurile: I have tried to do the same thing (i.e. build on remote machine and install on local) for 'elftools' and 'libxml2' packages from the main guix repo (to avoid conflicts I have to rename these packages). They both behave as they should, i.e. only single output is downloaded when package is installed. Hence something should be wrong with my <nckhexen>mehrad: How does it pollute the parent shell? <nckhexen>On second thought. I'm not actually sure what 'pollute the parent shell' means. <nckhexen>But since it can't affect the parent (as in, your, interactive) shell to begin with, 'pollute' sounds like an implied subset of that :o) <__cornflake___>Hey all, I'm still working on packaging catppuccin cursors. I fixed the issue of xcursorgen not working, but now when I try to install from file it complains that there is "no code for module (guix licenses)" even though I added it with use-modules. Link to full command outputs and package file: https://paste.debian.net/1331682/ <nckhexen>mehrad: Did you even try my example? :-/ <Rutherther>as for your error. Why are you even trying to put guix licenses to the build env? the licenses are usually used only in the package definition itself, no need to pass them to the daemon <Rutherther>it's possible that the latest guix revision is not yet built, so that's why it builds at your side and doesn't get substituted <potatoespotatoes>there is a workaround, but I have to disable things which nixos doesn't let me do (easily). I was hoping I could just write a variant of openssl and slap it in a config to override all of the dependent derivations <Rutherther>yeah, it is a thing, there are multiple ways to do it, if just once, then just use --commit parameter of guix pull <Rutherther>as for overriding guix revision with guix pull. That can be quite tricky, unfortunately guix pull doesn't seem to provide mechanisms for that, so you would probably have to use something else. I think the easiest could be to use the "guix-for-channels" function, where you would provide the guix channel and you could override that package with openssl as you suggested. <Rutherther>That would build the guix executable. In guix, as opposed to nix, the channels are actually part of the guix executable. Then you could change the profile to this one (in your /var/guix/profiles/per-user the channels reside, those are called current-guix, and the current one is symlinked to ~/.config/guix/current). To do that you should be able to just use guix install with --profile flag <Rutherther>but really, it will be easier to just use older commit <Rutherther>you shouldn't need to go to commit before openssl, you can just go to a commit that has substitutes <potatoespotatoes>ah... how do I find the latest release with a cache? I pulled a commit I thought would work but it just pulled in more commits >.< <potatoespotatoes>it still builds openssl : ( I'm just trying to roll back to something in August <potatoespotatoes>haha, okay I'm in august 2023 and it's still building this derivation. so it must be in /some/ cache. <Rutherther>(just to make sure, send with a paste site, not as messages) <Rutherther>I meant mainly the start of guix pull log, that doesn't seem like that <potatoespotatoes>I think I'll just find another box / vm to put guix on. this seems like it isn't the average experience -- something must be weird about this install <potatoespotatoes>ah shoot -- need to run to a meeting. thanks Rutherther for helping me debug this! <Rutherther>potatoespotatoes: hmm, so it seems like it builds the version for 1.4, the one you are currently at. Not the one you would be downloading. I guess it's maybe needed even for the download somehow? I am not sure honestly, I've not yet understood fully what is needed to build guix with guix pull. The newest openssl used is 1.1.1u. So I suppose no matter what revision you use you will get this error. Probably will be much easier to just change the date... <Rutherther>it's really an unfortunate situation to have a time bomb in the latest guix release :( Interesting that I didn't get hit with this when I installed guix on nixos <luca>Anyone got any tips to using neovim with treesitter? It seems as though to use treesitter it wants to compile some files, but on guix you need a couple dependencies installed to do so. I could probably install them, but is that the best idea? <efraim>there's a tree-sitter package and what I assume are a bunch of addons <Rutherther>luca: you could also install the tree sitter libraries with guix - there are tree-sitter-{lang} packages. These solve the problem of needing the deps. But I am not really sure where neovim expects the libraries to be, you might have to tweak a variable to say where to look for them. If you guix install them, they should be under ~/.guix-profile/lib/tree-sitter <luca>Ah it's called tree-sitter, not treesitter. I was greping for the wrong thing <Rutherther>tree-sitter itself is already an input to neovim, so I would expect only the individual languages to be a problem <__cornflake___>Rutherer: That's awesome! Glad you packaged it! I was trying to add the licenses to the build env because I though from the backtrace it was trying to build inkscape and that's what got messed up, so I wanted to make sure that the licenses were in inkscape's build env <Rutherther>__cornflake___: I see. Guix builds packages in separate sandbox environments. Your package cannot affect inkscape's package build. Are you on latest guix? For me inkscape is just pulled from substitutes. So if there was a problem recently it is fixed already on latest guix <nigko>futurile: It seems I've found something interesting! The source of problem seems to be 'cmake-build-system'. Could you try to install any package from the main guix repo with multiple outputs and cmake-build-system? There are some. I have checked 'x265' and 'libgit2'. When I run 'guix install <package>' it downloads substitutes for all package <nigko>outputs, not just 'out' output! <yelninei>nigko: Could this be related to grafts instead? <nigko>yelninei: It seems to be related. When I run 'guix install --no-grafts <package>' for x265 and libgit2 packages it downloads only substitutes for 'out' outputs. But why grafts are related? I don't understand yet. <PotentialUser-0>Rutherther I just installed it and tried quickstart and this error came or should I not expect it to work out of box? <PotentialUser-0>hmm, I dont seem to get anything to work. Can anyone help me please. Is that normal? Am i too stupid to have too less knowledge or no patience? would be really nice, if someone could lead me into the right direction of what I am doing wrong. <ekaitz>efraim: did you try to build guile in riscv? <ekaitz>efraim: i'm trying to finally add the JIT compilation library to it for RISCV <luca>Hi, I've added a package definition in rust-apps.scm in a local checkout of the guix repo (rust-team branch) and would like to fetch/fix checksum/build/install it. How do I do that exactly? <Rutherther>luca: when working with the guix checkout, it's usually best to use the pre-inst-env script. It is described under Contributing > Running kguix Before It Is Installed in the manual. After you have the setup, you would just use "./pre-inst-env guix build name-of-the-package" <Rutherther>nah, it's what happens when one has home row mods and shift is on k xD <zjabbar>Hey Guix! Does anyone here use GNU/Jami on Guix? It's been really buggy for me, but I see no errors on the bug tracker. It works fine for me on another non Guix system. <luca>Do I _have_ to `make install` in guix.git? It's taking for ever D: <ieure>luca, No, see the manual section "Running Guix before it is installed" <luca>Oh I misread "used even if you have __not__ run" <luca>also ahhhh I ran guix style -f on a file and it messed up all of it's formatting. I thought everyone was supposed to run that command on guix files D: <yelninei>luca: I think the error is cargo cant find the uom crate <luca>Ah right, I forgot to add it. Thanks! <luca>I see that uom-0.34 is packaged, but my crate wants 0.30. How important is that version difference? Or "it depends"? <luca>If it works I should opt for the newer version? <ieure>Rust (and many other languages) want to use precise dependencies in order to reduce potential for bugs reported against their project for stuff that changes in versions of their dependencies. <ieure>This is perfectly reasonable if you're a corporation deploying a monolithic binary to a production system, and works very poorly if you want to ship software as part of a distribution. <luca>Yes I've read the gentoo blog post :P <ieure>So, it could work, or it might have subtle bugs that aren't present with 0.30. No real way to know. If you build with 0.34, then ask for support upstream, good chance they'll turn you away for using a different thing than they shipped. <luca>So if unsure, I should opt for a specific version (0.30 in this case) <ieure>It's really a pick your poison kind of situation, both options kinda suck <nutcase>ieure: thank you for the LibreWolf context menu fix! Is the video acceleration issue maybe related to my today's observation that I couldn't see anybody in a Zoom (shame on me!) call? <Rutherther>I am having sound issues lately in LibreWolf, is this also issue for others? Not everywhere, just some sites <nutcase>Rutherther: yes, I also have some sound issues, also in LibreWolf. However, I think I have several ones and not only in LibreWolf. I observed an nosound-issue while watching some youtube video. Suddenly the sound is gone <luca>Anyone got any tips for `Error: Error { source: Os { code: 2, kind: NotFound, message: "No such file or directory" }, description: None }`? I am getting it when trying to package rust-battery. I assume it might be because it's checking for a battery and guix's chroot doesn't have one, but I really don't know <Rutherther>luca: that can be many things. What phase is this in and what are the crates it depends on? <Rutherther>I see you have libgit2 in inputs. Then I would try adding libssh2 as well <Rutherther>alright, so that means it could be what you already mentioned. It will be quite impossible to make it working if that is the case. The build environment cannot depend on a real battery info. Apart from that sometimes the crates on crate.io don't contain the tests and you have to use their original repo, ie. from github as source. But here it seems more probable to be the battery missing <Rutherther>if all tests depend on battery presence, then you can just add argument "tests? #f", because the other option would probably be making sort of a stub battery <Rutherther>anyway, got to go, hope someone else will help if more help is needed <luca>I guess there are 2 tests that pass. But otherwise idk enough about this crate to know if I should just skip all tests or not. But I'll do it for now and someone will complain under review if it ever gets that far <__cornflake___>Rutherther: I believe so, but I'll run guix pull and upgrade everything just in case