IRC channel logs
2025-09-10.log
back to list of logs
<bavier>fnat-xmpp: keeping two versions is definitely done in some packages. i don't recall if haskell has issues with that, I think it's probably fine. <FuncProgLinux>I couldn't find this issue on any search engine :L but how do you deal with builds that fail with OSError: [Errno 30] Read-only file system: '/gnu/store/zs7iww43srx9kl0rcyrhs39nnf04cgha-gtk+-3.24.43/lib/gtk-3.0/modules'? <gabber>FuncProgLinux: this build seems to try to add stuff to that modules directory, doesn't it? <gabber>we usually package these modules separately and disable the mechanism that tries funky store modifications <FuncProgLinux>gabber: Yes, apparently it happened on a libxapp version bump <FuncProgLinux>idk why but packaging libraries is way harder than desktop programs :s <FuncProgLinux>Digging deeper it seems that I need to set and env var PKG_CONFIG_GTK__3_0_LIBDIR to point out to #$output "/lib" directory <hanker>so much hinges so often on an env var <tux0r>NASA grade? you mean, each signal takes DAYS? <apteryx>how can I have goto-address-prog-mode enabled globally? I added (goto-address-prog-mode) to my ~/.emacs, but that's not enough. <apteryx>solution, in case someone was reading: (add-hook 'prog-mode-hook #'goto-address-prog-mode) <adanska>is there a way to turn the output of `guix system build` into a manifest? I'm trying to get the weather for all of the packages in my system <efraim>as46: you'll have more luck with some of those questions in the #nonguix channel <as46>efraim sad I cant even install guix due to no free driver <Deltafire>heh, system wouldn't boot to desktop after upgrading <Deltafire>could be the libxml2.so.2 issue which is affecting other packages (after rolling back system) <Rutherther>what issue specifically? did someone report an issue with ABI incompatibility? <Deltafire>libreoffice and flatpak are displaying the same error message for me <Deltafire>i can rollback.. but would like to try out the PR and see if it also fixes the other issue i have (not booting to gdm) <Rutherther>hmm, interesting. Yeah, could definitely be the cause <efraim>if I have a perl script and it only uses perl modules contained in the perl package itself, do I need to wrap the script with PERL5LIB? <simendsjo>Has anyone seen problems with Sheperd using the new guile-fibers? PID 1 runs at 100% CPU here, and that looks suspicious. I've reported an issue, so chip in if you notice anything: https://issues.guix.gnu.org/79418 <efraim>we don't have a perl team, do we? <simendsjo>cbaines: Thanks! Looks like civodul found the root cause already, so we'll probably get a patch soon. <levenson>efraim: maybe if you plan to use different versions of perl?) <nimous>Hi. Is there a recommend resource on learning the Scheme language? Or do I just read Guix and Guile manuals? <awth13>nimous: there is a crash course in The GNU Guix Cookbook that also includes links to further resources <identity>nimous: or here: (info "(guix-cookbook) A Scheme Crash Course"), evaluate this in Emacs, or in a terminal without the brackets <awth13>identity: right! how do I link from info? <identity>awth13: i press ‘w’ for ‘Info-copy-current-node-name’ and write out the (info ) <apteryx>is it me, or is bordeaux very slow for looking up substitutes at the moment? <awth13>c0co just complained about it to me too <awth13>could not init a system because of bordeaaux being slow <yarl>guix build --with-input=guile=guile-next works fine. <yarl>Now, I would like this guile-fibers-with-guile-next to be used in place for guile-fibers for every package that needs guile-fibers. <yarl>I created a local channel with one package: ((options->transformation <yarl> '((with-input . "guile=guile-next"))) guile-fibers) <yarl>but guix build -L . guile-fibers style uses guile-3.0.9 as input. <yarl>What can I do? How can I specify an "order" (make guix build choose guile-fibers from my channel rather than the default one) <Rutherther>there is no "order", the packages have coded in themselves to choose the guile-fibers from the specific guile module <Rutherther>if you want to change that, you need to change those packages, ie. map over them with package-mapping and rewrite the inputs, you can use package-input-rewriting procedure for that <mange>Another option is to create a custom Guix channel with your change applied to the guile-fibers package. <mange>Which is basically: clone Guix and apply your patch, then use guix pull --url to point to your patched repo. <Rutherther>for this, indeed, you would need to map everything in your system <nutcase>Hi, according to htop, my PID 1 (shepherd) uses 100% CPU. This happens since a system reconfigure yesterday. Does anyone else have this issue? <nutcase>ok, thanks and sorry, I didn't find the issue @codeberg <simendsjo>nutcase: The best part is that my home shepherd also uses 100%, so now I have two processes running at 100% continuously. Some coworkers seems to think my fans are a little noisy today. <Rutherther>doesn't rebooting help? Based on the issue this should trigger after suspending? <Rutherther>or does it trigger for you even without suspend? <simendsjo>I did try rebooting, and the issue appeared again shortly after logging in. I don't want to reboot many times a day. I tried restarting some services, hoping that would resolve the issue, but unfortunately I haven't found anything that works. I might have to patch shepherd until it's fixed as my laptop is running quite hot. <simendsjo>Rutherther: I thought it happened without suspending, but I see an uptime of 16:40, so maybe I didn't try rebooting. But if it happens again shortly after I don't want to reboot the machine several times per day. <Rutherther>sure I am working on the assumption that the issue won't occur without suspending/restarting services. If it does, that's a new trigger of that behavior that is not reported in the codeberg issue for now <simendsjo>Rutherther: I'll try a reboots then. If it only happens after suspend, it's only a few times per day I have to do it until a fix comes along. <Rutherther>nutcase is claiming that even after reboot it did not help, so I am not completely sure it will help... but probably worth a try if reboot is not a problem for you. Also I think it would be good if someone who gets the issue without suspending/(restarting services) reported it to the issue <apteryx>simendsjo: hi, the CPU usage issue is triggered by the commit bumping guile-fibers that I pushed earlier, which was supposed to fix a memory leak (it seems it helps with that, or perhaps it's processing things more slowly now...) <peterpolidoro>Hi all. I am running guix on top of debian and the latest guix pull just crippled me by messing up emacs. Now when I try to run emacs I get the error: "emacs: error while loading shared libraries: libxml2.so.2: cannot open shared object file: No such file or directory". Any suggestions? <kestrelwx>I'd just 'guix pull --roll-back' until the issue is closed. <peterpolidoro>emacs -q does not work. guix pull --roll-back does not work. I have rolled back over 10 generations and it still does not work for some reason <mange>You've reinstalled emacs after the rollback? <cricri>Is this a general issue with the system-config-printer package? <PotentialGuixUse>Hi, can anyone give me advice? I want to install GNU/Linux using Guix with s6 as init system, but i need to know .scm syntax and what all commands do and what *modules* do (like gnu, etc..). Where can i find manual about Guix? <ieure>PotentialGuixUse, Guix System uses Shepherd as its init system, I don't believe s6 is supported. <PotentialGuixUse>ieure, is it not possible to change init? Or use custom repository(so called channels) to include it? <Rutherther>anything is possible... given enough work. The amount of work will be quite large though <gabber>PotentialGuixUse: in general everything is possible <ieure>PotentialGuixUse, Guix is extremely hackable, so "possible?" Yes. But you'd have to write some significant code to do it. <identity>if you plan to bring everything you got used to Guix, you might encounter problems; Guix is not quite minimalistic as it stands, at least in my opinion. i recommend trying ‘The Guix Way’, including Shepherd, and see if you like it first <abbe>is waybar (or waybar-experimental) broken for someone else too, esp. w.r.t. recent libxml2 commits <abbe>waybar: error while loading shared libraries: libxml2.so.2: cannot open shared object file: No such file or directory << broke after recent guix system reconfigure <abbe>❯ LD_LIBRARY_PATH=/gnu/store/z44pnpm4b85rcdmqa0nkn4lwbcxc3gnk-libxml2-2.9.14/lib waybar <<< works <ieure>That's very weird, there's usually a phase in the package build that checks for unresolved libraries. <ieure>Maybe libxml2 has a graft that's broken or something. <Rutherther>yeah, that is exactly the case, it's already been reported and there is a PR fixing it <FuncProgLinux>I also experienced this since I applied the system fix yesterday that required the 6GB download. Got desperate and just passed "--keep-going" to the system reconfigure <FuncProgLinux>libxslt-1.1.37 failed to build because of this and thus a ton of python-something packages couldn't be built. Also some others like parted, eudev, lvm2, xkeyboard-config, etc :( <futurile>no substitutes? (or you're not using them by choice) <Rutherther>if you do --with-input=libxml2@2.9=libxml2@2.14, there are no substitutes for that :) <FuncProgLinux>futurile: Substitutes were going at 20Kb/s I ended up leaving things to build over night <abbe>is there a way to figure out why pid 1 is at 100% cpu <FuncProgLinux>python-black-24.10.0.drv': 15 dependencies couldn't be built <abbe> 1 root 20 0 310,6m 101,8m 113,2 0,3 259:15.30 R /gnu/store/37m0a0ydy74wl2qrf2w1jdgqhxwbaxac-guile-3.0.9/bin/guile --no-auto-compile /gnu/store/hybzcm1macv33msirh3x9aimnkp1g4hi-shepherd-1.0.7/bi+ <abbe>seems to be looping at: [pid 1] epoll_wait(12, [], 16, 0) = 0 <abbe>shepherd 1 root 12u a_inode 0,16 0 2111 [eventpoll:6,8,9,10,17,19,20,22,24,26,27,28,31,32,34,35,37,39,41,43,45,47,51,56,57,59,60,61,62,63,66,67...] <abbe>sometimes, I wish we could just have a systemd alternative, seems like reinventing wheel in guile scheme :) <futurile>FuncProgLinux: where are you based? You know there's some other substitute servers now right? <futurile>(just read a mastodon thingy about the USA one being back up) <FuncProgLinux>futurile: Latin America :) my system always searches at bordeaux first and then at ci.guix <FuncProgLinux>I don't use/have socials but I'll see if I can subscribe to this substitutes server if that could make life easier :) thank you! <futurile>the association is Colin works for Genenetwork, and they use Guix - all part of the 'science' users of Guix like INRIA <FuncProgLinux>futurile: I have seen guix involved in a lot of academic stuff. If it was easier to use I would love to introduce it to a couple friends that work with R and Julia <futurile>FuncProgLinux: there's a fairly active group of science people that use it for deploying things on clusters and doing reproducible research. There some links from the main Guix site and I think they do virtual meet-ups if you're interested <apoorv569>futurile: I have updated both PRs I mentioned yesterday as suggested. <futurile>apoorv569: OK, I can probably look at them by the end of the week. I have some work for Guix Foundation and a blog post in the queue. <futurile>apoorv569: give me the numbers again and I'll assign them to me <csantosb>Wow. One day I really do need to understand what exactly is going on with QA and emacs-team branch. It is still there after one week. <futurile>csantosb: if you ever figure out how it works, please tell me <csantosb>Rutherther: I understand it builds emacs (and family) and all emacs-build-system packages, which takes a few hours in my old laptop <Rutherther>csantosb: it builds everything that has been changed. It will stay there until someone closes the issue. But if you click on it, you can see what remains to be built on what architecture. Also when you click on it you can see why it isn't currently building: "Submitting builds for this branch suspended as master branch substitute availability is low for: " <Rutherther>apart from that as far as I know there have been some issues with qa, like sometimes it cannot evaluate guix, has to be reevaluated etc., so that slows things down as well <ieure>Just wait until you have to rebuild everything that's compiled with it. <FuncProgLinux>ieure: I'm afraid that, when the libxml incident is fixed I'll have to rebuild everything again :( <ieure>FuncProgLinux, Likely correct, but isn't a bunch of what you're doing working around that mess? <FuncProgLinux>ieure: It is but I don't know how to do the advanced stuff yet in guix like moving across PR's or locking my system in a stable state <FuncProgLinux>Last time I pointed my guix channel to a local checkout it took like an hour to index and failed with an unknown key error <Rutherther>you need to disable authentication when using commits from non-committer <ieure>FuncProgLinux, You can add (commit) fields to your channel configuration to pin Guix to a specific commit. <Rutherther>usually better to use "guix describe -f channels" rather than specifying manually. Although for guix channel it doesn't matter, for other channels it does as you should also pin the dependencies of the channels <FuncProgLinux>ieure: I have done that I think last week I tried to change the commit manually but it didn't work <FuncProgLinux>When libxml first broke mate I just went to "channels-lockfile.scm" changed the commit and ran guix pull -C ./path-to-channels-lockfile.scm <postroutine>Hello. With `home-bash-service-type`, why does the environment variables are set into `.bash_profile` and not `.bashrc` ? <postroutine>Each time I start a terminal software, it only source bashrc and not bash_profile, so my environement variables are not set. <Rutherther>postroutine: because that's how the files were meant to be used, you can find that in the shell's manual. Environment variables are meant to be sourced on login. Not on each shell run. <Rutherther>rc files are meant for configuration of the shell itself, not for modifying environment variables <Rutherther>to apply them to your DE/WM/compositor etc., yes <Rutherther>retropikzel: please elaborate on your problem, your question doesn't make sense to me <retropikzel>Rutherther: I have software written in C that depends on gtk headers. But "guix shell --development gtk gtk+" for example does not make an environment where the program can find them <retropikzel>But I think I found them under $GUIX_ENVIRONMENT/include, it's just that they are not in gtk/gtk.h but gtk-3/gtk/gtk.h. So propably adding -I${GUIX_ENVIRONMENT}/include/gtk-3 helps <ieure>retropikzel, The --development flag is for developing those packages, not for developing software which needs them. <luca>Have you considered using pkg-config? <ieure>retropikzel, It makes all the inputs of gtk and gtk+ available in the shell, so they can be built. It does not make gtk / gtk+ themselves available. <Rutherther>retropikzel: so are you compiling it? Then you would also need gcc-toolchain in the shell. Also, why are you adding development argument? That gives you the inputs of the package, not the package itself, it's for developing the package you put in. And with gcc-toolchain it still won't 'make an environment' with include path going to include/gtk-3, only to include. The package is supposed to use something like pkg-config to find where libraries are... <Rutherther>... located rather than rely on the default locations. And that would have to be provided as another argument to the shell to get the proper pkg config path <yelninei>i586-gnu substittues are out of commencement.scm finally after core-packages-team merge. I think they will hit the automake issue soon which took me a lot of retries to get past the flaky tests. <postroutine>Rutherther, but as Guix define a default PS1 in the .bashrc, I have no other choice than to define my own PS1 into the .bashrc ? Because if I set the bash config field `guix-defaults?`, it will disable way more than a default PS1 ? <retropikzel>Rutherther: I was adding the --development just to see if it helped <Rutherther>postroutine: seems so, yes, you have to define it in .bashrc because of that. But it's the correct place to do it, so I don't really see why it would be an issue <postroutine>But, you have say that environment variable are to be set in `.bash_profile`, as `home-bash-service-type` do. <Rutherther>yes, and PS1 has nothing to do with environment variables <Rutherther>ieure: that's not exactly right, the --development flag affects only the next package. gtk+ will be available in the shell when you do guix shell --development gtk gtk+ <postroutine>But `PS1` is an environment variable set as, for example: `PS1="\u@\h "`. <Rutherther>whether it's exported as an environment variable, that's another question, it's not visible in the example you gave. But it generally shouldn't be as it's only relevant for the shell itself <Rutherther>consider doing "MYVAR=1" and then "printenv MYVAR", you will not get 1, because it's not an environment variable. It's just a local variable for the shell <Rutherther>only if you first did "export MYVAR", you would get 1 on "printenv MYVAR" <simendsjo>Tried the new channel which rolls back the guile-fibers change, but now emacs cannot start because libxml2.so.2. Reconfiguring with --no-grafts solved the issue. <Rutherther>simendsjo: it's problem with libxml2 being grafted with libxml2-next that has removed libxml2.so.2, but apps are linked against it <awth13>how do I turn a `local-file` into a string that I can append to another string? <awth13>I want to intern a file in the store and concatenate its name in the store to a string I am constructing <Rutherther>awth13: that's not usually the right thing to do. You should typically use gexps and ungexp the local-file. Then you get its path when building. What are you trying to achieve? <awth13>I am writing a sway config using `home-sway-service-type' and I want to pass a background image (the local-file in question) to swaylock <awth13>for which I need to append the resolved filename of the local-file to a string <Rutherther>awth13: so you're passing this to startup-programs? <awth13>I thought `#~(string-append "the" #$(local-file "path"))` should work but it didn't <awth13>setting a variable in the list of variables <awth13>`((lock . ,(file-append swaylock #~(string-append ...)))) <Rutherther>yes, that won't work, file-append accepts string as second argument <Rutherther>you would need ",#~(string-append #$swaylock ... #$(local-file ...) ...)" <FuncProgLinux>I didn't know guile could translate S-expressions into HTML! <ieure>FuncProgLinux, SGML (thus: XML, HTML, and XHTML) are just more verbose s-expressions. <ieure>SGML/HTML to a lesser degree, because closing tags is optional, but XML/XHTML? Verbose sexps, 100%. <FuncProgLinux>ieure: I like it because it opens a posibility to make simple web stacks using HTMX + Missing.css on Artains <FuncProgLinux>btw I managed to "go back in time" (sort of) by rolling back my commits-lockfile.scm file and guix pulling there with --allow-downgrades. I haven't gc'd in like 4 months, that helped too