IRC channel logs
2026-01-10.log
back to list of logs
<umanwizard>Given a `manifest.scm` file, is there an easy way (other than opening it up and editing the lisp directly) to add/remove a package from it? I.e. something like `guix package -i` or `guix package -r` but operating on a manifest rather than a profile <umanwizard>Best thing I found so far, for example to add `emacs`, is the rather ugly: t=$(mktemp) sh -c 'guix shell -m ../manifest.scm emacs --export-manifest > $t && mv $t ./manifest.scm' <gabber>csantosb: do i remember correctly that you did some cleaning up in the engineering.scm vs electronics.scm matter? where are we with that project? is it still on? should i move gnucap to electronics.scm while i am at it? shall we move all EDA software to the electronics.scm module? WDYT? <gabber>sneek, later tell csantosb that i mentioned them in my previous message <PotentialUser-24>Hey! So I got a bit of an issue that may be hard to visualize but ill try: I got my dotfiles linked to .config using home-dotfiles-service-type, and i decided to try some quickshell configurations and noticed the following problem: <PotentialUser-24>- The service sends files to the store, renames them with a hash then symlinks them to .config, and something about this process messes with how quickshell follows modules and breaks the whole thing (for reference, normally quickshell configs look like "quickshell/test/main.qml" "quickshell/test/modules/a.qml" ...). <PotentialUser-24>I can think of a couple dirty patches that can fix this but i just wanted to check first if there was a convenient home service or something i didnt know about for a situation like this, maybe something that only sends the .config/quickshell/ folder to the store instead of the individual files? or something that just symlinks without sending to the <PotentialUser-24>Okay i dug into it a bit more and the best solution would be to link the "dotfiles/.config/quickshell/" folder directly to the store, preserving its structure and then link that to "$HOME/.config", any easy way to do something like that? any miracle service i dont know about around? <oliverD>I got in after uninstalling flatpak and updating guix (I'm not sure what fixed it I couldn't see anything obvious in the logs). Now Icedove won't start <dariqq>Am I imagining things or did webkitgtk used to use the cmake ninja backend? <sneek>csantosb, you have 1 message! <sneek>csantosb, gabber says: that i mentioned them in my previous message <csantosb>gabber: Sure, to me all electronics related packages should be moved to electronics module, engineering is a different business <gabber>csantosb: cool! should we open an issue on codeberg? <gabber>and maybe clarify what electronics.scm entails in a module commentary? <csantosb>Do we have module comments explaining the reach of the packages ? <csantosb>Would be nice to have, but I'm unaware of any other module providing this kind of information <gabber>cf python-build.scm after the module definition <gabber>Is there some sort of policy on milestones for Guix? Can we (e.g. as teams) just open them up or should we discuss them first on guix-devel? <csantosb>Oh, nice, I was unfamiliar with these ";;; Commentary:" in the code. I'm for including it in electronics.scm. <gabber>csantosb: good, i'll prepare a patch <csantosb>No idea regarding the milestones, though. Do you have something in mind ? <gabber>i do, indeed! but i guess i'll send my (/our) ideas through the guix-devel mailing list anyhow <csantosb>Regarding moving packages between modules, other than ensuring there is no conflict somewhere else, the complexity holds on keeping the copyrights right, which implies checking the log of the package. Booring, but dooable. <gabber>hrmpf, yes. i hadn't considered that. imho this is not an important goal, but i think whenever we get to touch packages, we can also do the tidying up, no? and maybe we find some motivated rookie(s) to help us with that? <ekaitz>csantosb: have you seen the ML email about Guix and F-Si? <ekaitz>gabber: it was you who sent the email? <csantosb>ekaitz: now, I have 😁; yes, Gabriel, gabber, sent it. <csantosb>Well, it comes to having more hands to do the job, as usual <csantosb>To complete, other than the F-Si, there is also the Fossi Foundation, closely coupled to YosysHQ and their Oss Cad Suite <csantosb>During Orconf 2025 people usually manifest the need for a complete set of tools, they currently do a lot of scripting, docker, etc. <zenmaya>does the grub keyboard layout option work for someone? I cant seem to get it to work properly, it just stays as us qwerty <identity>the grub keyboard layout option works *only* in grub. it *should* work in grub <apteryx>am I the only one to cope with network error messages on codeberg? <apteryx>I need to refresh the pages manually almost every time I want to post something. <gabber>same. i just copy, reload the page, paste, then it usually works <gabber>i sense codeberg's been having more troubles lately, i guess from scaling up since we moved there (and probably generally more traffic on their end)? <gabber>csantosb: have you opened a [wishlist] issue on codeberg for oss-cad-suite-build? <gabber>i've told my F-Si peeps about it but forgot just which suite it was/where to find it <gabber>ekaitz: yes, interesting and exciting indeed! <ekaitz>gabber, csantosb : I'm very interested on hardware, but I don't have the time for everything. But very interesting times here! <ekaitz>in fact, I want to get more into electronics and I'm starting to realize the only way I can do that is if I find some mentors... <gabber>WDYM "getting more into electronics"? apart from handling a soldering iron for my toy projects i have little to no clue about all that jazz <gabber>good thing: you can partake even with little time on your hands <ekaitz>i'm supposed to be an EEE, but I spent all my career in software... :( <gabber>IMHO telling people the right things can help very much! <gabber>and (like in my case) minimal/basic knowledge about electronics and more knowledge about guix already seems to do the trick <ekaitz>i'm now interested on electronics design, I'm making a quite complex (for me) circuit and pcb (it has USB2.0 and all!) and I'm struggling a lot with it <ekaitz>also the CPU design part is very interesting, I'd like to design a simple CPU <ekaitz>the last VHDL i got was like... 15 years ago <gabber>all i know is that Verilog looks *much* nicer (: <gabber>what software are you using for that? <gabber>are you intending to build and run it eventually? <janneke>about a month ago my simple home server would suddenly hang during boot. the last update was somewhere in summer. the most recent version that would boot was from november 2023. the server being >10y old and always having been unreliable, i assumed the hardware would really be dying now. today i booted the primary disk in a spare laptop, and it showed the same boot problems: only the nov 2023 installation would boot. hmm <janneke>so then i thought: ah, that must be store corruption! however, running "guix gc --verify=contents,repair" yielded no results. i'll probably just try another disk. ideas? <gabber>but you did reboot the system from time to time in between upgrades/reconfigurations? <janneke>gabber: yes, did ~8 upgrades in the meantime, and the system reboots spontanously every month or so (yuck!) <janneke>gabber: it's a "nice" puzzle, right? at least, /me has run out of obvious ideas of what could be wrong <gabber>what else did you consider apart from store corruption? have you done an fsck? how/where does the boot process fail? <janneke>it hangs just after the disk(s) is (are) checked and mounted <janneke>in any case, the disk is problematic; possibly installations after nov 2023 happen to access some file that makes the disk hang <gabber>so, the kernel boots, hardware is initialized, then the machine hangs? <gabber>i've had issues with non-booting machines when *another* drive than whatever mounts / can not be mounted (due to guix expecting it to show up eventually) <gabber>what kernel are you booting? the hurd? <janneke>hehe, linux-libre; so 6.5.8 boots, anything later doesn't boot <gabber>is there a significant configuration difference between that nov2023 and the later ones? <gabber>and by "boot" you mean the machine / guix os and not *just* the kernel, right? <janneke>but it gets worse: i've re-installed, using time-machine, the nov2023 6.5.8 version with a minimal change, that also hangs <gabber>on what machine did you do that? on the home server, the spare laptop or a third one? <janneke>yeah, the only change i made was adding an option to sshd (allowing rsa keys) <gabber>and you never exchanged the (main or other) drive (resulting in a different UUID or such)? <janneke>i did that all on the server hardware, only today i tried the disk in the spare laptop but didn't make any changes; apart from manullly fsck'ing (no results) and guix gc repair (no results) <janneke>anyway, i'll just re-install the server to a new disk and report back -- unless someone has a brilliant idea in the meantime :) <gabber>let's see if that works (: if this fails, too we gain a bit more insight <gabber>csantosb: yeah, i've seen that discussion. and i did not like the general tone. "allowing people" and such. that's kind of why i hope we can craft an official "milestone" (where i hope you will join us in defining cool goals) so "people" will not complain because we're like, uhhh, "officials" doing important work <gabber>and i don't like #4548 being closed without the package being there yet :( <andreas-e>gabber: I think the main point in #4548 was that there is no point in piling up package requests to an issue tracker that already keeps growing... Not enough people do not enough work to deal with the current backlog, so unless someone proposes a pull request for a new package, it is unlikely the suggestion will be picked up by someone bored for lack of work. <csantosb>Yeah, the tag was a bit misleading for me, my bad; now it clearly states "Feature requests", so I closed it. <andreas-e>No problem with creating a milestone, but it should be a concrete and finite endeavour (like making a release); otherwise a team would be appropriate, and there already is the electronics team. <PotentialUser-7>Did Anki get removed from Guix? I could've sworn it was installable earlier this week. <bdju>I can't seem to rsync from one of my Guix machines to the other, it thinks rsync isn't installed, but it is. Seems to be a PATH issue, but only with ssh and not the interactive shell. How do I go about solving this? <bdju>I have rsync installed in the user profile but not the system profile. Is that the issue? <bdju>That indeed fixed it. Neat. Will make sure to have that in my system profile from now on then. <ekaitz>gabber: i just answered in the thread but the "bundle" thing doesn't feel like a toolchain? <ekaitz>gabber: but I think we need several toolchains for your goals, because designing a PCB and a FPGA are quite different jobs that require different set of tools <ekaitz>for example, i'd say kicad doesn't include the symbols and the libraries by default, which doesn't make a lot of sense "kicad-toolchain" could be something, also including some circuit simulation software and such <Nessah>Full source bootstrap succeeded after 5 days. I noticed it didn't build linux-libre and maybe a few other parts that were already installed. <Nessah>x86-64, ThinkPad T14s Gen1, base install with GNOME <Nessah>(didn't build in my case means it didn't need to) <ekaitz>I was hoping you managed to merge the RISC-V one hehe <ekaitz>efraim: we have to manage to get that merged ^^ <Nessah>I don't have a working RISC-V setup here, but somehow it did cross builds for riscv for binutils, gcc, etc, during the install bootloader phase <Nessah>the worst offender was webkitgtk, needing 32GB to build at all <yelninei>i think thats because grub depends on qemu for tests and qemu depends on some firmware (opensbi-riscv64 ) <PotentialUser-24>in case anybody runs into an issue like this with quickshell in guix, ill leave this workaround in the archives for the divers of the future: <PotentialUser-24> ;; HACK: Problem is that home-dotfiles-service sends individual files to the store, <PotentialUser-24> ;; But quickshell resolves modules paths from the store, which breaks since the <PotentialUser-24> (simple-service 'quickshel-workaround home-xdg-configuration-files-service-type <vagrantc>struggling to figure out how to run guix-publish after some other service(s) are up... <vagrantc>i want to bind to a yggdrasil interface, but with a static (ipv6) address it fails to bind, and with a hostname it also does not appear to listen ... oddly, if i restart the service after name resolution and the yggdrasil interface is up, it still fails <vagrantc>but if i reconfigure the system and *then* restart the service, it works <vagrantc>suppose i should try guix-help (or is it help-guix? :) <vagrantc>my guess is the service checks for the address and/or hostname, and then simply restarting it does not re-check weather those exist... but that is a wild guess <csantosb>Any idea on why codeberg refs are guix/guix!5502 instead of guix/guix#5502 ? <Rutherther>csantosb: wdym they're that and not the other? You can use both <Rutherther>csantosb: true and in some projects it's #. Yeah, I don't know. Looking into settings of my own projects I don't even see that being configurable <zenmaya>Rutherther: stays us qwerty both in disk decryption before grub menu appears and even in grub menu <Rutherther>zenmaya: the first one is expected, grub doesn't have the keyboard layout loaded that early. The second not really. <zenmaya>well can I load it that early? I really need dvorak there <Rutherther>zenmaya: you can definitely try looking into getting that supported, would be great to have it in the guix's grub bootloader configuration :) <zenmaya>another nerdsnipe :grief: /j (i already have nebula vpn service planned) <Rutherther>csantosb: oh so I guess the idea is that PRs have "!" and issues have "#"? But in the end it is interchangeable due to how forgejo handles the .../issues/XXX and .../pulls/XXX routes <vagrantc>it seems the freeze on master is, somewhat predictibly, not working? <Rutherther>janneke: does Hurd not have file system UUID support / label support? Why does the installer put indices of the devices in the config? <csantosb>vagrantc: you're right; we should concentrate during next week exclusively on fixes, putting updates on hold <vagrantc>csantosb: i think there are people following it, but the ones not following it seem to be not reading or understanding such messages :) <janneke>Rutherther: no, it needs [mach] device/partition numbers <Rutherther>janneke: quick question... I installed Hurd from the installer and now I can't log in, getting invalid password. I am certain the password is the correct one. Any ideas? <nckx>zenmaya: (So nice to see you still around) I think that ‘all’ you have to do is have Guix copy the keymap to, say, /boot/grub and load it from there, instead of the current ‘keymap /gnu/store/…’. It's also called too late, after ‘cryptomount’. <Rutherther>janneke: nope, I set password of root to "root" and password of user ruther to "rutherther", neither works. Also tried putting in empty when logging in, didn't work <janneke>Rutherther: weird, possibly yelninei has an idea when they're around? <gabber>in what build phase are patches applied? <Rutherther>gabber: what patches exactly? There isn't a phase that would apply patches. If a particular package applies patches in a phase, a custom phase is made <gabber>in this case: "audacity-ffmpeg-fallback.patch" in the audacity package <gabber>do you know of a nice example i can look up (that doesn't take too long to build locally)? <Rutherther>gabber: no phase, that's an origin patch, it's applied when the origin output is made <gabber>which happens silently in the build process, i guess? <Rutherther>it happens when the origin's derivation is built <gabber>i mean: i do not see the string "ffmpeg-fallback" in the build log <Rutherther>ie. "guix build --source ffmpeg --check". I see the patch being applied... <gabber>i don't see it in `guix biuld audacity --check` <Rutherther>that is the package, not its source, its source is a dependency of that <gabber>thanks for the clarification! i hope i can fix the ffmpeg/audacity issue soon <janneke>Rutherther: come to think of it, i haven't seen any pam-style login on the hurd; all we had "until recently", is typing the "login root RET" <Rutherther>janneke: that is failing with invalid password, both with empty password and the password I set in the installer <janneke>right, so something changed since i last saw that <Rutherther>if you could check this it would be nice. It would be nice to get this right for the 1.5.0, since otherwise it makes the Hurd option quite unusable <janneke>it's quite some time ago i actually installed the hurd, using childhurds <Rutherther>does hurd utilize /etc/passwd and /etc/shadow at all? <janneke>Rutherther: yes, that would be pretty silly <Rutherther>because currently it seems the installer sets up only those files, for both hurd and linux <janneke>that is to say, debian's hurd uses pam and everything and is a pretty usable system in that regard <janneke>right, we'll surely need them some time... <janneke>in any case, i'll ask yelninei and civodul, one of them probably made a change here