IRC channel logs
2023-05-19.log
back to list of logs
<lambdanil>I wish Guix had a bigger community and more packages first and foremost <lambdanil>otherwise I don't think I'm personally missing anything all too important currently <lambdanil>I'd like to see a rootless Guix option eventually though; might try something myself <bjc>that'd be pretty cool <Guest709>hey guys, I have troulble getting the dropbox sync job to run with guix <acrow>Guest709: I haven't used it to know. <Guest709>is it possible to sync between aws s3 and your pc? <acrow>Guest709: I have not used it on guix but I believe s3cmd is a good package that would facilitate this. <Guest709>acrow what are you syncing your files to, if I may ask? <bjc>syncthing is peer-to-peer; you don't need a server to hold them <bjc>i sync between my phone and a couple computers with it <Guest709>bjc nice, I could also like to have my data in the cloud. <WilhelmVonWeiner>why is it that the homepage refers to Guix as a distribution of the GNU operating system and not a package manager anymore? <lilyp>well, because it is a distribution? <lilyp>that doesn't make it no longer a package manager, but it's important to recognize that Guix is powerful enough to become a daily driver <WilhelmVonWeiner>i see, there's no focus on marketing Guix as just a package manager along an OS designed to operate effectively with Guix <lilyp>If you think about it deep enough, every distro is just a package manager with along with supporting infrastructure designed to operate more or less effectively <lambdanil>I think Guix is more attractive as a package manager to most people though <lambdanil>And if someone is going to get started with Guix System the package manager is a good starting place <WilhelmVonWeiner>lilyp: I don't really think so, but if that's your operating perspective this marketing decision is fair enough <lemos[m]>hurrr, on a package, how can i change the destination dir of something i am fetching <elevenkb>hey there y'all's do you have any tips for using guix with an unreliable internet connectoin <lemos[m]><lemos[m]> "hurrr, on a package, how can i..." <- like, outputing this on a source dir <lemos[m]>i need to fetch a git repo and configure it as a subproject <PotentialUser-53>Hi, I'm trying to update Guix packages on a foreign distribution after a long time of not using it. `guix pull` errored so reinstalled using the script with the overwrite option. I got the root user's guix working well, but things are a little broken in my user profile now. If I run `guix package -u` it says `guix package: error: path <PotentialUser-53>‘/gnu/store/z7a6sbvqzb5zapwpznmjkq2rsxil6i67-glibc-utf8-locales-2.31’ is not valid`. Any idea how to get it happy again? Thanks! <apteryx>mirai: shepherd: Evaluating user expression (and (defined? (quote transient?)) (map (# ?) ?)), eh <PotentialUser-53>Moving my profile aside let me start a new one, which seems to be working... I guess I can just reinstall new each of the same packages as before anyway. <apteryx>mirai: I came up with 'current-service' as something a bit easier to access a single <live-service> object <mirai>does it handle a service with “virtual names”? i.e. NetworkManager / networking <mirai>from the comments, it looks like this shouldn't result in syslog pollution, is that right? <apteryx>also running the script with my user profile I also got the '?' in the output as shown above <apteryx>So I don't think locales have something to do with it, though I'm not sure. <apteryx>mirai: I think you have to use the canonical name, which seems to be defined as the first name it provisions <lemos[m]><lemos[m]> "i need to fetch a git repo and..." <- ok, i found something to refine my question <lemos[m]>how can i properly change the unpack phase? <lemos[m]>finding where the source is downloaded and changing it <lemos[m]>in my case, the method is git-reference, so i am looking for the necessary phase to change <lilyp>why change the phase if you can change the source via --with-source ? <ulfvonbelow>maybe it's one of those very obnoxious packages that insists on downloading software at compile-time <lilyp>in that case change the source so it doesn't <lilyp>i know, i know, easier said than done <mkj>hello. have anyone configured their guix setup to use cryptsetup with keyfile, gpg key, or similar instead of manully typing passphrase? <ulfvonbelow>hmmm, I think I screwed up my installation image (I manually partitioned and ran 'guix system init gnu/system/install.scm' ...). It's booting in legacy BIOS mode, and now when I try installing with grub-efi it complains that it doesn't support efi variables or something like that <jpoiret>ulfvonbelow: the install image is configured to use grub-bootloader <ulfvonbelow>hm. I had assumed that "EFI variables are not supported on this system." on a system that clearly supports UEFI (causes grub-install to fail btw) was because it hadn't booted in UEFI mode... <jpoiret>well, it didn't boot in efi mode, because it's not using grub-efi-bootloader <ulfvonbelow>but presumably others have been able to use The Installer to install for UEFI systems... right? <jpoiret>yes, but not with an installation image that you installed with `guix system init` <jpoiret>usually people do `guix system image -t iso9660 ...` <lambdanil>is there a way to get libreoffice in different languages? <ulfvonbelow>right, but I'm looking at the (gnu system image) code and I don't see anywhere where it modifies an operating system to use grub-efi <ngz>rekado: Hello! I think I've completed the changes on the texlive build system. I'm not sitting on almost 400 patches, most of them being boring package definitions rewriting. I would appreciate if you had time to review the initial ones. If so, how would you like me to proceed? <ngz>rekado: I'm also considering adding a phase to the build system to automatically generate font metrics when "fonts/source" directory contains ".mf" files. I have to experiment with it a bit, tho. <janneke>jpoiret: is (re)creation of glibc/hurd on your agenda, do you perchance have patches for that already? <ngz>Is it possible, during a build phase, to find out what the top-level build directory is? <ulfvonbelow>okay, looks like the disk image code special-cases iso9660 and automatically replaces the bootloader with grub-mkrescue-bootloader <jpoiret>you mean the bootstrap blobs, right? <janneke>jpoiret: no, i mean the packaging (define-public glibc/hurd ...) <jpoiret>ah, no, that's quite tricky and I haven't thought about it yet <jpoiret>ah, no, you need the patches for the rumpdisk to build, right :) <janneke>not sure what the best cause of action is, possibly civodul has a trick to add patches for hurd only that i don't see ... <jpoiret>yes, there is :) I did that also when updating the hurd to avoid a world rebuild, it's very ugly though <jpoiret>but I guess having glibc/hurd is maybe better in the long term <jpoiret>since then we could try out even newer hurds <ada_adanksa>Hi! I know everyone must have seen this exact message a thousand times before, but googling alone has lead me to nowhere significant. Would anyone happen to know how to get Guix booting with Secure Boot? I'm trying to get Guix installed alongside windows on my machine, but Windows 11 requires SB to function. Am I able to somehow sign Guix myself to <ada_adanksa>get it to work on a SB machine? Or is my only option to manually disable SB everytime I want to use Guix? <jpoiret>ada_adanksa: Does W11 really require secure boot? <jpoiret>I have a simple dual boot with both of them, and no secure boot enabled <jpoiret>currently Guix can't be run with secure boot afaik <ada_adanksa>I believe so, W11 requires the TMP chip and secure boot to install under normal circumstances <jpoiret>it's theoretically possible but would require some work <ada_adanksa>I've seen the NixOS devs working on this but reached a dead end on the pr thread, so Im not really sure how the problem in general is wrt being solved <jpoiret>ada_adanksa: from some official windows source, the PC has to be SB capable but SB doesn't have to be enabled <_graywolf>ada_adanksa: If the problem is win11 and SB (assuming it's mandatory), maybe you can downgrade to win10? <ada_adanksa>thats interesting. when i tried to boot w11 w/o SB, it booted to a screen complaining about bitlocker and stuff like that. i must have enabled bitlocker during install. i could probably use a vm to test if w11 without bitlocker could boot without SB <jpoiret>ah, maybe bitlocker does require SB to protect its decryption keys somehow <jpoiret>I think I've disabled it a long time ago on my machine <jpoiret>apparently bitlocker can support non-SB setups, but maybe you need to manually degrade it somehow <jpoiret>but it's definitely possible :) i'm running it myself <ada_adanksa>i'll try just turning bitlocker off on my w11 install. maybe it might work <jpoiret>ada_adanksa: stackexchange tells me that you should be able to input the recovery passphrase on first boot after disabling SB <jpoiret>now for where that recovery passphrase is :) <ada_adanksa>decrypting my drive is gonna take 1 and a half hours :D lovely <ada_adanksa>alright, thanks guys! ill probably be back once I get guix installed and everything, really look forward to the sheer tinkerage im in for <f1refly>when i invoke `swaylock`, i get '[source/pam.c:16] swaylock is setuid, but was compiled with the PAM backend. Run 'chmod a-s swaylock' to fix. Aborting.' <bjc>you need to install swaylock in the system config and use ‘setuid-program-service-type’ for it to work <ulfvonbelow>my reading of that error message is that it *is* setuid and *isn't supposed to be*. <bjc>oh, you're right. huh. i didn't notice that when i first saw it on my system. i *am* installing it setuid and it works <bjc>did swaylock-effects change recently? <bjc>ohh. i'm not using screen-locker-service-type. maybe that's the difference <f1refly>swaylock is /run/setuid-programs/swaylock <f1refly>i've been using an ugly hack before as well and only now switched to the service type <ulfvonbelow>anyway, it seems to want to either be compiled with the PAM backend or be setuid, and the two are mutually exclusive <jlicht>bjc, f1refly: swaylock was changed about 18 hours ago to include the PAM-related stuff <f1refly>and swaylock-effects probably inherits swaylock <f1refly>i dont understand why swaylock-effects doesn't want to be setuid :/ <f1refly>what is the correct course of action for me with my problem now? <ulfvonbelow>well, either make it not get built with pam, and keep it setuid, or make it not setuid <ulfvonbelow>the former would involve using a modified version of the package, the latter would involve moving it from the setuid programs to the system packages <f1refly>and since the new screen locker service type inherently sets setuid i'd have to remove the service type altogether, right? <ulfvonbelow>that service should probably be modified to reflect the recent changes <ulfvonbelow>but if you don't wanna be bothered with running a local checkout until that happens, you'd remove the service, yeah. <f1refly>i looked at the original swaylock project and it's using the exact same setuid check, does swaylock get any special treatment? <f1refly>i didn't find anything out of the ordinary in the swaylock package suggesting that it shouldnt be suid or something similar <f1refly>it's leaving out the setuid extension <apteryx>fun: 2023-05-19 03:15:00 28594 /gnu/store/zmb73m7qgqagpb4vjhdd9jgkwp716p0m-wireguard-wg0-monitoring: resetting `JPWIbC9qMlnTkWfq000000000/A/B=' peer endpoint to `myhost.duckdns.org:51820' due to stale IP (`66.158.X.X' instead of `66.158.Y.Y') <anemofilia>I am using sx to start x, but no matter which wm I boot to, my keyboard and mouse don't work <anemofilia>My user is already in the input group and I have installed xorg-xinit in system level <iyzsong[m]>anemofilia: does there exist some error in the Xorg log? i think sx put it under ~/.local/share/xorg <nckx>Ding dong. What does Aquilenet host for us, again? <nckx>Things? Stuff? Probably. <nckx>Thanks. That's what I thought (no, really) but it wasn't mentioned at /donate. <nckx>There will be downtimes on the 26th. I'll forward the mail to help-@. <cbaines>I think I've seen the email (to guix-sysadmin) <andreas-e>I thought a bit of French would do no harm :) <nckx>That's how it always starts. <nckx>I was going to transummarise the thing on the public lists, unless someone already has/was planning to? <andreas-e>It was easier to just let the message pass instead of translating it! <nckx>But has it been sent publicly? <nckx>Some people rely (only/more) on bayfront that berlin. <andreas-e>It has been sent to guix-sysadmin by Aquilenet. <nckx>I'll take that as a 'no' and forward it further. <bjc>anyone around who can take a look at my criu/sssd patches which fix core-updates breakage? (#63149,#63091) <bjc>they're the only remaining reasons i need to use pre-inst-env for system installs <civodul>bjc: hi, sure! what conclusion did we reach wrt. zip file timestamps? <bjc>i'm using the procedure from (guix build python-build-system) for sssd with the ‘@@’ notation. for criu, i copied in a modified version which is symlink aware, and added a TODO to it mentioning it should be removed when the version in python-build-system is fixed <bjc>the fix is in the python-team branch currently, i think, but it'll probably be a while before that's mainlined <anemofilia>Do someone know why it says things about systemd-logind? <bjc>i don't know for sure, but guix uses elogind by default, which is extracted from systemd-logind <anemofilia>Seems that some modules like fbdev and vesa fail to load, but IDK if that's the problem here <_graywolf>Is there any documentation on the topic of setting up caching proxies for the Guix resources (or running Guix in containers)? I would like to start using Guix in my CI/CD system, but without hammering upstream with requests. Any reading on this topic that anyone could recommend? <civodul>_graywolf: hi! you could run 'guix publish' on your network, and possibly enable auto-discovery <anemofilia>Idk what could be the problem with sx... If someone can take a little time to help me, I have sent the log to 0x0 <bjc>thanks for pushing those patches civodul <acrow>bjc: Why is pre-1980 defined to be 1980-01-02? <bjc>python's zip can't handle dates before 1980-jan-01. i don't know why the second of january was picked, but it doesn't really matter <acrow>Hmm.. confusing. Doesn't that make things dated 1980-01-01 a problem? <bjc>no, the first is fine, too. my initial patch used midnight on the 1st (utc) <acrow>It's a crazy way to define the beginning of an epoch. :) <next4th>anemofilia: did you start a X application at the end of 'sxrc'? you can try 'sx -- xterm' to see if keyboard/mouse works <jpoiret>would it be fine to have a world rebuild for i586-pc-gnu, be it native and cross? <cbaines>jpoiret, yep, I think i586-gnu (native) is pretty broken still, and rebuilding all of i586-pc-gnu (cross) is fine <jpoiret>nice! I still have some more clean-ups to get rid of hardcoded archs, and we could then also add a newer hurd glibc + hurd packages then <jpoiret>still wondering if upgrading the bootstrap blobs will resolve the m4 building issues or if the tests just don't work anymore on Hurd <unfroq>Hey Guix! When I configure a service with guix home, will it be able to access the wayland/sway session? I wrote a shepherd service and applied it with guix home, but it cannot connect to the display. It says: "failed to connect to wl_display". When I wrote a shepherd service without guix home, it just worked. <bjc>unfroq: no. you need to start shepherd from within sway to do that <bjc>i disabled ‘auto-start?’ for ‘home-shepherd-service-type’, then added a ‘swaymsg exec shepherd’ to my sway config <minima>hi, i started having issues with xelatex and a particular font (IBM Plex); the error i get is "xdvipdfmx:fatal: Cannot proceed without the font: <path-to-font>/IBMPlexMono-Regular-Pi.woff" <minima>this happens with a file that used to compile just fine until a few days ago <minima>i understand this might be a problem with xelatex not being able to work with woff files <minima>the file indicated above exists in my filesystem by the way <minima>are you aware of any change in guix that might be triggering this error? <unfroq>minima: Did you try another woff-Font? <minima>unfroq: i tried another font (without checking if it were woff) and the file compiled ok; latex folks on #latex confirmed that xelatex doesn't play nicely with woff files <minima>i'll try another woff font now though <minima>hm i don't seem to have any other woff font installed in my system <unfroq>bjc: Thanks for the hint. I'll try to spin up shepherd that way, then. <unfroq>minima: I would try to install one somewhere and try that particular font in order get a better view on the problem. <unfroq>minima: Or you try to find the font in ttf or other better supported format. <minima>unfroq: got it, let me try that, thanks <jpoiret>janneke: so, i'm only looking now at your rumpdisk attempt. I didn't really get the situation with the clock monotonic patches, don't we already carry them? <janneke>jpoiret: we carried an "adapted version" of them <janneke>those were just good enough to build python or python-minimal (pass its tests) <janneke>it's terrible, these patches were created ~10y ago <janneke>at the time we (I? I really don't remember) tried to be "smart" and looked at and adapted Debian's patches <janneke>possibly they were even changed, they were imported at glibc-2.31 iirc <janneke>anyway, using freshly imported patches for Debian fixes the rumpdisk panic <patrickt>Hello, I am looking to compile a software with a dependency on librt (-lrt), latest gcc-toolchain doesn't seem to include it. I have been searching in /gnu/packages/commencement.scm to see if there is a place to include it. I suspect Debian/Ubuntu includes it in the build-essential meta-package. A little guidance on how to add librt would be much appreciated. Thanks for your time! <acrow>'guix pull' fails with and undefined _libc_pthread.. error. /run/current-system/profile/bin/guix succeeds in the pull but, for reasons I don't understand, fails to install a working version to ${HOME}/.config/current/guix/bin/guix. How can I install a fresh reliable guix to my user profile? Why can I not do '/run/current-system/profile/bin/guix package -i guix'? <jpoiret>janneke: so should we remove the old ones we're carrying? <jpoiret>acrow: when does that message appear? as soon as you run guix pull? <jpoiret>patrickt: if you can influence that, you don't need `-lrt` with our glibc <jpoiret>if you can't really avoid passing `-lrt`, then you can add `glibc:static` to your shell in the meantime <jpoiret>ie. librt now has been completely integrated into libc, like many other libraries <jpoiret>acrow: can you do `env -u GUILE_LOAD_PATH -u GUILE_COMPILED_LOAD_PATH -- guix pull`? <jpoiret>somehow when guix starts it seems to load an old glibc <jpoiret>janneke: ah no, I see they're hurd-specific, sorry <jpoiret>acrow: can you do `LD_DEBUG=libs guix pull 2> log` and paste the contents of log? <patrickt>jpoiret: Thank you so much, I greatly appreciate your excellent prompt responses as always, much appreciated!! <jpoiret>acrow: how about `env -u GUILE_LOAD_PATH -u GUILE_COMPILED_LOAD_PATH -u GUILE_EXTENSIONS_PATH -- guix pull`? <jpoiret>i'm wondering how the gnutls extension is being loaded here, since it's the one which comes from the rest of your system, using glibc 2.33 thus causing the problem <acrow>jpoiret: Trying to think of anything related to that. In a past profile I had installed guile-gnutls during explorations. But it isn't in the current profile (but maybe still in the store). <acrow>jpoiret: Yeah, both guile-gnutls and gnutls (multiple versions of both) are in my store. <jpoiret>acrow: you forgot to escape the newline <jpoiret>either put everything on one line or add a \ before the newline (I think?) <jpoiret>it being in the store isn't important <jpoiret>acrow: what if you do a `/run/current-system/profile/bin/guix shell --pure -- guix pull`? <acrow>Hmm... do I need to provide a full path? <jpoiret>but there might be some other env variables missing from there as well for guix pull to work correctly, i'm not too sure <acrow>jpoiret: K, with a path a lot more debug info is provided. I'll fire it over in a moment. <jpoiret>ah. You might have to set the PATH first <jpoiret>but all those error messages are worrying <jpoiret>so `/run/current-system/.../guix shell --pure` then inside that shell, set PATH to ~/.config/guix/current/bin, then run guix pull <acrow>FWIW, here was the previous command line: /run/current-system/profile/bin/guix shell --pure -- /home/fpp/.config/guix/current/bin/guix pull >log 2>&1 <acrow>So, after some fiddling -- that is running. It's going to take a minute and then bc I needed to see what was happening it will not produce a log (but I can try again). However, watch it, I can tell you that all those worrisome messages reappeared but guix is continuing to download things and 'do the pull'. <acrow>At the end of this I expect there to be a good copy of guix in that container -- but, sadly not in a profile that I can adopt, no? <jpoiret>it's not a container, only an environment without environment variables that might interfere <acrow>Accepted. Ok, so it has successfully completed but running guix describe gets the same outcome. It's short. I'll post it again. <mirai>apteryx: I've noticed something interesting regarding utf8 <jpoiret>acrow: you might want to upgrade your system while in the pure shell <acrow>you mean system reconfigure? <johnabs[m]>Hi all! I've got a quick question. Does anyone here know how to load the julia-* packages when installed via guix rather than Pkg? <jpoiret>ah, is `guix describe` erroring out *in* the pure environment? <jpoiret>if so then that's a pretty huge problem <bjc>is there a branch git for the current hurd work? <johnabs[m]>I tried doing "using julia-qt5base-jll", for example, but it can't find it, and I'm not sure if there are any extra hoops I need to jump through to make it available <acrow>jpoiret: I'm happy to give that a shot but I'm still puzzled why I cannot just build guix note where it is in the store and then link it with .config/guix/current/bin/guix? <acrow>s/build guix note/build guix, note/ <mirai>apteryx: within a guix shell pure environment as described for “hacking guix” with 'less' included, if you do ``less gnu/packages/java-xml.scm`` the © symbol is displayed as <C2><A9> (it also prompts with "gnu/packages/java-xml.scm" may be a binary file. See it anyway?) <acrow>What do you think is in my environment that is creating this toxicity? <acrow>That core-update is the only recent activity I think I might have foolishly undertaken. <mirai>apteryx: naturally ``LESSCHARSET=utf8 less gnu/packages/java-xml.scm'' solves it but I should note that outside of the pure environment the © symbol is displayed correctly <acrow>I'll get that system reconfigure going and let people know how it goes. <acrow>jpoiret: Any tips on how to incorporate a sudo into that shell? <janneke>"<jpoiret> janneke: so should we remove the old ones we're carrying?" sure! <janneke>currently i'm reverting the old ones, and re-applying the new ones -- all exclusively for hurd <jpoiret>janneke: ah, don't worry about that, I'm testing a patch to carry a libc specifically for the Hurd <acrow>jpoiret: What I tried was: /run/current-system/profile/bin/guix shell --pure -- env PATH=/home/fpp/.config/guix/current/bin /run/setuid-programs/sudo /home/fpp/.config/guix/current/bin/guix system reconfigure ~/Sync/machines/optiplex/config.scm. It seems to have failed on making info-dir.drv .... hmmm... https://paste.debian.net/1280687/ <acrow>jpoiret: Could I switch-generations back to a profile from months back, run guix pull from the elder profile and recover the system? <acrow>jpoiret: Here is the error: install-info: /gnu/store/0i81lpfnn05pmjc5f43q4nfvd27r08f7-guile-gnutls-3.7.12/share/info/gnutls-guile.info.gz: empty file <jpoiret>acrow: this is getting weirder and weirder <jpoiret>yes, you could try rolling back the guix pull profile <bjc>guix gc: freed 211,194.66663 MiBs 👀 <bjc>finally cleaned up after all the c-u stuff =) <acrow>jpoiret: I'm also wondering if that particular guile-gnutls package ought be removed from the store. If we removed it I think the next reconfigure attempt would either gather a good substitute or build the package. Again, that's what I think but others may know better. <jpoiret>I don't think the package itself is defective <jpoiret>it's just built against an old libc but somehow gets picked up by guile <acrow>Then why did the reconfigure fail reporting that the gnutls-guile.info.gz file is missing? Hmmm... <acrow>jpoiret: When I --switch-generation=-10 (ridiculous) and run guix pull I get "Nothing to be done". So, I clearly don't understand who/what updates the instance of guix. Shouldn't the old profile contain links to an older guix instance that would then pull in an up to date substitute? <jpoiret>janneke: just got Hurd from february to boot :) <Guest28>GNU Artanis from Guix isn't working. art run returns "artanis/config.scm:549:10: In procedure parse-namespace-cookie: <Guest28>#<procedure parse-namespace-cookie (item)> "Config: Invalid item" (expires "3600")" on an empty project (art create test && cd test && art work) <anemofilia>seems the sx problem has to do with setuid-programs and stuff <anemofilia>How can I add xorg-server to the setuid-programs list?