<nckx>You can click through on each successive page, too. <nckx>This service, it sure has a lot of data. <nckx>Maybe because it no longer runs Cuirass and that link assumes it does? I'm just guessing by now. But the answer is still: not enough outputs to be able to say they match anything, and that's what you asked for with output_consistency=matching. <nckx><Maybe because> No, that's not it. I don't know then. Seems to be a bug in the GDS. <nckx>Sorry: ‘Guix Data Service’ — the thing you're using :) <nckx>The data. in the .guix.gnu.org. <nckx>It's run and maintained by cbaines. <pashencija[m]>Is there any sense for me to contact russian companies that generally mirror repos (like yandex or universities) asking if they want to setup a mirror? How much resources would it require to mirror channels/substitutions without rebuilding anything? <bjc>the good news is i'm getting substitutes. the bad news is: git-minimal-2.36.1 39KiB/s <bjc>ci.guix.gnu.org needs a good nap <cbaines>bjc, have you tried bordeaux.guix.gnu.org? <bjc>i have not. how do you specify it? <cbaines>are you using Guix on a foreign distro, or guix as a system? <vagrantc>unless you've configured guix-daemon substitute servers... <cbaines>if you're using a recent enough guix, as vagrant says, the guix-configuration should include it by default, both in the substitute urls and authorized keys <bjc>i /thought/ i'd seen bordeaux being used before. maybe the issue is that it was found on ci, so it's just using it <bjc>can i disable ci on the command line? <cbaines>my home internet connection is so slow, that they behave the same... <bjc>i'm having the same issue with bordeaux. this must be something on my end =/ <vagrantc>well, great news that there's a US mirror, anyways. <cbaines>vagrantc, it's a caching reverse proxy (with only 10GiB of cache), but it might help, I'm interested to find out anyway <vagrantc>i use a caching proxy at on the local network ... be curious how well the caching proxy of a caching proxy works :) <vagrantc>i vaguely recall getting the documentation for how to set up nginx to do this for guix somewhere, but can't recall where... <cbaines>the mirror responds to narinfo requests locally, which means there might be an improvement in latency for the operations that require lots of roundtrips <vagrantc>of course, i suspect "guix build guix" is still broken on aarch64 ... i haven't been doing much with x86_64 lately <cbaines>hmm, I reconfigured an aarch64-linux machine today, and I think that would have involved building guix <vagrantc>oh, did someone quietly fix it without closing the bug? <cbaines>so, the issue might have been fixed \o/ <cbaines>vagrantc, I think efraim closed it when the fix was pushed, but the guix package wasn't updated, so the bug got re-opened <vagrantc>though i'm surprised that it wouldn't also fail on all other unspecified platforms... <lechner>Hi, does anyone have a working nfs4 mount on a Guix client? <bjc>lechner: i have guix mounting nfsv4 currently <bdju>I've installed rakudo and moarvm but I don't seem to have the `raku` repl for some reason. anyone know what I'm missing here? I already had rakudo in my manifest from over a year ago and I thought it used to work <lechner>bjc: with rpc.gssd or with gssproxy? <bjc>i don't see either of them in ps <bjc>ditto:/var/lib/machines/guix on /mnt type nfs4 (rw,relatime,vers=4.2,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,timeo=600,retrans=2,sec=sys,clientaddr=192.168.1.251,local_lock=none,addr=192.168.1.5) <bjc>i didn't pass any options, just a straight "mount ditto:/var/lib/machines /mnt" <lechner>bjc: can you see the exports on the server? <bjc>ah, right, my exports are managed by zfs <bjc>everything's weird =) <bjc>dunno if it'll help, but here's the zfs options for it: babar/nspawn/images/guix sharenfs no_root_squash,rw=@192.168.1.0/24 local <bjc>nothing really special there, either <lechner>bjc: bingo! that's working. do you operate a KDC? <bjc>it's all very insecure, but it's lan-only <bjc>i was wondering about the gss stuff. i only remember seeing that from krbv <lechner>bjc: i cannot currently start gssproxy on the client, which is my favorite way. it replaces both rpc.gssd and idmap but is not well publicized. as my guile skills grow, i hope to contribute a solution <bjc>i'm just happy to see there's still an interest in kerberos! it was always pretty niche and just seemed to only get more niche as time went on <bjc>i guess activedirectory still uses it <lechner>what is people's favorite alternative? <bjc>i don't know that there is a decent one, tbh. but, at least in unix-land, everywhere i've worked has just embraced tofu ssh <lechner>are the public ssh keys propagated automatically? <bjc>i did have one place that did it, but only for certain machines. they eventually killed it when going to a stateless puppet deploy system. wasn't considered worth saving <lechner>bjc: while simple by itself, kerberos can be difficult to integrate (primarily due to PAM). for a while, my favorite way to deploy it was sssd, which is not yet in Guix. did your employers manage passwords centrally? <bjc>depends on what you mean. most places managed dozens of passwords centrally <bjc>sso for every system. truly a wonderland <bjc>i used to maintain kerberos v patches for sshd. it was not a fun task <bjc>this was long enough ago that, iirc, it was for tatu's original sshd, not openssh. it didn't have gssapi support at the time <lechner>yeah, aside from all the terrific work in Guix group authentication is a way a distro could make a true difference. i'll eventually get an MIT krb server working in Guix, and will then contribute <bjc>good luck! fwiw istr heimdal had become preferred <bjc>i don't have an opinion, really. past my time working intimately with kerberos. i *do* remember the mit code being impossible to understand, though =) <lechner>bjc: i don't mind switching. the tools are a bit nicer. i just got used to the warts, and upstreams take MIT-related questions more serious <lechner>as far as I can tell, however, neither is currently in Guix, right? <bjc>and mit-krb5, seemingly <bjc>and something called shishi that i've never heard of before <lechner>bjc: that's great! then i can convert my last machine to guix. shishi is incomplete <lechner>Hi, does anyone have a good way in Guix to get from gdm into sway? <horizoninnovatio>good morning. I have a couple of apps (calibre & qutebrowser) pulling up the similar error: Failed to import PyQt module: PyQt5.QtCore with error: libgthread-2.0.so.0: cannot open shared object file: No such file or directory - appears that pyqt5.qtcore is not being seen and is installed - any ideas for a solution? Thank you. <bdju>lechner: I ended up removing gdm and I just run `exec sway` in a TTY after logging in. not sure why gdm doesn't work right on guix or why it's the default despite that. <bdju>I used to use some strange hack I didn't even understand to just remove gdm from %desktop-services%, and then on a later install I just changed to using %base-packages% and building up from there <jackhill>bdju, vagrantc: If the story is that gdm doesn't work right, I have the same story here (and use `exec dbus-run-session sway`). But it's only on one of my computers, which is whacky, and why I haven't tracked it down. GDM woks everywhere else. On the problematic computer, sway also doesn't like it if the monitors go to sleep. I have to run it with WLR_DRM_NO_MODIFIERS=1 otherwise if a monitor is powered <jackhill>off, it will never come back on, and even with that sway often looses track of windows when the monitors go off and on. Sometimes it will even crash. It's wacky, especially since this isn't even that exotic of a computer, ~2017 Intel branded small form factor thing. <jackhill>lechner: on my other computers, I just install sway in the system profile and GDM picks it up. GDM does have to be running in wayland-mode to launch other wayland things, and GDM will (silently?) fallback to X if it thinks its wayland session didn't start, so that's something to look out for. <bdju>I run elogind, that may or may not affect sleep and such. I have it disable sleep on lid close at least <ennoausberlin>Hello. These days I played around with guix lint and while I like the idea, I got somewhat surprisingly a message from Software Heritage, that my private gitlab url was stored there with username and access token (read only luckily). Reason is, that my package definition uses https git urls with access token as password. Software Heritage promised <ennoausberlin>to take action, that this will cleaned up on their site. So far so good. But is it possible to not contact Software Heritage at all when the repo is set to private? <ennoausberlin>BTW: I do now use git://mygitlab urls from now on. But maybe others might be affected as well <jpoiret>lechner, bdju: did you enable wayland for gdm-service-type? <jpoiret>Wayland sessions won't get picked up by a gdm running in X mode <jpoiret>I'm not sure we switched to wayland mode by default yet, but the wayland support was added ~6 months ago so now could be the time <bdju>kinda gave up on gdm years ago, not gonna mess with it anymore <rekado>I had a couple too many issues with wayland so I switched back to X <rekado>apteryx: node 130 is now connected to the public switch, so we could assign the second public IP to it. <ennoausberlin>Hello. I again struggle with git-reference url. I want to reference my private repo in a package definition. On the command-line I can run: git clone git@gitlab.kokyou.de:enno/signalstatuslist.git <ennoausberlin>I am not asked for user name password so ssh is configured correctly. How does the url have to look like for this repo in a package definition <ennoausberlin>but if I try to install from my channel definition I get fatal: unable to connect to gitlab.kokyou.de: <PurpleSym>ennoausberlin: Afaik you cannot use private repositories in package definitions. The build environment has no access to your SSH configuration. <ennoausberlin>PurpleSym Bummer. How does people manage private packages then? <PurpleSym>You can check it out locally and use --with-source for example, ennoausberlin. <ennoausberlin>PurpleSym Can you elaborate this? Can this be inside a channel or does it mean I have to install package by hand than? <PurpleSym>Yeah, by hand. You could also use the `local-file` source instead of referencing git. <PurpleSym>Like so: `(source (local-file %source-dir #:recursive? #t))` <ennoausberlin>PurpleSym I have to think about it. Can the credentials somehow obtained by the build environment. I'd rather not using the local-file way ***wielaard is now known as mjw
<attila_lendvai>e-book reader (from Calibre?) displays a white page. i can kinda select text, but nothing is displayed, IOW, probably a font issue. any hints/ideas? home service might be involved. <PurpleSym>ennoausberlin: Everything in /gnu/store is public and build environments are sandboxed. I don’t think that’s possible. <f1refly>how can I rfkill bluetooth on boot? It consumes about 3W on my machine and I never use bluetooth anyways so I'd like to disable it <bricewge>Looks like it will be live recorded, for anyone interested it's at 17:15 CEST today <civodul>i think it's well worth watching (esp. for those who speak French or wish to practice :-)) <f1refly>I'm afraid my french is too poor for that <cbaines>civodul, regarding #55335, I'm fine going with switching to AF_INET6 if you are. Shall I go ahead and push the change? <civodul>cbaines: yes please! you can add a comment linking to the bug report <civodul>cbaines: ooh, that's right, or shepherd may crash on the first connection <civodul>so maybe we'll have to skip this step and directly go with the full-blown shepherd fixes i posted? <cbaines>I've only tested connecting from localhost, which does work, but it looks like jackhill has done more rigerous testing <cbaines>well, I'm just talking about the testing to find that bug <cbaines>I'll have a go at reconfiguring a system using the newer shepherd and see how I get on <bdju>I hit a blender build failure last night but the log seems to be too big to paste to ix.io <fiesh>yay me too a couple days ago <nckx>tail -n 1000 probably suffices. Or use the bug tracker, if ʿbigʾ is not many megabytes. <Guest55>What's the Guix way to set an env variable for all user processes (not just those started from an interactive shell)? .profile doesn't get sourced on login. <attila_lendvai>civodul, sorry about the expiration on the test pgp keys! i didn't think of that... <bjc>bdju, lechner, jackhill: i can confirm that toggling ‘wayland’ to #t in gdm-configuration has it picking up the sway desktop entry, so you can start it directly from the greeter <pashencija[m]>Let's suppose I have edited guix/build/gnu-build-system.scm on my channel. How do I apply that in the system? Is there some commit ID to be bumped? <yarl>I don't know what is Ludovic Courtes nickname, is he here? <yarl>Also, If anyone here wants to check? <bjc>yarl, can you submit the patch to 55324@debbugs.gnu.org? <bjc>that way it can be more easily tracked <bjc>fwiw, if you haven't set it up yet, ‘git send-email’ will always format things correctly for you <civodul>yarl: yup, i replied and pushed yesterday, thanks! :-) <yarl>I haven't and won't right now :( but I definetely will :) <civodul>oh actually no, the message was still in my draft folder <attila_lendvai>make check-system fails on master (jami service, "One does not simply initialize the client: Another daemon is detected"; actual error is coming from start: args: ("match" "no matching pattern" #f)) <attila_lendvai>...which is a pitty, because i have a patch that i would like to check not to break anything, and it involves jami changes. <unmatched-paren>i.e. if i have (package/inherit linux-libre) where do i go from there? <apteryx>rekado: re node 130, noted. I'll try to see how that's done (static network service config?) <pashencija[m]><unmatched-paren> "how would i go about creating..." <- For x86? <unmatched-paren>i know there's some interactive tool for configuring the kernel in its repo <GNUtoo>Hi, when using guix pull, where does the space is consumed? there is probably some in .cache/guix but is there space taken somewhere else? <pashencija[m]>And the suggested fix works on x86_64 machine, but fails aarch64 <unmatched-paren>actually, what _are_ the benefits of a custom kernel, aside from the obvious learning experience? some desktop users seem to actually use one as a daily driver, so there must be some advantage, right...? <bjc>if you have specialized needs, you need a special kernel. but for almost everyone, i'd say it's not necessary <bjc>i've had to build one for some newish hardware a couple years ago, but before then i hadn't built one since the early 2000s <jackhill>unmatched-paren: if you want to build locally instead of using a substittue, than not building drivers for all the hardware you don't have can be a real boost. Some people seem to like to build drivers they know they'll need into the kernel rather than as modules which might make it easier for them to manage what needs to go in the initramfs <GNUtoo>I tried with rm -rf and I get "Read-only file system", though I can delete other files not in /gnu <GNUtoo>ah I probably need to find how to remove the sticky bit <GNUtoo>no, to move /gnu to another partition, the mv worked but it didn't delete the old files *unmatched-paren technically wouldn't recommend uninstalling guix either :P <GNUtoo>for uninstalling I've a script but that's for another use case <unmatched-paren>GNUtoo: ah, i see. not sure whether that's safe, though it's certainly possible... i think there's some option somewhere to mount the /gnu/store rw... <GNUtoo>umount store worked so maybe I can delete it now *unmatched-paren looking through the manual <GNUtoo>I didn't figure out that something was mounted on top of store/ to make it read-only <GNUtoo>The bacground issue is that I want to use Guix on a device with not a lot of space, so it crashes and I want to see what happens if I give it more space <GNUtoo>If I can make it work it'll probably save me a lot of time for Replicant <GNUtoo>It's a long story, that phone is used for testing modem related code <GNUtoo>*The OS I use is Parabola and Guix runs on top because valgrind and gdb are broken on parabola armv7h or at least were broken when I started doing automatic tests <GNUtoo>And I also use a guix.scm to do automatic tests so it's not a bad idea to be able to also run that on the phone <GNUtoo>Once that's stabilized and so on I'll merge the code in Replicant <GNUtoo>That could also be used in GNU/Linux now that we have (deblobed) kernels that are based on upstream Linux <GNUtoo>Though it would require to hook the low level library (libsamsung-ipc) with a higher level daemon like oFono <omlet[m]>is there no other immutable gnu distro beyond the guix? <omlet[m]><unmatched-paren> "omlet: no, not at this point..." <- More gnu immutable distros <lechner>bdju bjc jpoiret jackhill: thank you for your replies! i will try that very shortly and report back <lechner>Hi, what's the preferred way, if any, to sign GPG keys in Guix, please? (a tool like pius, for example) <luishgh>hi guix, what does the cargo-build-system #:skip-build argument mean? <unmatched-paren>luishgh: i think it's for crates that don't get built on linux but need to exist to satisfy cargo's checksumming <unmatched-paren>but if it's not given in the inputs cargo will complain about a missing package <unmatched-paren>sneek: later tell maximed: will #:skip-builds be unnecessary in antioxidant-build-system? <luishgh>unmatched-paren: hm, makes sense. Thanks! *unmatched-paren thinks individual builds systems should be documented in depth <luishgh>currently trying to package helix, rust in guix is really weird compared to other packages lol <luishgh>guix is almost making me dislike rust <luishgh>unmatched-paren: thanks! i'll take a look at it, this channel seems really cool <unmatched-paren>luishgh: rust has excellent ideas at its core, but there are serious issues, especially with cargo <littlebobeep>Do you still have to compile Rust compiler like 7 or 8 times to bootstrap? When I looked into it, it seemed unlikely that every step in this process was reproducable... <luishgh>someone should do an unofficial list of cool guix channels <unmatched-paren>luishgh: btw, my helix package has an issue, which is detailed in the comment above its definition <littlebobeep>unmatched-paren: people in #javascript say that npm is the best package manager in existence, which is proven because it is the most popular package manager :D <unmatched-paren>i mean "it poplar there4 it gud" could also be applied to javascript itself <luishgh`>unmatched-paren: ignore my latest question, my network failed. found your answer in the logs <littlebobeep>I haven't studied npm enough to have a technically valid response, but I found a lot of links indicating problems with npm and they just said "too much to read" <unmatched-paren>littlebobeep: i believe all the supply chain attacks alone justify the criticism <jackhill>unmatched-paren: am I to understand that tree-sitter needs both rust *and* javascript‽ <jackhill>wow, what a mess. Thanks for working on it! <unmatched-paren>but someone else merged a different issue with it (yours, luishgh`?) and i removed it from guixrus <luishgh`>yep, i added it in my patch series to update neovim to 0.6.1 <nckx>lechner: You obviously want a different answer, but my preferred method is gnupg… on any distro. <unmatched-paren>luishgh`: it would be greatly appreciated if you sent the rust-tree-sitter and tree-sitter-cli packages to mainline guix <lechner>nckx: thanks, gnupg is always involved. it's about automating the distribution of the detached signatures. pius, for example, sends separate signatures in encrypted emails to each uid on the key. <luishgh`>that's great. i'll only check if this collides with this [1] issue and work on it right away! <unmatched-paren>nvm, you'll be fine just waiting for that (/begging someone to merge it after reviewing it yourself :) <luishgh`>unmatched-paren: you're probably right, i'll check your helix package and try to port it to guix then <nckx>lechner: Oh, for parties and the like? pius *is* in Guix by the way. <lechner>bjc jpoiret: Hi, is the wayland toggle documented anywhere? <bjc>yeah, it's in the entry for the gdm-configuration, and referenced from the desktop-services part of the manual <bjc>i only checked the devel manual, so maybe the stable one is out-of-date <djeis>Has guix ever had a package for openjdk-8? <nckx>dgcampea: Thanks for pointing that out. The second entry should no, I think, be there? <nckx>Which is weird because it was added *later* than the more complete-looking entry above it. <dgcampea>I'm not sure, loginctl shows the greeter session as having Type=x11 after setting display-server to wayland and logging into gnome(wayland) <nckx>That doesn't sound related. The manual is a bit confused, but the code is OK. <nckx>And I use (display-server "wayland") without issue. <nckx>Which loginctl command is that? <dgcampea>'loginctl' to see what sessions are around <dgcampea>one of them should have Name=sddm / Service=sddm-greeter <nckx>Thanks. I've never used loginctl. <nckx>Did you set the desktop file to use? (No idea what it should be for GNOME.) <dgcampea>ah, I assume you have a (auto-login-user ...) as well in sddm-configuration ? <nckx>I'd be surprised if that's somehow needed for wayland to work, but who knows, I've always had it. <dgcampea>I'm using sddm to choose which DE to start <dgcampea>the odd thing is that the sddm session (USER=sddm) starts with x11 even after setting (display-server "wayland") <nckx>dgcampea: I don't know… Here's my configuration <https://paste.debian.net/plainh/f4d4e291>, maybe you could try that, see if it works, if it does that points to a bug in the SDDM service (I guess). If it doesn't… uhm, it might still point to one, but a far more confusing one. <nckx>(Using gnome.desktop or whatever it's called; I don't suggest you migrate to Sway :) ***w1gz_ is now known as w1gz
<lechner>bjc: i just use (service gnome-desktop-service-type) Would you please post the gdm part of your config? <bjc>i'm doing it through ‘modify-services %desktop-services’, since gdm is in there by default <bjc>i posted it earlier. lemme see if i can find <lechner>bjc: thanks so much! sorry i missed the earlier post. i am on a reduced terminal. should wayland be enabled by default nowadays? <bjc>no problem! as for default, you're asking the wrong person. i've been using it exclusively for months, and it seems fine. but for the last few years i would try it after hearing people say "i use it exclusively and it seems fine" only to have it be a disaster for me <nckx>Does building emacs-deferred ever finish? <dlowe>hm. is there something I need to install to use u2f with yubikeys <vagrantc>lechner: i think there is a nix importer, but not sure of the quality <civodul>vagrantc: it's gone! it had bitrotted and wasn't very useful <civodul>lechner: however you can try: guix import go github.com/rfjakob/gocryptfs <vagrantc>i wonder with a fairly standardized debian package (e.g. debhelper with dh debian/rules) how hard it would be... <civodul>i suspect it'd be similar: might work well for simple cases, but simple cases are simple <civodul>more complex cases with custom scriptery wouldn't be handled <lechner>vagrantc civodul: Could Guix potentially subsume Dpkg by presenting a Debian unstable file tree in Guix? <lechner>nvm, that wouldn't work because of their preference for absolute paths <vagrantc>guix also prefers absolute paths, but coinstallable ones :) ***jess is now known as jess`
***jess` is now known as jess
<tricon>jess: welcome back. some imposter was posing as you. <nckx>£s tricon back where they belong. *tricon ducks back in his Hobbit hole <civodul>we're back to the level of 2013 or something <tricon>civodul: too much drinking for Libera's birthday, i bet.