IRC channel logs
2023-12-10.log
back to list of logs
<afm-blip>ieugen: did you set up the channel bin-guix? <PotentialUser-37>Does anyone know why the default image registries are missing from podman in guix and how to add them back? <xelxebar>Been using for a while and always have gotten a host of "error: plugin: filetype already registered: XXX" errors on startup <xelxebar>Wondering if others are seeing the same and, if so, how we can do anything about it. <xelxebar>afm-victoria: zathura is a pdf reader. If you're bored, would you mind spinning it up in a guix shell and sanity checking for me? <xelxebar>afm-victoria: Oh, right. I forgot. You need to also install zathura-XXX. For pdf, zathura-pdf-poppler should be enough. <xelxebar>Huh. Looks like zathura is hosted on pwmt.org – programs with movie titles. Go figure. <afm-victoria>It doesn't do anything--it complained about not being able to install the plugin but now it just opens a black window and then nothing <afm-victoria>Have you tried evince? It's a crappy pdf reader but does the job <\f>ACTION coughs Emacs with pdf-tools, cough, cough.. <xelxebar>afm-victoria: Yeah, zathura is like a pdf reader for vimmers. If you don't give it a filename at the command line, then you can open one by typing "o" at the window. <xelxebar>Do you see the errors I was saying? Zathura works just fine for me, too, despite spamming the CLI with errors. <afm-victoria>What's your setup? Are you on a guix binary or a foreign distro? <xelxebar>afm: Thanks for checking. I'm on a a Guix System. <xelxebar>Guess I need to track down what's leaking. <lalebarde>Hi, I have tried my first Guix install as an OS on a disk having already Windows 11. So I made room for a partition dedicated to Guix and installed it there. But when I boot, Windows is no more available. I understand that the Guix installer has overwritten the boot loader. How can I restore Windows with a dual boot please? <lalebarde>I am not proficient with grub nor Scheme nor Guix, so it is a big challenge for me <cnx>how do i call generate-*-documentation? <attila_lendvai>so, there's python-rich. it's too old for something i'd like to do. updating it is as easy as ./pre-inst-env guix refresh --update python-rich. but sending it as a commit in an email, and then someone applying it from email, and pushing it to master... is a disproportional amount work. because of that i'm not doing it, but that frustrates the cooperative idealist in me. <sneek>Welcome back attila_lendvai, you have 1 message! <attila_lendvai>M-x debbugs-gnu-guix-search "Sacrifice builders on OOM." => no match => user dismayed. <attila_lendvai>yay! applied my first patch from emacs/debbugs (pretty hopeless without a detailed documentation on the workflow, and i'm using emacs for some 15 years now, except for my emails) <jpoiret>hey guix, anyone ever use rsync over ssh to a guix system? rsync tries to run the remote rsync but doesn't go through the login shell so doesn't find it <jpoiret>i see some people have resorted to putting a hack in .zshenv to source /etc/profile when a ssh connection is detected <jpoiret>well, in the meantime i'm using the horrible --rsync-path='"$($SHELL -l -c '\''command -v rsync'\'')"' <futurile>jpoiret: set up a different user for the rsync, have that start a guix shell doing --fhs .... sounds awful as I type it :-) <attila_lendvai>nckx, thanks! FTR, what crashes my laptop is guix shell in a guile checkout, and then make. <attila_lendvai>PotentialUser-64, you need to add it into the list of services of an operating-system object. and also use-module a few things <attila_lendvai>PotentialUser-64, although udev-rules-service is defined in (gnu services base). i think it's available without any use-module <attila_lendvai>the package rtl-sdr comes from (use-modules (gnu packages radio)) <PotentialUser-64>added (use-modules (gnu packages radio)) and it seems to be doing something now. thanks. <theotherone>Hello, I'm struggling to setup bluetooth on my guix machine. From the internet I got that one has to add the bluetooth service to the system configuration and add the user to the lp group. But now I am confused as to how to connect a new device. From herd status bluetooth I get that the service is running and my user is also member of the lp group. Do I need some other programm to setup the connection? I'd really appreciate any hel <nckx>I have (service bluetooth-service-type (bluetooth-configuration (auto-enable? #t))) in my OS and have always used the bluez package's bluetoothctl to control devices (‘discoverable on’, ‘trust DEVICE’, ‘connect DEVICE’, …). <nckx>It's low-tech but the job gets done. I assume there are equivalent GUIs. <attila_lendvai>theotherone, you need some bluetooth manager app. i'm using gnome, and the default stuff that it provides works fine. <theotherone>Ah, ok. Thanks, nckx and attila_lendvai :) Then I'll look for a bluetooth manager app. <sneek>mfg was last seen in #guix 2 days ago, saying: that's a compatibility layer between oss4 and alsa, it seems. I found that via google and at least debian has a package libasound2-dev i don't know the guix equivalent of that though. <lilyp>btw. I figured out the problem with my laptop: Turns out that the Mesa driver segfaults some of the time <lilyp>What was the way to make it not do so again? I already rm'd the shader cache, but that's not helping <lalebarde>I have modified /etc/config.scm for dual boot, run guix system reconfigure on it, and it breaks with error: aborting reconfiguration because commit blabla of channel 'guix' is not a descendant of xxxx. Is it normal? Why that? <nckx>Why obfuscate the commits? <nckx>All it means is that ‘xxxx’ is not a parent to ‘blabla’, but without information nobody can tell you if it's normal or not. <lilyp>Is this a fresh installation? If so, did you run `guix pull' at least once? <nckx>It can also happen if you ‘guix install guix’ (which you oughtn't do!). <lalebarde>I did not ofuscate, it is not on the same system and was too long to rewrite. No, I did not pull. I am going to start with that, and yes, it is quite a fresh install (yesterday) <nckx>What I mean is it's normal in some situations and abnormal in others. <bienjensu>I'm trying to run an appimage with the example used on the FHS blog post. However, guix shell fails because I'm missing statfs (part of gocryptfs). Doesn't respond to adding it to the profile &c. <nckx>bienjensu: Can you (s)trace the programme to see what it's trying to exec? <nckx>attila_lendvai: Oh, then this patch won't help. You could use choom(1), maybe add it to an alias/helper/the Makefile…? <lalebarde>After `guix pull` I have run again `guix system reconfigure /etc/config.scm` and got the same error with the proposed solution to add `--allow-downgrades` <bienjensu>~/bin/appimage is a script that runs the example from the blog post <nckx>lalebarde: Did you open a new shell/session? Sounds like you're still invoking the older guix. ‘type guix’ should return /home/$you/.config/guix/current/bin/guix . <nckx>‘--allow-downgrades’ isn't a solution to anything; it's a way to knowingly downgrade your system. <nckx>If you want to do so, that's fine, but it doesn't fix anything. <nckx>Did you try ‘type guix’? <lalebarde>I have already runned `guix system reconfigure /etc/config.scm --allow-downgrades` <nckx>bienjensu: The only way I've ever ‘parsed’ strace is with ‘| grep some-remotely-interesting-string’ 😉 There's no mention of statfs, so maybe you need strace -f. <lalebarde>`type guix` returns `guix is hashed (/run/current-system/profile/bin/guix)` <nckx>Right. That's the old Guix. <nckx>If you've run ‘guix pull’, make sure you're in a fresh shell, and that ‘type guix’ returns ~/.config/guix/current/bin/guix, then run the system reconfiguration again. <nckx>That will ‘upgrade’ the system back to the latest Guix version. <bienjensu>nckx: I think I misread the error message from guix shell. Statfs is reporting an error with an empty string `guix shell: error: statfs: : No such file or directory`. <bienjensu>there's meant to be /something/ before that last colon. <nckx>Agreed that ‘guix shell: error: statfs: : No such file or directory’ is whining about "", not statfs directly. <nckx>but that's the statfs syscall, not the statfs binary. <bienjensu>That line at 66531 is calling statfs with an empty string. <nckx>Is this gocryptfs dependency something upstream told you or something you assumed? <nckx>But statfs(2), not statfs(1). <bienjensu>I'm just running the example from the guix fhs blog post. <nckx>You said above that ‘guix shell fails because I'm missing statfs (part of gocryptfs)’. <nckx>Which is sounding… implausible at the moment, and I just want to confirm who told you that. <bienjensu>I assumed that was the case but I misinterpreted the error from guix. I said so above, its the empty string no doubt. <bienjensu>I think it might just be my envs, haha. One sec. <nckx>Thanks for the URL :) Are you using the (old) VSCode version from the blog post, or (presumably) a later one? <bienjensu>My mistake, I was trying to `--expose=$XAUTHORITY` when that env isn't defined. <bienjensu>still something slightly weird with how lazily that's checked <nckx>Guix should probably warn about that, or even refuse to run (since I can't think of a valid use case). <bienjensu>`guix shell --container --emulate-fhs --expose=""` => `guix shell: error: statfs: : No such file or directory` <dissoc``>when i try to deploy to remote machine i get: failed to deploy . git- C $CHECKOUTPATH exited with status 128; output follows: error: unknown option `cruft' <nckx>bienjensu: I'm as surprised as you are. <dissoc``>its been a while since i deployed to this machine. any ideas how to fix it? should i move the files locally and try running reconfigure on the host itself? <nckx>Any idea where this ‘cruft’ string is coming from? <nckx>It doesn't seem like Guix. <nckx>(Based on a quick grep.) <dissoc``>nckx: on that checkout directory there is some guix code that rgrep has results for this cruft <nckx>bienjensu: It won't be for tonight, but I'll take a look & add a check for empty strings where this happens. Unless… you'd rather have an easy patch to hack on yourself? :) <bienjensu>nckx: I'm just trying to write one. Struggling to find where args are parsed/checked for that particular subcommand. <nckx>dissoc``: I found some unrelated phase names and a comment, nothing more. <nckx>I did use grep -w though. <nckx>Removing -w doesn't give me anything more suspicious. <nckx>bienjensu: (guix scripts environment), but it's probably better to validate later/lower down, wherever that is. <nckx>That's the thing I don't have time to dive into tonight. <nckx>specification->file-system-mapping maybe. <nckx>The problem is that Guix only ever sees --expose="", the bogus $XAUTHORITY is only seen & resolved by the shell, so there's only so much we can do to make this error friendlier. Basically, just draw more attention to its emptiness than was done by the unixy ‘statfs: :’ error, and hope the user finds the cause. <nckx>I'm starting to think we should perhaps ignore (and warn) about empty values, this would make scripting a bit more robust. Otherwise you'd have to write ${XAUTHORITY:+--expose=$XAUTHORITY}. Maybe that's OK. <bienjensu>It seems it doesn't even warn when --expose is used without --container. <dissoc``>nckx: seems like reinstalling git resolved that issue. not really sure what was going on <attila_lendvai>lechner, try to isse ,m (gnu tests reconfigure) again. make sure that geiser switched to the right module before the ,reload <attila_lendvai>lechner, i have a vague memory that it fails to change the current module when compilation fails, but does so at the second time <attila_lendvai>lechner, i think a meta-command is just something geiser prints (i think the ,reload stuff) <nckx>People of desktop-land, how again do I install org.gtk.Settings.FileChooser or its schema or whatever it is? <lilyp>Hmm, those schemas are set up by the glib-schemas profile hook, so I don't know which package has them naturally <lilyp>If you want a one-time shell, include glib so that XDG variables get set <msavoritias>following up on the xmpp proposal in the mailing list, would bridging be a requirment to be allowed to setup the xmpp server under the guix domain? <attila_lendvai>civodul, syslogd truncates long messages. would you mind if shepherd opened its own /var/log/shepherd.log file instead of writing to syslog? or maybe both? <msavoritias>i suggested hosting a server under the guix domain as i said in the email because apparently to list an xmpp room in the site it means to be official. and the official comes with requirments <attila_lendvai>civodul, how about both? set up so that debug messages only go to the shepherd.log file, the rest goes into both? <attila_lendvai>it has already confused me once, until i accidentally noticed that the entries are silently truncated <attila_lendvai>i think the current cutoff is at around 480 chars due to the use of UDP packets, or somesuch. <lalebarde>nckx Thanks, Guix updated. I have to do more work for dual boot. <nckx>msavoritias: ‘a requirement’ from whom? Guix? Libera? (I'm not sure if me representing both is a boon or a disadvantage here…) <nckx>lalebarde: If the ‘other distro’ doesn't have predictable linux/initrd file names, you can also have its GRUB load Guix's grub.cfg through the ‘configfile’ command. You don't necessarily have to dual-boot through Guix's GRUB. Just a thought. <evilsetg>Has anybody here ever tried packaging mastodon? <jpoiret>nckx: I didn't understand this snippet this way. They explicitly *want* Guix to update grub.cfg without installing GRUB i think <nckx>I'm not sure how else you understood my message. <nckx>As for the reason: my guess is Linode, like my old Scaleway server, has its own ‘out of band’ boot loader that doesn't like/need a bootable GRUB installed to storage. It just wants a grub.cfg that it can chainload. <nckx>The same trick is used for Libreboot machines that use the GRUB payload. <nckx>evilsetg: I can't answer a negative, but I've never heard of any attempts. <nckx>jpoiret: The reason I suggested this for lalebarde is that it might be more convenient to use the other distro's GRUB, because Guix's support for ‘dual boot’ (a hard-coded kernel + initrd file name) was quite limiting last I checked. <nckx>If you let Guix manage its own grub.cfg, you still get the nice generation menu. <msavoritias>nckx: guix specifically. for bridging that is. I have heard multiple responses the last few days so i wanted to be sure. <nckx>Maybe, but I'm also missing any context including your mail. So I'll read that thread first. <msavoritias>sounds good. i dont mean to put any pressure. my proposal was basically to host an xmpp server under a guix domain <nckx>My own red line would be a ‘bridge’ like I'm seen in some channels where you get this wonderful UX of <bridge_bot> msavoritas: Hello. <bridge_bot> nckx: Wat. <real-irc-user> This is annoying. <bridge_bot> nckx: Yes. <ieure>nckx, I agree, I hate those. <nckx>I think it was called choreogram? <ieure>cccccckgcuedfiliuehiritfhelhjljtuchkhbvfjvbj <ieure>(Yes, I know I can disable that, no I won't, it breaks my setup) <msavoritias>well personally i dont care about having a bridge at all <nckx>ieure: I thought for a moment it was an expression of rage until I recognised the format. <msavoritias>but if a bridge is a requirment i can see about setting up biboumi which afaik works quit well <nckx>msavoritias: There are advantages to both. Plus, I'm not dictator over the IRC channel or anything, so my opinion isn't final. <nckx>Personally, I'd *like* to avoid fragmenting the community, *as long as* the bridge doesn't cause more noise than is worth it. <nckx>Matrix was absolutely great at this until it wasn't. <msavoritias>yeah.. hence why i have gotten a lot of responses to host it ourselves instead of using a third party <msavoritias>personally i was thinking of having a biboumi and a pure xmpp room if bridging was wanted. <msavoritias>assuming i can even host the server under the guix domain that is <Kolev>msavoritias: glad to see you're using Guix. XMPP! <3 <Kolev>Wiah jabber.el supported OMEMO. <ieure>Kolev, dino seems buggy, have you tried running it under strace or ltrace to see what it's doing to cause the segfault? <Kolev>ieure: idk about those tools <ieure>Kolev, They trace syscalls (strace) and library calls (ltrace), and show you what a program is doing. It's an unsophisticated, but very reliable way to help figure out why a wide variety programs misbehave. <ieure>Essential tools for the discerning *nix hacker's. <Kolev>I will try once my Chromebook unfreezes. Compiling kernel. <pastor>Hi, I'm going to provide a patch for ImHex (https://github.com/WerWolv/ImHex). In which file should I put it? Should it be its own file? It does not have any specific dependencies that are not already present in Guix <nckx>We don't currently encourage single-package files unless said package carries arounds its own ecosystem. <nckx>Like some kind of ancient mastodon. <pastor>nckx: Isn't hexedit a package on its own? <nckx>Yes, but so are ht and bvi. <pastor>Alright. I will put it there then <Kolev>I made a package but didnt add it to Guix because licencing <nckx>(Is there any other licence statement I'm missing, Kolev?) <Kolev>nckx: So it's just flat out no license, nonfree? <nckx>If the same source was later opened, that would be fine. <ieure>Was the source that was opened the same as the source which was used? <nckx>‘the BSD copyright is pretty much the same thing’ — yes, but are you releasing your code under BSD? And BSD-what? As a potential distributor, this is frustratingly vague. <ieure>Kolev, FWIW, `guix shell dino -- dino` works for me. Do you have any dino stuff in ~/.config or ~/.cache? Could be reading config or state data which is triggering a bug. <nckx>ieure: Who knows? Hence my use of ‘same’. <Kolev>ieure: i dont want to lose my dino settings. My OMEMO keys are in there. <ieure>Kolev, Sure. But you could move those things somewhere else and see if dino starts up -- that'd isolate the problem to either the program or your configuration, which seems helpful to narrow the scope of your debugging efforts. <nckx>ieure: If the code that was opened under a BSD licence was based on X, and supdup was also based on X, we can conclude that the X authors would be OK with releasing supdup under BSD. This is both logical and the kind of clever reasoning that can get one into (very hypothetical) legal trouble. <Kolev>ieure: ah, ok. Ill try once my computer unfreezes. Building kernel. <ieure>Kolev, Honestly, it's likely a bug in dino. The suggestions you've gotten (run it under valgrind/strace/ltrace -- or even gdb) are your best paths forward for figuring out what's wrong. <ieure>Or contacting upstream to see if they have ideas -- but they're likely to give you the same advice, because a segfault with no further information isn't debuggable. <ieure>And if your config has key material in it, you can't really share that. So you're the only one who can get enough information to help figure out what's broken <nckx>I can always be wrong but without a less confus{ed,ing} licence grant I don't see this surviving a review for inclusion in Guix, sorry. <Kolev>nckx: ok. I'll keep it in my channel for myself then. <civodul>efraim: what’s the impact of objc support on closure size? <civodul>presumably due to the new search paths we now end up with the same thing multiple times on the search path <civodul>and thus include_next no longer works as expected <civodul>(“sensitive” prolly not the right word; i mean “brittle” or something like that) <civodul>there’s a typo in 2d8aa104aaf9a2e229e31d13f6cdabf57a2e2e9a: C_* and CPLUS INCLUDE_PATH are swapped <civodul>probably not the reason of this issue though <jaeme>Okay I'm updating ripgrep to version 14.0.3 and the way it generates man-pages is by invoking "rg --generate man" which will generate a man page in roff format to stdout. <jaeme>'rg' is the name of the binary fyi <attila_lendvai>jaeme, that won't trivially work when cross-compiling. you may need to conditionally add the native package as a native-input to itself, or something like that <attila_lendvai>jaeme, to call the binary that you are compiling, it needs to be able to run on the host that is building the package. when the package is being cross-compiled, then the host cannot run the target binary. <attila_lendvai>it's a similar issue with help2man, e.g. for shepherd when built from git <jaeme>attila_lendvai: hmmm, I see for 'fd' that it uses `(unless ,(%current-target-system))`, Could I use that as well? <jaeme>Also how do I redirect stdout to a file in Guile/Guix? <attila_lendvai>jaeme, you can check for (%current-target-system), but when it's not false, then you need to do something. not generating the man file will make the native and the cross-compiled package not reproducible build. <attila_lendvai>but i have to note that don't have a lot of experience in cross-compiling... <jaeme>attila_lendvai: I see, thanks for pointing that out. <jaeme>"Why did upstream have to do this to me?" <jaeme>Well I'm not going to deal with this today so I just put a TODO comment <nckx>with-output-to-file or call-with-output-file, depending on what fits best, even if it's irrelevant now. <nckx>(Agreement on upstream bein' silly willies here.) <attila_lendvai>jaeme, swi-prolog seems to be an example of how to do this (use itself as a native-input dependency when cross-compiling) <ieure>jaeme, It's a lambda which doesn't use any arguments. <ieure>jaeme, A bare _ generally means to ignore an argument. A lambda which took one argument and ignored woudd be (lambda (x _)) <jaeme>ieure: I see, I was reading the docs for with-output-to-file and it came up. <nckx>It's conventional: the _ is a regular variable name that is actually bound, and you *could* actually use it. Only don't, because that would be silly. <Kolev>Does Guile recognize the Greek letter lambda as a substitute for `lambda`? <nckx>Unfortunately for fun, it goes against Guix style guidelines. <zilti>Fortunately for fun, you can also use ligatures in your editor to at least make it look like there are actual lambdas :) <zilti>Is there a rough estimate when 1.5.0 will be out? Or whatever the next version will be? <nckx>I do so in emacs (not using ligatures; I forget how). If you want to know why my lines sometimes accidentally exceed 80 characters. <nckx>zilti: It'll be 1.5, but no target date. <jaeme>nckx: It better not be before rust-team gets merged >:( <nckx>That would be a perverse deadline, but a deadline nonetheless, hmm... >:) <jaeme>To be serious though, Guix 1.5 should have pipewire by default for audio in %desktop-services. <jaeme>pipewire just reached the mythical 1.0 release. <nckx>I'm all for pipewire & all but I'm nonplussed by the 'so you must switch it to be the default NOW why is it not the default yet it's 1.0 NOW hurry' vibe. <nckx>Not you personally, just in general. <jaeme>nckx: I get that, the 1.0 thing is an in-joke within the pipewire space because it was 0.3 for so long despite being production ready for a while. <jaeme>But yeah, Fedora/Red Hat deciding the tooling for GNU/Linux is a bit annoying <cow_2001>oh, it expects the patch file to be in the PWD <ieure>nckx, Showing fancy lambdas in Emacs is probably glasses-mode. <zamfofex>Hello, everyone! Does anyone happen to know what kind of encantation I have to perform to get PulseAudio working on a ‘guix shell --container’? I could get the program I want to run to use ALSA by exposing ‘/dev/snd’, but I want to be able to get it to use PulseAudio too.