IRC channel logs
2024-06-28.log
back to list of logs
<attila_lendvai>ACTION is so annoyed by debbugs. i've accidentally sent 2 out of 3 patches as new issues, and for several minutes nothing happened. i don't know the bug ids, i cannot close the second, unnecessary one, and i cannot send the two missing for the first one... <ieure>attila_lendvai, I hate the email patch flow. <ieure>I have been using and hacking on Free Software since the 1990s and can barely make it work. <freakingpenguin>I've learned to love it but I also irrationally hate the hassle of needing to fork a repo just to submit a few patches. <attila_lendvai>freakingpenguin, forking happens once. the hassle with debbugs is contignuous. (unless, i guess, you learn some hidden emacs something that is not documented anywhere) <freakingpenguin>If I had a large set of patches in various stages of acceptance I'd probably appreciate the forking+PR method more. <attila_lendvai>and pardon my voiced annoyance... it wouldn't be like that, if the email workflow wasn't defended sometimes verociously. <ieure>Different strokes 'n all, I try to get along, but I really really do not like the email flow. Hard to contribute, hard to review. <freakingpenguin>Haha, it definitely has it's problems, ease of use being chief among them since setups are so customized. <attila_lendvai>i've never spent so much energy on learning something that gave me back so little. the learning curve of debbugs is near flat for me. lot of effort in reading and configuring obscure stuff, and precious little gain in return. i deal with it just rarely enough to keep forgetting stuff. <attila_lendvai>i'm somewhat deligthed to report that `mumi current 71811`, and `mumi compose --close`, and sending emails from my emacs (that i configured ages ago!)... seems to have worked (hopefully, i won't see the updates for several minutes). <eikcaz>Anyone using a localy hosted email server? <eikcaz>I seem to be unable to send mail to localhost <eikcaz>I get "Mprog: A= argument required". Web search only finds this string in this irc's 2022 log <dmirth>Does anyone know if there is any problem with the Guix substitute servers? <dmirth>I ask because I'm trying to do a fresh install and it gets to the point where it attempts to connect to the Internet and I'm getting an error (across several systems) that the "selected network does not provide services to the Internet". This is despite the fact that I go into an alternate shell (ctrl-alt-F3) and confirm that I do have a valid IP and valid network connectivity <dmirth>Just thought I'd check to see if anyone else asked about this <freakingpenguin>dmirth: Is this on a enterprise-style network? There's sometimes odd behavior with TLS on those networks, see #71238. Your issue does sound a bit different though. <dmirth>Thank you very much freakingpenguin, peanuts, and jakef. Its certainly given me stuff to look at. <dmirth>freakingpenguin: did you ever get this straightened out? <Guest86>How do you declare packages to be removed? I want to get rid of a bunch of GNOME junk. <Guest86>How do you declare packages to be removed? I want to get rid of a bunch of GNOME junk. <janneke>Guest86: ...declare package to be removed? <Guest86>Like, how do you remove packages using config.scm? <PotentialUser-86>Hello. I wonder, what is the best way to work around a "temporary" breakage of a package. In my case I just use time machine for now. I could write an own package definition and put it on my channel, but I guess, the --with-input transformation can fix this as well. This commit <PotentialUser-86>Guest86: You are talking about system wide packages? guix system reconfigure should be fine and run guix gc afterwards. At least, that would be my guess <PotentialUser-86>Guest86: You might follow this thread to get an idea about profile packages and system wide packages. https://issues.guix.gnu.org/47748 - If you declared packages in your config.scm just remove it from that file and run guix system reconfigure. If you are talking about packages like wpa-supplicant that are used by your system desktop services, you <cbaines>any last changes for core-updates-new before I push it as core-updates? <ngz>(I sneaked in one already) <sneek>civodul, you have 1 message! <sneek>civodul, dsmith says: Might want to update the topic for 3.0.10 <cbaines>cool, I'll go ahead and push as core-updates then <cbaines>civodul, did you see my email regarding Cuirass and PostgreSQL? It seems like a config change could get things working again. <civodul>cbaines: i did! i started looking into it but i have a hard time focusing… <civodul>basically selecting jobs first, because there are many more jobs than evaluations <hungry_ghost>it's a bit confusing that guix-service-type and unattended-upgrade-service-type takes different (channels ...) argument <hungry_ghost>for the latter I had to use #~(load "/etc/guix/channels.scm) <dariqq>i built something on one machine and wanted to export it to a slower one with guix archive --export/--import. But it was only recognised on the recipient if i exported the --no-graft output and redoing the grafting on the other machine. Is that expected? <civodul>dariqq: ‘guix archive’ is pretty low-level, perhaps you should use ‘guix copy’? <dariqq>civodul: thanks will look into it. I have not set up ssh yet and have been transferring the nars manually so far <dariqq>probably should also look into setting the fast machine to serve substitutes/as a build-machine but had networking issues (the fast machine is running in a vm and it can access things on my local network but not the other way round ) <civodul>oh yes, ‘guix publish’ is another option <dariqq>guix publish works now. I (just) had to forward the guix-publish port <freakingpenguin>dmirth: Never solved it consistently, it just comes and goes. No clue on the cause. <redacted>I'm building a home service, but I'm not sure how I'll test it during development. Is there something analgous to `guix shell -f <filename>` for services? <redacted>Will I need to modify guix's load path somehow to (use-modules (my custom service))? <lechner>redacted / not if you use your custom channel <lechner>Hi, has anyone had issues with 'guix publish' lately? I believe it stopped working locally <dariqq>lechner: I am trying to set it up now. the weather says the substitute is available but it cant be downloaded <dariqq>on a system running an older guix i get a connection failed: Permission denied warning <lechner>i have been getting connection refused errors <dariqq>on a foreign system, daemon from 425cf1fbe2fff25bda1b5eb948ce01170d23ab6a <dariqq>on a system i reconfigured today i dont eve nget that error <lechner>btw, this bot message was useless because FSF hasn't update Cgit in eight years. newer versions send the commit message in the html title <dariqq>yes it just silently fails and tries to build <dariqq>i.e. it does not fetch the substitute and builds locally instead <dariqq>just tring to set it up today because building browsers on an old thinkpad is not fun <lechner>dariqq / please also consider 'guix deploy' from your threadripper <dariqq>i have avahi, and speciy --substitute-url, the key is authorized <lechner>ok, i'll be back later. please post to -devel if you figure it out. i have to reboot a server between you and i to figure out something with a development version of the shepherd <dariqq>hmm, maybe i am not getting the error message on the laptop because of the prot forwarding <redacted>`guix home container -L$HOME/guix-channel ~/config.scm` tells me that my service type is an unbound variable. But it recommends adding the package to `use-modules` (so it knows twhere the package is), and I've got the exact `use-module` invocation that it recommends. <redacted>The docs make it sound like -L is for *packages*, because it adds the directory to the GUIX_PACKAGE_PATH. It's less clear if that's the load path for services too. <redacted>Are there existing home services that depend on system services? <ieure>redacted, I think -L is a fairly generic switch used to add to the -load path, and works with pretty much any Guix command. <redacted>ieure, yeah, I think my problem is deeper than that switch. I may be misunderstanding the module system somehow. <freakingpenguin>-L does add to load path, it just also evaluates the files to search for package definitions. In most cases that's not a problem. <freakingpenguin>When you see an error like "unbound variable blah blah blah" that can often indicate there's an error in a module imported by (my custom service). <redacted>freakingpenguin, turns out there were errors in my service definition <redacted>I have now learned to do `guix repl -- my-service-file.scm` before trying to use the service <freakingpenguin>guix repl foo.scm seems to generally handle reporting errors better. <dthompson>it's slow in the sense that I'm still not rich ;) <aldum>it would still be something if it was foolproof <dthompson>"I gotta figure out how to make money on this thing. It's simply *too good*." <freakingpenguin>$ guix develop + distributed shepherd sounds fun. I'll finally be even more annoying to my Kubernetes friend and say "okay that's cool but Guix can do that too you know" <dariqq>lechner: I am not sure what the problem was Tried a bunch of things and i am now back where I started it works. <dariqq>the permission-denied was an selinux error for me <JetpackJackson>im getting a strange error trying to package arduino-cli, "error: go-google-golang-org-grpc@1.64.0: dependency cycle detected: <JetpackJackson> go-google-golang-org-grpc@1.64.0 -> go-github-com-envoyproxy-go-control-plane@0.12.0 -> go-google-golang-org-grpc@1.64.0 -> go-github-com-arduino-arduino-cli@1.0.1" <freakingpenguin>JetpackJackson: Not sure of how to solve it, but envoyproxy-go-control-plane depends on org-grpc which depends on control-plane which ... <freakingpenguin>I assume this was from guix import go, so maybe it's a version thing? Perhaps --pin-versions would help. <freakingpenguin>Anyone know if there's a way to view /just/ the current locks held by guix processes? guix processes spits out more info than I need so it's hard to parse, I just want to see what build jobs are running on a remote system. <JetpackJackson>hmm after a while it dies, spits out error stuff, and says "Git error: reference 'refs/tags/v0.2.3' not found" <freakingpenguin>JetpackJackson: I encountered that before and I think it was because the remote repo used 0.2.3 instead of v0.2.3 for the tag. Not sure of how to solve that. <ieure>I had a job where I wrote Go daily for 2.5 years. I do not think much of the language. <ieure>It's the finest programming language of 1983, problem is, it came out in 2009. <JetpackJackson>honestly at this point i could write a script that just downloads the binary into my project directory or smth <freakingpenguin>They release tarballs containing the compiled executable right? Maybe write a package for that in a custom channel or contribute it to one. <freakingpenguin>I'm not sure why some go projects provide compiled executable releases but won't provide a vendor directory in releases because of fears of supply chain attacks. <ieure>JetpackJackson, You'd create your own channel and add it to the set of channels Guix pulls from. <freakingpenguin>JetpackJackson: Any scheme file that returns a list of channels will work. I'm partial to (cons* (channel ...) (channel ...) ... %default-channels). <bigbookofbug>still having difficulty getting ft2build.h being recognized but rather baffled by the error itself, as its originating from one of the dependencies (libxft), rather than the package itself <bigbookofbug>if im understanding correctly, shouldn't libxft be able to resolve its dependencies internally? including freetype as an input does not seem to resolve the overall error, since the error is on account of a dependency of a dependency <bigbookofbug>i tried looking at the package code for libxft but wasn't able to find anything out of the ordinary in it, and it builds standalone, just not as a dependency when packaging warpd for x <freakingpenguin>I'm finding SDDM fails to start in QEMU when using guix system vm. /var/log/Xorg.0.log complains about "No devices detected" and all I have is a TTY. Sound familiar to anyone? <bigbookofbug>freakingpenguin: ive run into a few issues w/ sddm but not that one specifically <JetpackJackson>is there a guide for learning how to use pgp specifically for channel making <bigbookofbug>i struggled a bit with pinentry though, and had to set the pinentry path in my guix config as well or else it would just hang forever <bigbookofbug>i think it really comes down to generating a key, making a branch for your public key, and then including the fingerprint algonside an introductory commit in your channel specification <JetpackJackson>wrt pgp: running "gpg --export --armor --output JetpackJackson.key JetpackJackson" should export the public key right? which is safe to put in the repo? <ieure>--export-secret-key is what exports the secret key. <ieure>You can also look at the output to confirm, the ASCII armor means it'll have a --- BEGIN PGP PUBLIC KEY --- header (which will be "PRIVATE KEY" for, you know) <apteryx>ugh, how do I use 'bbdb-create' with an empty address? Ctrl+J doesn't allow me to skip the address fields anymore. <ieure>apteryx, Are you using company/vertico or similar? Those make it difficult to select an option not in the autocomplete list, M-RET will let you pick something else (incl. a blank string). <freakingpenguin>I think ivy is C-M-j is ivy's shortcut to use current input if M-RET doesn't work. <apteryx>ah, yes, probably C-M-j from ivy (ivy-immediate-done). <apteryx>I keep forgetting about that one, thanks <freakingpenguin>I haven't touched ivy in a year but the sheer oddness of that binding must have made an impression. <dariqq>I've been having issues with emacs and gnome+wayland were areas are not responding to mouse clicks. Is there a wayland version for emacs? <freakingpenguin>dariqq: I believe emacs-pgtk (pure gtk) is capable of running emacs on native wayland w/o x11 <gnucode>Well define-configuration is actually pretty cool. I've been having some fun playing with it the last few days. <dariqq>freakingpenguin: Thanks, that seems to work. Curious why this is only an issue now as i have been using wayland and the normal emacs for months now <JetpackJackson>drat i signed the commit that pushed the key but im comparing the way the repo looks to one that works and it doesnt look the same <freakingpenguin>There was an update to 29.4 5 days ago, but it was done as a graft on emacs-minimal. My guess is it's either a change in emacs from the 29.4 upgrade or something weird with grafts. <dmirth>I have a quick install question: am I correct that its not possible to do a manual guix installation and partition a drive with LVM? <dmirth>at least with the default install iso <dariqq>freakingpenguin: the ungrafted emacs works fine. Either a problem with grafts or emacs@29.4 <freakingpenguin>dariqq: If you wind up determining it's a guix bug from grafting probably worth a bug report. I'd be surprised if it was a 29.4 issue since the NEWS file only lists a small change to org-mode. <dariqq>cant reproduce the bug now anymore. Second time that happended today <Deltafire>anyone know what this error means, at the end of unattended-upgrades.log: "guix system: error: no matching generation" <dgr>if I guix system init I get a error building -info-dir.drv with the gz stating wget.info.gz being a empty file - how to debug? <freakingpenguin>Oh dear, StumpWM is failing to start and complaining about a memory fault, seemingly as a result of upgrading from e1a66d2ddd to ae55410e07. <lispmacs[work]>hi, do we have a program in some repo package that will remove ansi escape codes from input? I know I've heard of such programs but I can't remember any names <JetpackJackson>this is awkward, ive set up everything in my repo multiple times and i still get failed guix git authenticates and failed guix pulls <freakingpenguin>Hmm, it's not StumpWM itself that's the problem. For some reason it's throwing memory errors when loading any part of my (unchanged) configuration. <vagrantc>whoah. it's been about two years since i've seriously poked at reproducible builds issues in guix ... <apoorv569>Just reconfigured my system and home config and emacs fails to load, I get this error, `Cannot open load file No such file or directory /gnu/store/mfyb9kzkikmmbfi3dfin0panmlbarnb0-emacs-erc-image-0-3.82fb387/share/emacs/site-lisp/erc-image-0-3.82fb387/erc-image-autoloads` <apoorv569>If I remove this package I get similar for another package.. <freakingpenguin>For testing, try building without grafts and see if the error continues. The 29.4 upgrade was done as a graft but seems to cause some issues. <freakingpenguin>Sorry, reconfiguring. Just know that means you'll be missing various security fixes. I'm also guessing it's a graft issue, but it could be something totally different. <apoorv569>Oh! this is apply to all packages I am guessing.. should I try this in a `guix shell` instead? <freakingpenguin>JetpackJackson: The issue is in your .guix-authorization file. It's missing ending parenthesis. <freakingpenguin>Probably a bug that the error message doesn't say it's a file in the repo that's the problem and not the channels.scm file you passed. <apoorv569>Tried `guix shell --no-grafts emacs`, I get same error still. <apoorv569>OK, I get different error now, `Debugger entered--Lisp error: (void-variable for)` <apoorv569>not sure where its getting this `for` variable from.. I don't have anything named `for` in my config. <apoorv569>No.. doesn't look like it.. probably some package is causing that. <JetpackJackson>ive fixed my .guix-authorizations file but i still get "#<unknown port>:5:1: unexpected end of input while searching for: )" when i sudo guix pull <ieure>Why are you using sudo for guix pull? <JetpackJackson>foriegn distro, and normal guix pull cant get my channel for some reason <ieure>JetpackJackson, Guessing your SSH key is unreadable by your user. I think the error is a failure to parse the channel information in the remote repo -- probably .guix-channel or .guix-authorzations is missing a closing paren. <freakingpenguin>JetpackJackson: Is your introduction still starting at commit 2970fc3864? <freakingpenguin>It would use the .guix-authorizations starting from that commit even though you fixed the file later. <freakingpenguin>Option one: make sure you use 139d7d01 or later in the channel introduction. Option two: squash/fixup your .guix-authorization ) fix into the introduction (although that'll still change the commit hash) <ieure>JetpackJackson, Probably the simplest thing is to `git clone' the repo URI in your Guix channels. <ieure>The `git clone' fails? Or the clone works, but the pull still fails? <ieure>I swear, this channel has more nicks that are just English words than any other channel in history. <ieure>Weird! I think the Guix repos get checked out into ~/.cache/guix/checkouts (or something real similar) -- do those look good? Can your user read them? <ieure>Could be that if your initial `guix pull' was as root, they got created owned by root and your user can't read them. <ieure>Generally, `guix pull' as a user, and only escalate for `sudo guix system reconfigure' is the correct approach.