IRC channel logs
2025-06-17.log
back to list of logs
<RavenJoad>Where should new contributions be sent? I have like 5-6 tree-sitter libraries I want to add to Guix. <RavenJoad>Also, one of my patches, 78434, has been sitting for a while and required updating some lief (which is a pretty big library). How can I build packages which depend on that change? <RavenJoad>ieure: Ok, so Codeberg and PRs are the way to do things now? Sounds good. <RavenJoad>Ok. I still have 2 outstanding issue emails on debbugs that I want to close before the full migration date. #77448 and #78434. Could people look over those? I would like to close all of my things on debbugs before switching to Codeberg stuff. <livia>Does anyone of you remember there being a conference video of Guix being used directly in the guile repl for modifying package definitions, installing packages, etc? I found it neat because it showcased the implications of using a general purpose programming language for configuration. <livia>Unless I dreamed it up in my sleep, I once saw such a video in the past. <RavenJoad>livia: There is `guix repl`, which does the heavy lifting of importing all the Guix-specific stuff as libraries into a normal Guile REPL. From there, I don't know what video you're referring to, but you might be able to piece it together from there. <ieure>livia, Don't know the video, but see "Using Guix Interactively" in the manual. <livia>ieure: Hmm. If I remember correctly, the person wasn't making extensive use of REPL commands, but rather running macros and procedures <RavenJoad>livia: As for your build failure, it seems like that package's Makefile simply doesn't have a "check:" target. <ieure>RavenJoad, Yeah, I fixed it. <RavenJoad>Sorry, that wasn't for livia, but for ieure. <rkazak>Hi - trying to execute a guix pull and the command is aborting because it's claiming that the commit id is not a descendant of (commit id)? Has anyone seen this before? <rkazak>"This could be an indication that the channel has been tampered with" ? <ieure>rkazak, What commit(s) does it mention? <rkazak>the to commit starts with 9aad3c and the bad descendant starts with ad8cb7a <ieure>rkazak, Okay, yeah. So some commits accidentally got pushed to Savannah instead of Codeberg; you pulled those. Then they were removed by forcepushing, which causes the message you're seeing. <ieure>rkazak, tl;dr is give the --allow-downgrades flag it mentiones. <ieure>rkazak, Good time to make sure your channel config is pointing at the new https://git.guix.gnu.org address. It's the new default, so unless you've changed it, a reconfigure & restart of the guix daemon should do it. <rkazak>Ok fixed, I thought that the guix tool was going to warn us of the need to change that? <apteryx>ieure: and the funny thing is that firefox from nonguix doesn't have that same problem (it doesn't crash in that situation like librewolf) <ieure>apteryx, Hummm, I don't know. Never run into that, though I did notice that element-desktop wasn't opening URLs for me, which sounds like maybe the same issue. <untrusem`>can anyone share their guix-exwm setup, I want these things <untrusem`>- how to use separate emacs instance for exwm to avoid freezing up <untrusem`>I just started usin exwm a few hours ago, reading the guix manual on desktop environments <untrusem`>maybe I should ask this in exwm channel, but wanted to know how guixers setup their exwm <jakef>before i open a bug, trying my luck here <jakef>the hdf5 package has some funky paths in its cmake config file <jakef>setting dirs to ${PACKAGE_PREFIX_DIR}/${PACKAGE_PREFIX_DIR}/... <jakef>so we get: include could not find requested file: <jakef> /gnu/store/...-hdf5-1.14.6//gnu/store/...-hdf5-1.14.6/lib/cmake/hdf5-targets.cmake <jakef>that is from line 185 of $(guix build hdf5)/lib/cmake/hdf5-config.cmake <remyd1>Hello, I don't remember how to see the content of a manifest corresponding to a remote package (from a channel). <jakef>remyd1: do you mean --export-manifest arg to guix shell? <remyd1>jakef, I think this is for already installed packages <remyd1>If you want to see the manifest beofre installing it, I think you can see the content of the whole channels manifests <remyd1>I already did it, but I don't remember how <jakef>maybe guix describe -f channels? <jakef>manifest is throwing me off here a bit <remyd1>jakef, no, this is to export your list of channels with checksums <remyd1>maybe it was just a plain curl request ... <mawumag>hi guys, does anyone know if there is an ongoing effort to upgrade scipy 1.12? <zimoun>How do I review a snippet for one commit of a PR on Codeberg? I’m only able to add a comment to the final state of the files and not commit per commit. Do I miss something? <csantosb>zimoun: Commits tab, sha1 column, click commit, brings you to files changed ? This is what you mean ? <zimoun>Ahah! If I click on “Message” column, I cannot. But indeed I can with the “SHA1” column. Arf! <zimoun>Well, I understand why many things in life appears to me awkward. Because I’m definitevely not designed for the common interfaces. Anyway. :-) <jakef>i now appreciate why 3d slicer hasn't been packaged xD <andreas-e>zimoun: I understand - this is how one ends up feeling old! Graphical interfaces usually make it easier to discover things. But then if you do not see them (the other day I just did not see the button to reopen an issue on the webpage), they are difficult to document. Instead of just writing a text, one ends up in the best case with lots of screenshots. <zimoun>Well, for trivial patches, it’s okish… But it’s more than pain when it becomes more complex… <zimoun>Yesterday, using my good ol’ mail client, I closed ~40+ bug reports and commented or tagged on others; and probably processed between 100 and 200 messages. <zimoun>This morning, using the webinterface bah after processing 10 messages (3-5 issues and/or PRs), I just gave up; the cognitive overhead for parsing, for deciding to comment/reply or not, and then for going to next… is just… insane for me. <zimoun>Well, it’s a real question for me: How do people scale with the webinterface? <andreas-e>Thanks by the way for going through the old bugs! <andreas-e>Why is it more difficult? Can you not just click from issue to issue and write comments in the text field below? <andreas-e>Well, the commits are a bit more hidden, which is annoying for simple pull requests and maybe useful when there are long patch series. <csantosb>zimoun: in the same way as the do with windows os ... they just learn how not to <zimoun>andreas-e: For example, sometimes I do not decide which issue is next, it’s just the next thread (shortcut: Alt-N) and then press N for the next message orAlt-N. Sometimes, I tag one message in the thread and keep reading then quickly jump back, press R and comment. I do not have to click to read the patches, i.e., I can easily switch from one to the other. Etc. <zimoun>I mean it’s not about resistance of changement. It’s about scaling. <noe>zimoun: shouldn’t be to hard to write a debbugs-like interface to forgejo, right? <noe>with all the benefits of forge <zimoun>noe: it shouldn’t be too hard to write forgejo-like interface to mails. ;-) <csantosb>noe: tarsius, author of magit/forge, is a guix user, he plans to implement forgejo in forge <zimoun>My question – I’m trying to adapt so I’m really open to learn! –: How do people read e.g. 50 PRs? <csantosb>Just learned it is possible to edit and comment <noe>csantosb: yes, however it being mails in notmuch would be epic for tagging and search <zimoun>But I cannot process at the same rate. :-) <zimoun>For one example, there is not real cache. So when the network is unstable, it’s annoying. <zimoun>Whatever. My playtime is done… Back to “stuff“. :-) <csantosb>I know; little helpers to get you closer to what you already had before ... Let's be positive. <ieure>Good (insert region-appropriate time of day)! <noe>Today is a good day, isn’t it <wehlutyk>I'm developing a guile file, which runs fine in its "guix shell -L ... -f ... -- guile" shell (with proper compilation) <wehlutyk>but I can't find a way to run that properly inside emacs (doomemacs) <wehlutyk>I'm trying out "guile --listen" to run-and-edit my code in the shell, then connecting from emacs. That does show guile inside emacs, which runs in its proper "guix shell" environment, but the font seems mistaken with many ^A^B symbols <wehlutyk>and if I try "run-guile" in emacs, the only way I see to add the "guix shell" env is with "add-to-load-path" inside emacs's guile, but that doesn't include C libraries <wehlutyk>(so hangs whenever a C lib provided by the project would be needed) <wehlutyk>any hints to combine (1) doomemacs for guile+geiser, (2) "guile shell -L ... -f ...", (3) C libs developed in that project ? <postroutine>Hello. Do you know if someone have already used Guix to transform a PC into a media player for a TV with Kodi ? And if yes, have shared their operating-system script ? <identity>postroutine: i do not know about that, but i guess there is `home-kodi-service-type' at (info "(guix) Media Home Services") <postroutine>Is it possible to define a home service from an operating-system definition ? Or do I have to use Guix Home and an home-environment ? <identity>postroutine: i do not think you can, but you can try to convert the home service to a system service by tweaking the service's code and use that <postroutine>When a service-type have a configuration, and this configuration have a package parameter: Does that mean this package is automatically added to the list of installed packages when we define a service from this service-type ? <ruther>postroutine: no. The package won't be installed to the profile by default <yelninei>it depends if the service extends profile-service with its package or no? <identity>postroutine: normally it is not, but you do not have to install the package for the service to work <yelninei>then i think the no answer is a bit misleading <ruther>yelninei: I disagree, op is asking if it's done automatically. It's not. The service has to declare it. <postroutine>There is a parameter "openssh" with as description: "The OpenSSH package to use." <postroutine>I have seen it on multiple -configuration, but I didn't find an explanation on the manual. Maybe I missed something. <ruther>postroutine: that depensd on what the service does. In this case it provides a daemon service, so this daemon service will use this package to get openssh out of. <ruther>s/to get openssh/to get ssh command/ <spiderbit>Hi I want to use swank in stumpwm have installed it but it seems to not be found <spiderbit>do I have to restart stumpwm before I can reload the config or something? <identity>spiderbit: you do not need to edit the package, adding it to your profile (and probably restarting) is enough, i think <spiderbit>"edit the package"? do you mean with that what I have done with the pasted code? <futurile>Q: is there a clever way to get the output from CI - I'm trying to get it to show me build failures and it just keeps timining out <spiderbit>Well I just restarted stumpwm doesn't find it anyway. <identity>spider: to answer your question, yes, but i guess a better term to use is "transform"; no idea how to fix your issue with stumpwm, i do not use it <bavier>do we know of cases where the shepherd can start to use a lot of system memory? I have htop reporting 10G resident use for shepherd. <ajarara>hello #guix, after a guix pull, X now fails to start because org.freedesktop.Accounts was not provided by any .service. From looking at irc logs I see accountsservice provides it. It _is_ in %desktop-services and that's what in my config. It's not a shepherd service, so it doesn't show up in a herd status. How can I confirm that it's successfully activated? <nico_>In reconfiguring guix home on guix system, is there an easy way to withold application updated until substitutes are available? A few applications are very big and take too long to build locall. <ieure>nico_, Not really. You can `guix weather' to check for substitutes after pull / before reconfigure. <ajarara>oh, nvm, I found an empty org.freedesktop.Accounts.conf in /var/log/messages. Will come back once I follow this lead. <gabber>is there a button where i can start a review on codeberg? or do i just comment in the Conversation tab? <noe>gabber, In the files changed tab, click on the plus on the line where you want to comment <gabber>ahhhhh, the only tab i hadn't looked at (: <noe>Once you are done, either click “add single comment” or click “start review” then optionally add more comments and end with “finish review” on the top right <gabber>so only commenting the code actually starts a review? <gabber>commenting in the Conversation tab does not? <ieure>You can also review without leaving a comment by clicking "Finish review." <gabber>ieure: that button only emerges after i leave a comment in the code? or where can i find it? <ieure>gabber, Pretty sure it shows up whether you've left a comment or not. And it's in the upper-right of the files changed tab, as has been mentioned. <gabber>ieure: ah yes, i see it. i thank you two! <ajarara>hm I started a time-machine-based git bisect to find out when accountsservice started exporting this empty file causing xorg to fail, but ran into unrelated build issues polluting the results about 5 steps in. accountsservice itself hasn't been bumped, one of its dependencies has been, causing the empty org.freedesktop.Accounts.conf (which causes xorg to fail). <ajarara>I could skip but that would make this take a lot longer since I'd have to stick around and parse the results to mark skip or mark bad (a time machine build takes a while just to set up, even on a 9950x!) <ieure>ajarara, Sounds like this is a d-bus service? <ajarara>'[ -s "$(guix build accountsservice)"/share/dbus-1/system.d/org.freedesktop.Accounts.conf ]' is the test I used (wrapped in guix time-machine --commit) <ieure>ajarara, I'm not sure why this would make Xorg fail to start? I don't think Xorg talks to dbus at all. <ieure>ajarara, FWIW, I have a machine on commit 3bd7d7e from yesterday (accountssservice 23.13.9) and the file exists and looks fine. <ieure>Actually no, that's from earlier today. <ajarara>hm, thanks, I'll try that. I have 23.13.9 on both good and bad, just with different prefix <ajarara>I'm guessing it's not X that depends on dbus/accountsservice, it's gdm <ieure>Yes, that sounds more right to me. <ajarara>the relevant line in /var/log/messages is '... org.freedesktop.Accounts.conf, line 1, column 0: no element found' <ieure>ajarara, What commit are you on? There have only been four since the commit that works for me, none of them seem related to xorg/gdm/dbus at all. <ieure>3bd7d7e is "Add forgejo-runner" in the commit list. <ajarara>2e005e2784429c0be812f530864cff93972ef1c8 was the one that failed, that is on June 8th <ieure>But a newer commit doesn't reproduce your problem for me. <ajarara>hm, nope, still empty. I'm not using a modified guix-daemon (just some standard substitute urls/keys) <ajarara>Maybe a bad substitute? I've gotten a bad one before (ccid that expected glibc 3.28 at runtime when it was built against 3.29) <ajarara>Is there a way to force a local build for just this package in a configuration? I could apply a trivial transformation and feed it into the service definition. <ajarara>that's what I did at least for pcscd (the ccid consumer) <ieure>You could `guix challenge' the substitute. <ieure>But I think if it was that, someone else would have noticed that GUI sessions were broken some time within the last nine days. <ieure>What do you mean "it shows clean?" <ajarara>as in it does not differ from the substitute, identical. Is full store path consistent? For 3bd7d7ef19c199554c29925a84f53fcb4f4e9016 (your commit), accountsservices is 4rqr8r4pnj7cxy9b1xihvaiqfd1zw5if <ajarara>as in, /gnu/store/4rqr8r4pnj7cxy9b1xihvaiqfd1zw5if-accountsservice-23.13.9 <ieure>I understand. I have the same hash. <ajarara>sounds like a bad file system.. I'll try a live CD and report back. Thanks for the help! <ieure>ajarara, fsck your filesystem? Don't have any better ideas. <ajarara>I _just_ moved off a dual btrfs onto a single ext4 setup because I was like "there's no way I get file errors and I can always restore from backup" <ajarara>if that's it I'm forking over for a second and re-adopting. Spent half the day on this. <ieure>I don't know how you can see an empty file in your store item, but `guix challenge' succeeds. <ieure>ajarara, There are some commands in the manual under "Invoking guix challenge" that let you manually compare, that might be worth trying. <ajarara>yeah it does it instantly instead of building anything.. maybe it's like "oh well you have it locally so I won't bother building!" <ieure>ajarara, It doesn't build, but it compares what's in your store with what's on the substitute server(s). <ieure>Normally you'd force a rebuild by gc'ing the store item, but accountsservice is low in the graph, not sure how to manage it with something like that. <ieure>It's honestly pretty hard to force Guix to rebuild something, it'd be nice if there was a simple --rebuild flag or something. <ajarara>I thought that's what guix build --check was for <ajarara>at least for me. Suddenly very wary of this installation lol. <ieure>Like I said, it's hard to make it rebuild stuff. <ieure>ajarara, `guix build --check --no-grafts --no-substitutes accountsservice' <ieure>But that bootstraps the whole dep chain, not just the package. <ieure>Like I said... it's hard to make it rebuild stuff. <ieure>ajarara, `guix build --repair' might also be helpful here? <ieure>ajarara, Are you positive that: a) the file in your store item is zero bytes, and b) that file is what's actually being used by your current system generation? <ajarara>the first fails pretty quickly on a ./configure of gdl (as you say, the whole dep chain), guix build --repair prints out the store path happily. For question b, yes, that's why X fails and the path I had shows up in /var/log/messages <ajarara>For A, not sure how I'd tell if e.g. emacs and my boot process reads it as empty <ajarara>I think trying an fsck is a good idea. I _have_ had chunk errors on a previous ext4 install (which is what prompted me to go dual btrfs with data healing) <ieure>ajarara, I'm a little fuzzy on details, but there's probably a dbus-activation-service that builds a dir full of symlinks into the store. So the store item might be fine, but the thing linking to them could be busted. <ajarara>I mean the store item isn't, I'm looking at it directly with emacs <ajarara>I'm gonna grab dinner, then boot into a guix install iso and fsck. If that comes back clean reinstall from said CD. ieure, thanks for the help! :) <ieure>Well... what makes no sense to me is that Emacs sees it as 0 bytes, but `guix challenge' sees the same thing that's on the substitute server. And my system with the same store hash has the file and also passes `guix challenge'. <ieure>I don't know of any FS corruption issue that could cause that behavior. <ieure>Do you get 0 bytes if you `ls -l' or `cat' it? <ajarara>-r--r--r-- 1 root root 0 Dec 31 1969 org.freedesktop.Accounts.conf -- cat is an empty buffer <ajarara>this is on a system with ECC ram, pretty freshly built <ajarara>(confirmed clean/ECC capable with memtest86) <ajarara>anyway, dinner! I'll ping you when I figure out the problem