IRC channel logs
2023-12-29.log
back to list of logs
<csantosb>"--no-substitutes" doesn't imply no binaries at all. <ieure>Kabouik, Comments say they only observed it on Fedora, so... we don't? <ieure>(I don't know if Guix does/does not do anything for this) <Kabouik>Maybe I'm being overzealous and don't even have the issue indeed; I'm just trying to refactor my Sway config and clean things <Kabouik>Yeah Gimp starts pretty much instantly whether I use GTK_USE_PORTAL=0 or not, I guess we're fine <bdunahu>This is similar to a question I asked a few days ago, but I found more information: The standard emacs-29.1 package in the standard guix repository does not have working alpha-background transparency (I tried it on an arch-managed system and the full guix system). The emacs-pgtk package does have working transparency, but I am using X11. Has anyone else run into this issue and found a solution? <ieure>Does pgtk not work under X111? <bdunahu>No, it includes a warning which says it is unsupported. There were also some major visual bugs I ran into <Guest14>--rounds=3 does nothing since I already built it. How can I force it to actually build it 3 times? <Guest14>ah, --check --no-grafts work. did not know i need additional flags <bdunahu>To anyone who uses emacs alongside guix here, do you compile it yourself, use it from a different source, or something else? Or maybe the package I am trying works on your machine for your use case? <cobra>how can i build the guix binary tarball (for another architecture)? <bdunahu>Do you use the alpha-background frame parameter from v29? <Guest14>bdunahu: I can do test it in a VM, though it will take some time <Guest14>which commit do you use from guix? (guix describe) <bdunahu>That's okay. My main machine is busy right now, though this one is using commit 0d13d09. Does your VM use x11 with the standard package named 'emacs'? <Guest14>yes (gnome is the DE). I run those two cmds in the scratch buffer and nothing happened <podiki>the last time i tried alpha background it didn't work on x11 for me without pgtk and then it was buggy (didn't work consistently) <Guest14>For Wayland desktop environments (as opposed to X) you need a PGTK (Pure GTK) build of Emacs for this to work. ah nvm <podiki>there's emacs-pgtk (and other variants with pgtk) <bdunahu>Yes, I tried the pgtk version and it worked, but I am using x11 <bdunahu>Guest14: are you using wayland then? Maybe echo this variable: $XDG_SESSION_TYPE <Guest14>returns x11, hmm, normally wayland is used, guess that is just the case for Fedora and not Guix <bdunahu>Thank you both, maybe the problem is with this build then, because they feature worked fine coming from parabola. I'll start looking at my other options <Guest14>You could also try the guix-help@gnu.org mailing list and ask for help. Maintainers of the Guix package will probably read it and can better help you or may even fix the package itself <bdunahu>thank you, I'll check with them first. <tsmish>later tell euouae I've tracked down why test doesn't run correctly. The reason is guix has broken load paths which prioritize system directories over the repo directory. To solve this add "(add-to-load-path source)" to etc/system-tests.scm:78 (right after (define source ...))" <tsmish>sneek: later tell euouae I've tracked down why test doesn't run correctly. The reason is guix has broken load paths which prioritize system directories over the repo directory. To solve this add "(add-to-load-path source)" to etc/system-tests.scm:78 (right after (define source ...))" <tsmish>7 hours down the drain yet again, because something likes to put stuff before GUILE_LOAD_PATH <ulfvonbelow>the universe of things that can be broken by environment variables is truly limitless <tsmish>ulfvonbelow: the fun fact is envvars are correct. guile has %load-path corret, but guix repl has stuff prepended before them. <tsmish>sneek: later tell euouae or alternatively add "-L$(top_srcdir)" as an argument to guix build at Makefile:7039 (check-system definition) <tsmish>sneek: later tell euouae running "make scripts/guix" before running the test would have also worked, albeit for a different reason. <adanska>pretty good! trying to wrap my head around monadic parsers for a tool im making <adanska>its pretty austere,, nothing like megaparsec in haskell <adanska>have to build up most of the primitive parsers myself <ieure>Nice. I learned CL a couple years back, now I wish everything was more like it. <adanska>and monads in a dynamically typed language look a little different compared to something statically typed like haskell, so the examples online take a bit more decoding that usual <adanska>same! CL really is my favourite language. its more versatile than python whilst outperforming too (using something like SBCL). the repl really makes development that much easier <adanska>i really miss it when i have to use other languages <ieure>Yep. I really like the condition/restart system, and CLOS. Also TRACE/UNTRACE are really terrific. I think there's a Clojure lib that gives you that. <ieure>I bought the MOP book, haven't really dug into it, though. <adanska>yeah, i havent delved into the MOP at all, but reading about it would be super informing. <adanska>ive currently got hard copies of Practical Common Lisp, Common Lisp Recipes and SICP. PCL was a fantastic starting point, and i still refer to it for things like writing macros and the condition system <adanska>CLR is a fantastic book though. 85% of my CL questions are answered in depth in that book <adanska>maybe this holidays i'll go through and do SICP all the way through. i think it would be a great experience <rekado>there are only a handful of ongoing builds on ci.guix.gnu.org <rekado>efraim: I see that you and civodul are both logged in on ci.guix.gnu.org and following the cuirass logs. Are you investigating the poor build farm performance? <efraim>rekado: I tried a bit yesterday but I'm not now. I just have my screen session open not doing anything <efraim>I also tried restarting cuirass yesterday and restarting some of the cuirass-remote-workers but that didn't seem to increase the number of builds going out <efraim>I checked some of the build nodes and they didn't seem overly busy or sitting with dead builds taking space <rekado>yeah, I poked around and restarted cuirass and the cuirass-remote-* services as well <rekado>it looks like the workers are eager to do something; they ask for work, but are not given anything to build. <rekado>maybe there’s a problem with the postgres connection pool <efraim>maybe the sqlite db is too big? or the WAL is too far behind? <efraim>I was thinking of the one in /var/guix/db <rekado>shall we restart the daemon on ci? <efraim>I'm finishing 'guix gc -C 1' to see if that gave it time to do anything <efraim>I'll try vacuuming the db, not sure it'll help much <efraim>database shrunk from 19G to 4.7G, the WAL is still 38G <efraim>rekado: if we stop the daemon, run vacuum and start it again it'll absorb the wal file <efraim>I'll test it on one of my machines first <efraim>well it didn't absorb the wal on my arm board <efraim>I restarted the daemon and cuirass services <meorly>how come guix home ignores my channel definitions? is there something that needs to be enabled? <meorly>it generates ~/.config/guix/channels.scm correctly but the channel doesn't show up in guix home describe <jpoiret>that channel file is used by `guix pull` only <rekado>meorly: all it does is manage that file <vixus>hey all, I'm just setting up a fresh install (laptop, graphical installer, xfce desktop) and having some issues. a) laptop doesn't suspend when the lid is closed on battery, b) xfce can't lock the screen, c) xfce's pulseaudio panel plugin ignores my volume keys. I've tested the capabilities (loginctl suspend works, xlock works but xflock4 doesn't, <vixus>testing with xfce4-volumed-pulse the volume keys work) but something's not being wired up right :D <rekado>efraim: my contact in the network team says that there are still the same problems reported on the same switch ports — even after changing the switches <rekado>this suggests that something’s wrong with the server. <rekado>the iDRAC says “SYSTEM HAS CRITICAL ISSUES”, but that’s all. No details. <rekado>rebooted the iDRAC; now it says “SYSTEM IS HEALTHY” as if nothing had ever happened <rekado>I’ll have to reboot ci.guix.gnu.org to apply firmware updates for the network cards and BIOS <gabber>i can $(guix system image) my operating system/image definition, but i fail to $(guix deploy) to that machine: missing modules for #<file-system-label "Guix_image">: uio_pdrv_genirq -- any idea what's going on? i just added a kernel-module-loader-service-type (but this didn't change since flashing the image and now deploying) <bdju>the homepage listed for the pcsxr package (playstation 1 emulator) oddly is redirecting to the microsoft website... <bdju>seems to be the whole codeplex website redirects now. I'm not familiar with codeplex. did it die? <ssouth>"CodePlex was a forge website by Microsoft." <ssouth>Presumably abandoned after their acquisition of GitHub. <bdju>ah okay. so they killed it off after buying github <bdju>between the pcsxr package being broken (hasn't built successfully in months I think), dead homepage, and seemingly not getting updated anymore, maybe the package should just be deleted. looks like there's a more up-to-date fork called pcsx-redux also that someone could package to replace it <ssouth>bdju: You going to try packaging it? <vixus>I'm kinda confused if DPMS should be getting managed by xfce-power-manager or by elogind <ssouth>Is bordeaux.guix.gnu.org having issues? It's returning "502 Bad Gateway" errors to requests for substitutes. <sneek>wehlutyk, you have 1 message! <sneek>wehlutyk, nckx says: [I replied earlier but it seems to've been lost.] I suspect I approved your message and whitelisted your address. That doesn't expire, so something must have changed. <wehlutyk>In setting up a Framework 13 with Guix, I'm trying to get the recommended "power-profiles-daemon" running to make power usage lower <gabber>wehlutyk: peanuts is a bot that gets the title of a webpage and re-posts the link <wehlutyk>so my challenge is after having "guix install power-profiles-daemon", and/or mentioning it as a service in "dbus-root-service-type", running powerprofilesctl always gives the same error <wehlutyk>Failed to communicate with power-profiles-daemon: g-dbus-error-quark: GDBus.Error:org.freedesktop.DBus.Error.ServiceUnknown: The name net.hadess.PowerProfiles was not provided by any .service files (2) <wehlutyk>do people have any hints to get this working? <gabber>wehlutyk: did you just $(guix install power-profiles-daemon) or did you install the service in your operating-system config.scm? <gabber>you *are* on a Guix System, right? <wehlutyk>(looking for the git link for my guix-system configuration) <wehlutyk>there might be some leftover in that system config, as I'm transferring from an old x240 config <jpoiret>wehlutyk: on guix system you'll need a service definition for ppd, I wrote one this very week but haven't upstreamed it yet :) <wehlutyk>I'll be out now, but will re-read this later in the day in case you could share that <dm10>Hello, is anyone able to tell me where I can find configuration files for the pre-built qcow2 VM image? I don't have the actual image to hand, just a foreign GUIX installation. <jpoiret>dm10: usually the minimal one is at gnu/system/examples/bare-bones.tmpl <jpoiret>If you don't have a checkout you can look in ~/.config/guix/current/ <dm10>There is a vm-image.tmpl file, would that be it? <gabber>n.b. $(guix system image --image-type=qcow2 some-config.scm) <dm10>Does anyone have experience running the standard VM image on an OpenStack cloud? Is it straightforward? <gabber>dm10: i personally don't, but it might be relatively easy (there are some resources on deploying images on cloud hosts -- i think a blog entry and one in the cookbook?). if you can't make it and need some more specific answers than what can be provided here posting to the help-guix mailing list is usually a good idea. you won't get answers as quickly as you might get them here but chances are higher you will get some answers (: <Franciman>i'm using home-xdg-configuration service. to configure my configs, but sx requires an executable config file <Franciman>it doesn't seem possible to execute the symlink guix home creates, tho <Franciman>well, usually dotfiles are not meant to be executable, tbh <gabber>it might be a bug from sx (i don't know what that is) -- yes, usually config files shall be read, not plainly executed <mirai>Franciman: perhaps something like <mirai> (service home-xdg-configuration-files-service-type <mirai> `(("tmux/tmux.conf" ,(local-file "files/tmux.conf")))) <mirai>and have said file with +x permission <Franciman>hm i also have the said file with +x permission <mirai>is this with home-xdg-configuration-files-service-type ? <Franciman>and i think that the linked file is also executable, when reading the ls -al output <gabber>huh. some programs don't follow symlinks <Franciman>lrwxrwxrwx 1 francesco users 48 29 dic 14.57 sxrc -> /gnu/store/9sb8h7gjfcghsprnfqryjz6nw35l4ndd-sxrc <Franciman>i think also sh doesn't execute the linked file <jpoiret>I am doing exactly the same as you, running on the amd fw13 :) <jpoiret>I also run linux 6.7-rc7 with a couple of patches (the amd preferred cores one!) <jpoiret>fwupd is a bit more annoying to set up <rekado>ACTION sent mail to the guix sysadmins to discuss the state of ci.guix.gnu.org <podiki>rekado: thanks, was about to ask what was going on as I noticed build rate has really fallen off (mesa-updates is getting nowhere) <jpoiret>well, regarding happy news, GRUB 2.12 seems to work properly. I'll need to clean up my work in separate patches, we're also running more tests now. <cwebber>I've been out of the loop for a bit. With the newer change-id stuff, is there anything one needs to add to one's own git configuration to support change-ids? <cwebber>or is that all handled on the git server automatically? <cwebber>I see my own git is doing it automaticaly <lalebarde>Hi, I still try to perform my first guix install. I want to keep may home partition where I have already users with defined uid. I could see I can add a uid definition in the guix configuration file, but shall it be `(uid "1234")` or `(uid 1234)`? <jpoiret>cwebber: the Makefile update your repo-local gitconfig. a bit controversial, although it does provide better OOTB experience <jpoiret>lalebarde: both uid and gid should be numeric <Guest14>To the people that use BTRFS or bcachefs, how are you aware that the RAID is degraded? I normally use something like TrueNAS which has alerts <cwebber>submitted patches to update blender to the latest LTS :) <apoorv569>I just tested the `home-gpg-agent-service-type` in `guix home container. <apoorv569>and even though I set `GNUPGHOME` using `home-environment-variables-service-type` to be under `~/.local/share/gnupg` it still created the directory under `~/.gnupg` <Guest14>Doubt that this is possible since the .gnupg in home is the standard directory. You would need to edit the service itself or add it as a feature to change the default path <Guest14>and the env var is for the gpg daemon not the guix home service, therefore it still created it in the default location <apoorv569>Yea, but it would be nice to something like the `home-zsh-service-type` has `(xdg-flavor?` which uses the `~/.config/zsh` instead of placing it `~/.zshrc` <lalebarde>Hi, I have managed to install guix for the first time with xfce4 and i3, the last one for testing. Guix starts with i3. How can I switch to xfce4 ? <jpoiret>lalebarde: you should see GDM (a login manager) on start, no? <jpoiret>it should have a little drop-down to select the environment you want to sttart <lalebarde>jpoiret I did not see it. I have to disconnect, but I don't know how <jpoiret>if you have i3, you should be able to do Win+Shift+E <lalebarde>ok, disconnected, thanks, but there is nothing to select the WM <lalebarde>ok, the wheel after having selected me as a user :] <ulfvonbelow>hmmm, for some reason the user environment doesn't seem to be getting set up in xfce <guestinator>hi guix, has anyone seen Emacs think the time zone is UTC despite (getenv "TZ") returning "America/Chicago" or similar? <guestinator>(current-time-zone) returns (0 "America") which seems related, until I run (set-time-zone-rule nil) after which it returns (-21600 "CST") <guestinator>this wasn’t happening when I built Emacs 29 from source 🤔 <guestinator>even after adding tzdata to my home config, TZDIR doesn’t seem to be exported in the profile <wehlutyk>jpoiret: thanks! I'm curious about the other progress you mention: I haven't seen any ppd PRs, and didn't know about fwupd! <wehlutyk>do you think there'll be a point when this is merged into upstream? Or would you publish more docs / examples of your config? <guestinator>ok it looks like no packages define tzdir as a native search path, so maybe this is just broken outside of guix system… <wehlutyk>jpoiret: I see "error: ppd-configuration: unbound variable" and my guile habits are not developed enough -- does that "(default-value (ppd-configuration))" need something else, like an actual config definition? <jpoiret>you should be able to put them alongside the file and use the usual (search-patches ) instead <graywolf>Hi Guix. I discovered that libreoffice is crashing on my machine due to missing files. I would like to send a patch, but I am not sure how the fix should be made. Should libreoffice package set (native-?)search-paths, or whould I use wrap-program/wrap-script on the soffice&libreoffice binaries? <csantosb`>I can build, install, create a container env, etc. No problem. <csantosb`>Now, when I try to package 'ghdl', I cannot find gnat/gnatmake, <csantosb`>What am I doing wrong ? I see the python executable, but not the gnatmake binary. Any idea ? <csantosb`>And yes, I had to disable 'validate-runpath', for some reason that I ignore. <jfred>interesting that Guix doesn't fall back to another substitute server in that case, if I pass in `--substitute-urls=https://ci.guix.gnu.org` my build proceeds <lfam>graywolf: The answer depends on the crash <graywolf>lfam: (soffice:26384): GLib-GIO-ERROR **: 21:54:05.635: Settings schema 'org.gtk.Settings.FileChooser' is not installed <graywolf>Doing guix shell libreoffice gtk+ instead of just libreoffice solves it. <graywolf>I found that for example geany uses 'glib-or-gtk-wrap, so I am now trying that approach. <graywolf>However it sounds... somewhat subobtimal having to do this for pretty much every package that uses gkt... <apoorv569>To respond to a thread on the mailing list I reply to the person's email address right? <apoorv569>I'm talking about the guix-devel mailing list BTW. <graywolf>apoorv569: I am usually replying to all. Otherwise I believe it would be received just by the user. <graywolf>Technically you should be able to reply only to the mailing list, but reply all seems like safe default to use. <apoorv569>What would be the address of the mailing list? I don't want to open new thread. The thread already exists, I created it. But I want to reply to a person who responded to me. <graywolf>The list should be in the CC, so just pressing reply all in whatever editor you are using should do the trick. <lfam>graywolf: Oh yeah, that's a longstanding problem. It would be great to come up with a comprehensive solution <graywolf>lfam: was option of "reversed" propagated-inputs explored? That would allow the gtk+ package to declare that whenever something using it is installed, gtk+ would be installed as well. <lfam>It's been a while and I've been distracted from Guix for some time, but it doesn't sound totally alien to me <lfam>I mean, in Guix, the package dependency graph is totally controlled, and the sky is the limit to making it do what we want <aarcov>I have two open Guix issues/merge requests and was wondering if I may have replied to the issues incorrectly <aarcov>I replied to the issue thread, but never CC'd the person that replied to the thread; when someone replies to a guix issue thread are they automatically subscribed to it or should I CC them from now on? <aarcov>And what is the typical turn around time for minor patches that don't modify any other packages? <lfam>aarcov: I don't know what's the "right way" to reply, but I always use "reply all" <lfam>Regarding turnaround time, I also don't know. From my perspective, there is a flood of contributions but not yet enough people to review them and then either commit them or communicate effectively to a committer that the patch should be committed <lfam>I'm taking a look at the sirikali patch now <aarcov>Ifam ok, I'll probably use the reply all from now just in case (till told otherwise) <aarcov>Makes sense on the reviewers limit, making contributions is one level of comfort, but pushing straight to the project itself is something that definitely requires more familiarity <lfam>Yeah, it's not an easy problem to solve <lfam>I notice that the copyright headers in all of the cpp files I looked at describe the license as, to paraphrase, "GPL 2 or any later version". So I'd like to change the license field to just "gpl2+". Is that okay? <lfam>Just when I was going to help land their patch <graywolf>Hm, is there a good reason why %default-modules is not exported from build systems? <graywolf>I assume (@@ ...) would be frowned upon, correct? <graywolf>I am trying to understand how the build systems work (as far as modules goes). My understanding so far is that for any build system, default value for #:imported-modules is %xxx-build-system-modules and #:modules is %default-modules (for those build systems that define it). Also correct?