IRC channel logs
2024-09-08.log
back to list of logs
<cow_2001>i've sent a reply to a debbug replier, not to the debbug number address ~_~ how embarrassing <nalaginrut>hi folks! Is there available minimum qemu image for guixsd? just like netinst of debian <adanska>whats up with the libreoffice build failure on master? checking ci the logs dont reveal anything, and it only lasted 19 seconds <adanska>making me think that this was some issue with the build machine <adanska>like the log cuts off halfway printing a line during the unpacking phase <lilyp>is it desired that modular texlive has no proper dependency handling? <lilyp>I find myself adding packages that are required recursively by style files more often than i want <decfed>can anyone make the guix documentation? For me it errors with "cannot open ./doc/guix-cookbook.pt_BR.texi". <nckx>decfed: I can, and it gets remade when I rm, but the Texinfo targets are notoriously fragile. Maybe ‘rm doc/*stamp* doc/version*.texi’ could help without having to make clean & rebuild the entire tree. <decfed>nckx: thanks for looking into it. I fixed it by editing doc/local.mk and removing offending languages <nckx>Damn snowmake, so easily offended. <nckx>For the record, did you try the suggested rm? <jpoiret>lilyp: i think it's because texlive itself records too few dependencies <lilyp>How does it record dependencies? Could we parse the \RequiresPackage annotations? <efraim>the screen update broke my cargo-culted screenrc. now I actually have to learn what to do <apteryx>or 'C-h i m screen' in Emacs if you want a taste of the dark side <efraim>... and here I was looking through the source in reverse commit order to see what changed <apteryx>reconsidering your problem (where something changed in an incompatible way), a changelog or release announcement could be a nice read <efraim>I still have the release announcement, just I was handed this screenrc at least 7 years ago and have barely touched it since. I have a magic status line on the bottom that I never understood, so... it's going to be fun while I try to understand what it used to mean <efraim>hardstatus alwayslastline '%{= G}[ %{G}%H %{g}][%= %{= w}%?%-Lw%?%{= R}%n*%f %t%?%{= R}(%u)%?%{= w}%+Lw%?%= %{= g}][ %{y}Load: %l %{g}][%{B}%Y-%m-%d %{W}%c:%s %{g}]' <efraim>the most comprehensible part to me is the [] brackets, so I can at least know where I'm modifying parts of the config <efraim>the 'i' for 'bright' is a lie, it makes it italic <decfed>nckx: yes tried both rm commands but this did not solve my issue <decfed>how do I recompile all scm files to go files after running git pull in my guix checkout? <nckx>If you don't have make, or are missing something else, use ‘guix shell -D guix [maybe some other packages I dunno] -- make’, so you don't have to foul up your user profile with Guix build requirements. <Rutherther>is it better to do that, or to use shell from manifest.scm from Guix source? <Rutherther>okay maybe instead of asking I should've looken into its source first. It adds guix development manifest as well, but apart from that also few utilities <nckx>Both are fine. As usual, the answer is just ‘my habits long predate this fancy new thing’. <abbe__>is there a known issue with guix publish serving .xz archives ? <abbe__>in former case, connection stays open <abbe__>(service guix-publish-service-type (guix-publish-configuration (port 7888) (host "127.0.0.1") (advertise? #f))) ;; is my guix-publish configuration <nckx>I don't see what there is to misconfigure. <abbe__>nckx: do you have caching enabled, or some other options ? <abbe__>did you try querying guix-publish service directly ? <abbe__>I'm reverse proxying through Caddy, so I wonder if nginx somehow workarounds this, but Caddy fails to <nckx>I don't see what there is to work around. <abbe__>well, guix-publish does not disconnect <nckx>Anyway, I'm currently building Guix. When it's done, I'll run it locally. <nckx>My experience with nginx is not that it helpfully closes lingering connections, but who knows. <abbe__>or perhaps using http1.0 proxying <nckx>Yes, I can actually reproduce this on the same host that nginx's fine. Connecting to :3000 directly simply hangs after transferring the data. Huh. <nckx>Maybe, but that's all client-side stuff and we're using cURL. Could be some underlying problem of course. <abbe__>i noticed it with guix stalling on this <abbe__>i don't think anyone exposes it directly, everybody puts nginx in front of it <nckx>I'm not aware of anyone running an unproxied guix publish server (at least not publicly) so this might've gone unnoticed. <abbe__>i guess time to file a bug report <divya>I'd like some help on this because last time I tried dual booting with guix, it would always not add arch to the GRUB menu. <nckx>divya: ‘error: service: unbound variable’ doesn't look related to anything dual-boot. I don't think there's a link. <nckx>Have you tried rebooting the system? <nckx>(I don't know where that ‘unbound variable’ is coming from and if at all possible I'd prefer not to find out. Just pray it away.) <divya> I did reboot the system, the system hasn't updated. Neither the packages have been updated. <nckx>It says the reconfiguration went fine. <nckx>I wonder whether the grub.cfg that Guix is updating is the one that you are booting from. <nckx>Does ‘readlink /var/guix/profiles/system-6-link’ match the ‘gnu.system=’ value in ‘cat /proc/cmdline’? <divya>I'd have to check. But the thing is, if guix did get updated, then it would have the packages I installed. But I don't think I do. <meaty>what is the guix equivalent to the systemd journal? <hako>podiki: I haven't touched Hyprland packages for quite a while... I'll look into updating it next week. <sneek>hako, podiki says: are you still updating your hyprland packages? i think we should be able to move them to guix (possibly first on mesa-updates where newer libdrm and wayland-protocols are right now) <podiki>hako: actually i managed to do the updates locally and am running on the latest version now, 0.42.0; let me at least paste you a diff in a moment <nckx>divya: Each guix system reconfiguration creates a new generation that can be booted, but it leaves old generations untouched by default. So it's possible that Guix is building a new generation, reporting success (ignore the ‘unbound variable’ error for now), and yet a reboot is somehow booting an older generation. <nckx>meaty: Old-skool log files in /var/log. <nckx>And if something's not there, the answer most frequently and unfortunately is ‘somebody forgot to make the Guix service log properly’. <apteryx>with wayland, how do I achieve stuff such as starting a user shepherd, which I used to place in ~/.xsession ? <apteryx>ACTION is discovering a brave new world <Rutherther>no general way, depends on your compositor or display manager. For example if you use a display manager you could put it to your .profile, or if you do not use display manager, and don't mind it starting with tty, then also just to .profile. It's even where guix home puts it <nckx>There is no ‘Wayland’ programme, so the answer is however you configure your chosen compositor. I launch it in ~/.config/sway/config. <nckx>This business of typing what I'm typing but faster and more helpfully must cease. <apteryx>I'm currently using GNOME (will look into sway or other tiling 'compositor' later probably) <nckx>Right, then the answer is ‘however GNOME launches stuff at start-up’ which knowing GNOME will involve .desktop files of some description. <apteryx>does Guix Home attempts something smart to be compatible with both Xorg and Wayland? <apteryx>(such as generating both a .xsession and profile, with some variables used as guards to avoid doubly setting things, perhaps?) <apteryx>that's confusingly named if it has nothing to do with Xorg :-) <podiki>i get nice emacs bindings everywhere with it <podiki>which they have as examples in the repo (always a good sign if you are an emacs person), which i slightly modified <nckx>That's interesting. What do you use it for? <nckx>ACTION is basic ~/.xkb peasant. <nckx>Ignore me, I'm tunneling through DNS and have epic lag. <podiki>while you can get some default emacs movement across gtk pretty easily, having the same copy/paste everywhere, more complicated movement (like page up/down, moving by words), and even chords <podiki>e.g. to kill a browser tab I do C-x k like i would a buffer in emacs <apteryx>is ExWM working with wayland? you might like it <podiki>i also switch () and [], but via the wayland compositor config <podiki>i like stumpwm (big lisp fan here, big surprise) but nothing close for wayland yet <podiki>nckx: also, hi! good to see you around (i was also bit scarce for a bit) <apteryx>reading about GNOME/Wayland init, it doesn't seem like it'd source ~/.profile <Rutherther>it's not gnome/wayland that would do that, the display manager would do that <nckx>podiki: Very same! It's nice to see so many familiar faces that were ‘new’ a year or two ago still contributing. Or longer. Time flies. <nckx>It's still the least worst operating system and package manager I've ever tried. <podiki>oriansj: ah thanks, i think i had briefly heard of that. have you used it? <apteryx>perhaps dropping a .desktop file with Exec=shepherd in the XDG autostart user directory would be enough to get shepherd started <oriansj>well dwl based desktops generally are all very tweakable <apteryx>or I should migrate to Guix Home already and discover how it does things. <oriansj>so as long as you feel comfortable reading documentation it is fine <podiki>apteryx: yeah i moved a little to guix home and it starts user shepherd <podiki>otherwise i use my wm/compositor for some autostart, or via .desktop files and calling dex sometimes (depending on my set up i sometimes need that for the "autostart" stuff) <Rutherther>when you use gdm as display manager, it will just source the profile using your shell as login shell, so the .profile gets executed <oriansj>sadly no one bothered to package dwl-guile yet; as wayland and guile should be popular with the guix crowd. <apteryx>Rutherther: ah, I was reading something from 2017, initially it was not configured that way, but it was reinstated <podiki>ACTION must away for now, bagels <Rutherther>though to be honest I am not sure if WAYLAND_DISPLAY will get populated when you start it like that or not, probably not as it probably runs before gnome wayland starts its session. I personally start from tty, and have a service that I start to populate the env var in Shepherd <apteryx>weird; sticking 'shepherd' inside a new ~/.profile file didn't cause it to run in my session <apteryx>I should perhaps try in ~/.bash_profile? I do have it already. <Rutherther>yeah, when .bash_profile exists, .profile doesn't get executed. The guix home services put source ~/.profile to .bash_profile, that's why .profile works <apteryx>strange, my emacs service no longer starts, errors with: Another instance of Emacs is running the server, either as daemon or interactively. <apteryx>that's after migrating it from .xsession to .bash_profile, for use with GDM+Wayland <Rutherther>I don't know why it would error with that, though I am a bit afraid emacs is not going to work properly because of the WAYLAND_DISPLAY missing. If prior to this you started shepherd manually from within the session, then it already <apteryx>Rutherther: you must be right, launching 'shepherd' after loging in, where WAYLAND_DISPLAY is set, there is no problem <Rutherther>as an alternative to what I provided, one could also make a service that would function like the home-x11 service does. What it does it look into the directory with Xorg sockets to see which display got started. The equivalent in wayland is looking into "XDG_RUNTIME_DIR" for "wayland-X" file. Then setenv WAYLAND_DISPLAY to that file name. It just loops for like 10 seconds to see if that file gets created. Since you use a display manager it should be... <apteryx>I guess a shepherd.desktop inmy user XDG Autostart directory would be easier, might have WAYLAND_DISPLAY set by then? <Rutherther>I don't have experience with it, but my guess would be yes <Kabouik>Trying to reconfigure Guix after a long time, I'm getting "guix system: error: service 'dockerd' requires 'containerd', which is not provided by any service". What am I missing? <Rutherther>I don't understand your question. It tells you exactly what you are missing, or what are you asking exactly? <Rutherther>(just add containerd service type to your services) <Kabouik>I thought it did, but adding containerd as a service in my config.scm didn't solve it, so I probably misunderstood what it told me <Rutherther>no, really, that should be it. Even the manual mentions you just should add that <Kabouik>I tried editing my use-service-modules line with "(use-service-modules base cups desktop networking ssh virtualization xorg linux containerd docker)" but got "error: module (gnu services containerd) not found" <Kabouik>It is not clear to me what I should do, sorry. <Rutherther>just put containerd-service-type to your services list <Kabouik>Thanks, I just re-initiated the reconfigure and it looks happy so far <apteryx>weird, I could start shepherd by using XDG Autostart, but 'herd status', but somehow it doesn't create its socket file (/run/user/1000/shepherd/socket doesn't exist) <apteryx>ACTION has enough for today -- later :-) <apteryx>Rutherther: thanks for entertaining my mumblings <Rutherther>podiki did you work on the wlroots update? I could submit a patch soon if you haven't looked at it yet <podiki>Rutherther: ah no, i forgot/was distracted by other updates. but happy to look at a patch when you have it <podiki>it was to add a newer version (keeping previous ones for now) right? <Rutherther>should I submit it to guix patches right away or somehow send it to you first? <podiki>you can submit to guix patches and either cc me or just send me the number here <Rutherther>yeah, I've realized it would for sure break stuff. wlroots decided to rename the .pc file starting with 0.18 to have wlroots-version.pc, so everything using pkg-config, targeting "wlroots" would break. Though I am wondering about naming of the packages. I don't know who chose to set names of the packages wlroots-version instead of just wlroots for all of them <Rutherther>it's important to bind them to different symbols for sure, but the names I don't understand <Rutherther>and am unsure whether to name the 0.18 "wlroots-0.18" and keep 0.17 as "wlroots" or name wlroots 0.18 "wlroots" and rename wlroots 0.17 to "wlroots-0.17". (although I would prefer the most to just rename all of them to "wlroots", but I am not sure if it's fine to do that...) <podiki>could have the same name, just would git grep for anything that might try to do an input transformation or something based on name <podiki>i guess just stick to the naming we have and can separately change that at some point (maybe when all are on the same current version in future?) <rj-31>hi, I've downloaded the qcow image from https://guix.gnu.org/en/download/. I'd like to verify the image with GPG but I cannot find the public key anywhere. Does anybody have any idea where to import that from? <aswjrisp>I have desktop in my use-services and (service seatd-service-type) in my list of services. When I reconfigure I get "guix system: warning: service 'seatd' could not be found" <aswjrisp>How to make sure the seatd service is working properly? <aswjrisp>I do not see seatd in the output of `herd status`. <Rutherther>podiki it's issue #73124, you should also have it in your e-mail, at least if I've guessed the person you are correctly <mrvdb->hi guix. What is a proper way to express "use 'package-myvariant' whenever 'package' is needed/depended on" ? I've used a package-mapping construct so far, but this still requires going over each package one by one. <Rutherther>since you need to go over all deps in that case it's really best to use package-mapping. Though you can of course use the package-input-rewriting wrapper of package-mapping that you just give pairs of package . package-variant <Rutherther>mrvdb- there is no need for that package-mapping as package-input-rewriting already calls package-mapping, so replaces it through all dependencies <Rutherther>and you can really just call it on all packages. The packages that don't have the dependency won't be changed <mrvdb->Rutherther thanks, that gives me some new things to try. <lilyp>is any qt expert around? qtbase on gnome-team fails to build due to some error in tst_qsharedmemory <lilyp>(qtbase@6, that is – I'm trying to update gnome-shell and it's somehow pulled in) <lilyp>ahh, yes, through pipewire, why not? <Franciman>oh no, guix pull is giving me segmentation fault <Franciman>i remember some time ago you were speaking about this <lilyp>franciman: roll back, then guix pull again <ieure>When I have a package which inherits from another definition, is there a symbol I can use to reference the parent, or do I need a symbol bound to the original package definition? <lilyp>if you can't roll back, use the previous revision through /var/guix <ieure>ex. I want to do: (package (inherit (create-some-package)) (source (inherit (package-source the-inherited-package)) ...) <Franciman>this is the first time i rollback (just to highlight how stable guix is for me) <ieure>Would be nice if there was a symbol bound for me vs. having to do (let ((parent (create-some-package))) (package (inherit parent) ...)) <podiki>Rutherther: got it thank, will look <podiki>Rutherther: i think there are no changelogs in commit messages; and now no version of wlroots needs the hwdata patch right? <Rutherther>sorry, what do you mean by no changelogs in commit messages? yeah, exactly, no version needs it since it was all waiting for the update to use pkgconfig. It can be checked from the build log - if hwdata was missing configure would say it didn't find hwdata, and that DRM backend has been disabled, since it needs hwdata <podiki>Rutherther: ok, then i can apply patch to remove hwdata patch first <Rutherther>yeah, though the one I sent won't apply cleanly since it removes it only after wlroots-0.18 has been added. I can move over the commits, add change logs and submit v2 <podiki>but that's fine, i can put those in and slightly rearrange these changes <podiki>i can also update libinput on mesa-updates <podiki>any other related changes anyone have pending? <fnat>I've got a marionette test that fails, any trick I can use to inspect the issue (e.g. print a log file as part of the tests result or potentially connect to the live marionette VM)? <ieure>In my personal channel, I can build any package with `guix build -L. some-package-name'. If I try this in the Guix channel, it spews a bunch of errors. Why is that? <ieure>It would be very convenient if I could build changes in Guix without needing the pre-inst-env. <aswjrisp>I am going to try rebooting and check if seatd is running after reboot. <aswjrisp>herd status tells me that seatd is running after reboot. <Rutherther>podiki what is the plan for merging mesa-updates btw? I've submitted new version. Let me know if something is still not fine (I would like to know since I am quite new to this, so there may still be some stuff I would be missing) <lilyp>ieure do get used to ./pre-inst-env it's a better approach <ieure>lilyp, I am used to it, but I don't like it much -- I end up frequently having to nuke my state and recompile everything, which takes a good amount of time and can't be offloaded to my build machine. <aswjrisp>When I try to start sway with the command `sway` from the linux console as root, I get this error about not being able to find i915_dri.so here https://termbin.com/d55y <aswjrisp>I do not know what I should try do deal with this error message. <aswjrisp>I can get a little further by starting i3 with startx from the linux console and able to open a terminal emulator, but when I try to open a graphical webbrowser qutebrowser in a terminal I see a similar error message about not being able to find i915. <aswjrisp>I have tried many different approaches including sway, i3 with startx, sddm to start sway but it hung after entering the user password, using %base-services and %desktop-services. <aswjrisp>I am stuck and not sure how to get either i3 or sway to start and work from the linux console. <aswjrisp>I would prefer to get sway to work but have not been successful with either. <aswjrisp>I shared a link to the config I am using above. <Rutherther>aswjrisp that probably means your hw needs i915 driver, but mesa in Guix is compiled without it. You may need to change mesa to build with i915 driver <podiki>Rutherther: will let you know! as for the merge, i think nothing controversial in there, just takes a while to get subs on non-x86 arch <podiki>aswjrisp: we should have i915 drivers, we build just about everything in mesa that is supported (the only thing we don't currently is nvk, that i know about) <Rutherther>i915 is not in the list of gallium-drivers in configure flags <Rutherther>I've just tried compiling with i915 in the list and then it outputs i915_dri.so, so mesa really seem to need it in this list so it's included inside of it <podiki>weird, i was pretty sure "auto" built it, as i had checked based on a previous patch/update i did <podiki>or was it removed from mesa and only in their older versions/fork? <Rutherther>podiki maybe it was removed from the default drivers from mesa, I am not sure tbh, I don't really follow mesa <podiki>docs say i915 is not maintained and to move to other versions. but if current version (24.2.2) can build those drivers still, we can enable <Rutherther>yes it seems so, built for me. But I don't have a way to test if the drivers really do work. Btw I've tried adding -Damber=true to configure flags and I get `Unknown options: "amber"', so it seems this branch is not available anymore <Rutherther>maybe it would be good if aswjrisp verified locally first if the i915 is working <podiki>odd, meson file for mesa says it should have i915 on x86-64 <podiki>along with llvmpipe and softpipe, none of which show up on configure <podiki>i think in update to 22 I did check when changing to auto that i915 and crocus were configured/built, not sure what happened <podiki>oh whoops i don't see gallium set to auto in our build <Rutherther>yeah, but that's under an if that is satisfied only if there is auto in -Dgallium-drivers, and that's not there in Guix mesa - why not btw.? <podiki>bah it must have been lost somewhere <Rutherther>it was changed in 9b2a3e023f7090d19c98e7582220340df3962555, but doesn't seem to give explanation <aswjrisp>Rutherther: podiki: Thank you for looking at this. What can I do to test on my side? <aswjrisp>When I run `guix locate i915_dri.so` it tells me it is part of mesa 21.3.8 <podiki>i'll try to remember to ping you when i make the change to mesa on mesa-updates and you can use that branch if you'd like (will need to built subs) <podiki>i suspect from mesa 22.1.7 onwards in guix i915 was dropped due to the change above <Rutherther>of course if this is the problem, this will fix only sway, you still won't be able to run the browser for example, for that the browser would have to be also recompiled <Rutherther>but now that I am thinking about it... would grafts work here? it tries to look for the file, if the file was there, would it just work? <podiki>i think the issue was that in merging/updating some of the changes were duplicated/dropped, but i will restore on mesa-updates <Rutherther>aswjrisp you might also try this. https://termbin.com/vg3d only mesa will be built, and sway and qutebrowser will be just grafted. I am not sure if it will break somewhere, it's very well possible. The first way is much safer, but also means recompile of all packages that use mesa <aswjrisp>Rutherther: I will try the first one that recompiles all the packages that use mesa, as I think it is just sway and qutebrowser. To speed things up I will comment out qutebrowser the first time and if sway works then I will uncomment out qutebrowser so it can be built as well. <Rutherther>there is no qutebrowser defined in the first file, you will have to add it yourself there, but I've made sure to add the w-i915 function separately so it's easy to add more packages. Alternatively you could also "map" over them <Rutherther>it doesn't recompile all the packages by itself, it's just that if you wanted the packages to work, you would have to recompile them <aswjrisp>Rutherther: podiki: Would this problem go away if I installed guix on another computer? (different hardware with different graphics card) <Rutherther>Yes, if that one is included in the mesa drivers <aswjrisp>Rutherther: I am brand new to guile so not sure how I would add more packages, but I will try the first code snippet you shared now. <aswjrisp>Rutherther: I will get another computer ready to install guix on. <aswjrisp>I included the parts of the first code snippet in my config and started the system reconfiguration. <galois`>How do I use glad in my openGL project? I installed it with guix package but the header can't be found. <ieure>So is git send-email still busted? #73030 is unresolved, so, seems like it? <robin>galois`, how are you trying to use it and what error do you get? (for reproduction) <robin>(random guess) if it's a code-generation tool (i see a "templates" directory) it might need additional packages, perhaps propagated-inputs even <galois`>Yeah it generated a src and include dir. I compile it with: g++ -Iinclude/ src/glad.c main.cpp -o main -lglfw -lGL <robin>galois`, how do you run the tool to generate main.cpp etc.? <galois`>main.cpp is my own file. The src/include-dirs have been generated by the glad-website and packed in a zip archive <robin>ACTION pokes around the glad package <galois`>I mean it's ridiculous for me to ask here as this question is not tailored for this place.. <redacted>My system can no longer build successfully because python-apprise is failing to build. While I could roll back to a previous generation, that means my *other* software will also not be updated, right? <redacted>If a single new package version or package dependency version fails to build, how can I update my system?