IRC channel logs
2025-12-20.log
back to list of logs
<vagrantc>it doesn't get far enough to run lsmod, but i can 'cat /proc/modules' from ,bournish in the initramfs <Rutherther>I mean when you actually do boot to linux. It's pointless in the initrd anyway as it doesn't have the module <vagrantc>ah, you mean with and without a usb stick plugged in? <Rutherther>ideally without the whole device and with it + with the iso on it, yes <vagrantc>ACTION suspects udev loads some extra modules regardless <vagrantc>Rutherther: once the installer starts, there are no differences in modules before and after inserting a usb stick ... i suspect udev has already loaded all the other modules <Rutherther>yes, it will probably locate it according to the device and bus it sees, that's why I asked for without it completely <Rutherther>could be ehci_pci ehci_hcd. At least in the configuration I sent. Possibly might need also xhci_pci, xhci_hcd, usb_common, usbcore <vagrantc>not sure i understand what you mean by "without it completely" ... i booted without the (virtal) usb stick plugged in ran lsmod, plugged the usb stick in, ran lsmod ... got identical results <vagrantc>though it did in fact recognize the usb stick after plugging it in... <Rutherther>I don't know what sw you use, but from the softwares I am used to you configure the computer, give it devices. Then after you give it the device you can plug something inside of it or not. If you use qemu directly, it's for example adding "-usb -device usb-ehci,id=ehci -device usb-storage,bus=ehci.0" <Rutherther>yes, I expected it will recognize it. It's just the initrd that doesn't have the modules necessary for it added <Rutherther>those modules either need to be compiled in the kernel (which I am not sure how to do well) or put to initrd-modules <Rutherther>I am pretty sure virt-manager allows to add or remove individual devices, ie. a usb storage device <Rutherther>anyway I've gotta go, I have meant to go to sleep 2 hours ago 😅 <vagrantc>yeah, i think you are right about xhci_pci, xhci_hcd, usb_common and usbcore ... no ehci or ohci modules present <vagrantc>now that i can build images myself i can more easily test things <Rutherther>you can try putting all of those to initrd-modules and maybe it will just work. Maybe we should add them to default-initrd-modules conditioned by aarch64-linux for now. Not sure how to reconfigure the kernel so that they're just loaded like with x86_64 <vagrantc>yeah, those are some of the thngs i'll try :) <Rutherther>I have tried recently reconfiguring the kernel with the options you suggested and sd_mod / sr_mod were still needed in initrd-modules to recognize cdroms / harddisks <Rutherther>but I have to build the kernel for like ... 5 hours? so it's not easy to iterate on that <vagrantc>i can manage a kernel build in ~45 minutes so that's not too bad <cwebber>emacs on aarch64 isn't building? now that one's a surprise <cwebber>how many people are actually using guix with aarch64? I tried using guix home on top of debian on the mnt pocket reform but it's been a pretty disappointing experience so far, more than I expected <ieure>cwebber, I have no idea how many use it, but it definitely feels like a fraction of amd64 users. <ieure>Is emacs itself broken, or is one of its inputs? It seems fairly common for stuff low in the package graph to break on aarch64, which makes tons of other packages fail. <vagrantc>after the great ungrafting of late 2025, emacs took a while to build <vagrantc>ah, i am also using emacs-no-x ... regular emacs presumably has even more dependencies <vagrantc>rust was the big ugly, and the substitute servers are apparently having difficulty building it <cwebber>is there a good date to time-machine back to for now? <cwebber>to be fair emacs hasn't finished compiling yet <vagrantc>somewhere between december 5th and the 18th, for sure ... i suspect it is a closer to the 18th than the 5th <vagrantc>i had to pretty agressively prune generations to make room to build 20 some versions of rust... <vagrantc>ACTION wonders the status of mrustc trying to cut out more rust generations ... <cwebber>vagrantc: ok, I'll roll back till start of the month <vagrantc>i can probably narrow it down a bit further <cwebber>knowing that there was an ungrafting thing and that's what messed it up is already helpful <vagrantc>8cef389d35c105653c3f1a330004b0ca7d4c2478 seems newer <vagrantc>f20e7c6eeddaba20b334975633927b472400005f slightly newer, has substitutes for "emacs" <vagrantc>heh, the very next commit after that is cd4a0e2b7b4188a08e3ede6704705a7ed4988f4d gnu: gtk+: Ungraft. <vagrantc>so guessing that is the currentest pre-ungraft commit :) <vagrantc>"git merge --no-ff" seems like a really harmless way to make these sorts of things easier to find <vagrantc>at the expense of a single "empty" commit <basicnpc>Help! I use emacs30.2 from the official channel, but there doesn't seem to be GUI emacs, only terminal emacs. <ieure>basicnpc, The emacs package has the usual GUI support, I'm using it right this second. <ieure>Woefully insufficient information to give an answer. But you don't have to do anything special, just run it. <ieure>In a graphical environment of some sort. <basicnpc>`emacs` in my terminal just open the terminal version directly.. <ieure>Then $DISPLAY is not set, so it thinks it's running in a text environment. <ieure>When you say "in my terminal," what terminal is that? <ieure>You can't really just set it, it should be configured for you. <basicnpc>It's alacritty, running under the mahogany WM. <ieure>Are you on Wayland? $DISPLAY is for X11. I have no idea how Wayland manages X compatibility. <ieure>I've only run Wayland with GNOME or KDE, it's not my daily driver setup, and has always worked, whether running Emacs from a terminal or the DM menu. <ieure>Sounds like you're on a more exotic setup, looks like this is one of the more exotic bits you get to figure out how to make work. <ieure>I've genuinely no idea where to even start. <basicnpc>ieure: Solved! Turns out I need emacs-pgtk <basicnpc>I want to test working with a new version of 'sbcl' (a common lisp implementation) with a contrib support. I also want to customize the standard common lisp build system (i.e. the asdf/build-system) to use my newly packaged sbcl. What's my easiest bet on this? Should I fork the official guix, and start my own branch? I don't see any easy way to tweak that build system. <tschilp>this is guix 8fcaf10c383b7649a42c4607817c7cc72fbd5b85 <cbaines>tschilp, the variables field doesn't exist in the Guix version you're using <tschilp>cbaines: thank you! happy to see I haven't completely misread it :P. Pulling... <ximon>Which desktop or tiling window manager has the least attack surface <identity>i would expect that to be one that uses wayland, maybe cage, but that is a weird question to ask <tschilp>removing ~docker-compose~ from my package list lets the reconfigure succeed. mhm <elevenkb>Why would `guix home switch-generation -- -2` error out with the message: `guix home: error: -2: extraneous command` <Rutherther>elevenkb: I don't think it knows how to handle '--' <identity>the documentation implies it does, «When specifying a negative value such as ‘-1’, you must precede it with ‘--’ to prevent it from being parsed as an option. For example: guix home switch-generation -- -1» (info "(guix) Invoking guix home") <basicnpc>I want to test working with a new version of 'sbcl' (a common lisp implementation) with a contrib support. I also want to customize the standard common lisp build system (i.e. the asdf/build-system) to use my newly packaged sbcl. What's my easiest bet on this? Should I fork the official guix, and start my own branch? I don't see any easy way to tweak that build system. <Rutherther>easiest is definitely just doing it on the guix repo and modifying the sbcl package definition itself. But build systems should generally allow you to change the package without any modifications. asdf-build-system has lisp keyword for that, so you can try adding #:lisp your-sbcl-version-here argument. If that won't pick up your sbcl, something is likely wrong in the build system then and should be changed at some point <basicnpc>I miss some fonts I had on my macbook. The easiest way to me to start using it, is to copy and paste? I don't have to really package it yet, right? (I will consider contribing later.. after I setup an env that's comfortable.) <Rutherther>sure you can just copy them to one of the locations that are used, I think ~/.local/share/fonts/ <basicnpc>I'm trying this command from the manual `guix graph coreutils | dot -Tpdf > dag.pdf` <basicnpc>However, dot is not found. I `guix search dot`, and the closest I found was xdot. Installed but dot is still not found. <basicnpc>How can I mitigate this independently? Is guix search the right tool? <Rutherther>guix locate is better tool for that. However it can currently only search local store, so it will probably not help you <basicnpc>Ah, that sounds right. How can I mitigate this issue independently in the future? <Rutherther>unfortunately there's not much better options than those two. Some commands are mentioned in guix search. guix locate will help you as long as you have that package already, but forgot the name <Rutherther>this definitely should be improved, ideally the substitutes could build database for guix locate <Rutherther>s/the substitutes/the build servers for substitutes <basicnpc>Thanks! I add that to my note, and hopefully one day I can contrib. <trev>has anyone experienced system configs clobbering each other somehow, specifically with shepherd services? <trev>ex. reconfigure to a whole different system config and then still see the other services <basicnpc>Is there any command that list all packages I have/available, and listing their source channel too? `guix package -I` doesn't tell me whether it's from guix or nonguix. <Rutherther>trev: did the reconfigure finish fully? if you still have the log, can you inspect the upgrade-shepherd-services (there is a print saying what .drv is built for the upgrade-shepherd-services script <trev>Rutherther: yes it did. I think it may have actually been a stray symlink left behind by a service (etc-profile-d) <Rutherther>trev: symlinks will be left there, that's correct, but how does that relate to shepherd servicesd? <yelninei>could someone reschedule the gcc-final builds for i586-gnu and x86_64-gnu on berlin? They block everything else and i am not sure why they failed <trev>Rutherther: cause i was using one of the services to create /etc/profile.d/ scripts <Rutherther>I see "guix offload: error: corrupt input while restoring archive from #<unknown channel (freed) 7f3599317f20>". This could signify that there aren't signinig keys properly set up <Rutherther>I will restart them, but I suspect it will fail with the same error. Someone will need to set up the signing keys <yelninei>this is also the error when the connection drops when they reboot during a build and there is a timer set up to do that regularly <Rutherther>btw do you happen to know what to do when something is already built, but cuirass is saying it's scheduled for days? <yelninei>thanks, I should have kept an eye on them, appearently they never build the new commencement after updating hurd and mach <yelninei>sorry I cant help with that, the substitute is availble but cuirass thinks it is not? <Rutherther>sort of? It's just saying 'scheduled' and yes, it's substitutable from ci, but I am afraid it might be causing other packages to not build. I can't say for sure, though <ieure>Sounds like it, I've run into some similar issues with my own. <basicnpc>Is there a way to list all packages and see which channel they came from? <yelninei>guix package -A, but it only gives the source file and not the channel <basicnpc>!! Does the source file tells about the channel? <basicnpc>Namely, can 2 different channels use the same directory for source files? <basicnpc>Woah :-) `guix package -A | rg nongnu | wc -l` => 174 <basicnpc>Ah hem.. oh, that doesn't mean I have installed all of them. <Deltafire>weird, i thought i had alt-tab completing symbols in emacs, but now it's switching apps in niri <basicnpc>`info guix` is nice.. but I wish that allows reading in emacs/org-mode. Any method? <identity>Deltafire: you can use ESC TAB or C-M-i instead, or just TAB if you enable ‘completion-preview-mode’ <kestrelwx>Oh, are the substitutes for the guix modules a new thing? <mroh>basicnpc: "pandoc -f texinfo -t org guix.info" should do this. <basicnpc>yelninei But that's not really org-mode.. In particular I cannot search the whole document. <ieure>basicnpc, The normal Emacs commands work in info buffers, so C-s will search. You can also set bookmarks. <basicnpc>I use tput to setup colorful bash prompt.. which guix package should I use? <basicnpc>guix package doesn't seem to have terminfo either. <noe>basicnpc, There’s a keybind you can use: C-h R guix RET <identity>no mention of it in the manual, at least <FuncProgLinux>untrusem: I've got a similar thing packaged called "Warpinator" it builds but the binary is installed wrongly <ieure>untrusem, Looks interesting. Some related things already in Guix are magic-wormhole (but uses the network / relays) and nncp.