IRC channel logs
2022-12-17.log
back to list of logs
<rekado>I booted the rockpro64 again from USB into Guix System. Wanted to install grub-efi-bootloader to /mnt/boot/efi (where I mounted the EPS partition), but the installation fails. <rekado>I checked and /sys/firmware/efi does not exist <rekado>does this mean that I cannot use grub-efi-bootloader on this system? <bost>Hi. Is there any activity regarding node.js upgrades? I mean some tickets, issues, or any other effort? <nckx>rekado: On x86 it would. If CONFIG_EFI_VARS_FS is enabled. <KarlJoad>rekado: Without my personal channel, the error disappears. It is just with my channel. <mbakke>rekado: grub-efi needs to modify EFI variables, in the embedded firmware, which only works if the kernel is already booted with EFI (and /sys/firmware/efi is available) <KarlJoad>Is 6.0.12 the default linux-libre kernel now? Did something change in the Linux build system? Could I find that info with guix info/news? <nckx>Well, GRUB sure thinks it ‘needs’ to. If you haven't booted with EFI, try using the removable- variant, rekado. <KarlJoad>nckx: I just read the news from the Guix pull. Apparently there is a customize-linux procedure now. I'll have to investigate that, instead of the jankiness I am using now. <nckx>Oh good, I wrote all that for nothing. <nckx>And there was some fall-out from the change, so it's not implausible you found more. <rekado>I guess the cookbook entry “Customizing the Kernel” needs updating then <oriansj>rekado: I'm starting to think requiring documentation updates to go with all code updates <oriansj>even if it is just adding an example <rekado>nckx: grub-efi-removable-bootloader worked! <rekado>yes, in grub, loaded from the SSD <nckx>oriansj: They *are*. It's just not enforced. <rekado>it doesn’t boot, but this is success <lechner>rekado / for a first EFI boot, you may find the EFI stub in the kernel handy, together with the kernel option init=/bin/bash (or as appropriate) <rekado>I never fixed the CPU reset thing on “pci enum” + “scsi scan” <rekado>so I patched u-boot’s arch/arm/lib/interrupts_64.c … to ignore the error <kori>i'm trying to use the radicale service, I tried logging in for the first time, and now it's throwing internal server error, strace isn't helping too much either <kori>anyone have any clue how I could debug this? <rekado>sad, though, that I still can’t boot ¯\_(ツ)_/¯ <rekado>from within grub’s console I can read the SSD <rekado>gotta try with a serial console again; I’ve been doing this with a HDMI display all this time <nckx>lechner: How do you set the command line? <oriansj>nckx: as with dealing with a small child; any rule that isn't vigorously enforced, isn't a rule but a suggestion which will generally be promptly ignored. <rekado>the grub console is finally some comfort and it feels like progress. <nckx>lechner: That link just suggests typing it in by hand. Does not sound useful for anything but a toy OS. <rekado>lechner: I find this page overwhelming. <nckx>You mean you're not eager to type store items by hand into a pre-boot EFI shell? Why do you hate Progress? <lechner>it's only one time until you can do 'modprobe efivars' <nckx>But still a huge step down from GRUB, which is already working. <lechner>you can use the kernel image instead of grub.efi (or boot.efi at the removable media path) <lechner>nckx / grub is not booting the kernel, is it? <lechner>anyway, sorry. just trying to help. please ignore <nckx>init=/bin/bash won't do anything, there is no /bin/bash. You'll be manually typing in store paths from a piece of paper or other PC, maybe booting Linux, and then what? You've bought yourself a boot, but no progress towards a fix. <lechner>there are static executables one can copy to / <nckx>All true, but you don't need to boot the installed system to do that. <lechner>no, but you may need to modprobe efivars in order to to install grub-efi <kori>pls respond ;_; i've been grepping everywhere trying to find where these logs could be sent to <nckx>kori: If nobody knows, nobody can respond. Sorry. <lechner>nckx / rekado wrote he cheated. that's never worked for me when installing an EFI bootloader <nckx>lechner: How did it break? <lechner>i really can't remember, but there is no point in arguing. you and rekado are more capable than i. it was silly of me to interject. sorry <nckx>I did not consider this arguing. Sorry then. <nckx>We certainly don't have much to go on! If this were a real support request, and not someone falling asleep, I'd ask for useful details 😉 <nckx>‘I still can’t boot’, OK, + ‘[there is a] grub console’ — I picture normal mode, but maybe it's rescue mode (oh god no). <lechner>there is something about the boot order variables that's never worked for me, but i cannot remember why the installation failed. i think grub-install or one of it's alternatives simply failed. it would be helpful to know if rekado cheated when copying the EFI files to the boot partition <nckx>I don't think he did that by hand. <nckx>That's the whole point of the removable- variant. <nckx>And it does not affect Linux, how she is boot, only the name of GRUB on the ESP. <nckx>GRUB works, we know that much, so UEFI's job is done at that point. <lechner>everything you write makes sense. i was only trying to help in a situation when someone may have a limited set of options. <nckx>Agreed, but being able to move the kernel to the ESP implies having a working (rescue) system, and if you have that, you might as well do whatever other stuff (installing a different GRUB, copying these static binaries you mentioned, …) whilst you're still in it. <lechner>as mbakke wrote, that tends to work only if the EFI variables are accessible. it's a heightened requirement from copying files, which could happen in a rescue system booted via legacy MBR <lechner>but i have not had those issues in several years, so maybe something has changed <nckx>You don't need EFI variables (=booted in UEFI mode) to install an EFI GRUB. <lechner>like mbakke, i am not sure about that <nckx>That is not how I read their message at all. <lechner>rekado wrote: Wanted to install grub-efi-bootloader to /mnt/boot/efi <lechner> (where I mounted the EPS partition), but the installation fails. <nckx>OK, can we not do this. Now I am becoming annoyed. <lechner>okay, i am going to sign off for about 25 hours. have a good saturday! <nckx>Enjoy your 25-hour bender! Jelly. <apteryx>cbaines: I just found out a tool, 'patman', which produces patches from the current branch commits automatically, runs check scripts on it, and format git-send commands. a bit like our etc/teams.scm on steroid. It can also interact with patchwork <oriansj>apteryx: the greatest joy of software is learning that those that came before us were also brilliant and knew a few things which in retrospect; we probably should have eagerly listened to... <KarlJoad>nckx: You were right about the kernel. It was the new customize-linux interface. I needed to switch to using it, and I needed to change my usage of another procedure which internally called customize-linux too. <nckx>apteryx: Does the cookbook need to be updated? <nckx>(KarlJoad: Did you rely on the cookbook at all?) <apteryx>nckx: yes it does; I have some patch in the works <apteryx>was the patch removing the input labels of linux-libre reinstated? <nckx>I've lost track of patches vs. applied commits, sorry. <KarlJoad>nckx: apteryx: The cookbook needs an update about this, yes. I patched what I needed together from the docstring of customize-linux <apteryx>KarlJoad: it could be that you're not on the latest commit <apteryx>the change was reverted and not yet re-applied <apteryx>giving us more time to fix the cookbook before we do <apteryx>there was also a NEWS entry to advertise the change, but that is perhaps easily missed <nckx>(Assuming you mean --news) it was. <nckx>(I don't think anyone reads NEWS after installing.) <nckx>(I'm very parenthetical today (Sorry)). <nckx>KarlJoad: Right, but only afterwards, IIRC? <vagrantc>ACTION wonders if it wouldn't be ok to "lie" and have guix 1.4.0rc3 actually release including guix 1.4.0rc3-1commit actually be the guix-daemon that is used <nckx>I guess there's no way around that. It was only 6.0.12 looking like a kernel version that made you look. <nckx>Did _not_ mean to imply you should have known. <KarlJoad>I pulled and then read the news through "guix pull --news". I am using Guix c193b52 right now. <KarlJoad>nckx: Don't worry. There was not way I could have known. I just started working on my NixOS->Guix migration again after a full university term. I missed a lot. <nckx>I'm subscribed to guix-commits@ and still wasn't sure what ended up being (un)merged :) <apteryx>hmm... seems patman just led me to send patches to guix-patches instead of u-boot, apologies ^^' <nckx>Hm, my /var/guix/db/db.sqlite got completely corrupted again. No I/O or other errors in dmesg. <nckx>ACTION shrugs, restores. <apteryx>nckx: hehe, these are mine, per ~/.gitconfig <vagrantc>that patman lives in u-boot upstream is awkward ... but apparently it is independently useful? recently got a request to package it... but ... have not yet fallen into that trap <apteryx>haven't sent it yet; I wanted to feed some fixes to upstream first <apteryx>yes, it can be used for at least linux and u-boot, and it could be adjusted for guix too, I guess <nckx>Oh no. I just typed ‘/gnu/store’ in an e-mail with company-mode enabled. <nckx>ACTION goes to get coffee. <acrow>I guess you cannot C-g your way out of that? <nckx>Nope. But it wasn't too bad in the end (SSDs). Still, I jump at any excuse to get coffee. <KarlJoad>libtirpc is failing to build with an ld error for the hurd-vm-service-type. It is saying libgssapi_krb5.so has an unrecognized file format. <apteryx>sounds like a cross-compilation problem <KarlJoad>Agreed. I am not sure where it is, but I can submit a bug report if that makes things easier. I need to start working on getting system configs working in my channel right now, instead of debugging things. <KarlJoad>Is the system hosting the Cuirass manuals down? Or did the nginx process crash? I keep getting an nginx 404 on the Cuirass manual. <apteryx>vagrantc: would you like to be cc'd for the patman submission? <apteryx>rekado: would you like to be cc'd for the u-boot-documentation review? <apteryx>KarlJoad: you don't have it installed locally per chance? (info cuirass) <KarlJoad>apteryx: I can build it locally and read it that way. Just the website is down. <KarlJoad>When pulling my channel, Guix freaks out about deploy stuff being in a file with an operating-system configuration. <KarlJoad>Guix says (exception syntax-error (value operating-system) (value "invalid field specifier") (value ((filename . ".../system.scm") (line . lno) (column . cno))) (value (operating-system %system)) (value %system), but I can build %system just fine. <vagrantc>apteryx: if nothing else, i'm curious ... sure :) <apteryx>hm, patman uses the linux style signed-off <taoqy>hello, I am staring to use guix and setup locales, I install tmux and find it use glic-2.23, so I want to install glibc-utf8-locales-2.23, but `guix search` can't find this version. <taoqy>The manual says: Note that the glibc-locales package contains data for all the locales supported by the GNU libc and weighs in at around 917 MiB. Alternatively, the glibc-utf8-locales is smaller but limited to a few UTF-8 locales. <taoqy>Is there a way to install glibc-utf8-locales? <taoqy>And in the store, I see a directory `/gnu/store/fnr1z6xsan0437r0yg48d0y8k32kqxby-glibc-utf8-locales-2.33`. Can I install it from the store? <AwesomeAdam54321>The current manual is a bit out of date, but you can find the information in the devel manual <taoqy>Ok, thank you. AwesomeAdam54321 <taoqy>AwesomeAdam54321, After define my-locles, how shold I install it, I try the command `guix package -i --install-from-file=my-locels.scm`, but not work. <taoqy>Oh, the params order is matter. <taoqy>It should be `guix package --install-from-file=my-pack/locale.scm -i` <taoqy>said, build of /gnu/store/68znigiq1gy3kw0sry91arkwf86wk56y-glibc-utf8-locales-2.33.drv failed <jaft>Anyone know why, using the ~python-build-system~, I'd get a "no setup.py found" error when the "setup.cfg" file is getting pulled down from the repo. and is in the root of the project? <jaft>I tried checking out some other existing packages and it doesn't /look/ like anyone has to do anything different for projects using a "setup.cfg" file over "setup.py". <taoqy>AwesomeAdam54321, It's work now. It a typo, en_US -> en_US.utf8 <taoqy>Hello, I build a golang project that use cgo, but it faild with error `fatal error: linux/errno.h: No such file or directory`. Seems I should install `linux-libre-headers`, but my kernel is 3.10. There is no such version in guix, also I find a fuction named `make-linux-libre-headers` that is not public. How can I insall a 3.10 kernel header? <civodul>taoqy: hi! you're building it as a Guix package or interactively? <sneek>civodul, you have 1 message! <sneek>civodul, yarl says: Thank you for disabling the tests when crossbuilding texinfo. I was try silly things in order to fix it : add itself as native input so the tests are run natively, which is mostly irrelevant. I guess that's the answer to every crossbuilds failing on tests depending on running the program itself : disable the tests? <civodul>yarl: yes, tests are disabled by default when cross-compiling, but here i had erroneously enabled them <taoqy>civodul Hi, just a raw go package. <civodul>taoqy: you mean you wrote a Guix package definition that uses go-build-system, right? <taoqy>no, I use `go build` command '=D <civodul>taoqy: ah alright; in that case, make sure you have "gcc-toolchain" in your environment <civodul>e.g., guix shell gcc-toolchain cgo go -- go build whatever <rekado>I reckon my kernel boot problem is due to the fact that I’m using linux-libre instead of linux-libre-arm64-generic <rekado>(the kernel on the USB drive that boots correctly is linux-libre-arm64-generic) <rekado>problem is: the modules for that kernel don’t build right <rekado>says there is no “ahci” or “usb-storage” module <rekado>these are part of the default set of modules <rekado>it’s possible there’s some corruption on this old USB thumb drive; but I’d be happy to hear from others who use that kernel. <jonsger>civodul: will you merge the version-1.4.0 branch to master or the other direction before 1.4.0 release? <lembrun[m]>Despite my best attempts I can't get the mouse cursor to change theme/size in alacritty and other native wayland apps <kreved>seat * xcursor_theme Adwaita 24 I have this in config file <yarl>What is the "@" in "3.10 Building the installation image for ARM borards" ? <cdo256>Hi Guix! I'm writing my first package (an updated version of emacs-tramp). But the compiled sources aren't ending up in site-lisp. I want to inspect the derivation using `guix repl`. But I'm stuck after (open-connection). I can't find the documentation for how to continue. Could anyone provide a hint please? <yarl>cdo256: From my notes, you can do something either like "(with-store store (valid-path? store <yarl>"/gnu/store/zzr7hqy2xp9zjrrwsgypbhyzpxbsg97q-guile-gcrypt-0.3.0.drv"))" or "(run-with-store (open-connection) (lambda (store) (values (valid-path? store <yarl>"/gnu/store/zzr7hqy2xp9zjrrwsgypbhyzpxbsg97q-guile-gcrypt-0.3.0.drv") store)))" <yarl>aah, quotation marks... You can do either: <yarl>(with-store store (valid-path? store <yarl>"/gnu/store/zzr7hqy2xp9zjrrwsgypbhyzpxbsg97q-guile-gcrypt-0.3.0.drv")) <yarl>(run-with-store (open-connection) (lambda (store) (values (valid-path? store "/gnu/store/zzr7hqy2xp9zjrrwsgypbhyzpxbsg97q-guile-gcrypt-0.3.0.drv") store))) <yarl>note that in the repl you can also use ,run-in-store and ,enter-store-monad. See "Using Guix Interactively" in the manual. <yarl>For the operations you can do in the store, look at guix/store.scm <civodul>jonsger: post-release i'll merge version-1.4.0 into master (like i did post-RC) <civodul>certainly not the other way around :-) <cdo256>perfect. I'll be reading for days :) <rekado>a *lot* of modules seem to be missing. Perhaps they are all built in? <rekado>things like uas, usb-storage, usbhid… <f3n1x>hi guixers ! i thought that they were found in the 'net-tools' package, but they are not ! Where do utilities like 'dig' packaged into ? <jlicht>f3n1x: dig (and friends) are found in the "bind:utils" package+output <f3n1x>ACTION loves the gnu project in general and the guix community in particular <jlicht>regarding the inevitable 'how was I supposed to intuit that', some work is being done to support an index of sorts from filename <--> package <f3n1x>ACTION runs 'guix install bind' ... <f3n1x>but 'dig adomainname.coop' leads to command not found. What am i missing ? <jlicht>guix packages have outputs, they show up when you `guix show bind'. <f3n1x>uh... which sorcery is this ': * ' ? xD <f3n1x>some regixpy thing maybe ? Oh <mubarak>for a month and half now I see that in "Latest" guix x86-64 images doesn't have ext4(mkfs.ext4) filesystem support by default. Is btrfs filesystem going to be the default filesystem? <civodul>mubarak: hi! there was a bug whereby mkfs.ext4 was not in $PATH by default but it's fixed now AFAIK <sughosha>Hi all, I am trying to update zrythm to the latest beta version, but the problem I am facing is that it requires [libpanel](https://gitlab.gnome.org/GNOME/libpanel) which is available for GNOME 43. I would like to know if GNOME 43 update in Guix is planned yet. <mbakke>I think the only major difference from the current kernel is CONFIG_PREEMPT_NONE <the_tubular>Where can I read how to use the targets keywoard and not target ? <haugh>(guix records) is really an amazing lib that changed the way I think about the role of the record type. I wish it was part of Guile core. <haugh>civodul, you worked on this module for 10 years?! It shows. Inspiring stuff. <KarlJoad>I know I have asked this a lot, but I still have not found a solution yet. Can you put guix deploy descriptions in a channel? <jgart[m]>KarlJoad: In a Guix channel you mean? You want to put new guix deploy machine implementations in a Guix channel? <jgart[m]>If so, you can put any code in a Guix channel so yes <KarlJoad>Yes. I have a system defined for my cuirass server, and set up to be deployed. I have defined it as a module too (using (define-module)). But I get an exception about a syntax error for the operating-system field in the machine list. <jgart[m]>Can you share the git repo for it or a pastebin? <KarlJoad>I can share a pastebin for the failing deploy file in question; 0x0.st/onhK.scm. I shouldn't share the channel here. Non-free stuff is in there. <KarlJoad>The rest of channel works, because in another branch I have the deploy systems removed, and the channel pulls fine. <KarlJoad>Another question about channels. How can I get a channel to build an operating-system configuration? I want my CI server to build my laptop's configuration, but am not sure how to go about it. <lembrun[m]>Anyone got experience using sysrq on guix? I'm trying to debug why my system hangs on reboot/shutdown <lembrun[m]>When I'm pressing trying to dump the current list of process with sysrq nothing is happening <nckx>Guix's MAGIC_SYSRQ options are identical to mine, which work. <nckx>Is /proc/sys/kernel/sysrq 1? Can you trigger events through /proc/sysrq-trigger? <lembrun[m]>Well I do use the magic shortcut Alt+Print+{reisub etc} <nckx>Have you used the same key combo successfully on this exact keyboard before? <lembrun[m]>it prints something like dumping list of processes <nckx>L should show backtraces per CPU, not a list of processes. <lembrun[m]>yeah or whatever it was, the point is it's not showing anything <nckx>Does this apply to all commands like W, T, etc.? <nckx>I mean, maybe your kernel really is hung completely? <lembrun[m]>it really is some kind of process that is stuck i think <lembrun[m]>it's just that any sysrq command that is supposed to log/dump <nckx>Hm. SysRq+T works on tty2 here (and hangs my machine and disconnects me from IRC :), but not on tty1… <lembrun[m]>and for whatever reasons I don't get the hang anymore .... <lembrun[m]>note that it's the latest kernel from the thing not to pronounce here <nckx>But that won't help you debug a shutdown. <nckx>Or do your logs still get saved? <nckx>Is there a simple (elf-file? "/foo") predicate? <cwebber>there has to be a better way to do this ;) <cwebber>what's the best way to say "hey I want to download this git repo and find out what the hash is for this dev version I'm defining right now" <singpolyma>cwebber: make package, copy any hash, replace bunch of digts with 0, try to build. Error message gives your correct hash ;) <jlicht>cwebber: I usually do some local perusin' with some license detection tooling as well, so for that situation, you can just `guix hash -xr .' <jlicht>but that assumes an clean checkout at the commit/tag you're packaging already existing <mubarak>I'm thinking of mounting /dev/sdb2 to ~/my-files directory, instead of mounting it to /home. My question is can guix create my-files in my <KarlJoad>How can I get a channel to build an operating-system configuration? Do I need to make a package out of it? I want my Cuirass server to build my laptop's configuration, but am not sure how to go about it. <lembrun[m]>it's a shame the log don't tell me much about the hang : ( <mekeor[m]>KarlJoad: i wonder if reading through the "maintenance" repository could help you. dunno <mekeor[m]>mubarak: i doubt that ~ will be expanded correctly <mekeor[m]>mubarak: check out create-mount-point? in the manual. it's an option of file-system <KarlJoad>mekeor[m]: I read through it, and as far as I could tell, the maintenance repo is just behind-the-scenes host orchestration. I think Guix hosts the information to build operating-system images inside Guix itself. <KarlJoad>I am glad to see Guix's CI is failing to build Hurd images with the same error I am. <oriansj>KarlJoad: always makes me wonder how the developer did the build before doing their commit <oriansj>and did they remember to do guix build --rounds=10 to check for nondeterminism <KarlJoad>oriansj: As someone said last night, it might be a cross-compile thing. libtirpc fails on linking because of a file format error. <ham5urg>What do you think hinders Guix to gain total world domination? <civodul>lack of ambition would be a hindrance, but there's no such lack <singpolyma>civodul: I don't mean to say guix community lacks ambition certainly, but ambition for world domination perhaps <KarlJoad>ham5urg: I would also say that Nix has a more dominant position (being introduced a decade earlier), with a less pragmatic approach to packaging. That is not to say Guix's packaging is wrong, just Nix's can be more convenient for the average user too. <ham5urg>I like the way Guix handles packages. <jeko>how can I tell guix home to not alter my .profile ? haha <KarlJoad>I like the way Guix handles them too. More the _types_ of programs packaged. I prefer Guix's way, as I found good replacements for some of the non-free ones I used to use, but other people may not be so willing. <ham5urg>The only think Guix misses from my point of view are service definitions. Maybe I'm wrong and have yet to discover more. <jeko>singpolyma: that's a way ! thank you