IRC channel logs
2023-12-22.log
back to list of logs
<qrpnxz>no idea what it's doing. I'm thinking something is wrong. On my laptop it was really building, like maxing out CPU. It used some substitutes, but also built stuff and it's done. Here my only indication that it's alive is the terminal ascii spinner. Barely any CPU activity. <rekado>building the early bootstrap stuff takes a lot of time <qrpnxz>but a lot of time doing what? It's activity is not appreciable on my system monitor <qrpnxz>i think something else is messing with guix, but idk <recj>i am having trouble getting the setup for emacs <recj>for some reason, it's not autoloading the pakcages even though I added the guix profile subdirectory to my load path <recj>and installed emacs-guix <recj>I also added (guix-emacs-autoload-packages) to my init.el <recj>not sure if this is more of a question for #emacs though <ulfvonbelow>qrpnxz: there have been cases in the past of some tests of various packages getting stuck nondeterministically, though I don't know how many of those are in the closure of guix itself. It's not certain though, it could just be I/O-bound or latency-bound somehow. I'd recommend looking for processes in top running by a 'guixbuildNN' user, and maybe strace them to get an idea of what's going on. If the spinner is going, it means that the <ulfvonbelow>build log is getting output lines, so it probably isn't stuck. <qrpnxz>ulfvonbelow, ah yes, I found there's many process by guixbuildNN. Most are idle, but there is that one process with varying CPU activity, usually using 1-2 cores. I feel a little better now. Still a super long wait. <qrpnxz>now it has been building gcc-mesboot, it looks like <civodul>qrpnxz: are you testing core-updates? <qrpnxz>dunno if that answers your question <civodul>i mean, you’re rebuilding everything on purpose, right? <qrpnxz>yeah, i told it to not use substitutes. I am regretting it. On the other hand, it's pretty awesome it build EVERYTHING from source xD <qrpnxz>i need to see how to enable substitutes because apparently my 5950x CPU might as well be a pentium in this application <qrpnxz>can't find what i would need to do <davidh38>what is the easiest way to start pavucontrol at the start of guix <davidh38>I need this as a workaround for the chromium bug <ulfvonbelow>probably depends on the desktop environment / window manager <davidh38>I need help, please. I had dbus-run-session i3 and dropbox client from dropbox worked, but now it stopped working at dropbox: load fq extension '/app/extra/.dropbox-dist/dropbox-lnx.x8 <davidh38>dropbox: load fq extension '/app/extra/.dropbox-dist/dropbox-lnx.x86_64-189.4.8395/PyQt5.QtDBus.so' <davidh38>What am I Doing wrong with the dbus session, it worked flawlessly, but now it stopped suddenly <davidh38>does anyone have any idea. I am starting the dbus-seession-run i3 from .session <davidh38>I thought this solution would be run well forever <qrpnxz>mkay, so about 9 hours in xD processed stopped. Says I actually found a bug :o gonna report it <podiki>qrpnxz: what architecture are you on? <podiki>oh, well unless you are on a slow machine (and non-ssd) it should only take a few minutes at most <podiki>what commit are you on, from guix describe? <qrpnxz>SSD w/ Fedora 39 running on an 8c16t 5950x w/ 128GiB ram. It's not the hardware :) I'm building everything. idk what commit it is. I installed guix with the script from the site (seems to be guix 1.4.0) and what was talking 9 hours was "guix pull" podiki <podiki>and you were trying to do a guix pull? <podiki>which script from the site? under download the "standard" version, binary? <podiki>certainly report the error with the output/log; if you have that handy can you share it (via paste site) <qrpnxz>podiki, sure i'll make a paste. One min. <podiki>and no issues running that script to install? <podiki>and does fedora have selinux rules by default on? <qrpnxz>yeah no issues at all with the install script. selinux would not let guix function so i did setenforce 0 which puts it in permissive mode. After that I did "guix install hello" which took a while, but it worked! Then I decided to do the guix pull <qrpnxz>I actually saved the install log just in case. <qrpnxz>i'll send a paste of that as well <podiki>i'm not familiar with selinux but I know people have reported things before and had to work around it, but seems like you did that already <podiki>that's okay, i switch them all the time too, both ways :) <podiki>well you said no to substitutes in the install script (though why guix still polled the servers I'm not sure) <podiki>so we would expect long times for everything since you'll build from source <podiki>still, things should build unless something is broken, which it certainly could be <podiki>shouldn't since we have a build farm and people doing guix pull all the time too (though usually with substitutes) <podiki>so I would certainly submit a bug report, but did you want to build everything from source? <podiki>looks like something happened with trying to have too many build jobs running, I might run with less max jobs so that all can get enough threads and memory <qrpnxz>this is why heroes like me (:P) build from source like maniacs: finding the bugs 😎 <qrpnxz>Gonna try restarting see what happens <podiki>i think saying 16 cores and 16 jobs is not the most efficient, but it shouldn't fail <apoorv569>When there is a change in package definition, not the software just the definition, like I added libglvnd instead of mesa for opengl lib, I increase the `revision` by +1 right? <podiki>i would do "guix pull --max-jobs=4" maybe? dunno, 4 threads each sounds good, and if only 1 job can be running it will still use 16 threads if it can <qrpnxz>most of these jobs are gonna single thread often so i rather have 16 jobs even if sometimes they spawn more work. I have 16 logical threads! I want to be alive by the time this finishes. <podiki>apoorv569: for a git version package? I don't know, I think it is only when updating the commit but not the version otherwise? <podiki>qrpnxz: if you definitely want to be alive you would just use binary substitutes :) <qrpnxz>loool, well idk how to enable them now so ;;) <podiki>still, I would try with less as a test, I don't think you're hitting 16 jobs at this point something happened where it couldn't get another build worker <qrpnxz>but now that it failed building i want to build it even more, cause it should build <qrpnxz>gonna tone down the jobs and try again <podiki>so try with like 4 or 6 jobs, gives plenty of threads for each and enough for more things to build at once <apoorv569>podiki: Yes, its from a git repo. I have a `let` binding for `commit` and `revision` on top, just below the `define-public`. The revision is `0` ATM. <podiki>multiple jobs helps at times, but often at some point it goes down to one or just a few <qrpnxz>gonna do one job just to get rid of thread weirdness for a sec and make sure it still errors <podiki>apoorv569: you can check our git log for package updates; I always have to look for git package versioning as I forget :) <apoorv569>I only changed a dependency and added `PREFIX` in make flags. <podiki>qrpnxz: max-jobs without specifying threads is what I would usually do, I think specifying max threads is only useful if you want less than your cpu max, as most builds default to using as many as nproc <podiki>apoorv569: my instinct is you don't update the revision then but don't hold me to it <qrpnxz>podiki, makes sense. Default is to use however many so i should just remove that flag <podiki>yeah, if i remember from other conversations, max-jobs is to increase (from 1 default) while max threads is to decrease (uses max default) <podiki>everything we can builds with max threads, but some packages or tests don't like that so we have to use less; of course some things are single threaded before/after actual compiling <qrpnxz>seeing a lot of "updating substitutes 100%"... <podiki>so max-jobs I like to do more when I know I'll be downloading a lot or building quick little packages <podiki>yeah, not sure why that pops up since you disabled them in install; that sounds like a bug to me <qrpnxz>so far it seems happier with just one job. Either that or it's just taking longer to get to the error <podiki>your error suggests something wrong with how build workers were allocated; i'm not familiar with the details, but it seems like something that shouldn't happen <qrpnxz>indeed. What it looks like is that i said 16 workers, but iirc the install script didn't make that many users. Seems it's trying to make more workers than is possible with the current amount of users <qrpnxz>and it found some opportunity where it coulda done more than 10, and i allowed it by saying -M 16, but then it fails. (This is my speculation.) <qrpnxz>yeah, it has been running fine now <qrpnxz>also, looks like the worse part was the bootstrap. After gcc-mesboot et al I saw a lot more activity and resource use <podiki>yeah, the early stuff is slower and shouldn't have to do it again (until that is updated, which is not often) <podiki>that sounds right, ran out of labor <qrpnxz>Oh. Yes. Baby! 100% CPU usage. That's what i like <podiki>so the rare occurrence of actually needing more than 10 workers at once; still that should default to at least as many workers as logical threads available <apoorv569>How much time does it take for a mail to appear in the archives? <podiki>first messages from a sender go through moderation i believe for most lists <apoorv569>Its been more than 15 minutes, I don't see my mail in there yet. <podiki>your first message to that list? <podiki>so, as I said, there is a moderation queue probably, so you'll have to wait for someone to do that (likely in a few hours as most people are in europe and it is not daytime there) <qrpnxz>build of /gnu/store/cr5cmzdjs74s32m6yzfl076l292cr7vf-docbook-xsl-1.79.2-0.fe16c90.drv failed <qrpnxz>bad dependency? seems it can't find a bunch of files <adanska>is there any reason in particular why the home-zsh-service-type doesnt have a dedicated alias field like bash and fish? it seems like an odd omission <apoorv569>If I use the wireguard-service-type, is the VPN is always gonna be conncted? <alexey`>Hello there! A simple quiestion. I'd like to build the Guix system locally from sources on a separate partition. I've noticed that the installation image uses the `/mnt/tmp' directory. Suppose that `/mnt/tmp' is a separate partition (or logical volume). How big should it be? What's the recommended size? Is 10G enough? <lilyp>That depends on the size of your profile. I'd not recommend using big binaries like icecat or chromium from Guix if you do. <lilyp>Other than that 10G ought to fit a few generations. <alexey`>One should never have to touch files in /etc or to run <alexey`>commands that modify the system state such as useradd or grub-install. In fact, you <alexey`>must avoid that since that would not only void your warranty but also prevent you from <alexey`>rolling back to previous versions of your system, should you ever need to. (GNU Guix manual, Chapter 10, Section "Instantiating the System"). Well, am I right in saying that the user has no default protection from changing contents of the `/etc' unintentionally? <isaneran>alexey`: other than being only writable by root I dont think so <alexey`>Yes, that's what I implied. OK, no problem. <alexey`>/mnt/tmp/guix-inst/<hash-string-1>-libgit2-1.3.2-checkout/ <alexey`>/mnt/tmp/guix-inst/<hash-string-2>-libgit2-1.3.2-checkout <alexey`>Why two instances of the same package? <alexey`>Shouldn't it be just one directory for the `libgit2-1.3.2'? <alexey`>May this be a result of running `guix pull' twice? <snape>Hi, on the master branch Guix seems to want to build unusual stuff like gcc, mesboot, etc <snape>And i have those messages of <civodul>snape: try “rm gnu/packages/bootstrap.go && make” :-) <civodul>an ABI change that went undetected by the machinery <snape>civodul, it works, great, thank you! <Franciman>hi, does anybody have a zotero package definition, by chance? <efraim>rust-ring is just ... ugh. I might actually end up with a computed file for the sources <davidh38Guest3>running the dropbox client from flatpak gets stuck here: app/extra/.dropbox-dist/dropbox-lnx.x86_64-189.4.8395/PyQt5.QtDBus.so' <efraim>I've been using dbxfs for dropbox, not sure if it's missing any features though <alexey`>All right. Where is the default configuration file of the installation image itself (the one written to the usb stick)? There are 4 files in `/etc/static/configuration': bare-bones.scm, beaglebone-black.scm, desktop.scm, lightweight-desktop.scm. Are they what I am looking for? <efraim>it makes the files available through a fuse filesystem <alexey`>efraim: So, they are not available when I just booted the GNU Guix system from the stick, are they? <efraim>I don't remember where the installation file os image is located <alexey`>efraim: Ok, no problem. Thank you. :) <lechner>Hi, does the FSF mail server add the Debbugs stanza "Package: guix-patches" when the recipient is guix-patches@gnu.org, or is that done by 'git send-email'? <theesm>lechner: I don't think git send-email would add something debbugs specific; so it's probably something on the receiving site doing this (though I don't know enough about debbugs/the fsf mail infrastructure to pinpoint where this happens) <civodul>simendsjo: hi! the first message takes a while to reach the mailing list (there’s a “quarantine” period), but subsequent messages should arrive quickly <simendsjo>Thanks. I've sent several patches before, but it's been a while, so maybe I have a new quarantine period. <lechner>simendsjo / did you use a new sender address? <lechner>civodul / using the fcgiwrap-service-type on a remote machine seems to have broken 'guix deploy' <simendsjo>I don't think so, but I sometimes use a "tag" in my email address, so I might have sent other patches from simendsjo+guix@gmail.com instead of simendsjo@gmail.com. We'll just wait and see. If it doesn't appear early january, I'll ask again. <lechner>simendsjo / well, it's not on help-debbugs, which I now monitor for submissions gone wrong. i think you did everything right <euouae>I'd like to ask: how do packages change versions? Do they track git commits from upstream? or does the corresponding guix package maintainer put up a new release every now and then? <ieure>euouae, Someone sends in a patch for the new version. `guix refresh' can help detect new upstream versions. <ieure>euouae, Whether it's a Git thing or a tarball release thing depends on how the upstream handles releases. <euouae>ah so it's custom depending on the guile package definition <euouae>For example for grub I see mirror://gnu/grub/grub- ... <euouae>ending in tar.xz, and there's no primary url. I'm guessing that means it grabs from a mirror that (probably) syncs with grub's releases <euouae>okay yup, it tracks gnu releases. the mirrors in guix/download.scm:%mirrors show it <euouae>ah huh, there was a grub release 16 days ago from the previous one (2 years ago) <apteryx>ieure: by the way, Xfce already honors things installed under etc/xdg/autostart/ (via packages added in the operating-system 'packages' field), via the XDG_CONFIG_DIRS <gabber>i have this service definition: https://termbin.com/treo . Shepherd tries to start the process (which seems to work) but then reports it fails to start. the log does not show any issues (https://termbin.com/41ue) and starting the process manually also works fine. how could i possibly debug such an issue? <civodul>gabber: could it be that the PID file doesn’t show up? <gabber>should i delete the #:pid-file keyword argument? i guessed that was the facility to make (make-kill-destructor) work <civodul>#:pid-file instructs ‘make-forkexec-constructor’ to wait for the PID file, which should be created by the daemon <civodul>if the daemon doesn’t create such as PID file, then you can remove #:pid-file <gabber>i see.. but why wouldn't the daemon create the file? is it unable to fork? the exec part certainly seems to work <civodul>you have to check its documentation and see whether it’s supposed to create a PID file <gabber>ah, wait. you mean shepherd expects the subprocess (nMASC.py in this case) to create that pid file? <gabber>and is the process supposed to daemonize itself? <civodul>anyway, #:pid-file tells shepherd to wait until that file is created <gabber>does make-forkexec-constructor expect the process invoked to be run as a daemon? or does it expect the process to run in "foreground"? <gabber>like `emacs --daemon` starts a daemon in the background whereas just `emacs` runs a "foreground" process <civodul>foreground, unless passing #:pid-file <Kabouik>I just installed Guix on a new machine and rsynced my ~/.config to it from my previous Guix laptop (I know this is not the most guixy way to do that but I don't know how to use guix home yet, nor do I know how to set my configs in a declarative Scheme file). What would be the best way to list all the emacs- packages I have on the other machine to easily install them all on the new? <Kabouik>`guix package -I | grep emacs` gives me a list, but it's not exactly formated in the way I need it to run `guix install listfile` on the new machine. <euouae>Kabouik: guix package -I | awk '/^emacs/{print $1}' <euouae>you can convert newlines to spaces with `tr` or guix package -I | awk '{printf "%s ",$1}; END {printf "\n"}' <Kabouik>It does help, thank you! I'm sorry it turns out to be non-guix related, but I was wondering if there was a fancy guix command to just print a simple list. <Kabouik>Do I get a medal for running Guix system on the smallest x86_64 devices around? https://0x0.st/HgLR.jpg (the monitors are drivven by the smaller one on the left) <euouae>are these old machines of yours? <ieure>Kabouik, Nice! I've been thinking of putting it on my GPD Pocket (the original). <Kabouik>Those are as modern as it gets! The tiny thing on the left is a gaming handheld with an AMD 7840U and 64GB of RAM (yes :eyes:), the other on the right is a couple years old, with an i7 and 16GB, but easier to type on of course. <Kabouik>Dewit ieure! Those are GPD devices too. <ieure>Kabouik, I can tell. :) I used to have a GPD Win 2 as well, but it died after a year and GPD doesn't really offer support. Pity, it was a nice little thing. <Kabouik>I actually have had a relatively good experience with GPD support, but admittedly no machine really died on me. The Mini on the left is kinda the same size as the Win 2. <ieure>Yeah. Some other manufacturer finally got around to making another model with a physical trackpoint, and they're pretty inexpensive as well. Been thinking about one. <ieure>Problem with these sorts of things is that none of them support proper sleep, which is really unfortunate. Have to turn them completely off, which is a lot of friction to use vs. a smartphone, or they drain the battery in 24h. Irksome. <ieure>It is very nice to have a real computer the can do real computer stuff and also fits in your pocket. Phones are not real computers. <Kabouik>Yeah. The laptop on the right is my daily driver but I put it in Hibernation all the time, no sleep unless it's just for a couple hours and not in a bag. <ieure>Yes, I'll buy 10 of whatever the first device to handle that stuff properly & be reasonably supported by Linux. <ieure>MNT Pocket Reform looks interesting, also a fully open hardware design, which is terrific. <Kabouik>My phone kinda acts as a sub-real computer for me, running Droidian (Debian onto the Android kernel) + Phosh (Gnome) + Guix package manager, and it has a built-in keyboard (I'm more comfortable and faster with it than with the GPD Win Mini's actually), and DP Alt-mode. But it's aarch64 and a lot slower, of course. <zilti>I was hoping for the astro slide, but turns out it is apparently trash <ieure>But from what I hear about the full-size ones, not likely to have great support for what I do. I believe the full-size only properly supports Wayland, which means no EXWM, which means I won't use it. <ieure>zilti, I bought the Gemini PDA, and regretted it. Awful. <Kabouik>I have the F(x)tec Pro1 (and Pro1x), which is the main competitor to the PlanetCo devices, and I love them. Been using them daily since 2019, and even more now that they run Droidian for a full Linux experience. But it's still not mainlined. <ieure>Kabouik, I gotta try UBPorts on my Pixel 3a again. I was using that for a bit, and it was just about reasonably daily-driveable. I switched back because I had a bunch of text messages go missing, which meant I couldn't 2FA into stuff, but I'm convinced this was a carrier issue now. <Kabouik>I just dislike how UBPorts doesn't really support apt out of the box due to the read-only way it works. <ieure>Also have an original (Braveheart) PinePhone. It's not usable and I don't think it ever will be. The Pixel 3a with UBPorts is like 100x better. <zilti>ieure: yea, it's unfortunate. Their lineup looks so great at first glance, then keeps getting worse. <Kabouik>Yeah my Pinephone (and its keyboard) are in a drawer too. They don't compete with my Pro1, despite the more open hardware and kernel. <ieure>zilti, Yeah, I think I'm done buying their ewaste. <ieure>They also rolled out some cryptocurrency payment nonsense. I despise all that trash, it significantly lowered my opinion of them as a whole. <zilti>I'm holding out with my fairphone 3 until there is a usable phone without Android <ieure>zilti, And it was like, this year! Long after the hype bubble had burst and everyone understands what harm the stuff does. awful <Kabouik>I really like Droidian, but admittedly I have a keyboard on my phone, and that is night and day in terms of experience with an OS like that. I can live mostly in the terminal for many things like file manager, emails, irc, telegram, etc. But it does have all Debian GUI applications too of course, as long as they're compatible with aarch64. Many of my packages on the phone are actually installed from Guix. <Kabouik>Droidian is still very experimental, though. <Kabouik>I could actually make a shot of a Guix neofetch on a 5.9" keyboard phone, the 7" GPD Win Mini, and the 8" GPD Pocket 3. :> <Kabouik>I'm looking for examples of how to configure the nbfc-linux service in config.scm, if anyone is using that. <euouae>apteryx: do you have an encrypted guile system I could look into its config.scm? <euouae>I'm struggling to get it working, I thought changing to PBKDF2 would work but I still can't get it working <apteryx>a fully encrypted disk install configuration for Guix System? <apteryx>PBKDF2 being the older format more widely supported by GRUB? <apteryx>there's nothing special in my config; that should be supported out of the box using mapped-devices with the type luks-device-mapping <apteryx>it's complicated because of 5 drives in a Btrfs RAID <apteryx>well, not really complicated, but verbose :-) <apteryx>I'd just recommend to double check the UUIDs, these are easily confused <euouae>I'm putting in the uuid from cryptsetup luksUUID /dev/sda2 <euouae>and then grub complains about not being able to find the uuid of my ext4 partition <apteryx>putting that UUID in the mapped-devices ? <apteryx>luks UUID goes for the mapped-device entry of type luks-device-mapping <euouae>instead of defining %mapepd-devices I have the form mapped-devices inside operating-system <euouae>my setup is a DOS partition with /boot and / <euouae>maybe it fails to find device via file-system-label. <apteryx>i'd recommend using a UUID instead of a label <apteryx>uncomment mapped-devices, change its source to something like (uuid "..."), uncomment (dependencies mapped-devices) and modify the file-system device field to also use (uuid "...") <apteryx>the UUIDs will be different; for the luks-device-mapping it's that of LUKS; while for the ext4 file system it's that of that partition. Both can be viewed with 'blkid' when the fs is mounted <euouae>what's your invocation to cryptsetup luksFormat? <euouae>I think I just did `cryptsetup luksFormat --type luks2 --pbkdf pbkdf2 --hash sha512 --key-size 512` and now I'm wondering if those parameters don't work <nckx>podiki: <guix-mirror> Oh, dear, you're right. <sneek>Welcome back nckx, you have 1 message! <sneek>nckx, podiki says: I think guix-mirror stopped mirroring a couple of weeks ago <Kabouik>Has anyone gotten https://github.com/FlyGoat/RyzenAdj to work on Guix? It's for setting the TDP on AMD Ryzen devices, among other things. I successfully compiled it but it still shows many errors when I run `./ryzenadj --info`, I'm guessing it's missing kernel modules (I'm using the Linux kernel) or fails to find things where Guix puts them? <Kabouik>I could adjust values to 20W using `sudo ./ryzenadj --stapm-limit=20000 --fast-limit=20000 --slow-limit=20000 --tctl-temp=90`: https://0x0.st/Hgp8.txt But I'm not quite sure how to control current TDP to see if everything works as expected, if --info shows errors <euouae>well it gives failed to map /dev/mem: Operation not permitted so that's that issue <nckx>What the hell. I don't actually remember how the GitHub automatic mirror was set up, but either I've gone blind (not implausible in the convoluted labyrinth of this unusable Web site dot com) or the functionality was… removed? Surely the latter can't be true? <Kabouik>It does say "If you don't want to change your memory access policy, you need a kernel module for this task." euouae, but I don't know where to look to add that to Guix. <euouae>Kabouik: apart from /dev/mem the other error is related to a missing pm_table, do you have the ryzen smu driver? <Kabouik>I don't think I do, couldn't find any doc on how to add them <euouae>also you're opening up yourself to attacks with these, so I hope you trust what you're doing <Kabouik>I don't know much what I'm doing there, no, but having a handheld with a 7840U without being able to adjust the TDP would be too bad <euouae>what can I say -- I can't even get an encrypted partition to work <Kabouik>not* actually. Damn keyboards are difficult. <euouae>I thought you need setcap for that <euouae>and loading the kernel module can be done out of boot I think you just need to compile it and use modprobe <Kabouik>I'm really in unknown territory there, no idea how all this works. I also need to find how to set up nbfc in Guix. But when I do get past that, the machine should be all set. <euouae>I don't know how to add a kernel module the proper guix way <Kabouik>I don't know how to add a kernel module, period. :> <euouae>I spent 7 hours getting nowhere today ffs <lilyp>btw. before I ask on guix help, how do I get networking in guix system containers? <euouae>nckx: I would imagine you'd define some channel with modified kernel build instructions? <euouae>apteryx: yeah I can't get it to work <Kabouik>euouae I see lm-sensors can show PPT in W and average PPT, which seems to correspond to the values I set with ryzenadj --stapm-limit=xxxxxx --fast-limit=xxxxx --slow-limit=xxxxxx, so maybe I can just skip adding the kernel modules if I can check current TDP with that <Kabouik>It shows the actual consumption though, not the TDP, so I guess if there is no demanding task, it won't correspond to the tdp <apteryx>euouae: perhaps have a look at system tests under gnu/tests? <apteryx>at least one tests an encrypted root, if I recall correctly <apteryx>you'd see the commands it uses to encrypt the disk and create the partitions as well <nckx>euouae: I meant it literally: add a module how? Add a new module to the Guix package collection (the linux-module-build-system makes this trivial)? Make a module available for (auto-)loading (kernel-loadable-modules in your OS definition)? Add a module to the initrd, which will forcefully load it at boot (initrd-modules to the rescue)? <euouae>nckx, Kabouik is trying to use some software that requires a kernel module, and I recommended using mobprobe, I don't know how to install it properly with guix with the methods you described <nckx>podiki: I have no idea how to re-enable mirroring, the suggested workflow (‘GitHub Actions’) is downright ridiculous, expecting you to add files to your repository AFAICS. I'll try again tomorrow. <euouae>apteryx: great, thank you. those tests help <nckx>euouae: If you want ‘sudo modprobe foo’ to work and there's a package that provides foo, you can use (kernel-loadable-modules (list foo …)). <euouae>I think there is no package for it nckx, that's my understanding <nckx>Then the first ‘add’ is as a package using the linux-module-build-system. <euouae>great, so linux-module-build-system is where you'd start with that -- but still you can use modprobe for a quick test <nckx>I've changed plans & will return in half an hour or so. o/ <podiki>nckx: if it is too much trouble we can abandon it of course, there are a few already on gitlab I think and is easy to setup or on another git host <nckx>euouae: Is that the same module with which you wrestled for 7 hours? <euouae>nckx, ah no, I've just not been very productive this day. I looked GRUB trying to figure out how Argon2* can be added. Eventually I decided to use PBKDF2 for now and then I got stuck with that, a bunch of attempts to deal with it on my VM <euouae>(I was trying to figure out how to patch GRUB or understand the situation around it) <euouae>I sent an e-mail to grub-dev maybe something will come of it <nckx>Kabouik: For ryzen-smu, I expect adding it to kernel-loadable-modules should suffice. Did you try that already? If you're running the default linux-libre you can directly modprobe the result of ‘guix build ryzen-smu’ too, but not otherwise. But IIRC (kernel-loadable-modules …) are automatically built for your (kernel …). <nckx>If this is not about ryzen-smu then I got confused. <nckx>podiki: I'm not actually sure how much trouble it would be, yet; it's either terribly documented or I'm terrible at finding what to search for. <Kabouik>euouae, nckx: sorry I was away from IRC trying things. Turns out I may not need ryzen_smu at all (or I don't exactly know what for, which means I should probably not dig too much). I added `iomem=relaxed` to my GRUB kernel parameters in config.scm, and now `ryzenadj -i` works just as it should (I think): https://0x0.st/HgpY.txt <Kabouik>This means I should be able to adjust my TDP and check the current value using ryzenadj to set stapm, slow and fast limits. <Kabouik>I'll look into making a Guix package for Ryzenadj, if I can manage to write the definition for it (built it in guix shell using the build instructions). <euouae>ah nice, so iomem=relaxed is another way to get access to /dev/mem <Kabouik>Now I need to find how to use nbfc-linux on Guix, florhizome is the one who wrote the package, but I haven't seen them in a while. <nckx>Kabouik: It does, but it's technically less secure (assuming that the ryzen-smu module exposes the required range in a more controlled way). <euouae>sigh I keep falling into grub rescue <nckx>However, it's a grand workaround. <Kabouik>nckx: Hum. You might be right, but I think I'll have to cope with the extra risk for now (unless it's huge). I admit I don't know how much of a vulnerability surface that is (at all). <nckx>Kabouik: Yeah, me neither, I just felt obligated to point it out. <Kabouik>And thanks for that, I totally overlooked it (and now I will voluntarily ignore it by lack of knowledge to counter it, but hey) <euouae>I should maybe remove all the uuids because maybe dos mbr can't deal with <nckx>euouae: Why comment out (dependencies mapped-devices)? <euouae>ah, huh? I thought that was commented-in <nckx>(I'm not sure what you mean by ‘DOS MBR’ but it doesn't deal with, well, much of anything, and certainly not UUIDs.) <euouae>I don't understand early boot very well so I may not be coherent <nckx>This installation predates LUKS2 though, and I don't personally recommend using LUKS2 with Guix System at all, there are just too many missing pieces. Yes, I'm aware that it should® work™ for certain option subsets. <nckx>But that doesn't mean I go around recommending it 😉 <nckx>At least one person followed the advice in the manual and still couldn't boot. <euouae>LUKS1 vs 2 is just a format right? <euouae>the encryption parameters of pbkdf2 are still the same <euouae>installing for the umpteenth time <nckx>It would at least rule out $everything_else as culprit and if it does work, you can focus on your LUKS settings rather than your Guix System configuration. Seems like a bisection win. <euouae>I was hoping to get Guix working so I could at least submit some minor doc phrasing correction patches <euouae>isn't /etc available in guix too? <euouae>yup, anything I do with LUKS2 lands me in grub rescue. SIGH <Kabouik>Oh, so just editing /etc/sudoers would work in Guix too? Nice. <nckx>Kabouik: Not really, /etc/sudoers is not persistent. Instead pass your file to ‘sudoers-file’ and reconfigure. <nckx>E.g. (sudoers-file (local-file "sudoers")) ; in the same directory as your system.scm <nckx>You can also use ‘plain-file’ to declare it in-line. <Kabouik>I see, thank you. I will try that. Getting close to having my machine fully set up. I still need nbfc and taps to be recognized by the trackpad (xf86-input-synaptics is installed byt that was not enough; I'm on Wayland, maybe that's why). <nckx>Uhm, yeah, X drivers are pointless if you're not running X(org). <qrpnxz>got what looks like a reproducibility error? "/sha256 hash mismatch for /gnu/store/qyd6h3h0piaxj1wpibar66wjm8vycvdg-apr-1.7.0.tar.bz2" <euouae>oh, so it was LUKS2. OH lord have mercy <nckx>Taps are always ‘supported’; what the compositor [tells libinput to] does with them is controlled by it, and how you configure it. For example for Sway you'd use something like ‘input type:pointer {\n tap enabled\n … }’. <euouae>Hm, now I get the decrypt screen, I enter password, it shows me grub menu, but after choosing a kernel it goes back to showing me the decryption screen and it is stuck there <nckx>qrpnxz: The Internet is notoriously irreproducible. What was the complete output (or at least the misbehaving URL)? <nckx>euouae: The exact same screen as the first time? <euouae>"SeaBIOS ... Attempting to decrypt master key... Enter passphrase for hd0,msdos2 (...)... Slot 0 opened" is what I see <nckx>euouae: ’Slot 0 opened’ means the key was correct, so it looks like you're just looking at uncleared output? Guix will prompt for the passphrase twice (because there's no support for ‘key files’ or other ugly hacks to avoid that), but the two prompts are quite different and even blindly typing the passphrase a second time should boot normally. <lechner>ACTION thinks half of all problems here relate to LUKS <euouae>nckx: why should there be two requests for password? I only have root encrypted <nckx>Because both GRUB and Linux need to decrypt root. <cobra>The pipewire SPA headers (Simple Plugin API) don't seem to be shipped in the `pipewire` package. Am I missing something or are they not in guix? <Kabouik>nckx: Hum, but I use Sway on both my Guix machines, and taps work out of the box on one, not on the new one. <nckx>qrpnxz: That is potentially fascinating, because the host www.eu.apache.org does not exist. What's in the /gnu/store/qyd6h3h0piaxj1wpibar66wjm8vycvdg-apr-1.7.0.tar.bz2 file? <nckx>Kabouik: Err, OK. Still, there's no such thing as a ‘Wayland driver’ (or a ‘libinput device driver’, libinput being what most if not all compositors use under the hood). So if your compositor really isn't seeing any events when you tap, I'm forced to blame… the kernel driver? That would be weird… <nckx>If catting the corresponding /dev/input/eventN device doesn't barf garbage when you tap, but it does when you press a button or stroke the pad, then that's not something I've encountered before. <qrpnxz>nckx, file seems not to be there at the moment... May be getting cleaned up <Kabouik>sudo cat /dev/input/mice does throw some characters when I move my finger around the trackpad, but nothing when I tap (and the trackpad has no buttons, so I cannot see what clicks would show) <Kabouik>On the other issue before, my /etc/sudoers now looks like https://0x0.st/Hgfb.txt, but that doesn't seem to do what I want. That command still requires being run with sudo or it shows errors. <Kabouik>Oh wait, I think I get it now. It still needs sudo, just won't ask for a password. I'm dumb. <nckx>What were you expecting? <qrpnxz>i made a loop that tries to copy the file and reran guix pull. It errored again but seems i was able to copy... the empty file. Perhaps it was not able to copy before truncation? <nckx>qrpnxz: Admirable, but probably not a good use of your time 😉 Does curl not return anything? (For context: my current suspicion is some dodgy router or ISP hijacking NXDOMAINS or other such madness, because that hostname really ought not to exist.) <qrpnxz>i'll try guix shell curl, but i don't already have curl installed in guix so that could mess up idk. But i'll try it <nckx>You can use ‘guix download’ too. <nckx>curl's just my muscle memory. <cobra>ah, nevermind, ive found them <qrpnxz>guix downloading that url returns: <qrpnxz>0mdqa9w1p6cmli6976v4wi0sw9r4p5prkj7lzfd1877wk11c9c73 <qrpnxz>file: /gnu/store/yr88y4s17j38m991djzzddnchwp3kq0c-apr-1.7.0.tar.bz2 <qrpnxz>different hash there on the store <nckx>See, that shouldn't succeed. Shenanigans! Afoot! And the contents of that file? <nckx>ACTION places bets on HTML. <nckx>Double my money if it's an ad. <nckx>Yeah. You beat me to beating you to it. I'd just figured out that's the hash of the empty file. <nckx>Your ISP(?) returning an empty file is less funny and a lot less explainable. <qrpnxz>nckx, thing is that i tried sha256sum on empty file and it's a different number than guix says <nckx>Yeah, don't worry about that, Guix uses it's own special thing. It's not even the same base32 as real base32, and sha256sum uses base16. <qrpnxz>if i go to the link on my browser it also downloads 0 bytes <nckx>I'm feeling ready to blame your network, somehow, intercepting at least HTTP requests to nonexistent domains for unclear motives. <nckx>An arguable shortcoming of current Guix is that it takes HTTP 200 sucesses at face value, and throws an error instead of trying the next URL. That can sometimes be helpful (to sniff out rotten mirrors), and sometimes it's not. <nckx>Since the mirror is dead I'll remove it. <nckx>But it's something to keep in mind about your network if you notice other weirdness. <nckx>What's the IP adress you're getting for www.eu.apache.org? <nckx>Something like ping will print it if you don't have better tools like drill/dig. <nckx>Is that the 66.110.96.22 in your paste? I'm not actually familiar with tracepath output. <qrpnxz>www.eu.apache.org. 9 IN A 143.244.220.150 <nckx>HOW HELPFUL (╮°-°)╮┳━┳ ( ╯°□°)╯ ┻━┻ <nckx>I guess you're on AT&T? (Because I'm not! :) <nckx>If so, it seems that they are using a DigitalOcean droplet (lol) to hijack NXDOMAINS so they can brand your experience. <qrpnxz>and the IP gives 200 on http GET as you said <nckx>Unbelievable but not unheard of. <qrpnxz>host command reports the IP and also says: <qrpnxz>Host www.eu.apache.org not found: 3(NXDOMAIN) <qrpnxz>so, sounds like you know what you are talking about :) <nckx>It happens. Enjoy while it lasts, results not guaranteed, see your doctor if it persists. <nckx>However, I seem to be unable to download that file from any of the listed mirrors in Guix, so there may be a bug beneath a bug. <qrpnxz>i should say section on hijacking, the article is on the NXDOMAIN <nckx>This might be an EU vs US thing, or even part-of-EU thing, because I'm very grateful never to've encountered it. <nckx>My ‘I can't find that file on any mirror’ above was just PEBCAK (I blame distractedly yakking on IRC) because it's available, just moved to the archive domain: ‘guix download http://archive.apache.org/dist/apr/apr-1.7.0.tar.bz2’ should put this file in your store, where subsequent Guix commands will find it until you GC. <nckx>I'll still remove the nonexistent mirror domains but only after I've found out what happened to them and whether we should add new ones. <vagrantc>it's been a few weeks since i've updated my guix system ... have there been world-rebuilding events this month? 
