<nckx>If you could send a bug report to bug-guix at gnu dot org with all details (minus your IMAP password, I'd say) it would be much appreciated. <bluekeys>Thanks nckx, NieDzejkob, mroh. I'll start by sending the bug report later gtg <nckx>Nice. Simply adding (outputs (list "out" "debug")) was enough. <nckx>Had to guess that by looking at the source, could be documented better that it's automagic. <nckx>Yes, clear as day, if buried at the bottom. <nckx>mroh: I just never read the manual, only docstrings đ <alextee[m]>nckx: oops just saw it, thanks! gonna try that now <NieDzejkob>it would be nice if docstrings and the manual were one and the same in this regard... <nckx>For services that would be great. <NieDzejkob>nojr: I couldn't find a blog post nor anything in the cookbook, but there's digital-ocean-configuration in (guix)Invoking guix deploy. Not sure if you already found that <alextee[m]>hmm guix system: error: service 'guix-daemon' provided more than once <mroh>would be nice to have substitutes for an (implicit) "debug" output for every pkg (or some expensive (c++) junk^Wlibs). <gnutec>nckx: guix pull and guix package -i unknown-horizons. Nice job! <NieDzejkob>we don't have the capacity for debug symbols on every package, so we have to rely on people nominating things where they're actually useful <nckx>...I have no idea what I am doing dot gif dot gdb :-/ ⌠<nckx>I've added symbol and source directories. Why doesn't âwhereâ show me line numbers or anything useful? <nckx>It does that for main() but not for subsequent functions in libmailutils, but that's part of the same package... <nckx>Oh, I'm pretty sure it's dead. <nckx>That would explain the blood. <nckx>GNU manuals are friendly but boy do they assume you're ready to sit down for an hour with a nice pot of tea. <nckx>I'm not hear to learn! I'm here to copy/paste garbage into my terminal until something useful comes out. <joshuaBPMan>I did the same thing with someone's email footer in guix-devel. It had some disgusting looking C code. and I thought, I'll just run this random code and see what happens! <nckx>joshuaBPMan: ...and now your bank account is mysteriously empty and you get strange phone calls in a foreign language? Don't keep us guessing. <nckx>Hm. Second random search result answering âHow do I use GDB?â with âUse valgrind.â <nckx>Woohoo, babby's frist working breakpoint. <nckx>All I had to do was recreate the build-time hierarchy, so create /tmp/guix-build-mailutils-3.9.drv-0/ & unpack the --source there. <jsoo>butterypancake: I have some patches to expose those definitions. I should submit them <joshuaBPMan>nckx: Well I did get some random "your apple product was shipped". Click here. <nckx>So they knew your secret purchasing habits? Creepy. <joshuaBPMan>nckx: I actually haven't made any apple purchases...maybe that was obvious. <terpri>about to reconfigure my multi-ssd setup. any advice on using disk images inside of btrfs for VMs (with CoW disabled for them of course), vs. using LVM for VM disks? <terpri>btrfs used to have (reportedly) terrible performance for VM images even without CoW, dunno if that's changed <terpri>lvm would obviously be a bit faster...but also way less convenient <terpri>hm...fallocate and using raw disk images instead of qcow2 would probably help <bdju>so guix package -I works for checking the version of a user package, but how do I check for a system package? I tried just adding sudo and I tried replacing 'package' with 'system' <nckx>Probably by parsing /run/current-system/profile/manifest. <terpri>or guix package -p /run/current-system/profile -I <nckx>terpri: That VVFAT thing is both insane and cool. <nckx>bdju: guix package also takes a --profile argument, try that. I don't know if it works on systems. <nckx>...according to terpri it does. Nvm. <bdju>nckx: your first suggestion worked well enough actually <bdju>terpri: for some reason I got no results for what I was looking for with that command <bdju>but just browsing the manifest in less I was able to find it <terpri>also guix package --list-profiles is supposed to list currently active profiles, but isn't recognized as an option on my system <terpri>or...nm, must've made a typo or something ***rEnr3n0 is now known as rEnr3n
<terpri> https://askubuntu.com/a/1074499 hm, maybe i should just put guix on a terabyte ssd for now and see if i can pcie-passthrough a separate nvme ssd in the future. i don't need *that* much space now that i have a sata hdd for large archives and backups, just getting a bit cramped with guix + lots of giant monorepos under $HOME on a half-terabyte drive <terpri>it's also a bit strange that the uefi system partition needs to be so large, when all it really needs to contain is grub (~10 MiB) ***Guest58294 is now known as daviid
<nckx>I see. Apparently the minimum size on 4K-sector drives is ~260M(!), up from the previous 32M (which mine is). That's wasteful indeed. But it's a requirement of FAT32, not something we can change. <terpri>at that point one might consider using efistub to boot linux directly from efi-space :p though i suppose you'd still need grub for menus etc. <nckx>Apparently some (most?) machines can boot FAT16 and even FAT12, making the minimum size 32K đ Go nuts. (GRUB is 128K, not sure where you got ~10M). <joshuaBPMan>nckx: I didn't realize that grub was 128K! That's super tiny! <nckx>N.B.: that's only the EFI executable (grubx64.efi). /boot/grub/x86_64-efi is 3.3M here but that's on a different partition. <terpri>locales, modules, a unicode font -> ~10mb <nckx>Considering that's the kitchen sink, 3.3 for all possible drivers is not too bad). <terpri>(locales like...en@piglatin.mo, haha) <terpri>oh wait, of course grubx64.efi is the relevant file, i was just confused poking around /boot (which is just on the main filesystem on my machine) <nckx>My initrd + kernel is 40M... <pkill9>how do i add clang version 10 to a package's inputs? <nckx>I don't think that's typical, but don't have a typical system to compare. <nckx>pkill9: ("clang" ,clang-10) ; but I assume you mean something more. <pkill9>ok, i'll try that, it's just that `guix edit clang` returns a function <pkill9>a function that generates a clang package definition for different versions <nckx>Right, that's how the clang-Ns are created. <nckx>Doesn't mean they're not regular packages though. <pkill9>it's just it didn't return a variable <pkill9>i guess it generates the variables? <pkill9>as in, guix edit didn't send my editor to a variable <nckx>By âreturnâ you mean guix edit didn't jump to what you expected, or do you mean that âclang-10â in the REPL didn't? <nckx>I can see the logic in both. <terpri>so maybe i'll start with just moving guix to the 1tb until i run out of space <terpri>while i'm fiddling around with this...is there much utility to using btrfs subvolumes beyond separate ones for /home and the rest of the system? (for backups etc.) <terpri>maybe a third for /gnu or /gnu/store? <nckx>Unless you have a use case in mind I wouldn't. You can always boot from USB later & copy /gnu to a subvolume if you change your mind. <nckx>Otherwise keep / and /gnu on the same subvolume so snapshots have matching /gnu & /var/guix/db. <terpri>i don't really have anything important and non-guix-controlled outside of $HOME, except for /etc/config.scm itself, and some other trivial customizations in /etc <nckx>sneek: later tell bluekeys: Can you update mailutils & try again? Here, it now (seemingly) freezes. Progress! <sneek>Welcome back peanutbutterandc, you have 2 messages! <sneek>peanutbutterandc, nckx says: Look at the â'("FOO" ":" = ("BAR")))#t)))â line. This returns #t and ends the procedure. Then, under that, you try to return #t again outside of the procedure body. It's basically gibberish at that point. If you fix that up it should run fine. Good luck! <sneek>peanutbutterandc, nckx says: You don't have to use emacs but I recommend using an editor that at least highlights the matching â(â when you hover over a â)â. Highly-paid professional Scheme programmers don't count parentheses; they ignore them. <peanutbutterandc>nckx, I've got that package definition working and have figured out some more about the nature about the bug. BTW, emacs does not highlight matching parens for me when I am reading (not writing). guix installed emacs. `emacs -nw`... strange (vim does, BTW) <peanutbutterandc>It turns out, (wrap-programs defined in the package definition do not over-write each other. However, wrap-program does over-write if there already was a 'wrap phase in the build system. And that is the problem. <peanutbutterandc>I will be putting up the package definition to prove it in a few minutes. I'm trying to create a stand-alone version, too. ***rEnr3n1 is now known as rEnr3n
<peanutbutterandc>Okay, for some reason I can't have attachments on my email... so the bug report has not yet been filed. But here is a stand-alone definition of the package that can be used to see the bug like so: `guix build --verbosity=2 --file=this-definition.scm` https://termbin.com/on7c ***apteryx is now known as Guest11016
***apteryx_ is now known as apteryx
<abralek>@andrew it depends, but you need to run it anyways. If you use gnome you can add it to aurostart, if you use xsession you can run setsid shepherd from the . xsession <rndd>hi everyone! i'm trying to build emacs from tarball in guix env. during configure i got this error "The following required libraries were not found: gnutls". but i pul "gnutls" in manifest. am i missing something? <rndd>raghavgururajan, nope it is same <nefix>hello! I've created a docker image with docker system docker-image. How can I login as a specific user: the manual says to login with 'docker exec -ti $container_id /run/current-system/profile/bin/bash --login' (it logins as root) ***zap1 is now known as zzappie
<zzappie>nefix: hello. I'm not 100% sure but think you just need to replace /run/current-system-profile/bin/bash --login to your users's shell specified in /etc/passwd <zzappie>oh no acrually I think I wrong since they pintinng to the same bash binary anywa <nefix>zzappie: yeah, it points to the same binary *zzappie omg I just broke own typos in a single sentence record <nefix>zzappie: found, I had to pass the -u parameter to the 'docker exec' command <joesmack>when i look under 'latest' downloads i can choose 'guix on linux-iso' or 'vm-image of guix on hurd'.. is the stable version download of 1.1.0 then guix on linux or on hurd? <joesmack>oh... i could also just read what the page says... nevermind .. <zzappie>Hey anyone knows how to go about packaging projects that use git during build? As I undderstand git-fetch provides tree snapshot, no the repository itself. <zzappie>nckx: Guixoids is the official name? :) <nckx>zzappie: What does the package want to do with git? We almost always solve it by patching/otherwise disabling calls to git (many packages use something like VERSION="$(git describe)"; we just substitute ,version directly). <nckx>zzappie: Ha, no, not at all. <nckx>Users of Guix are guix, I'd say. <zzappie>nckx: Guixheads would be another one <nckx>Users of Guix on Hurd are guix/nurds. *janneke feels sooo welcome <zzappie>Mabe also Guixers since Gui users are Guilers also <zzappie>nckx: I'v packaged project called zenroom recently. And yes I did also supstituted bunch of stuff there. <zzappie>this means that one needs to update all the substitutes every package update manually, not very convinient <nckx>zzappie: Can you elaborate? Or point to the code if it's obvious? <zzappie>wait a sec I havent yet published It anywhere <nckx>It's still down? âš It's still down. <nckx>zzappie: According to an actual lawyer who makes actual money doing actual lawyer things, GPL means you have to publish all code you ever write immediately!! brb imma sue you now. <zzappie>there are hardcoding things like "VERSION_COMMIT = \"1ad133d\"\n" <zzappie>this means that every time I update package i need to update these string also <nckx>I see. Remember, everything is a code, you can use ,(string-take commit 7) and Guix will do it for you. <nckx>E.g., (format #f "VERSION_COMMIT = \"~a\"\n" ,(string-take commit 7)) should work here, though I didn't test it. <zzappie>nckx: thanks. But do you know what the advantage of not having .git during build? Is it for reproducibility purposes? <nckx>I think there are reproducibility concerns, yes, and size (the first .git directory here I randomly du'd is 2.0G, although that's Linux so probably cheating) is another one. <nckx>And, speaking personally, what's the *point*? I mean, why do upstreams do this? I look at that code & I don't get it. <zzappie>You mean why do they $(shell git stuf)? <nckx>Yeah. Debugging/bug reports I guess. <nckx>Making your version control system du jour's metadata integral to building your software without patching just doesn't sit well with me. But maybe that's a me-problem! ***zap1 is now known as zzappie
*nckx is an old and must accept that gits are the new tarballs. At least it discourages code drops. <zzappie>nckx: I aree. New projects are wild in terms of building workflows. <nckx>raghavgururajan: I can't completely deny that đ I didn't change it just to change it (I'm not that daft), but didn't indent it back to the original after playing with some other flags, it's true... <nckx>zzappie: Oh, your paste vanished already, I was going to modify it. <zzappie>im new to autotools, but despite its precieved fragility projects using autotools are the easiest to build from my experience. ***nckx changes topic to 'GNU Guix | get Guix at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs & patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: http://logs.guix.gnu.org | 1.1.0 is out! https://guix.gnu.org/blog/2020/gnu-guix-1.1.0-released/'
<nckx>peanutbutterandc: Thanks! What was the trouble with attachments? <peanutbutterandc>nckx, Oh it was a gmail thing, really. Nothing to do with our system. (: <zzappie>nckx: hm I thought I set paste to one month <nckx>zzappie: Aha! It's not just me. This happened to me as well, I blamed myself, now I blame the gnome-bin. <raghavgururajan>nckx: Hahah. But mine was OCD. I always wanted the *list* to *actually list* the items, :-D <nckx>peanutbutterandc: Just a style suggestion. Your previous code is correct, but the style I suggest here is more... Guixy? Modern? <peanutbutterandc>nckx, May I learn that please, sensei. I don't understand it. Is that G-expressions thingy? <nckx>Nope, G-expressions #~(look #$like this) and aren't related. <peanutbutterandc>all that being able to be written as (no %foo 'globals') part goes off my head... <nckx>peanutbutterandc: If this confuses you I shouldn't have posted it at this time; it's not that important. It really is just a style/idiom thing. <nckx>%build-inputs and %output aren't going anywhere, they will work probably forever, it's fine. <peanutbutterandc>nckx, But I am already interested. It is because of your suggestion why I am wrapping programs to not use propagated-inputs in the first place. (: <nckx>zzappie: Guixotic. (That was actually a distro name suggestion once...) <nckx>peanutbutterandc: Then I'll try to explain it better another time đ <peanutbutterandc>nckx, Okay! (: Please do tell sneek to tell me, if I am not around. Your method does seem to be a bit more succinct. <raghavgururajan>ghc-8.6.5 has been building for 4 hours, despite -c24 and -M1, on bayfront. <zzappie>nckx: Guixotic -- awesome this is actually captures the feel somehow. Very guixotic distro <nckx>zzappie: âThe page you are looking for does not existâ again. I wonder if it tries to do something clever with cookies that IceCat blocks. Just use paste.debian.net now that it's back. *nckx was trying to *save* you time, believe it or not. <nckx>raghavgururajan: I'm afraid I don't have any of those .drv. And g.t.gr is fully loaded today, I don't think it would outpace an idle bayfront with a head start. <nckx>raghavgururajan: The running ghc build does use 24 cores sometimes. Other times it's limited to ~5, due to dependencies between files. Nothing you can do about that. It's using as much as it can. <nckx>Wasn't there a recent change *in the kernel* to speed up âmake -jAlotâ? It's probably arrogance to think we'd benefit from that, but bayfront is âonlyâ on 5.4. Who knows... <efraim>Is there a patch flag I can add to ignore differences in line endings? *raghavgururajan sees all 32 cores lighting up on bayfront and screams "Not me!" <nckx>(This *wouldn't* noticably speed up GHC, which is legitimately dependency-bound, or almost anything else. I was just idly staring at âtopâ wondering about make overhead.) <nckx>raghavgururajan: It's fine if they're all used đ Just not by 4-8 processes *each*. I.e. if âN runningâ in top goes far over 32 for more than an instant. <DesperateKimapr>i tried both 'grub-install --boot-directory=/mnt/boot/grub /dev/sdb2' and 'grub-install --boot-directory=/mnt/boot/grub /dev/sdb' <nckx>Otherwise, we're just getting our money's worth. <DesperateKimapr>First failed, complaining that btrfs doesn't support blocklists and second claimed to succed <DesperateKimapr>yet when i tried to boot it i was met with screen stuck with "GRUB" in top-left corner <nckx>DesperateKimapr: I thought you were way past this but did you create a âBIOS Bootâ partition? It can be small (~1 MiB, it fits in the alignment gap at the start of modern discs) but it has to exist. <nckx>The blocklists error sounds like it's using the right drive but can't continue due to this missing partition. The succeeding one sounds like it might succeed because it's installing to an irrelevant drive. Although that's guesswork. <nckx>DesperateKimapr: Could you mount everything where it should be (/boot, etc.) and post the output of â[sudo] lsblk -fâ and âfdisk -l <for each relevant drive>â? <nckx>DesperateKimapr: That's ridiculously huge but that's all right. <nckx>Well, I see sdb, I think multiple drives, and if I can get confused so can GRUB. <nckx>Oh. Did you already handle this stuff raghavgururajan? <raghavgururajan>nckx: DesperateKimapr (or) Kimapr[m] (or) KimaprOnPhone, were asking about the GRUB stuff. I explained somethings so that they could get a clear picture. <nckx>Not to run away from a KimaprInNeed but I'd be very happy if you could continue that conversation đ Seems like you already know their situation and I should really get to work. <raghavgururajan>> âDesperateKimaprâ: sdb because sda is guix installation image on flashdrive <raghavgururajan>It is a problem. You cannot use sdb in boot-loader config, because, it next reboot (without any attached devices), you HDD/SDD will become sda. Then GRUB will grub will look for sdb which is not there. <raghavgururajan>So keep rebooting your installation-image, until it becomes sdb and your main HDD/SSD becomes sda. <raghavgururajan>Usually, your HDD/SDD becomes sda. Sometimes it happens to become sdb. No idea why. <raghavgururajan>See, the internal HDD will always be detected first and will be assigned as sda. But sometimes, with other USB devices attached, it could be assigned sdb, but upon reboot, it should be fixed. <raghavgururajan>If you are installing guix on external HDD and never gonna attach internal HDD, then reboot until external HDD is assigned sda and use sda in bootloader-configuration. <raghavgururajan>But if you are going to attach and use internal HDD along with external HDD, then reboot until you get sda for internal hdd and sdb for external hdd; and use sdb in bootloader-configuration. <apteryx>the web manual corresponds exactly to the info manual <DesperateKimapr>But `info -k Device-syntax` didnt (and `info -k Device syntax` along with `info -k "Device syntax"`) <raghavgururajan>DesperateKimapr, Lets start from clean-slate. Could you boot the installation image and do `lsblk`? <apteryx>DesperateKimapr: when you want to search in detail the manual the same you'd do in your web browser, 'info guix' to open the manual, then Control S to search for some text. <apteryx>the 'i' key is often to find indexed etnries (but relies on them having been indexed by the authors), the 'm' key to find menus, and 'g' for any node (section). <DesperateKimapr>i already did both 'lsblk -f' and 'fdisk -l', their outputs are in two separate files and i'm thinking how to send them here <raghavgururajan>So you have two things connected to your device. [1] Your installation-image on USB stick [2] external USB HDD. Correct? <DesperateKimapr>It is present in the computer case, but sata wire is disconnected from it <raghavgururajan>Are you gonna do guided/graphical installation or manual installation? <raghavgururajan>Okay. You will have to re-do it, as we gonna change bootloader configuration. <DesperateKimapr>Doesn't bootloader configuration in guix use grub-install under the hood? <DesperateKimapr>then why can't i fix it with grub-install and then change the configs in the booted system? <raghavgururajan>Because you are not in the installed system, where you could just to `guix system reconfigure`. <raghavgururajan>Restart manuall installation with formating sdb (external HDD) with GPT. <DesperateKimapr>It has some important data and there is no spare drive to copy it on <raghavgururajan>Okay. Now, with cfdisk, delete partitions that doesn't contain your important files. <raghavgururajan>No I mean. there are 9 partitions in that output. Which one has your important files? <DesperateKimapr>dima-guix (dima-files is captured into image file on the first partition and deleted) is my previous guix-system installation <raghavgururajan>Delete non-critical partitions (including previous guix installation) with cfdisk and redo lsblk. <DesperateKimapr>dima-guix-new and dima-files-new are partitions of the new installation, on btrfs for fault tolerance, the dima-files that is right now there is a dummy partition just to let the old installation boot <apteryx>bah, my autossh service on the remote has stopped or something :-/ <peanutbutterandc>question: If I use #:modules in arguments, does it behave as if ONLY the explicitly declared modules are imported? I ask because a working package definition that I added #:modules ((modules here)) just started throwing 'Unbound variable: modify-phases' error <nckx>peanutbutterandc: Yes, you need e.g. `((my kool module) ,@%gnu-build-system-modules) <DesperateKimapr>sdb2 is bios boot partition and sdb3 is the root partition of new system <peanutbutterandc>nckx, Whoa. So basically re-import everything I need if I happen to need one extra module.... that is strange. I will give it a shot then. Thanks for the pointer! <raghavgururajan>DesperateKimapr, Did you create then just now or was it part of previous installation? <raghavgururajan>If it is the latter, then delete sdb2 and sdb3, so that only sdb1 is left. <DesperateKimapr>I created them when i installed the new installation, except the bios boot partition, which i created today <raghavgururajan>DesperateKimapr, Okay, for clarity (as we are starting fresh), delete them. <sneek>KimaprOnPhone, you have 1 message! <raghavgururajan>KimaprOnPhone, Woah! No worries. After reboot, make sure there is only sdb1 in lsblk output. *terpri wishes there were a simple way to use virt-manager/etc. more like WSL2 for, say, running single debian apps <pkill9>terpri: there;s probsbly some chroot script sthat do that <terpri>where the guest has access to almost everything on the host system, including the fs, gui, etc. <terpri>pkill9, true, i've been thinking about fixing up debootstrap and using that instead of kvm etc. <terpri>(and it's just a SMOP to set up that sort of thing with qemu ofc, but still yet another SMOP to actually implement) <terpri>reading up on btrfs as storage for VM images, i wonder if anyone's benchmarked "diskless" VMs, where virtually all guest storage is accessed via nfs/9p/whatever <nckx>terpri: You might be aware but this is what Guix already does for all of /gnu &c. <apteryx>for VM as well, to share the full /gnu/store <apteryx>*as well* not sure about containers actually *nckx reads about virtio-fs. It's like the Wayland of file storage. Network bad. <apteryx>I'd think the namespace would just be setup correctly given there's no disk image involve *nckx back to work, don't tell anyone I was here. <terpri>oh right, it would make more sense for VMs than containers of course <raghavgururajan>Tobias Geerinckx-Rice (nckx) was here at 09:50:34 AM EST on 2020-08-20. <hendursaga>raghavgururajan: About mailing lists, I just got myself a Yahoo address with its 1 TB storage. That suits me well. <raghavgururajan>hendursaga, I see. That's great. I do not trust Yahoo though. But I trust Muppeth at disroot.org :-) <hendursaga>I don't see the point, personally. Since it's dedicated to mailing lists, pretty much everything's already (semi-)public record. <hendursaga>Eh. I guess we have different threat models, then. Good for you though! I've heard good things about disroot :) <nefix>is it normal that takes +2h to build a 'guix system docker-image'? <rekado_>nefix: the docker format is nested tarballs, so a lot of the time will be spent repacking stuff thatâs in /gnu/store into a tarball. <rekado_>2+ hours sounds unusually long, though <nefix>and it's a really simple image <raghavgururajan>Kimapr just left without notice, while I was helping them step-by-step. :/ <apteryx>raghavgururajan: could be another power outage <nefix>rekado_: it "hangs" in the 'building /gnu/store/5qycscin4y2w2cm3xigsrh6fx8lvygsv-guix-docker-image.tar.gz.drv...' step for like 2 hours :/ <apteryx>is it reproducible? If so, sharing the operating system declaration would be useful *nckx simply runs sed on the logs to remove all raghavgururajan's messages. *nckx points out that they are joking just in case. <nckx>nefix: Can you (sudo) strace the tar process or whatever's hanging? <nefix>nckx: I have no idea what are you suggesting xD <nckx>âstrace -p PIDâ will print a live log off all system calls performed by a process such as tar. It's a cheap & easy way to get a rough first impression of what's happening inside the programme without setting up a debugger. <vapid>does guixsd still not have lvm support *raghavgururajan wonder why we don't make ci.guix.gnu.org to point to both berlin and bayfront, via load-balancer or something. <nckx>nefix: Is it producing new lines the entire time? At least it's not frozen (and it doesn't *seem* to be looping, not tighly at least). Use -s 1000 to show more string content. <raghavgururajan>May be we can deploy berlin's CI at berlin.guix.gnu.org (like bayfront.guix.gnu.org) instead of ci.guix.gnu.org. And then use bayfront.guix.gnu.org and berlin.guix.gnu.org as mirrors of ci.guix.gnu.org <nckx>vapid: I'm not aware of anyone working on it. <nckx>raghavgururajan: Several, but later. *nckx goes bye-bye again. <nefix>nckx: yeah, it keeps adding new lines, but it "hangs" every time in 'read(14, ' and stays there for 30s - 2 mins *raghavgururajan says ta-ta to nckx ***zap1 is now known as zzappie
<apteryx>nefix: perhaps hung up trying to get substitutes? <nefix>apteryx: what do you mean with that? <apteryx>Guix will try to fetch binary substitutes to speed up tasks. It could be a networking problem, although it should fail and quit if it was that. <nefix>apteryx: I doubt it's that, since it hangs in the 'building *-guix-docker-image.tar.gz.drv' and it fetches the substitutions before <apteryx>if you look at top, is there a process taking up your CPU? <nefix> 6314 6300 guixbuil R 1734m 1% 7 10% qemu-system-x86_64 -nographic -no-reboot -cpu max -m 256 -nic user,model=virtio-net-pci -object rng-random,filename=/dev/urandom,id=guixsd-vm-rng -device virtio-rng-pci,rng=guixsd-vm-rng -virtfs local,id=store_dev,path=/gnu/stor <peanutbutterandc>I have something that can be resolved using: cat gemspec | sed 's_\(<mustache>.freeze.*.*~>\)\(.*\)\("])$\)_\11.0\3_'g on the command line <apteryx>nefix: I'm building it locally right now, and it's at the step of using QEMU to produce the image as well. <apteryx>One of my CPU core is peaked though, so if yours is idling at lower digits, perhaps something's wrong. <nefix>i've checked now with htop, there are 4 threads of qemu-system and one is at 140%, another with 80%, and the other two between 20% and 10% <nefix>maybe is it because I'm building it inside Docker? <nefix>and doesn't have /dev/kvm mounted inside? <apteryx>oh, that's possible! That'd make QEMU much slower. <apteryx>Mine has KVM support (still going at it). <guixy>hi guix. Is there any plan to enable passfiles in alpine? <guixy>I see Arch doesn't implement them because of a security issue. <guixy>But it is not used automatically even when enabled. <apteryx>nefix: it finished here, after about 24 minutes. <nefix>apteryx: mine is still going, and I've started it one hour ago :/ <apteryx>that's still a bit long, simply to generate an image. There were discussions of using genimage to speed this up instead of relying on virtualization, IIRC. <nckx>nefix: A tool to generate disc images from a directory tree. The exact tool isn't important, but that it would run directly on the build host. Guix currently uses an âelegantâ but slow method: it actually boots the system you've built in a VM, then creates the images from inside of it. <nckx>IIRC someone (apteryx?) already modified the creation of installation images to use that more direct method. <nckx>nefix: One can use native tools (guix) and not a totally separate one (genimage). Fewer code paths and subtle differences to keep in sync. I can't say more than that as a total Docker novice. <nckx>Oh, the other Frenchy M-name, sorry... <peanutbutterandc>ngz said something about putting the definition for ronn in groff.scm instead of ruby.scm <nefix>nckx: but in the end it has to generate the image, doesn't it? <nefix>Or haven't I understood what's really happening? <nckx>I assume that part is Docker-specific and won't guess. <nckx>peanutbutterandc: Do you agree with the move? <peanutbutterandc>nckx, I am too inexperienced/not-knowledgeable to have an opinion on the matter, sir. I will do as you and/or ngz say. (Preferably both) <peanutbutterandc>hmm... guix lint says my lines are too long though.... wonder how I should go about fixing that.... any tips? <peanutbutterandc>or maybe I should just have procedure call on one line, and each argument separate <nckx>peanutbutterandc: Are you asking whether to move or how to move? I agree with ngz. You simply move the entire package definition to groff.scm đ You might need to add some module imports to groff.scm. <apteryx>the code base can be used for inspiration <nckx>#:use-module (gnu packages ruby) ; etc. <nckx>And #:use-module (guix build-system ruby) <nckx>peanutbutterandc: So I thought you said that multiple wrap-programs overwrite the previous wrapper instead of appending the new variables. <peanutbutterandc>The more I spend time with guix, the more I realize it is way more resilient that I thought. The first thing that made me realize that was the different profiles for `guix` itself and guix installed packages... and now things like these <nckx>It's resilient considering how complex it is. <butterypancake>It became much more resilient for me once I found out that I set my grub timeout to 0, meaning I couldn't revert to previous generations on startup. I'm a much happier guy now. <butterypancake>I could've prevented two reinstalls had I just left the timeout damn well alone :P <butterypancake>I'm either going to have to finally install fonts, or start memorizing utf-8... *butterypancake looks up 01F603 <butterypancake>wait, what was the keybinding for looking up a char you taught me the other day nckx? I think I can use that <nckx>peanutbutterandc: C-u C-x =, if it's still relevant. <peanutbutterandc>what does C-u C-x = do? o.O It didn't do anything for me... Is that some hacker joke? o.O <butterypancake>darn, emojis in erc aren't in a format emacs likes so I can look them up with that command... I guess I'm memorizing utf-8 after all <butterypancake>peanutbutterandc: I think out names are too similar :P that was meant for me <nckx>Honestly, you two and your nicks. *nckx should use less pedestrian smileys if people are going to take the trouble to look them up đ¤ <butterypancake>nckx: You absolute monster. first open mouth smiley, then slightly smiling smiley, now cowboy smiley? I'm starting to think utf-8 was a mistake. <peanutbutterandc>nckx, BTW, I stole your commit-message (from 2283cf4c1d5db8ed9330c1d1b6fd6b0e46857086) for the patch. :D ***butterypancake is now known as pancake
<rekado_>raghavgururajan: these build farms are independent <rekado_>raghavgururajan: thatâs why you canât just pretend at the DNS level that they are anything alike ***pancake is now known as pancak3
<rekado_>for the record on LVM: âLVM supportâ here means âboot from an LVM volumeâ, not âread or write an LVM volume after bootingâ <nefix>how can I specify a manifest for a user in the operating-system file? <rekado_>because read and write works since forever <rekado_>I mean, you can, but thereâs nothing in âguix systemâ that would instantiate the manifest for you. <rekado_>nefix: for global packages use the âpackagesâ field. <nefix>and neither can't I specify that user's .bashrc and I have to use the skeleton field? <pancak3>nckx: is this name less confusing? I'll compromise on butter, but not pancakes <nckx>Don't drop your butter on my account. I just knew this would happen one day. <pancak3>honestly, It would be cool to have a built in guix user dotfiles and package deployer. I wouldn't want to packages to be system packages because then I can't remove them without a reconfigure. I'm not sure how deploying the dotfiles would work (a lot of guix stuff is read only and I don't want that). But if we had those two things, I could run an installer from a usb, and instantly have a fully configured computer <nckx>peanutbutterandc: That's what the log is for đ Adding âpublicâ is pretty nckx, I don't think many people do that. <nefix>pancak3: there's guix-home-manager! <nefix>not sure if you could automatizate the installation at 100% but you could do most of the job *pancak3 failing to find documentation <nckx>pancak3: It's not part of Guix proper. <peanutbutterandc>nckx, Well, you're the maintainer. And I'm stealing as much as I can from you :D <nckx>peanutbutterandc: I am *a* maintainer, and that doesn't make me an authority on anything. *zzappie people put emojis in git commit nowadays... <nckx>I tried putting an emojo in a bank transfer but it didn't work. <peanutbutterandc>nckx, Yes, I am aware. Still... (You may not be Gerald sussman, but you most certainly are my Prof. Brian Harvey. :) ) <peanutbutterandc>pancak3, Hmmm....... I have got a dot-file repo (local git repo), and all I have to do for it is `make` and poof! It takes care of everything... well almost.... <pancak3>yo, guix-home-manager looks like an absolute pain. read only home dir? That's madness ***zzappie is now known as butternutsquash
<peanutbutterandc>.... I still need to do `make bootstrap` for something like `guix install all essential packages` or something <nckx>peanutbutterandc: Thanks. I'm going to steal that line to thoroughly confuse and disturb my partner next date night. <nckx>pancak3: It's an extreme approach for sure. ***familia_ is now known as familia
<pancak3>peanutbutterandc: installing make is a pain. I have slow internet. It's actually mostly the downloading that I hate. If I could download all the packages ahead of time and have them on the USB installer, that'd be ideal. And then it'd install those packages, but not under the system profile. So I could run guix install git, and it would be like "oh I already have it" but it wouldn't be in the system path <pancak3>is there a way to install something but not under a profile? <pancak3>oh ya, that does do that. So what I'm walking about is possible. I just need to figure out how to store and fetch packages from the USB instead of the internet <peanutbutterandc>pancak3, There's `guix pack` if you want binaries to carry around with you.... <pancak3>I'm thinking something closer to guix copy. except that command takes two machines. I want to specify two stores located on the same machine <nefix>how can I install a package from Nix? Should I add the nix service and what after? <KimaprOnPhone>raghavgururajan: I installed bare-bones guix system with bootloader to /dev/sdb and it still doesnt boot <sneek>KimaprOnPhone, you have 1 message! <sneek>KimaprOnPhone, raghavgururajan says: What happened? <kristofer>nefix: check the manual for "Invoking guix import" <peanutbutterandc>All right then, I'll have to leave. Thanks everyone. Hope you all have a good time. <nefix>kristofer: and could that be inside the packages < operating-system definition? <apteryx>pancak3: perhaps guix archive, guix import? <raghavgururajan>KimaprOnPhone, Btw, did you re-create the BIOS Boot Partition by setting correct "TYPE" in cfdisk? <raghavgururajan>KimaprOnPhone, I wanted to do it step-by-step with you. So that we can retrace things easily. Shall we start from just having one partition sdb1? <pancak3>apteryx: I was thinking of that but it would be nice to also have these tools accesible from the boot media. Having emacs in both the installer store and on an archive on the installer seems like a waste of space. Maybe later I'll look into guix copy to see if I can poke things until they do what I want <janneke>building a system downloads hundreds of rust packages -- never seen that before, did i miss something/what could have happened? <KimaprOnPhone>I think i'll remove the btrfs root partition and create an ext4 instead <raghavgururajan>KimaprOnPhone, Okay. Please delete all partitions except sdb1, which contains your important. Ping me after doing that. <nckx>pancak3: You wouldn't store an archive on the boot medium. You'd make sure the items you want are in the medium's store (using GC-ROOT-SERVICE-TYPE or whatever; if you want them to be available at installation time just add them to PACKAGES), then copy them from there to the target system. <nckx>raghavgururajan: Could janneke's question be related to your changes? <raghavgururajan>nckx: I was thinking about it, but my new rust packages are only referred in librsvg-next <pancak3>nckx: but how should I do the copy to the target system? <raghavgururajan>nckx, janneke: It could be that graphviz (used by guix) depends on librsvg, which in-turn depends on my new packages. <nojr>Hi, I just noticed that python-django got updated to 1.11.29 but there's still no indication that django3 will be available... is there a reason for this? is there something about Django that's making it hard to package? <bluekeys>Hi guix. How do i update a single program to another version? <sneek>bluekeys, you have 1 message! <sneek>bluekeys, nckx says: Can you update mailutils & try again? Here, it now (seemingly) freezes. Progress! <nefix>bluekeys guix package -u <package> <nckx>bluekeys: guix upgrade THING. <nckx>(I always forget if âguix upgradeâ takes packages or regexes so I cheated.) <KimaprOnPhone>raghavgururajan: I deleted the previous root partition and created a ext4 partition <nckx>pancak3: Er, that's out of the scope of my answer, I just wanted to point out there's no need for a redundant opaque archive on the live medium itself. *janneke tries reverting to august 11 (short lived "favourite" commit 12bb72fb0f8b8c35d46e439ba79f0a3a9f7703db) and rebuilding -- possibly "only" a substitute is missing <raghavgururajan>KimaprOnPhone, Okay. Please also delete previous boot partition and re-create it for 1MB with TYPE "BIOS Boot", in cfdisk. <nefix>is there any reason to keep libvirt at 5.8 and not upgrading to 6.x? <nckx>I'd use gc-root-service type in the *destination's* system.scm, although that will keep upgrading those packages at every reconfigure. But why would you not want that if your point is to âpre-seedâ users running âguix installâ later. <nckx>If you want a one-time copy at installation time, that's probably a bit more complex. <pancak3>nckx: ya, that's why I said I might look into guix copy later. Right now it looks like it only copies the store from one machine to another, but maybe I can extend it to accept paths for a one time copy <NieDzejkob>nefix: it's probably just that no one updated it yet <KimaprOnPhone>I checked and it was already of "BIOS boot" type, and it is 4 megs <NieDzejkob>nefix: could you bump the version and hash, and check if it still works? <nckx>I've never actually looked at how âguix system initâ does this. It has to keep the DB on the target system in sync, so it's not just a blind cp -R. <nckx>No time like the present to learn. <nefix>also, I'm at Nix with a Guix docker container trying to reconfigure itself while building another Docker guix image xD <raghavgururajan>KimaprOnPhone, Okay. What is partition-number for BIOS Boot partition? <raghavgururajan>KimaprOnPhone, Not sure what that is. BIOS boot is legacy boot. Let me know after you did `parted /dev/sdb set 2 bios_grub on` <KimaprOnPhone>I need to overwrite sdb3 so it doesnt identify as old already deleted partition i created just now <raghavgururajan>You have only 3 partitions, correct? sdb1 your data files, sdb2 bios boot partition and sdb3 root partition? <raghavgururajan>After doing, `mkfs.ext4 -L my-root /dev/sda3`, let me know what name you used. <raghavgururajan>nckx: if i understand correctly, after wipefs, there will be no file-system correct? <nckx>Not one that will be detected by tools like blkid. <nckx>It only deletes a few magic bytes (e.g. âThis is a btrfs file systemâ), not the actual metadata trees or data, so you can in theory undo it and restore the contents. <nckx>But of course you should never count on that. <raghavgururajan>KimaprOnPhone, Better to delete sdb3 and re-create sdb3 with type "linux file system" in cfdisk. And then do, `mkfs.ext4 -L my-root /dev/sda3` <KimaprOnPhone>Actually i did it twice, first right after the big data partition, then i deleted it, created bios boot partition on its start and after that i created it again next to new bios boot partition. at <raghavgururajan>Could you please do `cfdisk /dev/sdb` and see what TYPE is mentioned for sdb3 ? <raghavgururajan>Now proceed with `mount LABEL=dima-guix0-root /mnt` and `herd start cow-store /mnt` <raghavgururajan>KimaprOnPhone, After the above, `mkdir /mnt/etc` and `cp /etc/configuration/bare-bones.scm /mnt/etc/config.scm` and `nano /mnt/etc/config.scm`. <KimaprOnPhone>Yes, i replaced all fields where needed, including timezone, hostname and user accounts <apteryx>I think we should make this message more obvious: To complete the upgrade, run 'herd restart SERVICE' to stop, upgrade, and restart each service that was not automatically restarted. => The following new services: x, x and x were started. Existing services must be restarted manually. *alextee[m] loves CPU optimizations <KimaprOnGuixInst>Installing for i386-pc platform. /gnu/store/pc8whppnq03ab3naw21zjq5w9l6nx2r9-grub-2.04/sbin/grub-install: error: unable to identify a filesystem in hostdisk//dev/sdb; safety check can't be performed. <KimaprOnGuixInst>this is the command it apparently executes: /gnu/store/pc8whppnq03ab3naw21zjq5w9l6nx2r9-grub-2.04/sbin/grub-install --no-floppy --target=i386-pc --boot-directory /mnt/boot /dev/sdb2 <raghavgururajan>KimaprOnGuixInst, Do this, [1] re-do `parted /dev/sdb set 2 bios_grub on` [2] Change /dev/sdb2 back to /dev/sdb, in config.scm and [3] `guix system init /mnt/etc/config.scm /mnt` <nckx>Yes. It should be a full device not a partition. <raghavgururajan>I might have been wrong about using parition instead of whole device. <raghavgururajan>I was told that. May be wrong info or I must have misunderstood. Sorry about that. <KimaprOnGuixInst>Can't i just execute the very grub-install command `guix system init` uses but without the last '2'? <raghavgururajan>It does, how else grub.cfg contains info to where to look for boot files. <apteryx>mcron insists on the #:user specified having a $HOME. <nckx>KimaprOnGuixInst: You can. <apteryx>perhaps I could get away with a /nonexistent extra-special-file service. <KimaprOnGuixInst>Is it unsafe to cancel guix system init while it is copying to the target? <nckx>Depends on what you mean by unsafe. A next âguix system initâ will complete just fine. <nckx>But the current (cancelled) one will be incomplete. <nckx>Even if you ran a completed one before it. <nckx>âguix system initâ is not transactional in the same way the final system will be. <KimaprOnGuixInst>by unsafe i mean will it be left in a consistent state if i run a guix system init after guix system init and then abort the last one <raghavgururajan>Ah, you are right, grub.cfg is generated separately via grub-mkconfig command <nckx>KimaprOnGuixInst: No, it won't. <nckx>On other distributions raghavgururajan. Guix doesn't use grub-mkconfig either. <KimaprOnGuixInst>by another automated proccess which derives grub config from system config *nckx can only pop in briefly once in a while, sorry. <nckx>raghavgururajan: By Scheme code in (gnu bootloader grub). <KimaprOnGuixInst>it may not boot again, and i'll have to use touchscreen keyboard again until i install irc client into profile on running guix installation image again <bavier>attempting to build libmhash package for i686 I get segfault in malloc() during testing :( <raghavgururajan>In my experience, the blinking GRUB_ might indicate it is unable to load correct drivers/modules. <KimaprOnPhone>Previously i used DUET on a flashdrive to swipe BIOS madness under the rug *raghavgururajan searches about DUET <raghavgururajan>The installation image boots, so your BIOS is loading grub correctly there. <raghavgururajan>I think installation image is designed to boot either from BIOS or UEFI. <KimaprOnPhone>When i tried to install DUET on the hdd i used here it didnt boot either <KimaprOnPhone>This may be a coincidence (i installed DUET differently on the hdd than on flashdrive) crappy BIOS, damaged hdd, <raghavgururajan>KimaprOnPhone, The last idea I have for you is to try using grub-uefi-bootloader with ESP parition instead of BIOS boot parition. <KimaprOnPhone>One (risky!) idea is to install the efi bootloader and then overwrite installation image with DUET <ss2>oh, finally got it.. I cleaned my bash profile. So whenever I startx (with i3), and `eval "$(guix package --search-paths=prefix)"', the i3 session is broken such that themes and preferences are not found anymore. <ss2>disabling it lets it work again, but only after moving ~/.guix-profile and ~/.config/guix to another name. <KimaprOnPhone>The computer is sometimes used by people besides me, and they use Windows on the internal hdd <nckx>KimaprOnPhone: Can you install a GRUB2 -- any GRUB2, from any source -- to this system and boot something with it? Basically: have you ever run Linux on this system before? <NieDzejkob>I don't know what to put in the version field, though <nckx>KimaprOnPhone: Basically, once there's a GRUB on your disc that actually loads, you can modify its grub.cfg to run âconfigfile /path/to/guixs/grub.cfgâ and never touch it again. You can configure Guix to leave it alone. <KimaprOnPhone>On this machine - yes, the guix installer and my guix system on efi (with uefi duet) <nckx>It doesn't matter how you install it. Use a different distro's live CD. Whatever. *nckx reading about DUET. <KimaprOnPhone>Grub from guix installer doesnt recognise any partitions but first partition with important data <raghavgururajan>nckx: I suspected it. But they booted guix installation image. That's something right? <nckx>Oh, it's a legitimate UEFI, just loaded from disc through your old BIOS instead of a ROM chip? *raghavgururajan tries to stay away from UEFI stuff *KimaprOnPhone wants to forget BIOS forever <nckx>raghavgururajan: Good point, although the Guix installer is generated by grub-mkrescue which uses a lot of tricks (and some partition table magic that's hard to do on a ârealâ system) to boot on as much as possible. <bluekeys>I'm running guix pull and then guix upgrade mailutils but I keep getting a warning to consider running guix pull. How many times will guix ask me? Does it not just pull to the latest? <nckx>Gah, I have to go again, sorry. But honestly I don't understand why you can't wipe all partitions you created for Guix, create a regular ext4 one for Guix (and a FAT32 EFI System partition if you don't have one yet), download a random GRUB2 EFI binary (or build one), boot that through DUET, and point it to Guix's grub.cfg. Configure Guix to never install its own grub through the (installer #~(const #t)) trick. Wouldn't that work? <KimaprOnPhone>Some info: DUET boots on the flashdrive guix installer is on right now, but it didnt boot on the hdd with my systems. This could be because i installed DUET differently for each one (I installed DUET to flashdrive from Windows as it was the only thing i could boot, for external hdd with my systems i used guix installer instead, roughly following ***dddddd_ is now known as dddddd
***apteryx_ is now known as apteryx
<rekado_>bluekeys: what does âtype guixâ say? <rekado_>bluekeys: are you actually using that newly pulled Guix? <bluekeys>I dont know if I am. I've run guix package -u. I thought that meant I was using it? <bluekeys>did a hash guix and then type guix says 'guix is hashed (/home/user/.guix-profile/bin/guix)' <lfam>bluekeys: Sorry if this has already been mentioned (I just joined the channel) but your `guix` command should be `/home/user/.config/guix/current/bin/guix` <lfam>Did you install the Guix package with Guix? For example, did you do `guix install guix`? <lfam>If so, you shouldn't do that, and you should run `guix remove guix` before continuing ***evhan` is now known as evhan
<bluekeys>lfam thx i don't remember doing that, but I tried guix remove guix and was told error: package 'guix' not found in profile. My .bashrc could be setting the path wrong. <lfam>bluekeys: Okay. Something is messed up if `which guix` or `type guix` shows that path you mentioned *raghavgururajan makes some GagnamStyle moves after sucessfully building gnome-minimal and gnome <lfam>bluekeys: If you like, we can help sort it out <lfam>Awesome raghavgururajan! <lfam>I noticed that `guix package --verbosity=9` doesn't print the URL of things it is downloading, which makes it difficult to deal with "guix substitute: error: host name lookup error: Name or service not known" <lfam>Is that a recent change? I remember it showing this info <lfam>Huh, now it works as expected <lfam>One fewer bug to file :) <nckx>That's not... much... gooder. <lfam>I think that the point at which it failed was not exactly "downloading a nar" but earlier <lfam>I noticed some DNS flakiness with that server today <lfam>It would still be nice if it printed the URL <lfam>I'm not sure where it failed, though, and it's hard to reproduce <lfam>Well, that error is only in the code in one location. So I can probably write a patch <bluekeys>lfam I'd love if you can help me sort it