<mark_weaver>civodul: maybe we should just switch to importing (guix licenses) with a prefix in linux.scm. wdyt? <mark_weaver>(one of these days I need to review the guile module handling code in detail) <civodul>mark_weaver: yes, i was looking into that <civodul>there's another one with haskell/xml and the 'expat' binding <civodul>mark_weaver: could you rebase security-updates on master and reevaluate? <mark_weaver>civodul: it seems wasteful to build it without the ruby update included, no? <mark_weaver>okay, I'll rebase ruby-update on master and then rebase security-updates onto ruby-update. <paroneayea>francis7: just booted up the minifree x200 (I was busy the last couple weeks) and it's INCREDIBLE <paroneayea>plus it booted with a picture of a penguin hugging a gnu! <paroneayea>iyzsong: you got a copy of your guix (and grub?) config somewhere I can peek at <lfam>That's the end of the log <lfam>I tried again and it failed just one copy earlier <lfam>I can successfully do `guix system vm` <mark_weaver>paroneayea: on Libreboot machines, I make /boot/grub/libreboot_grub.cfg a symlink to grub.cfg <mark_weaver>Libreboot includes its own copy of GRUB and a grub.cfg burned into the boot firmware. the first (default) menu item in Libreboot's grub.cfg will look for /boot/grub/libreboot_grub.cfg and load it if it exists. <lfam>Okay. I'm trying with a larger image-size, then I will send mail <mark_weaver>or rather, that sounds plausible to me, but admittedly I rarely use VMs <lfam>That was it. The result is ~1 gigabyte <mark_weaver>a better error message would be nice though, so maybe still worth sending a bug report about that. <lfam>mark_weaver: The weird thing is that a subsequent run did report "no space left on device". I'm not sure what the difference was or why the process failed at slight different points along the way each time I ran it. <mark_weaver>lfam: I guess that there's some concurrency involved in how the kernel does things, e.g. there are several kernel threads that do various tasks. <mark_weaver>and of course thread switches are triggered by clock interrupts, and thus non-deterministic <mark_weaver>the kernel panic you got before was because PID 1 exited. we ought to be able to avoid that, at least in cases like ENOSPC <lfam>`guix system vm` boots to a WM and `guix system vm-image` boots to a guile repl <mark_weaver>lfam: if you get a guile repl during boot, that indicates that something went wrong, e.g. it couldn't find the root partition, probably because you are launching the VM incorrectly (or not how it expects). <lfam>Ah, I see that `guix system vm` sets a whole bunch of command-line options go `qemu-system-x86_64`. I hadn't set any of them. I'll try again <mark_weaver>iiuc, "guix system vm" is essentially the same as "guix system vm-image" except that "guix system vm" does one additional step: it creates a script to run the VM. I would look at that script for clues. <lfam>I have a question for later once I am actually up and running. While creating these VM images, I have the operating system declaration in the host system, of course. Once I am in GuixSD, is the effective OS declaration available anywhere? For example, under /etc? Or is it up to me to rsync it into the VM for editing and reconfiguration? <mark_weaver>lfam: the OS configuration is not stored anywhere. it is up to you to manage it. <paroneayea>anyone have any experience booting the usb key off of libreboot? <paroneayea>mark_weaver: I'm trying ot install my x200 from minifree <paroneayea>but it's not bootin', though I really don't know what to do from the grub menu <iyzsong>paroneayea: hi! I don't use libreboot, and my grub config is almost same as the desktop.scm template. <paroneayea>iyzsong: main worry is I won't figure out how to get LUKS working <iyzsong>paroneayea: ah, that one, found it in commit 316d65be, just make sure /boot is on a seperated partition. <paroneayea>it seems like a good idea, but I hate having to enter my passphrase twice <iyzsong>paroneayea: no, I just use for root. I haven't try to use LVM in GuixSD yet. <iyzsong>the problem is that grub haven't any related configurations for LUKS, so after boot, I need to do some 'insmod' and 'cryptomount' by hand. <paroneayea>iyzsong: ah... can you possibly paste what you use for that? <mark_weaver>paroneayea: karhunguixi uses GuixSD with encrypted root partition on a Libreboot machine, as I recall. <iyzsong>"hd0,msdos2" is my encrypted root partition <paroneayea>the end of that email made me think it didn't work <mark_weaver>paroneayea: last I knew, GuixSD does not yet support LVM. if that's no longer true, than section 7.1.1 (Limitations) of the guix manual needs to be fixed, because it says there that support for LVM is missing <mark_weaver>paroneayea: no, it's a limitation in our initrd and maybe also filesystem-mounting services. <mark_weaver>paroneayea: I would post to the ML about it. my knowledge in this area is weak. <paroneayea>mark_weaver: I'll post to the ML and ask, thank you :) <paroneayea>ifconfig shows nothing when in the guix live usb <lfam>The USB installer live environment couldn't resolve hydra.gnu.org. Is this expected? I edited /etc/hosts and that worked. <paroneayea>lfam: hydra.gnu.org is being very slow for me at t he moment <paroneayea>lfam: also did you do the ifconfig / dhclient dance? <lfam>Me too. But I couldn't even resolve the name. <lfam>I also used git on my LAN <lfam>Yes, I'm downloading substitutes now that I edited /etc/hosts with the correct IP <lfam>Oh wait, I didn't try to ping google.com. I did `ping 8.8.8.8`. <lfam>Strange. I can't resolve google.com but I can resolve a personal domain of mine, not on the LAN. I don't know why that would work and seemingly nothing else. <paroneayea>lfam: I think your LAN might do local broadcasting <paroneayea>I know that I can sometimes find other machines on my network by name without specific machinery I laid down to do so <lfam>One of these domains is not on the LAN though. And I can't resolve another name that points to the same machine. <lfam>One of the domains that I can resolve, that is <fps>i noticed name resolution failures on the installer image before <lfam>Oh yeah? Is there a bug report? <fps>restarting dhclient usually fixes it for a while <fps>no, i think the channel always said "it's your network that's failing" <lfam>Seems unlikely considering the total lack of name resolution problems on my other machines <lfam>Any suggestions for debugging? <fps>sadly not from me as that's the part of linux that usually just works ;) <fps>one more observation though: usually it starts out fine resolving names, but then stops after a while <fps>let's see, maybe i can reproduce it now. i wanted to setup a qemu image on my desktop anyways :) dl'ing installer... <lfam>I didn't have any problems in the VMs. This is on bare metal. But if it manifests on VMs that will make it much easier to debug. <fps>iirc it happened in my vm's too <lfam>I guess that's a good thing :/ <lfam>Wait, it happened in the running GuixSD system? I'm just using the USB installer to initialize a new machine right now. Bummer if it happens in installed systems, too. <fps>just in the installer <fps>the installed systems worked fine afterwards <fps>lfam: can you confirm though that rerunning dhclient fixes it for you? <lfam>I'll wait until `guix system init` finishes. It took a long time to download the list of substitutes and I don't want to wait again. <fps>hmm, right now i can't even get a ping through though names resolve fine.. <lfam>Are you in QEMU? I don't think that it supports ICMP. At least, not in the default `guix system vm` configuration <lfam>Do you have i686 binaries? <fps>wanted to install links on the live image.. <lfam>Oh well, the machine I'm installing onto is i686 <lfam>Did you distribute the key anywhere yet? <fps>but something is not working right there anyways, guix package -i links wants to pull and compile gcc :) <fps>lfam: i posted it to the ML <fps>um, maybe it['s a bad idea to guix package -i anything on the unionfs of the installer image anyways :) <fps>but i thought that maybe some net activity might trigger the name lookup fail <lfam>I installed vim while in the installer <fps>cause iirc it usually happened at some point during the system init in between downloading two packages <fps>hmm, i'll setup a disk image to install to, too <lfam>I authorized your server on my x86_64 machine. Do I always have to use --substitute-urls in order to use it? <fps>i'm not sure about that part. maybe you can add it in the system config <lfam>How do I challenge your server? I tried it for the vim package but I got "no local build for..." <lfam>Do I need to force a local rebuild? <fps>might be. it's all these little things i want to find out about it, too :) <fps>i started building everythign from source on that box at one point in time <fps>but then to finish it off i just pulled the rest as substitutes from hydra <lfam>I thought you'd know ;) I'll read the ML announcement of `challenge` <fps>you do have vim installed though, right? <lfam>Hmm, guess not. It's from Debian <fps>install it from hydra <fps>then challenge fps.io <fps>or the other way around :) <lfam>I'm challenging hydra and fps.io at once. Your substitutes list arrived in a second or two... can't wait for the donations to add so we can buy a new hydra. <fps>once i'm done with this soul crushing dev project i'm working on, i'll go back to looking at gnunet/torrents/etc... to decentralize things more.. <fps>one more note about fps.io:9999: it's an ssh tunnel port forward to the real vm that sits in a different data center alltogether from the fps.io vm :) <fps>so it's two VMs and an ssh tunnel and it's still quite fast :) <lfam>hydra, fps.io, and localhost all build the same output for 'chess'. <fps>for n in `guix package -A | cut -f1`; do guix challenge ... "$n"; done :) <fps>to test all packages :) <fps>with ... replaced by options you know better than me right now <lfam>Haha, no! I'm trying to initialize a new machine right now. I don't want to add to hydra's burden too much ;) <fps>if i understand that right, it should be possible to do: <fps>trying that right now for this install <fps>interesting. it pulls guix-0.9.0 from hydra already <fps>seems there's still stuff missing.. <fps>hmm, it pulls everthing from hydra.. weird.. <fps>i wonder how to look into this further... hmm <fps>resolv.conf still has 10.0.2.3 [which is the qemu] <lfam>Are any of the system services not working? I don't know how to look that up in dmd. In systemd you would do `systemctl status` <lfam>I should have installed tmux before doing the system init so I could poke around while waiting. <fps> /var/log/messages has nothing interesting.. hmm <fps>deco is the thing to control dmd <fps>route says the gateway is 10.0.2.2 <lfam>Does Guix have any special logging service like journald? Or is it just in /var/log? <lfam>What does it mean for the "Running value" of a dmd service to be an integer? The value of the name service cache daemon service is 366. <fps>one hint is maybe that it usually happens after installing the first few packages, right? <fps>was it the same for you? guix got installed, maybe another one after that and then the name resolution stopped working? <lfam>Well, rather than go straight to the system init, I installed vim so I could edit desktop.scm. That worked. Then later on it failed when I tried to init. But there was at least an hour in between, I'd estimate. <lfam>So, I can't say it stopped working after a certain point. <fps>when doing init it failed to fetch the first package already? <fps>ok, that shatters my theory that it depended on installing the first few packages to the target fs <lfam>I'd also tried to install tmux and rsync but gave up due to slow hydra and not really needing them. I also installed git. <fps>i wonder if something about nsswitch.conf changed <fps>i guess i can boot another instance of the installer <fps>ok, when initially booted /etc/resolv.conf is non existant <fps>after doing the ifconfig/dhclient dance resolv.conf and route look identical to the one where it failed <lfam>So, you mean that the dance didn't fix the problem? <fps>no, that's not what i meant <fps>i booted another instance of the installer image in qemu <fps>and compared the files and routes of the two systems <fps>i didn't redo the dance on the failed one <fps>to keep it in its pristine failstate :) <lfam>On mine, resolv.conf is just "search lan<newline>nameserver 192.168.1.1 <fps>"nameserver 10.0.2.3" <fps>i'm not too experienced with name resolution debugging though <lfam>It works if I do `dig @192.168.1.1 hydra.gnu.org` <lfam>That's really at the limit of my current knowledge on the subject <fps>i just restarted the init on the failed box <fps>and now it chuggs on <fps>without doing another dance <lfam>It was a mistake to use desktop.scm. So many substitutes... <fps>wanting to rebuild from source though <fps>i've no idea really.. <lfam>Still initializing. There are collisions from different builds of gmp. <fps>yeah, i always start with the bare bones.scm :) <fps>killall dhclient; dhclient ens3 and then restarting init made it go on for real <lfam>I usually do that but I just want to test a grub patch so I'm not going to "settle in" to this installation. I figured it would be easier to test with all the affordances built-in <lfam>those steps made the name resolution work again? <fps>yes, it repulled the list of subs from hydra and went on <lfam>restarting init: is that `deco restart dmd`? <fps>nono, restarting guix system init.. sorry for being lazy :) <lfam>I'm not going to try it now. My hardcoded /etc/hosts will have to suffice for this project <fps>yeah, we'll have to wait for more knowledgable people to help with the name resooution thingy <fps>iirc though it's not even qemu specific. had it happen in kvm, too <fps>so it'll be one little weird thing on the installer images :) <fps>time to shower and get some rent earning work done ;) <lfam>Good luck! One day closer to the end of the soul-crushing gig, right? :) <sneek>Welcome back fps, you have 1 message. <fps>lfam: yeah, thanks. i was faintly aware of that. also filed a bug about the name resoution thing just now <paroneayea>can grub mount logical volumes with cryptomount? <paroneayea>maybe I should have had it do it *not* as a logical volume... <paroneayea>iyzsong: heya, could you possibly paste me your finished grub.cfg? <lfam>After leaving my GuixSD laptop on all night, I pressed 'ctrl' to wake the screen up. It flashed, then nothing. It only works during the BIOS bootup. Once grub or linux takes over (not sure which), it's gone. Luckily I can connect an external monitor. So weird. <lfam>When I say it flashed, I mean I saw the console for a second or two <civodul>and you can't get to normal state again? <lfam>Looks like a hardware failure. Or at least, a failure that manifest before grub. It's happening even in the BIOS and boot selection menus. Very old laptop <lfam>Annoying because it doesn't "see" the external monitor until after grub. <lfam>I'd like to set up authentication so I can press backspace 28 times and get the grub rescue shell ***user5252` is now known as user5252
<civodul>unfortunately no, there's nothing beyond that <lfam>Oh, well in that case, our grub isn't vulnerable to the CVE! ;) <lfam>I'll still try to patch it. <lfam>So, I'm accustomed to installing things on foreign distros from a guix checkout using ./pre-inst-env. Does that process also work on GuixSD? My intended method would be: git clone, apply the patch, `./pre-inst-env guix build grub` as a sanity check, then `./pre-inst-env guix system reconfigure config.scm`. Is that correct? <jadedctrl>How can one edit file in /gnu/store (such as configuration files for a package), or the X desktop files in /run/current-system? Any help would be greatly appreciated. <lfam>jadedctrl: You don't edit files in /gnu/store. They are supposed to be immutable. Are you using GuixSD or Guix on top of another distro <jadedctrl>If the files are immutable, how would one configure a program such as Apache, then...? <lfam>Okay, I'm only just starting with GuixSD so my knowledge is pretty rough. But you should edit the Apache service in the config.scm you use to configure the whole system, and then do `guix system reconfigure config.scm` <lfam>Each time you make a change like that, a new directory is created in /gnu/store, and the current system points to that new directory. Just like installing new versions of packages. <jadedctrl>Ah, that makes sense. Thank you, lfam. I've been looking in the wrong area of the manual, then. <lfam>I only just installed GuixSD so I don't know the details of how to configure Apache in particular. If you hang around here, somebody else will chime in eventually. Or, you can send an email to the mailing list (guix-devel@gnu.org) <jadedctrl>Ah, thanks again, lfam. :) I'll have to set up that service, then. <lfam>I'm not sure if the extensions described in my last link have been merged or not <davexunit>jadedctrl: yeah, we don't have an apache service yet, but the general idea is that the OS config would specify the configuration files somehow so that the system service knows which one to use when it starts. <lfam>davexunit: Got any thoughts about the question I asked ~90 minutes ago? About using ./pre-inst-env to test a Grub patch? <aeva>hiii - I'm going to try to install guixsd on my x200 today - is there any information on ideal partitioning scheme + setting up luks with guixsd? <civodul>aeva: at this point LUKS doesn't quite work for the root partition, or you have to make it separate <aeva>I had heard it was possible to do, with the caveat that you had to type the unlock password twice <civodul>yes some people did that apparently, but that requires tweaking here and there <aeva>I guess I'll wait until I can find out what that entails then :/ I want very much to have useful security lol