***taylan2 is now known as taylan
<mala>Kolev, is it building it locally? If you type "guix weather icecat" you can see whether there is a substitution available that should be downloadable <ytc>hello everyone. i'm thinking about switching to guix from traditional distros. but i'm using hdd on my laptop and guix was terribly slow while it was booting, so i have avoided it until now. <ytc>does it worth to buy an ssd for guix? is guix on ssd still slow compared to the other distros? <ytc>i am familiar with guile scheme. but boot time and lack of the ability to hibernate to an encrypted swap disk makes me think it again. <singpolyma>I haven't run system and also my workstation is silly, so I'm not a good judge on that <singpolyma>I know there are structural things that will always make boot a little slower, but I expect on a solid drive you won't notice too much. But I don't know for sure <GNUtoo>hi, what is the best way to test guix on qemu arm? <GNUtoo>Is it to reintroduce vexpress ca9x4 support? <GNUtoo>AFAIK it was removed and re-added in u-boot <apteryx>guix import pypi pytest-asyncio@0.8.0 perhaps? <apteryx>hmm, I was looking at some older history; nevermind <apteryx>GNUtoo: what do you mean by "test guix on qemu arm" ? What is vexpress ca9x4 (a kernel driver, perhaps?) <GNUtoo>vexpress-ca9x4 is a machine (-M vexpress-a9) <GNUtoo>I was wondering if there was an easy way to build an ARM image that could be booted with qemu-system-arm <GNUtoo>and u-boot is a popular bootloader used on ARM devices <apteryx>that would be a 'guix system image' of a Guix System operating system, but I don't think that's supported for ARM yet <apteryx>OK! Then it must still work, and perhaps you can use 'guix system --target=your-triplet image ...' to attempt cross-building the image for the target <GNUtoo>but then it stop booting, so I was wondering if I could instead run an image in qemu to see if it worked there too, if it was the amount of RAM that make it not boot, etc <GNUtoo>Hence my question, I want to run it in qemu not a real board <GNUtoo>And I need to know which bootloader to use for that <GNUtoo>And if restoring support for that vexpress-a9 makes sense or not <apteryx>the QEMU manual says that if your exact board is not there, chances are slim that your OS will boot <apteryx>should be cheap to try a few of the closest ones though <GNUtoo>I'm assuming that some people use qemu for testing <GNUtoo>So that's also why I'm asking here, for the people that know about it, what is the advised way to run Guix in qemu-system-arm <apteryx>good question. I'm curious to know too. <GNUtoo>Else I could just add vexpress-a9 back but it requires testing, building images, and I've already too many bug reports open in Guix <apteryx>and there's nothing on the guix-devel archives? <GNUtoo>I prefer to ask as things changed over time with that board <opalvaults[m]>can anyone tell me how to get sway to show up in GDM loging menu? unfortunately .guix-profile/share/xsession is read-only. Any other spot that GDM checks? I've also guix system reconfigured with sway in the package list <opalvaults[m]>so in theory it should have popped up if it was packaged with a .desktop file <GNUtoo>I had to test that but I've some updates issues at the moment <GNUtoo>(service gdm-service-type (gdm-configuration (debug? #t)(wayland? #t))) <GNUtoo>You need to configure gdm with wayland? #t at least ^^^ <GNUtoo>If that doens't work then we need to make other patches to make gdm pick up the sway session file <opalvaults[m]>The desktop environments in Guix use the Xorg display server by default. If you’d like to use the newer display server protocol called Wayland, you need to use the sddm-service instead of GDM as the graphical login manager. You should then select the “GNOME (Wayland)” session in SDDM. Alternatively you can also try starting GNOME on Wayland manually from a TTY with the command “XDG_SESSION_TYPE=wayland exec dbus-run-session <opalvaults[m]>gnome-session“. Currently only GNOME has support for Wayland. <opalvaults[m]>but I'm not exactly sure where I'm supposed to put that declaration <opalvaults[m]>my file doesn't have anything in particular regarding GDM, only gnome-desktop-service-type <opalvaults[m]>Yeah I think having GDM be compositor/server agnostic would be nice <GNUtoo>I'm assuming these compositors can start both session types (wayland and Xorg), though I've no idea why it does't look up for wayland session files by default (at least for gdm) <opalvaults[m]>well it certainly yelled at me for having gdm-service-type and gnome-desktop-service-type lmao <nckxmas>opalvaults[m]: You're trying to apply (call as procedure) gdm-service-type. <opalvaults[m]>man how anyone groks this stuff so easily I'll never understand lmao <opalvaults[m]>guix system: error: service 'xorg-server' provided more than once <opalvaults[m]>so there's a xorg service module is there a wayland one too? <opalvaults[m]>I swear I smack my head at this thing every day and it doesn't get any easier <opalvaults[m]>okay, so I'm trying to get Sway installed and showing up in GDM. I have sway installed, and I removed xorg from the service modules. What is the correct syntax to add gdm to services without xorg? <opalvaults[m]>because i'm just running into repl error after repl error. i think im going crazy looking at the word 'unwind' and 'with-exception-handler'. <opalvaults[m]>AIM: unfortunately that dudes config doesnt' work for me. It throws unbound variables (to what, I'm not sure, usually its a parens issue but check-parens didn't catch it) <nckxmas>opalvaults[m]: What's a config idec? <opalvaults[m]>i already had a sway config laying about so its backwards compatible and all that jazz <nckxmas>(I didn't actually test that; I use SDDM.) <nckxmas>I'm not sure you'll find many Sway users using GDM, they are rather different beasts in spirit. <opalvaults[m]>Is that just (services (append (list (service sddm-service-type) %desktop-service)))? <nckxmas>You need to import at least (use-service-modules desktop sddm xorg) to use those variables. <nckxmas>I don't use any of the %default-things myself so sharing my configuration would not be that useful. <nckxmas>Plus I'd have to remove all the swear words which is $work. <opalvaults[m]>I've also been staring at text all day, so I'll take the fault on that one. Anyways, good to know sway works <nckxmas>was supposed to contain 100% more ‘but I do use Sway’ ***phf-1 is now known as phf-1[afk]
***phf-1[afk] is now known as phf-1
*phf-1 is back (gone for 00:02.59) ***phf-1 is now known as phf-1[afk]
***phf-1[afk] is now known as phf-1
*phf-1 is back (gone for 00:02.04) ***phf-1 is now known as phf-1[afk]
***phf-1[afk] is now known as phf-1
*phf-1 is back (gone for 00:00.50) <lilyp>phf-1: don't go afk-spamming <phf-1>Oups, sorry. I misconfigured my client. <awb99>I have the same problems as opalvault when trying to install sway on my machine. <awb99>it would be great if anyone could post a config that works. <awb99>I had a look at all configurations that I could find, and I dont think anyone of the publicly available configs for sway work. <awb99>I am trying to rebuild my system, and I am getting a lot of node dependencies that I have no idea which package brings them. I tried to graph the packages but it seems that I can only graph the packages of one package. <pkal>Is there some way to figure out why guix upgrade is building derivations for packages I have uninstalled? <pkal>Or more generally why packages are being built that should have subsitutes (in my case ungoogled-chromium) <cehteh>could be that supstitutes are not yet available (buildfarm needs time as well) <pkal>Hmm, I'll try pulling again to see if that changes anything <cehteh>wasnt there some website showing the build status <cehteh>there is guix weather too, but by itself its not fast <cehteh>emacsclient can start emacs when its not running <cehteh>i sometimes restart it for some reasons, and wuld want it as user service too <pkal>cehteh: Sadly no, pulling again didn't help (I think my last pull was yesterday) :/ <lilyp>I too am wondering why there's so many stuffs missing substitutes currently <lilyp>e.g. mutter ought to exist according to CI, but a different one's being built locally <lilyp>perhaps the build farm is behind in terms of evaluations? <vldn[m]>substitutes are weird since yesterday <pkal>Still I wonder, why are substitues as dependencies being built for package I have uninstalled? <lilyp>Kabouik: gdm *ought* to spawn soon after you see that line, but you could try with a minimal installation if your graphics drivers make issues and debug from there <lilyp>pkal: Guix builds the full closure (both native and otherwise) of everything that is referenced. <lilyp>I don't know what you consider uninstalled, but chances are you have a wrong concept in the back of your head somewhere <pkal>Kabouik: I had a similar issue once, where the GDM user didn't match the UID of GDM's files. <pkal>lilyp: Yes, but nothing should be referencing these packages (various node packages) in my case? <lilyp>You could theoretically also come to it from the r side. <rekado>gc lock has been held since > 16 hours. <rekado>I’v been waiting a long time to finally copy the completed aarch64 build fo linux to reconfigure these build nodes. <cehteh>rekado: its still the delete thing? <rekado>no, it’s ‘guix gc’ running, holding the lock <pkal>lilyp: guix graph --type=reverse-package does indicate that node depends on ungoogled-chromium and js-mathjax, but neither of the two are listed by guix package -I <rekado>the other way around: ungoogled-chromium and js-mathjax both have node among their inputs <rekado>none of this suggests that node would be installed when either of these packages are installed, though. <Kabouik>lilyp, pkal, thanks for your answers. My problem is I can't type anything, it looks like I'm really stuck. Or can I go in tty and start from there maybe? <lilyp>but it would still be built/substituted <lilyp>that's what confuses most people I guess <apteryx>rekado: where are you at with the network storage array experiments? <lilyp>Kabouik: if you can go to tty, do that <lilyp>it might be stuck and not react to keyboard input tho <Kabouik>The install was made from the Guix System installer iso. I couldn't install using the GUI (got backtrace issues) and I struggled with the manual install, so what I did is go in the shell for manual install, and from there use `guix system image --image-type=efi-raw /path/to/some-config.scm`, and then I dded the image file to a drive. <Kabouik>It does boot, but leaves me there (picture posted above) <Kabouik>Yeah I have to try tty. Problem is I was travelling so I took an old 40% keyboard I never used, I have to find how to emulate F2 :p <pkal>I reused my installation disk, mounted the disk, prepared a chroot and then I could fix my issue <Kabouik>Honestly I have nothing set up so I could start from scratch instead of trying to save my install, but I wonder why it doesn't just work since I didn't customize anything <Kabouik>Even the GUI installer should work I guess <dissent>hey guix, how would I go about using `guix time-machine` to install an older version of texlive. <Kabouik>pkal do you think I could do what you did without chroot? I just plugged my drive to a another computer and from there I can see /etc/passwd and /varrun/gdm <Kabouik> /var/run/gdm appears to belong to root:cups-pk-helper according to ls -la <Kabouik>In /etc/passwd I see that the gdm UID is 983. Can't I just `sudo chown -R 983 /path/to/drive//var/run/gdm`? <Kabouik>I tried but that didn't solve the issue <pkal>Kabouik: If you don't have anything else installed, I think it is better to just do a clean reinstallation. <Kabouik>Yes, but I don't know how that would chance since this is already a new install from last night with 0 change <Kabouik>What's weird when booting is I am prompted for login and password before the issue occurs, and if I'm fast enough, I can even login. But regardless, the booting continues and will eventually get stuck (see login at the top): https://0x0.st/orAK.jpg <pkal>Well you could install the system without GDM, or chroot to reconfigure the system to replace GDM with something like slim <Kabouik>I'm trying to chroot but fail due to /bin/bash not found. I tried `sudo chroot /path/to/guix-fs /gnu/store/hash-bash/bin/bash and it did went through, but then it can't find any command, I guess it doesn't know where to look for executables <Kabouik>Also weirdly the hostname was that of my host distro, not the guix hostname, so I'm not sure chroot really worked <Kabouik>How did you manage to chroot into your guix filesystem pkal? <Kabouik>su: Cannot make/remove an entry for the specified session <fiesh>jackhill: so actually, I take back that modify-sysfs-in-some-way is a feature that's missing. it can all be done with udev, more or less at least, only hindered by its awful syntax and debugging options <rekado>apteryx: stuck :-/ We’re out of space on the SAN. We’ve copied 4.2TB and the SAN has hit its reserve, so we can’t even get the allocated 5TB filled. <rekado>so, currently, there are no more experiments from my side. <jackhill>fiesh: yeah, all we might need is something in the cookbook or manual <rekado>also: the discussion at https://issues.guix.gnu.org/51787 has provided enough evidence that there isn’t all that much we can do to improve performance for /gnu when we insist on keeping it flat and huge. <rekado>use of btrfs does not alleviate concerns over fragmentation, and its baseline performance is considerably worse in our initial tests. <rekado>xfs might be better for /gnu, but it’s unclear if it matters at our scale <rekado>xfs cannot easily be grown, so it would be messier to maintain in the long run <rekado>ext4 has good performance on the SAN and it’s easy to grow as we get more hardware. <jackhill>rekado: are you sure? I though xfs could be grown, but not shrunk (while ext4 can be offline, and btrfs can be online) <rekado>the biggest impact perhaps is the use of tiered storage; SSDs for caching boosts performance noticeably. <rekado>I’d like us to get unstuck soon :) <rekado>it’s a pity we can’t even get /var/cache moved off of the storage array <jackhill>rekado: ok, I'll take a look, but (/me looks at the calendar) I have lots of other stuff going on this weekend. I may not have any great insight, but can at least proofread out assumptions <karrq>hey I'm setting up guix again and bash is struggling sourcing the environment, apparently $profile/etc/profile is owned by root only... for reference it's /gnu/store/....-guix-md5/ <karrq>did I misconfigure something? is $profile supposed to be only readable by root? <karrq>I solved my issue: the script to source the profile was incorrect :P fixed now and it works <pkal>Kabouik: Have you prepared /dev/, /proc/ and /sys/? <Kabouik>I mounted them from my host distro using `sudo mount --bind /dev /path/to/guix-fs/dev pkal <pkal>Then I really cannot say what the issue could be :/ <pkal>I guess the user mailing list would be the next stop after this channel <opalvaults[m]>merry christmas everyone! I made a desktop wallpaper for you guys <Kabouik>Interesting pkal, this time I got all the same errors during boot, but eventually could see my mouse cursor, and then gdm started, I could log in and I'm now in X <Kabouik>I hope it's reproducible. Boot was rather long though. <pkal>Let's hope it is a one-time thing. <Kabouik>So I actually hope the boot going through is not a one-time thing! <Kabouik>Now I'll still have many things to fix I think, the first being my wifi card not being working, I don't know if it is supported by free driver <pkal>What wifi card you you have? <Kabouik>I am not sure what card it is pkal, but it's a laptop and I expect the gard to be mainstream <Kabouik>Can't really check until I reboot in my other distro and have packages to investigate the hardware. <ss2>packages, like lspci? <Kabouik>I was thinking inxxi, but maybe there's something in a clean Guix install that I could use? <ss2>which is actually available in Guix. <Kabouik>The card is an Intel Wireless 8265/8275 <lilyp>Kabouik, pkal: As perverted as it sounds, but you might have an easier chance switching VTs earlier than later <Kabouik>I rebooted but I think there was still something wrong in the filesystem, after I booted, the terminal didn't know any command except `guix` (not even ls or cp or mkdir), even after I sourced /etc/profile <Kabouik>Well that is weird. Guix was installed on a USB thumb drive, and now that I have powered Guix off and rebooted in another distro, the thumb drive doesn't show up anymore. I'm afraid it might be lost. <Kabouik>lsusb lists it, but nor parted nor lsblk see anything (they used to). <podiki[m]>that could explain it, if the drive was failing :/ <Kabouik>Hum that's weird though, I always got to the grub menu at least and then Guix started to boot. After I tried something like you did (changing /var/*/gdb UID to 983), I got the same issue once again, and then on the next boot it did work and started X. I couldn't do anything in there because no internet connection and no commands found, so I powered off, and now the drive shows in lsusb but not lsblk. I can't see how it would have worked "this much" if <podiki[m]>could have been its last legs? or filesystem corruption (would that mean it wouldn't show in lsblk?); not sure <Kabouik>Yeah I really don't understand why it no longer shows in lsblk, but I'm very doubtful about it failing just right when I try to install Guix while I have been using it for so long before with no issue, don't think it's a coincidence <Kabouik>Even the BIOS stopped listing it as a boot device <podiki[m]>boot sector/partition problem maybe, could just take one bad write or block <KE0VVT>I pulled and reconfigured, but I see no GNOME 40. <lilyp>Because we're on GNOME 41 actually :D <Kolev>lilyp: Strange. Why am I not seeing it after pulling and reconfiguring as root? <mekeor>can i run "guix pull --commit=HASH" in order to pull a commit with a better weather than current master? <mekeor>how can i find out a recent commit-hash with a good weather? <lilyp>there's channels-with-substitutes-available IIRC <lilyp>though substitutes seem kinda hosed currently <jpoiret>I may have kept a lot of locally built stuff from c-u-f so I've never noticed <lilyp>so bad, ci.guix.gnu.org fails as indicator <jpoiret>jobset overviews give you a better idea of which commit introduced lots of rebuilds <mekeor>it's a little pity that channel-with-substitutes-available doesn't support multiple URLs since IIRC there's also some bordeaux-URLed-cuirass available :) <mekeor>could be a nice PR to get done for beginners :D -- by adding a reduce/fold over a list :D <Kolev>Tying to update to GNOME 40 now. <rekado>so nothing is built since about 24 hours <rekado>I just killed the ‘guix gc’ processes. <jonsger>roptat: thx for the peer/pointopoint implementation in guile-netlink. But I wonder how I could try out the unreleased version :( <nckxmas>Thank you, raghavgururajan :) My best wishes to all of you! <awb99>I am trying to run "guix system reconfigure" for 36 hours now. And it is always wanting to rebuild linux-libre and node <awb99>And the substitutes are all there <awb99>I dont know why my machine is building. <awb99>Interestingly, guix weather tells me that there is no substitute for node, but that there are for linux-libre <awb99>But it still is building both. <awb99>Does anyone know how to sort that out? <awb99>I just cannot wait a few days for a simple guix system config change to go through. <awb99>I just wanted to update a mcron job. <awb99>Now I am stuck for almost 2 days with this. <nckxmas>Not only does guix weather find a substitute, but ‘guix build linux-libre’ downloads it. Are you sure you're talking about the same linux-libre derivation in both cases? <nckxmas>Which architecture is this that rebuilding Linux takes days? <rekado>awb99: not sure about 36 hours, but for 24 hours ‘guix gc’ was holding the GC lock, so nothing was built in the meantime. <rekado>I killed it about an hour ago, so if you’re still having problems fetching a substitute you may want to clear your substitute cache. <awb99_>Can anyone tell me why guix reconfigure is trying to build linux-libre when guix CI build farm has it built already ? <awb99_>This issue is going on for perhaps 36 hours. <nckxmas>Are you sure you're talking about the same linux-libre derivation in both cases? <awb99_>But this didn't change anything. <nckxmas>Did you read the previous answer to your question? <awb99_>I ran guix-weather before doing the guix System reconfigure <nckxmas>The cache is in ~/.cache/guix/substitute. <awb99_>So perhaps I should delete this directory <nckxmas>If that doesn't help, share the output of guix system build --dry-run. <awb99_>Sorry.. I am not understanding you ...what should I do with the cache directory? <awb99_>Do you have any idea what happened? <nckxmas>awb99_: Negative responses are cached for a while, so Guix was not ‘seeing’ the fact that ci.guix had a substitute for you. <nckxmas>guix weather does not use this cache. <awb99_>So now it is only trying to build node. <awb99_>But I also found node in the CI farm as built <nckxmas>I don't know if you saw rek ado's reply to you, but the CI was basically down all day. <awb99_>But it's built was finished just a few minutes ago <awb99_>And I thought the build farm has no jobs to build <awb99_>But it now seems to be built again with 248990