IRC channel logs

2024-05-26.log

back to list of logs

<rekado>is it possible to generate multiple files with "guix archive --recursive --export /gnu/store/...-profile"?
<civodul>rekado: no; it generates a single file that is the concatenation of each nar, with a word in between to indicate whether we’ve reached the end
<apteryx>'make distcheck' fails for me, due to a 'dot' error: Format: "pdf" not recognized. Use one of: canon cmap cmapx cmapx_np dot dot_json eps fig gd gd2 gif gv imap imap_np ismap jpe jpeg jpg json json0 mp pic plain plain-ext png pov ps ps2 svg svgz tk vdx vml vmlz vrml wbmp xdot xdot1.2 xdot1.4 xdot_json
<apteryx>graphviz build says: checking for ps2pdf... no and checking for pstopdf... no
<apteryx>seems PDF backend of graphviz uses cairo
<apteryx>is QA down?
<apteryx>nope: https://qa.guix.gnu.org/
<apteryx>although: Exception fetching branches:
<peanuts>"Guix Quality Assurance" https://qa.guix.gnu.org
<apteryx>cbaines: ^ in case you hadn't seen it yet
<apteryx>about my dot PDF problem, seems I had to include graphviz in my inpure guix shell
<apteryx>dot was found from my user profile, but somehow it was not finding the cairo backend then...
<abbe__>is issues.guix.gnu.org down ?
<abbe__>well, stalling it seems
<abbe__> https://issues.guix.gnu.org/ timing out https://www.irccloud.com/pastebin/FK4m1FNH/
<abbe__>results in 502
<Isaz>Does anyone have experience with guix-home-service-type? I created a guix image with an OS declaraction using this service type, and it downloaded all of the user packages correctly, but the actual image did not initialize the user's home directory or symlink any packages in (the directory was empty and owned by root)
<peanuts>"Snippet | IRCCloud" https://www.irccloud.com/pastebin/FK4m1FNH
<freakingpenguin>Isaz: When you run $ sudo herd status, do you see a guix-home-<your_username> one-shot service? Can you post a pastebin of your config?
<freakingpenguin>IIRC that service runs as <your_username>, so if you're trying to put files in a directory it does not own you may have permission issues.
<freakingpenguin>/var/log/messages might say something.
<apteryx>guix shell --pure git git:send-email openssh does *not* use the local manifest.scm file, does it?
<Isaz>Thanks freakingpenguin, the service is running. I could reconfigure, but I am booting from a usb, and would run out of space if it redownloads everything. This is the base config that my image record was created from https://termbin.com/nl5n
<freakingpenguin>Isaz: Have you actually booted with that image or are you just inspecting it?
<freakingpenguin>The home environment won't activate until Shepherd runs the activation script
<Isaz>I booted with it
<freakingpenguin>Isaz: Does anything happen if you manually rerun the service? $ sudo herd start guix-home-bdunahu
<freakingpenguin>I suspect the problem lies with /home/bdunahu being owned by root, curious why that isn't changed.
<Isaz>wow, yes, that fixed it. Maybe it's because I changed the ownership of my home directory when I first logged in
<Isaz>when I logged in, there was one file in the directory, ~/.cache/mpd or something, maybe that set the permissions incorrectly
<Isaz>not sure why that file was there
<freakingpenguin>So when you created the image, /home/bdunahu was owned by root until you logged for the first time?
<Isaz>when I logged into the bdunahu user it was owned by root and was an empty directory, besides that one file
<Isaz>I manually fixed the ownership
<freakingpenguin>Gotcha. No clue what caused that, very weird.
<Isaz>thanks for your help! Maybe I'll see why that one file was generated, it must have incorrectly set the permissions
<freakingpenguin>Probably something with mpd-service-type, maybe that service tried starting before the accounts were created
<Isaz>I do have the mpd system service linking into my home directory
<Isaz>I logged in as root first to set the password of the bdunahu user, and the mpd service uses a directory in bdunahu's home as the music directory
<freakingpenguin>mpd-shepherd-service depends on 'user-processes, and 'user-processes depends on 'user-homes, so /home/bdunahu should have been provisioned before starting mpd. Might be something off there.
<Kolev>Anybody set up Nextcloud?
<Kolev>I can't get the login to work.
<apteryx>oh, there's a graphviz-minimal and a graphviz
<apteryx>the minimal variant lacks PDF support
<apteryx>does nss-certs@3.99 works for others?
<apteryx>apparently git:send-email requires openssl
<theruran>any idea why I am getting these missing derivations errors after installing Guix from AUR?
<theruran>guix pull: error: reading file `/gnu/store/1xh5b299d09nrpgjpwa0wxpz0fs83d7p-guile-gcrypt-0.3.0.drv': No such file or directory
<theruran>also missing patch and glibc apparently
<theruran>during make check: FAIL: tests/builders.scm
<theruran>FAIL: tests/gnu-maintenance.scm - seems important
<theruran>FAIL: tests/pack.scm
<theruran>hmmm
<adanska>Hi guix!
<theruran>I am giving the guix-installer from AUR a try now.
<theruran>ha yeah it locks up like it was before. this time while downloading guile. I am running this in an Arch Linux StandaloneVM qube
<PurpleSym>The CI seems to be broken again. There’s 500k pending builds on the status page, yet most workers are doing nothing. Can someone have a look?
<thunktone>On
<rekado>PurpleSym: I don't have time to check this, but I've restarted cuirass and the cuirass-remote-server; I've also run VACUUM on the db. No impact.
<rekado>ACTION wrote the sysadmins
<PurpleSym>rekado: Thanks. Yeah, no progress on the haskell-team evaluation (https://ci.guix.gnu.org/eval/1348713) unfortunately :(
<peanuts>"Evaluation" https://ci.guix.gnu.org/eval/1348713
<PurpleSym>I’m also seeing backend timeouts quite frequenly :/
<jpoiret>glibc-2.35 build error figured out, I'm expecting to reboot on cu- soon :)
<civodul>jpoiret: wo0t! well done!
<jpoiret>it wasn't too hard, just realized something that was already written down for glibc-2.32 💩
<jpoiret>alright, see you on the other side :)
<jpoiret>hello from glibc 2.39 :) no issues yet
<janneke>yay, welcome
<jpoiret>i'll merge and push
<civodul>wb :-)
<civodul>exciting news
<PurpleSym>Is there anything that prevents us from updating grub to fix issue 70480?
<peanuts>"Guix bios installation: Grub error: unknown filesystem" https://issues.guix.gnu.org/70480
<PurpleSym>Most of our system-level tests currently fail because of the issue above and everyone installing Guix on a fresh ext4 partition will also run into this issue.
<attila_lendvai>the bootstrap guile (%guile-static-stripped) dies in getrandom. reproducer: guix shell -e '(begin (use-modules (gnu packages make-bootstrap)) %guile-static-stripped)' and then start guile. root issue: this guile is used in initrd, and when things fail, then it just sigsev's.
<attila_lendvai>make-guile-static-stripped only tests for `guile --vesion`, which works. but e.g. this dies: guile -c '(use-modules (ice-9 readline))'
<attila_lendvai>a wild guess: might be due to remove-store-references removing something needed? (there's no check for a #f return value, meaning it did not remove anything)
<janneke>now let's hope we get some substitutes for core-updates
<attila_lendvai>PurpleSym, IIRC no, it's just someone needs to do it and test it.
<attila_lendvai>i've tried to build %guile-static-stripped without the (remove-store-references guile2) line, and it dies with: "...is not allowed to refer to path `/gnu/store/3zl03prdg07ax4dny78hrzykillr6vcy-glibc-2.35'"
<attila_lendvai>i.e. there's a reference to glibc in the guile static binary, which is silently nuked by remove-store-references
<attila_lendvai>is this a bug? should the static guile in initrd be able to open a repl and a debugger?
<PurpleSym>attila_lendvai: I’ll add a patch from upstream to grub first and then we’ll see how much time is left for the upgrade…
<attila_lendvai>ACTION wonders what and how could be improved to avoid such issues falling through the cracks
<PurpleSym>Running the system tests before merging? They’ve been red since March.
<PurpleSym>#71210. Would be nice if anybody could give a thumbs up before I push.
<peanuts>"[PATCH] gnu: grub: Ignore metadata-csum-seed feature of ext4." https://issues.guix.gnu.org/71210
<attila_lendvai>PurpleSym, this is just a backported commit from grub, right? i don't see how that would break anything if it compiles (famous last words)
<umanwizard>Hello, is there a good reason why gcc or clang on Guix doesn't include .note.gnu.build-id in the compiled binaries by default, whereas on most other distros, they do?
<PurpleSym>attila_lendvai: Correct, it compiles and makes tests green, but yeah, famous last words.
<attila_lendvai>umanwizard, i'm not sure what that is, but if that is a random, then it breaks reproducibility.
<umanwizard>I think it's not random, it's a hash of some sections of the binary)
<attila_lendvai>yep, not ranom: "The build ID is a 160-bit SHA1 string computed over the elf header bits and section contents in the file. It is bundled in the elf file as an entry in the notes section."
<janneke>trying to build system on core-updates
<janneke>In procedure git_libgit2_init: Function not implemented
<janneke>ACTION isn't using that directly; possibly only a git-daemon service
<civodul>janneke: probably because your guile-git is trying to use an incompatible libgit2.so (like built against another libc)
<janneke>civodul: ah
<cbaines>janneke, do you know what the status of the mes bootstrap is on btrfs?
<dariqq>is there an easy way to list all dependencies of a profile? Something better than just the list of installed packages with guix package -I?
<ngz>dariqq: AFAIK, it is possible with a manifest file, not a profile.
<ngz>However, it is possible to change a profile into a manifest.
<ngz>err to export a profile into a manifest file.
<dariqq>hmm, something like 'guix size $(readlink -f /run/current-system/profile)' seems to do something similiar to what I want, though it also has not only packages and is quite slow. but something i can work with i think
<janneke>cbaines: no; we fixed a bug with stat64 a year ago or so
<janneke>but i don't remember getting any feedback on that yet
<cbaines>janneke, I was just looking at this failed core-updates build, and I wonder if there's still problems https://bordeaux.guix.gnu.org/build/47830aa1-6c8f-461b-bbf2-d5bdb56a13ce/log
<cbaines>the problems might have changed though, as it seems to be making stat64 syscalls successfully, but still failing
<janneke>ekaitz: looks familiar to you? ^^
<cbaines>I've found one issue again https://issues.guix.gnu.org/41264
<peanuts>"Bootstrap packages fail to build due to mes-libc lacking 'stat64' etc. syscalls" https://issues.guix.gnu.org/41264
<cbaines>I guess that is somewhat fixed, but there's maybe still problems on btrfs
<cbaines>there's also another issue here https://issues.guix.gnu.org/53416
<peanuts>"mesboot derivations, build phase errors, No targets specified and no makefile found" https://issues.guix.gnu.org/53416
<ekaitz>yes
<cbaines>the bordeaux build farm is a good test for this since milano-guix-1 uses btrfs
<ekaitz>that's something we fixed during the process
<janneke>cbaines: right
<ekaitz>stat and dirent
<janneke>not just for riscv64?
<janneke>ah, dirent too
<ekaitz>but that wasn't merged when core-updates was added
<janneke>we need a mes release i guess
<ekaitz>the dirent case is weird, because we only touched x86_64
<ekaitz>but we need to make the bootstrapping process use x86 in x86_64
<ekaitz>because it will automatically find the 64 bit case and fail
<ekaitz>janneke: all this is from the core-updates update I did? or more stuff was added on it?
<janneke>ekaitz: dunno, hard to tell...
<janneke>that is to say, until now it just worked (unless using brtfs or possibly very large disks?)
<ekaitz> https://issues.guix.gnu.org/68222 this is what we did
<peanuts>"[PATCH 0/3] RISC-V support to commencement until tcc-boot0" https://issues.guix.gnu.org/68222
<ekaitz>also mes-boot is in 0.25.1
<ekaitz>we should take a look
<ekaitz>and janneke we found setjmp and longjmp are broken in riscv and we need to fix
<ekaitz>the fix involves writing them in assembly directly
<ekaitz>we need to talk about that, as i don't know how to include them in the build system we have
<Googulator>Wow, nss has looooooooooooooots of tests...
<Googulator>Several hours spent already just running nss's check phase
<apteryx>Googulator: and it runs the test suite with slight variants 3 or 4 times...
<apteryx>see 70950
<apteryx>see #70950
<peanuts>"[PATCH] gnu: nss/fixed: Hasten test suite." https://issues.guix.gnu.org/70950
<Gooberpatrol66>throwing this idea out there: make the build directory copy-on-write using overlayfs or something before the check phase, so one can turn tests on or off without breaking hashes and substitution
<apteryx>that's an interesting idea, but you'd have to make even the install prefix an overlay as files can be mutated there
<Guest13>I guix pulled and system reconfigured after 8 months. When StumpWM starts, I get " Error loading /gnu/store/[...]-init.lisp: Error opening #P"/gnu/store/[...]-stumpwm-[...] Read-only file system".
<Guest13>And thus, my configs like fonts in ~/.stumpwm.d/init.lisp are not being loaded.
<Guest13>The content of my ~/.stumpwm.d/init.lisp is http://paste.debian.net/plain/1318185
<freakingpenguin>Guest13: Try building with --no-grafts
<freakingpenguin> https://issues.guix.gnu.org/62890
<peanuts>"StumpWM fails to start - Read only file system" https://issues.guix.gnu.org/62890
<freakingpenguin>Alternatively Guix Home is installing your stumpwm home configs and addons and Guix System is installing StumpWM itself, reconfigure guix home to use updated addons.
<Guest13>Should I do `guix build stumpwm --no-grafts`?
<freakingpenguin>guix system reconfigure <system.scm> --no-grafts
<freakingpenguin>The issue should go into more details on two possible causes
<Guest13>Thanks!!!
<Guest13>IIUC, I should install stumpwm in system and its modules in home; then reconfigure both system and home without grafts, correct?
<freakingpenguin>Guest13: Reconfigure both system and home with --no-grafts. That should solve either cause for the problem, which is either 1) Reconfiguring with grafts or 2) Reconfiguring system and not home IF system is installing StumpWM and home installs addons
<Guest13>Okay
<freakingpenguin>Guest13: I think even if the problem is occuring you can run loadrc (C-t ;) and it seems to still evaluate the configuration properly
<freakingpenguin>As a workaround
<Guest13>Ah, thanks!
<freakingpenguin>This problem annoyed me to death a while back, wish I knew the underlying cause.
<jpoiret>freakingpenguin: stumpwm probably relies on the full path of some files in a roundabout way, which then gets broken by grafts
<jpoiret>Building without grafts is discouraged, you will be missing important security fixes
<Guest13>freakingpenguin: Is `(,stumpwm "lib") considered an addon or no? Also, is technically right to be in system or home profile?
<Guest13>*is it
<freakingpenguin>Guest13: When I say addon I mean anything loaded by init.lisp. Don't know the answer to that one
<freakingpenguin>jpoiret: Are you saying that StumpWM capture a path such that Guix doesn't recognize/fix it as part of the grafting process?
<freakingpenguin>may capture*
<freakingpenguin>I'm not really familiar with how grafts work, I assume it's largely "scan for store file paths in output, replace"
<freakingpenguin>Although in this case the path does change when grafts are enabled, just not the same way as addons
<jpoiret>freakingpenguin: yes, you got it right
<trev>hi folks, trying to write a guix manifest for a game that uses guile-ncurses...it appears that guile-ncurses isn't being built with unicode support, so i tried to manually add the configure flag, but it gives me a nice error. can someone tell me what i'm doing wrong? https://github.com/trevarj/gdotris/blob/master/manifest.scm
<peanuts>"gdotris/manifest.scm at master ? trevarj/gdotris ? GitHub" https://github.com/trevarj/gdotris/blob/master/manifest.scm
<trev>i can post the error if needed, but i think i may be doing something wrong with the manifest alone
<freakingpenguin>trev: You're overriding the arguments in guile-ncurses with your own
<trev>freakingpenguin: right. i want to build guile-ncurses with unicode support
<freakingpenguin>Gotcha, but I bet missing one or more of them is causing the install phase to fail.
<freakingpenguin>Doesn't guile-ncurses in Guix already have the --with-ncursesw flag? Is that insufficient?
<trev>i think it does: https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/guile-xyz.scm#n2721
<peanuts>"guile-xyz.scm\packages\gnu - guix.git - GNU Guix and GNU Guix System" https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/guile-xyz.scm#n2721
<trev>but when i do a `guix shell guile guile-ncurses -- guile` and then evaluate %wide-ncurses it is #f
<trev>so that leads me to believe it actually isn't working, or i need to go to the ncurses level and enable it
<Googulator>Is there an official place to find nightly(-ish) builds of the guix-system installation ISO?
<Googulator>I'll likely soon  need one for reference to compare my own bootstrapped builds against.
<Googulator>looks like https://ci.guix.gnu.org/eval/1350308 might be what I need
<peanuts>"Evaluation" https://ci.guix.gnu.org/eval/1350308
<PotentialUser-72>Hi there. I've been trying to use OpenMW on Guix but keep getting a segfault. I
<PotentialUser-72>My crash log is https://paste.debian.net/1318199/
<PotentialUser-72>Any ideas what might be causing this?
<peanuts>"debian Pastezone" https://paste.debian.net/1318199
<PotentialUser-72>I'm using a copy of Morrowind I obtained from a disc, running OpenMW from a t450s, and the crash occurs as soon as the first loading screen is over but before a menu is loaded.
<PotentialUser-72>If I wanted to open an issue for OpenMW on Guix, do I just email bug-guix@gnu.org with what's going on?
<lard>yep
<PotentialUser-72>Alright, thanks!
<lard>mention your guix version and if you're on guix system or a foreign distro also