IRC channel logs


back to list of logs

<sammi>hey so like, how important is getting the name service cache daemon (nscd) working for guix to run on a foreign distro?
<sammi>I'm on fedora here, and for the life of me, I can't find any package for it!
<vagrantc>the problem is getting name resolution with a compatible libc
<vagrantc>the daemon kind of works around the issue ...
<vagrantc>there was some talk of it getting dropped from glibc ...
<vagrantc>i am not sure how to work around it not being available...
<sammi>getting nscd dropped from glibc?
<vagrantc>yeah :(
<sammi>(also tyvm 👐)
<vagrantc>but then guix and nix folks piped up ... but not sure where that conversation ended up
<sammi>I mean fedora can't be the only one that this is a problem for, right?
<vagrantc>debian still has it with glibc 2.36 and 2.37
<vagrantc>huh ... i would think it might also affect flatpaks or similar distribution methods ...
<sammi>those reach out to the system glibc? I thought those were sandboxed to prevent that?
<sammi>or is that one of those "yeah it's not good practice, but the right way is extremely inconvenient" kinda deals?
<vagrantc>just a guess that it would need to do something similar ... but maybe i am totally wrong on that one :)
<sammi>lol idk either X)
<sammi>in any case, thank you vagrant!
<sammi>I think the previous times I've set up guix I just winged it and installed the locale, x11 fonts and called it a day
<sammi>haven't run into any problems yet, but fingers crossed I suppose 🤞
<vagrantc>wonder if systemd-resolvd really can replace it?
<vagrantc>anyways, good luck!
<vagrantc>ACTION waves
<psycotica0>Let's say I have a package `a`, but I want to make some changes. So I make a new package `b`, and that package inherits from `a`. But one of the changes I want to make is to add a new build phase. I don't want to replace the phases, I want to add a new one to the ones already present in `a`.
<psycotica0>I know I can use `package-arguments` to get the arguments of `a`, and in theory I could get the set of phases from `a`'s arguments, use modify-phases on those, and then override the phases value of `a`'s arguments with these new phases in order to set the arguments of `b`. But I haven't found any guile functions that allow me to get/set/modify the
<psycotica0>type of datastructure that arguments is. Am I just going about this wrong? Or am I meant to build this myself?
<HiltonChain[m]>psycotica0: Something like "(cadr (member '(quote pre-configure) (car (cddadr (member '#:phases (package-arguments glibc))))))" ?
<AwesomeAdam54321>psycotica0: Packages that modify the package-arguments use (substitute-keyword-arguments (package-arguments PACKAGE) ...)
<HiltonChain[m]> <>
<AwesomeAdam54321>it can be used to add new build phases without replacing
<HiltonChain[m]>Oh, forget about that. <>
<HiltonChain[m]>But as the arguments are inherited, phases should already present.
<psycotica0>Ok! That can give me somewhere to start looking. Oh, but I assume `modify-phases %standard-phases` should probably be the phases of the parent, right? Rather than %standard-phases?
<HiltonChain[m]>Sorry I didn't fully understand your question......
<HiltonChain[m]>"(substitute-keyword-arguments (package-arguments <your-package>) ((#:phases phases) #~(modify-phases #$phases ...)))"
<HiltonChain[m]>The "(#:phases phases)" part is to bind #:phases (the one from the arguments to substitute) to phases, then you can use it as a variable.
<psycotica0>Ah, neat. Thanks! I'll try that out later
<Kolev>Does everybody type their LUKS passphrase manually, or is there an easier way?
<atka>hi guix, is it possible to use FDE with luks2? I used luksFormat --type luks2 --pbkdf pbkdf2 /dev/sdx (due to grub2 and argon2)
<atka>installation completes successfully but grub can't find my crypt volume
<atka>could've sworn I had this working before
<atka>but maybe --type luks is still required
<HiltonChain[m]>atka: <>
<atka>HiltonChain[m]: heh, thanks
<atka>crazy that fix has been pending for ~2 years or so
<ngz>sneek: later tell cbaines According to <> and <>, the branch has not been processed in the past 24 hours. I'm sorry to ask again and bother your with that, but is there an issue with it? It might be stuck.
<sneek>Will do.
<ngz>sneek: later tell cbaines Aaand it looks like it started now. I'm very confused about the time frames.
<rekado>with nscd being removed user accounts can still be detected via sssd. Guix needs to be able to talk to the sssd socket for that to work.
<yarl>Hello guix.
<yarl>Just checking, does `guix shell -D guix --pure -- make recheck TESTS="tests/store.scm" SCM_LOG_DRIVER_FLAGS="--select=^verify-store + check"` fails for you too?
<yarl>Hmm, apparently, I needed to run `make clean-local`. I don't think this is specified somewhere. Wouldn't it be better that `make check` or `make recheck` do this?
<ennoausberlin>Hi guix. I am playing around with wsl2 images. I could create the one provided by the guix system/images/wsl2.scm sources. But when I try to load it for testing purposes with docker load < path_to_tarfile it claims about a missing /var/lib/docker/tmp/.../bin/json
<ennoausberlin>Is this a bug or is my workflow wrong?
<ennoausberlin>Never mind. docker import was the way to go, not docker load
<ennoausberlin>Ooops. I just realize, that wsl2 images are not docker images at all. So I can not test the result under GUIX before copying it to a Windows machine
<cbaines>ngz, you can see the queue of jobs here and the events here
<sneek>cbaines, you have 2 messages!
<sneek>cbaines, ngz says: According to <> and <>, the branch has not been processed in the past 24 hours. I'm sorry to ask again and bother your with that, but is there an issue with it? It might be stuck.
<sneek>cbaines, ngz says: Aaand it looks like it started now. I'm very confused about the time frames.
<cbaines>there's quite a large backlog, but jobs are being processed at an OK rate
<ngz>cbaines: These pages are interesting, thanks.
<nutcase[m]>Hi. I removed my ssd from computer, reattached it and reset my BIOS. Now my guix system does not boot anymore. What could be the reason?
<nutcase[m]>I'm using grub-efi-bootloader
<ennoausberlin>nutcase: Some bioses can boot from the efi partition directly. Did you try that? F11 or similiar key combos
<jpoiret>nutcase[m]: you'll probably need to recreate an EFI boot entry manually
<jpoiret>not too hard using efibootmgr
<nutcase[m]>Is efibootmgr on install iso?
<ChocolettePalett>You could always install it there after booting I think, it depends on the amount of RAM though
<nutcase[m]>jpoiret: I booted the install medium, installed efibootmgr via guix and it finds my boot options, but no OS
<nutcase[m]>ennoausberlin: the boot menu does not show the entry
<ennoausberlin>nutcase: But the drive is shown as present? The EFI-Boot menu or the grub boot menu?
<jpoiret>wdym it finds your boot options?
<jpoiret>can you see the entry to boot guix's bootloader?
<nutcase[m]>ennoausberlin: the drive is shown, and I was able to create a new boot entry that shows up in bios boot menu. After selecting that, the bios boot menu disappears and reappears. Probably the efi entry is wrong
<nutcase[m]>How does the efibootmgr -c line have to look like?
<ennoausberlin>You can try to download refind and boot this from usb
<nutcase[m]>I tried with efibootmgr -c -d /dev/nvme0n1
<attila_lendvai>can someone send us a recentish commit that has full substitute coverage? or tell us a way how to find such a commit?
<vputz>OK, trying something again... I'm trying to write a package description for a binary that's downloaded as a tarbomb, but it's in .tar.bz2 format. shows that the definition of url-fetch/tarbomb only includes gzip as an additional module, not
<vputz>bzip2, and so the derivation of the file (not the package) can't proceed (lbzip2: no executable found, etc).  Adding (inputs) or (native-inputs) doesn't help because that adds lbzip2 to the containing derivation (the package).
<vputz>Should I abandon url-fetch/tarbomb?  If so, how do I manually download a tar.bz2 file and then uncompress it?  This is turning out to be trickier than expected!
<nutcase[m]>jpoiret: ennoausberlin: I think I'm missing the --loader option. Do I have to mount my disk's efi partition to /boot/efi and am then able to refer to grubx64.efi?
<jpoiret>nutcase[m]: you need to add an entry for /EFI/Guix/grubx64.efi
<jpoiret>not just for a drive
<jpoiret>yes, you need --loader
<attila_lendvai>vputz, this may help:
<nutcase[m]>jpoiret: Solved! Thank you so much!!!
<attila_lendvai>vputz, you may need to add something to native-inputs for bzip2, though
<vputz>Hmm, will give that a go
<vputz>So attila_lendvai the idea is to use url-fetch to download the raw bz2 file, use the nonguix binary build system to operate on that file, and replace the unpack phase so the unpack now happens in the package derivation rather than the origin derivation?  (first package I've tried to write so I'm pedaling fast, and familiar with lisps but not used
<vputz>much guile so there's some good-natured flailing here!)
<attila_lendvai>vputz, i don't understand the last part (i.e. when the unpack happens), i'm not familiar with the url-fetch/tarbomb stuff. but otherwise it's a good bird's eye view of what needs to be done. also note that discussing nonguix/nonfree stuff here is impolite, so let's head over to #nonguix for that.
<attila_lendvai>guix proper rejects the idea of binary packages altogether
<vputz>understood (although I've made use of the gnu build farm packages).  This is BSD-3 so source is available, but I'm learning the packaging system conceptually.
<HiltonChain[m]>vputz: You can see the definition of font-sarasa-gothic, it replaces unpack phase.
<vputz>Thanks!  Will look at that as well
<Kolev>Evolution won't work. It says a service is not running. Icedove won't work. It says `/home/caleb/.guix-profile/bin/icedove: symbol lookup error: /gnu/store/5h2w4qi9hk1qzzgi1w83220ydslinr4s-glibc-2.33/lib/ undefined symbol: __libc_pthread_init, version GLIBC_PRIVATE`. Claws Mail is slow.
<rekado>Kolev: is this on Guix System?
<Kolev>rekado, yes.
<rekado>2.33 is pretty old for glibc
<rekado>we’ve been using 2.35 for a while now
<rekado>where does 2.33 come from then?
<rekado>perhaps you inadvertently make a fresh application (using glibc 2.35) use plugins or libraries that were linked with glibc 2.33?
<Kolev>rekado, I don't know. I just did `guix install icedove`.
<jpoiret>Kolev: when was you last system reconfigure? what's `guix system describe`?
<civodul>Kolev: also, does it wor after typing "unset LD_LIBRARY_PATH"?
<Kolev>civodul, icedove: same message
<jpoiret>Kolev: you're using a way out of date guix
<jpoiret>at least your system generation was created with the 1.4 guix
<Kolev>jpoiret, ok, I'll update
<jpoiret>how are you running reconfigure?
<Kolev>jpoiret, `sudo guix system reconfigure /home/caleb/Projects/dotfiles/guix/hosts/bender.scm`
<jpoiret>looks good though, is your `guix describe` on the same commit?
<jpoiret>did you somehow pin the guix channel to the 1.4 branch?
<attila_lendvai>last time i had such strange issues i forgot that a channel was pinned to a specific commit in some channels.scm file.
<hwpplayer1>hi GNU Guix people !
<anemofilia>Is there any reason for texlab not being packaged for guix?
<anemofilia>I found it kinda strange since you can just `guix import crate texlab` to get the package definition
<jpoiret>anemofilia: does it work?
<jpoiret>I guess most people use AucTeX
<anemofilia>Didn't tried it yet because I didn't learn too much about channels
<anemofilia>hmmmmm, nice, I didn't know about auctex
<Kolev>jpoiret, I've done `guix pull` quite a bit.
<nckx>Hi all.
<nckx>anemofilia: You needn't worry about channels if you don't yet want to. You can use ‘guix build -f FILE’ on something like and even install it if it were to work.
<jpoiret>Kolev: what does `guix describe` say?
<jpoiret>nckx: hello
<mirai>is it doable to run an offloading guix daemon with least-authority-wrapper such that /gnu/store/… is bind-mounted to /media/… ?
<jpoiret>Kolev: and if you run `sudo guix system reconfigure ...` does `guix system describe` give you the same commit as in your `guix describe`?
<jpoiret>if so you should reboot, if not then that's weird
<bobbma_>Hi, I have a problem with package compilation. I got an error: "gdbus-codegen: command not found". Is this program available in GNU/Guix, if so in what package? Thanks!
<jpoiret>bobbma_: in what context are you getting this error? while guix is building a package?
<jpoiret>gdbus-codegen should be available in glib:bin
<nckx>That command is in glib:bin, but if it's ‘Guix’ throwing that error, you'll have to amend the failing package expression.
<nckx>Or what jpoiret said ( ͡° ͜ʖ ͡°)
<jpoiret>I gave the lazy answer 🙃
<nckx>The edit would look something like adding `(,glib "bin") to native-inputs. Which package is failing, bobbma_?
<Kolev>jpoiret, reconfiguring
<bobbma_>nckx: I'm definig my own package.
<nckx>Ah. Then you'll know what to do.
<bobbma_>Okey, thanks.
<elevenkb>Can anyone let me donwload a guix archive of ghc-next... I'm too lazy to wait for it to build.
<nckx>Not unexpectedly, texlab fails its test suite:
<hwpplayer1>ACTION likes GNU Guix but can't install it in his computer !
<nckx>ACTION .oO ‘CONFIG_HID_BPF’ — Linux sure is going in… a direction.
<aitzkora>hi? is there a argument to change directory when configuring (for example go to a subdirectory src) when using cmake-build-system or I must add a stage before configuring to change directory ?
<mirai>nckx: can you look into <> when you get the chance to?
<nckx>mirai: Oh, thanks.
<mirai>might be related: <>
<aitzkora>I think I found a solution with an example like "(add-before 'configure 'change-directory (lambda _ (chdir "DevIL") #t))
<nckx>I'm runnning OpenSMTPd 7 but not with TLS AFAIK.
<nckx>aitzkora: That's the standard way. The only build-system-level option vaguely related would be out-of-source?, but there's no build-directory or similar.
<aitzkora>@nckx : thanks for confirming. I was thinking there was a builtin "option" in the cmake-build-system, but the add-before solution satisfies me
<nckx>ACTION removes libasr
<mirai>aitzkora: no need for #t
<nckx>Oops, missed the #t.
<juliana[m]>I wonder if adding a build-directory to the standard arguments list might be worthwhile. I frequently find myself needing change directories during builds. Everything else inherits from gnu-build-system, right?
<aitzkora>@mirai : i stole that from "devil"'s recipe, i will simplify it. thx for the feedback
<nckx>There's a difference between #:build-directory and, e.g., #:tests? that I'm too lazy to explore right now but makes me prefer not to.
<nckx>Not that my opinion counts for anything :)
<janneke>if you need a separate build directory, there's #:out-of-source?
<nckx>Which defaults to #t.
<nckx>elevenkb: OK, now I understand why. :)
<nckx>Test suite is at 10%.
<podiki[m]>hi guix-ers
<nckx>ACTION is guix'd.
<nckx>‘guix time-machine --commit=82e2e3dd53fe0093700f69f434af4402dacf69ba -- shell --pure pdfarranger {adwaita,hicolor}-icon-theme gsettings-desktop-schemas -- pdfarranger’ crashes when clicking the top-left ‘Open’ icon with ‘Settings schema 'org.gtk.Settings.FileChooser' is not installed’. Any off-head ideas?
<nckx>Adding gtk+ on a whim starts building Mesa, which I'd rather not.
<nckx>Time machine because pikepdf is broken on master.
<futurile>ACTION waves hi to podiki
<ngz>nckx: Shipped pikepdf is very old (latest is 7.2.0, while ours is 2.16.1) and needs to be updated. This couldn't happen before core updates merge, tho.
<ngz>Fortunately, it has only 2 dependents.
<nckx>Yeah. I looked up the latest version & sighed.
<nckx>‘This couldn't happen before core updates merge, tho’ no comprendo.
<ngz>I cannot remember which one, but updating pikepdf required an up-to-date dependency that had many dependents, hence in the "core-updates" branch.
<nckx>I'm extraglad I didn't bother then.
<nckx>So there's a workin' pikepdf on c-u? I didn't think to check.
<nckx>ACTION checks.
<ngz>Nope. CU was merged, but pikepdf stayed old.
<nckx>IC 370d1cdb61e1b9b5467d2aed4cab89c2a8e2a491.
<nckx>No ideas what could provide org.gtk.Settings.FileChooser ?
<ngz>Hmm. It somehow rings a bell. I think you need to add glib-or-gtk-wrap phase from glib-or-gtk build system.
<nckx>Worth a shot! Thanks.
<Guest28>shepherd has this issue that it uses 100% of a core. where can i track the issue to know i can update my system?
<nckx>ngz: It worked! Thank you ♥
<ngz>Now I remember. Pikepdf requires an up-to-date qpdf package aaand ours is outdated… (11.1.0 instead of 11.2.0).
<ngz>Ah. It seems to build even with 11.1.0, contradicting its own documentation.
<podiki[m]>for the "request to merge branch" issue procedure, just need to make one? or is there some procedure for blockers (for mesa-updates it doesn't depend on any other branches/patches other than what it has on the branch already)
<ngz>podiki: I think that in your merge request, you need to reference the last branch in the merge queue list. I don't know if there's some special syntax for it. Probably.
<johnabs[m]>Hi Guixers! Does anyone happen to know of any resources on packaging and installing pre-built binaries for guix? I'm trying to get a really important, FLOSS research package working, that someone else was able to get working with guix-shell, but I'd like to be able to install it and run it as I would with any other package. I understand we usually prefer to build from source when possible, but this thing relies on so many ancient libs that m
<johnabs[m]>given up on that route.
<nckx>ACTION happily merges PDFs hours before a deadline instead of minutes, feels all grown up.
<podiki[m]>what sorcery is that nckx :-P
<podiki[m]>(the pre-deadline, not the pdf merging)
<podiki[m]>ngz: thanks. I guess it is ruby team and tex-team-next in the queue?
<nckx>The ‘not having to shave the Guix yak first’ one.
<podiki[m]>can't start my day without some (emacs/guix/etc) shaving first! that's just proper hygiene
<ngz>podiki: tex-team-next (my precious!) is the last in the queue.
<nckx>I should probably keep an important_work_stuff profile around next time.
<podiki[m]>ngz: got it! how is that looking, that's to update the full texlive distribution?
<ngz>Well, I compiled texlive-bin so many times that my computer doesn't need to look at its source code anymore to build it =/
<civodul>ngz: thanks for all the work on texlive!
<podiki[m]>I was thinking "mesa-updates" might stick around like the "kernel-updates" one as mostly a way to build substitutes, but not sure what others think
<janneke>johnabs[m]: un-repproducible research as a feature? ;)
<podiki[m]>yes, thanks ngz, just being an end user of tex packaging is hard enough sometimes!
<johnabs[m]>janneke: Well in this case the package is zotero, so not critical to my research results, just to my ability to ingest papers quickly into org-roam 😂
<johnabs[m]>Well, I have it set up with a lot of cool features in emacs! It auto builds citation keys, adds to my global bib file for easy referencing via org-ref, and allows me to insert pdf links into my notes digest so I can rapidly revisit papers when I'm in the appropriate context to do so, and those same links compile to citations in latex! (And right now, they're all hiding in a set of zotero folders that aren't easily searchable and it's a huge p
<johnabs[m]>(And sorry, this is off topic, I'll stop 😅)
<ngz>Oh! I encounter that "no code for module (guix ui)" error when trying to use "git send-email". cbaines helped me fixing it a couple of times already, but it just doesn't stick.
<hwpplayer1>hi people !
<swaying_reed[m]>I'm ready to give guix a shot. If I have an Intel WiFi card in my laptop, my best bet is using an ethernet dongle? (e.g. something listed in h-node as F/OSS compatible)
<nckx>Second to ‘simply’ replacing the built-in card if possible (and allowed by your vendor): yes.
<paulapatience>ngz: I just run it in ./
<swaying_reed[m]>nckx: thanks!
<nckx>Good luck!
<nckx>So close.
<johnabs[m]>Is there any advantage to building a package over running something in a container (it seems like there is, but I want to make sure I'm not putting in more effort than I need to)
<swaying_reed[m]><johnabs[m]> "Is there any advantage to..." <- not a guix user (yet) but from my experience, it depends. For example, there is a performance penalty to running a database in a container versus bare metal.
<swaying_reed[m]>i usually try whichever is easier, then only switch if i encounter a good reason to.
<johnabs[m]>swaying_reed[m]: Oh lovely :( I really need to get this zotero package working then, because my database is massive (And the shell version won't work as gcc:lib is no longer a thing 😭)
<johnabs[m]>Thanks for the input btw :3
<podiki[m]>johnabs: re: gcc:lib you can use -e '(list (@@ (gnu packages commencement) gcc) "lib")' as the input for now
<podiki[m]>which reminds me to nudge along
<johnabs[m]>podiki[m]: Even for the guix shell command?
<swaying_reed[m]>johnabs[m]: I love Zotero. Didn't know there was a container image for it.
<johnabs[m]>This is what I'm trying to run, roughly: guix shell --container --emulate-fhs --network --share=/tmp/.X11-unix --share=/home/$username libx11 gcc:lib gtk+ dbus-glib libxt -- ./zotero
<podiki[m]>johnabs: yes, that is for a guix shel
<johnabs[m]>OH Gimme a sec, lemme try it
<podiki[m]>gcc:lib is not accessible directly otherwise, currently
<nckx>What's ‘a container’ here?
<podiki[m]>yup, just replace "gcc:lib" with that for now, that snippet directly references the hidden package (the "@@" does that)
<nckx>ACTION wishes it were called @_@
<johnabs[m]>podiki[m]: That almost worked, but I'm getting this hilarious error: guix shell: error: ./zotero: command not found
<johnabs[m]>hint: Did you mean 'derb'? Is derb a thing, or has the absurdity of life decided to kick down my sandcastle? XD
<RavenJoad>I have a "2" GPU desktop. An ASPEED Video AST2500 on the motherboard (its a server board) and an Nvidia GPU. Currently only the AST2500 is being used by X. The GPU is seen by lspci, and xrandr sees it as a possible provider, but I cannot seem to use the monitors connected to the GPU. Has anyone solved something similar on Guix before? My other Linux distro (on the same setup) does not have this problem.
<johnabs[m]><swaying_reed[m]> "I love Zotero. Didn't know there..." <- I also love it, but getting it to work at all has been a struggle. I'm tempted to just install nix on guix system and use their version, even though I'd prefer to keep everything "guix-y"
<johnabs[m]>Actually, didn't there used to be a nix->guix converter or am I mis-remembering things. And if so, does it still exist?
<nckx>elevenkb: The test suite failed.
<johnabs[m]>nckx: My disappointment is immeasurable, and my day is ruined 😭 (But at least my memory is sharp, lol)
<nckx>My work here is done.
<podiki[m]>johnabs: likely means you are missing a dependency (or not in an FHS container but you are)
<podiki[m]>you can try an strace or LD_DEBUG=libs
<podiki[m]>(there's also zotero on flatpak, if that is helpful)
<johnabs[m]>podiki[m]: Okay, I will try to figure this bit out. Apparently guix doesn't have flatpack anymore unless there's an importer? I couldn't find the flatpack package...weirdly enough because the flatpack website has installation instructions which I can't reproduce?
<podiki[m]>we do have flatpak
<podiki[m]>guix install flatpak
<johnabs[m]>I tried that, it said package not found?
<johnabs[m]>"guix install: error: flatpack: unknown package"
<johnabs[m]>there's no c
<johnabs[m]>Pardon my illiteracy
<podiki[m]>johnabs: on the shell front, try with -D ungoogled-chromium which is likely overkill but gets you a ton of packages that are often expected for desktop programs
<johnabs[m]>podiki[m]: Trying it now :)
<podiki[m]>guix shell --container --emulate-fhs --network --share=/tmp/.X11-unix libx11 -e '(list (@@ (gnu packages commencement) gcc) "lib")' gtk+ dbus-glib libxt -D ungoogled-chromium --preserve='^DISPLAY$' --preserve='^XAUTHORITY$' --expose=$XAUTHORITY -- ./zotero
<podiki[m]>that works, or at least opens a program
<podiki[m]> with the exception of the recent gcc:lib change, that should still be a helpful reference for some tricks
<johnabs[m]>podiki[m]: A) Bless you, and thank you kindly for your assistant, it's open!! 😁 B) Is there a (preferably simple) way to go from container to package or is that a bridge too far?
<podiki[m]>not that I know of, though I suppose you could make that a simple script and package that with the inputs needed?
<podiki[m]>still, it would be in a container and likely you'll want to expose/share other things to use it
<johnabs[m]>Crud :(
<cbaines>oh no, "Signals delivery fails constantly" is now affecting the qa-frontpage
<Guest28>i made a guix system image but only need to overwrite partition 2 of my sd card with that image.  partitions of the iso are how would i now use dd to only copy the 2nd partition to dev/sdc2?
<nckx>Guest28: I'd use ‘losetup -P’ to create a partitioned loop device. That will scan the ‘iso’ (really: a hybrid) image for partitions create /dev/loop0p2 (IIRC) devices that you can then dd.
<nckx>s/create/and create/
<johnabs[m]>Quick question about deleting old system generations: does it need to be done as root?
<Guest65>Looks like poetry has been unable to build since may...
<johnabs[m]>Nvm, it's working :)
<attila_lendvai>johnabs[m], yes, except home generations.
<johnabs[m]>attila_lendvai: Ty! I (somehow) ran out of inodes, so now I'm clearing everything out so I can try to build a new package xD
<attila_lendvai>johnabs[m], this might be an unnecessary ritual, but after a gc i usually run a reconfigure, just to be sure everything needed is there in the store.
<johnabs[m]>attila_lendvai: I've not done that before, but I may start, just in case. Thanks for the tip! :D
<attila_lendvai>johnabs[m], there used to be at least one bug (now fixed), but who knows... it was much more headache to restore a non-booting guix than an extra reconfigure running in the background.
<johnabs[m]>Send help xD
<johnabs[m]>I went from 100% inode usage to 5%😭😂
<podiki[m]>Guest65: known problem with a pending patchset, but requires some more work (changes to python build system):
<Guest28>nckx thanks
<nckx>frog: Have you perchance tried updating opensmtpd?
<frog>nckx: i saw there were commits. updating my system right now
<RavenJoad>So looking at the Xorg log, I see all GPUs and the 2 monitors hooked up to the Nvidia card. It seems like X is choosing to start on the wrong GPU/Screen, and I cannot seem to change it back.
<jlicht>hey folks
<tex_milan>Hello, what path does (patches (search-patches ...)) expects? I have my override and it always tells me it cannot find the patch.
<rekado>tex_milan: in the guix repo it’s gnu/packages/patches
<tex_milan>I have my set of package definitions in ./ext/my/packages and using guix -L ext ....
<nckx>Does ext/ not work?
<nckx>(gnu/packages/patches is a hac—er, special case:
<frog>opensmtpd works for me again. thanks tobias
<nckx>You're welcome, and sorry for breaking your mails.
<nckx>Upstream released the fix at the exact same time I pushed a work-around.
<tex_milan>nckx: thanks! major issue was that created a patches folder in ext/ and that does not work. now I copied patches directly to ext/ and it works!
<nckx>This fun rule applies to channels as well: even if you use, e.g., <root>/tex_milan/packages/*.scm for patches, you'll have to put your patches in <root>. No tex_milan/packages/patches/ for you.
<tex_milan>nckx: don't want to judge :) but that looks funky
<nckx>It smells like fine French cheese.
<tex_milan>nckx: even more funny is that now when I have patches in ./ext and running guix shell -L ext python (original guix python) it wants to use my patches, not the guix's. funny. (i am trying to package python-3.9.9)
<tex_milan>nxkx: I had to rename them so they do not collide.
<Guest28>(swap-devices (list (swap-space (target "/run/swapfile")))) if I have something like this, do I need to create the swap device or does it just work out of the box?
<nckx>I guess you could change search-patch to be more clever, if you'd like.
<nckx>Guest28: You must create it with the desired size.
<Guest28>nckx so I can't define it in the operating-system section and need to manually create it?
<nckx>It's like partitions in this way.
<nckx>Here too, Guix could be extended to support creation at boot ☺
<Guest28>ah perfect. thanks
<nckx>The activation probably happens too early to do so via a regular system service, but I didn't check.
<agnem>guix home users: how do you propagate environment variables such as WAYLAND_DISPLAY into home shepherd processes?
<Guest28>there is  linux-libre-arm64-generic and linux-libre. what would be a reason to use the first one?
<vagrantc>Guest28: not having to track down exactly which modules to hard-code in your initramfs
<vagrantc>Guest28: and they change across kernel versions ...
<vagrantc>Guest28: the arm*-generic variants have more stuff as built-ins.
<Guest28>so basically better hardware compatibility?
<vagrantc>Guest28: they also follow upstream kernel configs, so are more likely to enable new features that were not present on older versions ... and the linux-libre kernel configurations are manually curated by guix folks
<vagrantc>Guest28: sort of ... but not guaranteed
<Guest28>ah perfect. this answers my question. thank you very much
<vagrantc>i would prefer the regular linux-libre if you can figure it out, in general, but for quickly trying a new board, i would also first try with the -generic variants
<nckx>IMO ‘curated’ implies a vision & diligence not necessarily present.
<vagrantc>manually processed?
<nckx>‘Committed to git’ 😛
<vagrantc>nckx: nice to your nick again ... been a while ... or maybe i haven't been paying attention
<nckx>Nope, I was away doing life.
<nckx>Poorly. Would not do again.
<vagrantc>ACTION misplaces a "see"
<johnabs[m]>Is there any chance that someone is thinking of casting the dbus-glib package into the fiery depths because it says it's obsolete? Because I really hope it'll stick around 😅
<nckx>johnabs[m]: removing it would currently break 4062 packages, so you're probably safe for quite some time. Why?
<nckx>We could add a comment if it's irreplaceable for some use case.
<civodul>cbaines: hey! i fixed a "funny" bug in squee:
<nckx>vagrantc: By the way, nice to see you still around as well. :)
<nckx>This applies to everyone without exception.
<Guest28>Somebody using Emacs with TRAMP and eshell? If I connect to a machine and insert commands like guix describe it reconnects all the time.  It says in the echo area: "Tramp: Opening connection guix for " but ls mv and so on works without that annoyance
<Guest28>I just don't understand why it needs that reconnection.  It should be cached by default
<vagrantc>nckx: without tracebacks too, hopefully
<nckx>ACTION applies vagrantc as a procedure for that horrible pun.
<nckx>Heh, so you can't reconfigure a freshly installed system without network access. At all. AFAICT.
<nckx>Guix, I missed ya.
<Guest28>Why should it work without network?
<nckx>Mainly because the machine currently lacks it.
<nckx>But also because it's counterintuitive for it not to.
<Guest28>Ah, you mean kinda as a joke?
<nckx>And that's coming from someone who's familiar with it :o)
<Guest28>or because you need network to download network drivers but you can't because you have no network?
<nckx>No, I made no changes to the required packages at all, that's why it's counterintuitive.
<nckx>One expects fixing a typo in a configuration string not to require downloading fresh packages. But it does, because Guix is a nesting little doll.
<nckx>This wouldn't happen later, after the first pull.
<johnabs[m]><nckx> "johnabs: removing it would..." <- I just built a new package that only wants to work with it, so make that 4063 xD
<nckx>The reason I can't manually configure the network is another gem: the ‘Web console’ offered by this back-alley VPS provider can't handle the shift key, so I can't type ‘.’.
<Guest28>I remember trying out Debian on my HP server to find out it requires drivers to make ethernet work I can't download because the network is bricked without it. That was so stupid and I lost faith in humanity.
<nckx>There's probably some obscure way I could echo one but meh.
<nckx>Guest28: Well… doesn't Guix have that same issue?
<nckx>Or were these drivers free?
<Guest28>No, it wasn't meant against Debian but against HP.  Why would you do that? I understand a proprietary OS like z/OS and so on but come on it is just Ethernet.  What is so highly confidential they need to hide?
<Guest28>couldn't update firmware either since that 10 year old machine required "service subscription" to download. Yeah enterprise market is fun. You can't even play around with old machines
<Guest28>It is a VPS? How did you managed to install Guix System on it? Isn't only possible if the company supports it?
<nckx>I've installed Guix on about 10 different providers and none of them have ever supported it 😅
<nckx>This one was very easy: it offers a SystemRescueCD virtual ISO, so you can just install Guix to RAM and init that way, and it doesn't do any weird out-of-band boot stuff.
<Guest28>Ah, didn't know that is possible.  Do you have a tutorial?
<Guest28>So some rescue iso, guix binary, normal guix init?
<nckx>And with it's much easier nowadays.
<nckx>Now, each VPS provider has at least one weird and incomprehensible quirk that requires a bit of extra knowledge of Guix/GNU/Linux administration, but the concept is simple.
<Guest28>Ah, I always thought it only works on Linode since there is a cookbook for it. But what I don't understand is the network issue.  Shouldn't this just work out of the box with virtualized hardware or are they proprietary, too?
<nckx>It wasn't related to drivers. It would have worked out of the box if I'd known that the interface name would change between SysResCD and Guix System. But since I specified an ‘invalid’ interface name, the network wasn't configured, I couldn't SSH, and the HtMl5 CoNsOlE is so janky I couldn't ‘ip’ my way back to sanity.
<Guest28>Ah, now I gotcha
<nckx>I do myself no favours by having a custom keyboard layout. It's probably borderline usable for people who don't swap control and caps lock, for example.
<nckx>Anyway, thanks for helping me kill time during attempt #2 ☺
<Guest28>Shouldn't the interface name always the same in newer Linux? They have now this feature I don't know the name of it at the moment.  Kinda strange since that is the reason theyi mplemented it
<nckx>‘Git error: the SSL certificate is invalid’ oh come on.
<Guest28>try date
<Guest28>does it say 1970?
<nckx>Nope ☺
<nckx>The date is correct.
<Guest28>ah okay, had that issue on a machine without ntp and thought it is 1970
<Guest28>and you mentioned network issue so I thought you know
<nckx>Yep. It was a good tip!
<Guest28>forgot nss-certs to install?
<rekado>our JDK cannot be extended with jmods. jlink complains about hash mismatches. I wonder if that’s due to the stripping of timestamps. Perhaps it happens after the hash was computed…?
<rekado>nckx: sometimes le-certs helps
<rekado>our nss-certs package is a bit old
<rekado>3.88.1 vs 3.90.0 (LTS) or 3.91.0
<rekado>(well, a few months old)
<Guest28>is udisks-service-type used to detect sd card reader changes in real time? I either need to reboot or run udisksctl status to detect it. kinda silly
<nckx>It was stupider than that. I hadn't installed nss-certs at all.
<Guest28>yea, happened myself 2 weeks ago
<nckx>I had le-certs in the store, so could probably have hacked my way towards a working Guix myself, but since I can now scp I just cheated and copied my laptop's…
<nckx>Whenever you're feeling clever and good about yourself, do some computer.
<nckx>ACTION doesn't use udisks.
<nckx>Guest28: I'd missed your message about nss-certs, sorry.
<Guest28>no problem
<Guest28>does nouveau require some special stuff in gnu guix or something? if i want to use it, it doesn't even start x
<vagrantc>wow, opensbi has strangely grown a sizeable reverse dependency graph
<vagrantc>gnome... gnome? for building some riscv64 firmware. really? really?
<vagrantc>astoundingly only 13 packages, though ...
<Kolev>vagrantc, that's crazy.
<vagrantc>it's because of gnome-boxes
<vagrantc>sure would be nice to decouple that from the rest of gnome
<vagrantc>i was not nearly as surprised to find qemu and such
<vagrantc>wait, what ... memtest86+ ... how on earth?
<vagrantc>that makes approximately zero sense.
<juliana[m]>gnome-boxes doesn't even work half the time, which is fun. I set up all the virtualization stuff and when i asked it nicely to make a VM for me it responded "no <3"
<juliana[m]>and that's not even a guix-specific issue; I think the only distro I've ever gotten it to work on was Fedora
<juliana[m]>anyway, i may have a small vendetta against gnome-boxes
<daviid>juliana[m]: don't know anything about gnome-bozes, but fwiw, maybe you could ask for help on #gnome-boxes on as well
<juliana[m]>oh it's okay i just use virt-manager or guix system vm
<apteryx>juliana[m]: for what it's worth, GNOME Boxes works fine for me on Guix System
<apteryx>(with the required services such as virtlogd or something added)
<atka>hi guix, how can I stop guix home from mangling my .bash_profile?
<atka>I have (guix-defaults #f) for the bash home service and my .bashrc is left alone now but .bash_profile keeps getting partially duplicated and spliced
<Guest9>Hi guix
<Guest9>I think the hurd and guix IRC log archives have stopped updating
<nckx>That would be because the month rolled over.
<janneke>rekado: seems to be there, but under yesterday's date?