IRC channel logs

2023-09-04.log

back to list of logs

<ulfvonbe`>> propagation only matters wrt to a resulting profile?
<ulfvonbe`>Propagation also matters for determining what the complete set of inputs to a bag's builder is - see bag-transitive-*-inputs in bag->derivation and bag->cross-derivation in (guix packages)
<nikolar>how do i correctly use can someone tell me how to use %current-target-system, it always returns #f
<podiki>anyone have any handy scheme/guix-fu for finding all packages that propagate another particular package? (in this case libx11)
<iyzsong>podiki: i get this: https://envs.sh/2T_.scm
<podiki>lizog: oh nice, thanks!
<podiki>that's not as much as I might have thought, though I guess a lot when we think of how sparingly we try to be with propagated-inputs
<bailey_zappe> there's a lot of past mailing list chatter about Ibus in GNOME
<bailey_zappe> but I haven't seen anyone's anecdote that it's fully functional
<bailey_zappe> in 2023. Is anyone using Ibus with IMEs like Rime or Anthy
<bailey_zappe> right now? Thank you
<podiki>iyzsong: thanks! (whoops tagged wrong person)
<iyzsong>bailey_zappe: well ibus should work after you have setup some environment variables (eg: GUIX_GTK3_IM_MODULE_FILE, GTK_IM_MODULE, etc) mannually.
<apteryx>or add gtk+ to the user profile?
<apteryx>shouldn't these search paths be added to ibus itself though?
<iyzsong>i think so, a search-path-specification could be added to ibus and fcitx. also, install gtk+ won't help since we haven't added it to gtk+.
<apteryx>ah, it's weird to have a patch adding support for GUIX_GTK3_IM_MODULE_FILE but no associated search path?
<apteryx>ibus for me regularly stops working, I'm not sure if it has to do with caching or something more serious
<apteryx>(with anthy)
<Rekt_By_Guix>Hey just found this IRC. I have been relentlessly attacked by, I believe, a combination of GUIX, Googles LLVM database, buildroot/bot, and, from what I can tell Apache IoTDB. I am desperate for a solution, although I'm fairly certain it may not exist.....But, lets keep it simple with IN THEORY, If somebody were to use APACHE to take control of my
<Rekt_By_Guix>local host, the essentially clone the firmware, injecting unsigned/malicious packages, then rewrite with RO permissions, can I recover? Or at the very least prevent further damages? It's a lot to explain but my cellular signal has been tampered with, APN does not align with character, and I just recently, for the first time, broke the connection by
<Rekt_By_Guix>button mashing a proxy and entering non existent DNS addresses/passwords for people that live nearby. I have video evidence of them approaching my house with a small unidentified device, pressing a couple buttons, and walking off. Too long of a story so, yeah....theoretically. Anyone?
<Rekt_By_Guix>Carrier**
<iyzsong>apteryx: um, i thought ibus's cache should work, but it doesn't. maybe i'll send a patch to disable its registry cache later..
<RavenJoad>What is the more general way to validate gexps in a define-configuration field's value? file-like? gexp? package? I have an alist of events to handler scripts '((annoyme . #~(const #f)) (doshutdown . #~shutdown)). Should the gexp be checked with (gexp?... )?
<apteryx>would it make sense to accept other types as well such as plain file objects?
<apteryx>if not, gexp? seems the correct choice
<RavenJoad>No, these must be scripts, so only program-file makes sense (from a *-file perspective).
<RavenJoad>But a gexp may be more elegant for simple things. I could then wrap the gexp in a program-file for the user, but I am getting ahead of myself. I need (gexp? #~(const #t)) to return #t first.
<RavenJoad>Is there a better way to make an alist of symbols & gexps than `((annoyme . ,#~(const #f)))?
<podiki>is ci.guix.gnu.org having some issues? does seem to respond eventually but sometimes times out?
<apteryx>load is high (20)
<tylerm>hey guixers! just got guix installed but i dont have any soundcards :(
<tylerm>i installed alsa services, pipewire is setup, and i also added alsa-firmware though no luck
<tylerm>wondering if anyone has any ideas, and thanks for any help. (also im running a dell laptop with a intel soundcard, and i have the nonguix channel setup)
<RavenJoad>tylerm: First thing to ask is "Is Linux the kernel seeing the sound cards?" If the kernel does not see them, nothing else will. lspci and lsusb will be your friends here.
<tylerm>ok thansk, lspci gives the device "00:1f.3 Multimedia audio controller: Intel Corporation Tiger Lake-H HD Audio Controller (rev 11)", so the kernel does see it
<apteryx>podiki: I think perhaps it's struggling with too many pending builds to keep track of (> 40000), per https://ci.guix.gnu.org/metrics
<podiki>hrm. yet all the x86 workers are idle?
<podiki>pending arm builds?
<apteryx>I guess so
<RavenJoad>tylerm: I'm unfamiliar with linux audio, but does the next layer up see the audio card? So Alsa in this case?
<apteryx>it's a bit hard to say from the CI
<apteryx>i mean web interface
<podiki>yeah, I just browsed the metrics and builds, but can't be sure
<tylerm>RavenJoad: alsa does not see the sound card, so the issue im assuming here is that there is no sound driver that is found for the soundcard
<RavenJoad>lspci can tell you the sound driver in use (if there is one).
<tylerm>usually packages such as sof-firmware are fine, and have worked on other distrubutions, but for some reason here it is not
<apteryx>podiki: I guess I could reboot the box, there have been other oddities recently
<apteryx>such as 'guix-daemon' failing to run 'guix gc' with a memory busted message
<podiki>the blunt tool of rebooting, yet (often) so effective
<podiki>don't do anything on my account at least, just wondering :)
<apteryx>I just don't like to reboot it because there's always some shenanighan to manually fix with anonip and some rsync thingy to start manually for mumi
<apteryx>the rest is fine (other than Dell firmware taking ages to post)
<tylerm>RavenJoad: ok so there is a driver in use "in use: sof-audio-pci-intel-tgl"
<RavenJoad>Next up is to track down why the audio card/driver are not seen by alsa/pulse/pipewire/whatever.
<tylerm>i am looking at my pipewire service right now
<tylerm>Ok so i have figured out how to get sound drivers working on intel laptops
<tylerm>Warning this is non free (so im sorry for those who prefer libre)
<tylerm>But you have to install BOTH, the intel-microcode and the sof-firmware packages
<tylerm>And append them to the firmware in your system configuration
<tylerm>Just wanted to pass on my knowledge after my 5 hours of pain
<the_tubular63>Can a service be part of a manifest ?
<jpoiret>the_tubular63: no
<the_tubular63>Thanks jpoiret
<the_tubular63>Also, how do I use a pipe inside my cron job ? It doesn't say in the example :/
<PotentialUser-84>Hi all ! I am getting an "unsupported manifest format" when I run a guix package --install-from-file
<iyzsong>PotentialUser-84: hello, are you using btrfs?
<PotentialUser-84>No, ext4
<PotentialUser-84>Last night I tried to build a package that I built before a few months ago, but I thought I would try the latest version, it failed on a sane phase build environment, so I did a guix pull, now I get the manifest format error.
<iyzsong>PotentialUser-84: okay, what's the version (should be: '(manifest (version ?) ...)' in the ~/.guix-profile/manifest file?
<PotentialUser-84>version 4
<iyzsong>seem fine, and what about 'which guix' and 'commit' in 'guix describe'?
<civodul>Hello Guix!
<PotentialUser-84>failed to determine origin, perhaps this guix command was not obtained with guix pull ? It's string version is 1.3.0
<PotentialUser-84>hi civodul :)
<PotentialUser-84>which guix goes to /usr/bin/guix
<iyzsong>PotentialUser-84: oh, that's not right. 'which guix' should be ~/.config/guix/current/bin/guix
<iyzsong>PotentialUser-84: are you on a foreign distro or guix system?
<PotentialUser-84>I have this in my bashrc export GUIX_LOCPATH="$HOME/.guix-profile/lib/locale"
<PotentialUser-84>a foreign distro, I am not hardcore enough to run guix :')
<iyzsong>PotentialUser-84: okay, you also need put PATH from current guix before the local one.
<jpoiret>you should have env-setting stuff in .profile, not in .bashrc
<jpoiret>otherwise it's sourced for every new interactive shell and will mess up eg. `guix shell`
<andreas-e>Hello civodul. Long time no see!
<PotentialUser-84>hmm, this is strange, if I use a gnome-terminal it uses my ~/.guix-profile guix binary, if I use a vterm or eshell, it uses /usr/bin/guix
<PotentialUser-84>Thank you so much all :)
<PotentialUser-84>This seems to work now.
<iyzsong>🥳
<PotentialUser-84>ah well, almost ... need a python package that doesn't exist ... can I install-from-file it and then use it in the main scm file that I am writing ?
<PotentialUser-84>it's python-num2words
<andreas-e>Maybe "guix import" it and send a patch to guix-patches@gnu.org ?
<iyzsong>PotentialUser-84: how it (the main scm file and python-num2words) going to be used?
<PotentialUser-84>so the main file I'm trying to create is python-odoo.scm version 16, and it seems to need num2words (on version 15 that I did last year I just commented it out and it worked). So now I need num2words but it isn't in guix.
<PotentialUser-84>ok, I did a guix import pypi num2words and it generated a guy.
<iyzsong>PotentialUser-84: not 'install-from-file', but that's a question for guile module. you can put it (define-public python-num2words (package ...)) into python-odoo.scm, or any other file (which then need add a #:use-module in python-odoo.scm)
<PotentialUser-84>ohhhh, ok. So if I put the define-public python-num2words above the define-public python-odoo it will pick it up?
<iyzsong>yes, and you also need add python-num2words to python-odoo's 'inputs' field.
<PotentialUser-84>Great, it wants to try now :) Thanks.
<civodul>time-machine’s broken, oops! https://issues.guix.gnu.org/65352#13
<civodul>howdy andreas-e!
<jlicht>civodul: howdy! I just came to report the broken time-machine, but you were faster :-)
<cbaines>morning all :)
<bonz060>2023-09-04 10:13Hi Guix! I have guix pinned at a4e9842a70775a54bbe1369881b739e7ea9a6432 which has both Python 3.9 and 3.10. In your package inputs, how do you specify the version 3.9? Even better, how would one make python-wrapper default to 3.9?
<bonz060>cbaines: morning
<rekado>silly idea: “guix search --interactive” to interactively filter the returned packages
<rekado>another silly idea: use embeddings instead of keywords to find matching packages
<nikolar>how can i force a rebuild with `guix build`
<nikolar>i am running `guix build -f package.scm`
<cbaines>nikolar, --check will have that effectr
<cbaines>*effect
<nikolar>thanks
<charlotte1123581>hey guys, what do you use for auth/ldap on guix? im looking for something like freeipa i guess
<ngz>Hello Guix!
<PotentialUser-84>hi :)
<nikolar>hello ngz
<PotentialUser-84>13 new packages so far :O
<ngz>What are you packaging?
<PotentialUser-84>odoo 16 , it's a python based ERP system
<PotentialUser-84>but so far 14 of the packages didn't exist in guix and I have to pypi import them and create define-publics
<ngz>At least, there's an importer :)
<PotentialUser-84>That's really true :)
<nikolar>i wonder if i can add GUIX_PACKAGE_PATH for guix home
<nikolar>for custom packages
<xelxebar>Sometimes I wish there were a way to containerize sheperd services.
<ngz>nikolar: I think GUIX_PACKAGE_PATH is deprecated. At least, each time I mention I still use this environment variable, I have the feeling I'm being frowned upon! :) Channels is the way to go, apparently.
<nikolar>well it's easier to just modify a file than to have a whole repo to manage, but i guess that's my best option
<xelxebar>Dang. And the texlive packaging effort continues. Oodles and oodles of them getting added every time I pull.
<ngz>xelxebar: The TeX Live packaging effort is over.
<ngz>There will be more changes to the packages, but they are all in now.
<xelxebar>Oh sweet! Solid friggin' work.
<PotentialUser-84>20 new define-hellos and no end in sight ;(
<PotentialUser-84>define-publics I mean :D
<ngz>xelxebar: packaging almost 4k packages was... interesting, for sure ;)
<PotentialUser-84>wow. 4k ??
<ngz>Yup. TeX Live contains roughly 4200 packages and a few hundreds were already packaged.
<PotentialUser-84>I wonder about the pypi importer, since it mostly works (it sets a null homepage and #f license about 60% of the time), can't they just write a scraper to automate it for all python modules?
<xelxebar>ngz: What did that look like in the trenches? Bulk imports and then a bunch of sanity checking and hand cleanups?
<PotentialUser-84>it also doesn't put in the define-public and just leaves it starting with (package
<ngz>xelxebar: Mostly. I first tweaked the importer and the build system. Then bulk import, license fix, build everything, edit descriptions and synopses, then push. Rinse and repeat until exhaustion of either the packages left or myself.
<ngz>PotentialUser-84: It's a small issue for all importers. I think using the "recursive" option (-r) adds the proper `define-public' wrapper.
<PotentialUser-84>did you forget a use-modules form ? No, I forgot how to spell X)
<xelxebar>That's an impressive grind, man. Single-handedly putting a dint in the gap between nixpkgs and guix packages deserves some accolades.
<PotentialUser-84>yep that must be tiring. I'm even getting a bit tired after 25.
<ngz>xelxebar: Thanks
<apteryx>civodul: hi! I thought I had fixed the main issue with the (gnu packages avr) module in 32792d2073, but I'm hitting yet another cycle when attempting to use (gnu packages avr) in (gnu packages firmware)
<apteryx>I've tried turning the avr-* components into procedure, but it seems that's not enough
<janneke>apteryx: are you able to comment-out some of the components to try to bisect which ones are problematic?
<elevenkb>Is
<elevenkb>Sorry, first time using raw IRC in a while. Ignore that.
<elevenkb>Anyway.. is there any way one can see why a certain home environment depends on a package?
<apteryx>janneke: I think it has to do with cycles introduced at the top level (guix records uses with inherit)
<apteryx>and I need to run 'grep --include='*.go' -rl avr-toolchain | xargs rm' between edits to really see the change, which is adding to my confusion
<apteryx>civodul: I think even (gnu packages cross-base) is faulty in not using packages at the top level
<apteryx>e.g. cross-binutils uses binutils as default argument
<apteryx>all macros (such as package) are evaluated eagerly (except for thunked fields), no matter where they are, right?
<civodul>apteryx: yeah, cross-base.scm is prolly faulty… but it’s “a bit special”
<civodul>hmm this is quite gnarly
<mirai_>apteryx: Did you get my message at <https://logs.guix.gnu.org/guix/2023-09-03.log#224302>
<apteryx>mirai: yes :-)
<mirai>Emacs (?) users, what/how are these “--8<---------------cut here---------------start------------->8---” lines inserted
<apteryx>C-c M-m
<apteryx>in a Message buffer
<apteryx>it's called 'message-mark-inserted-region'
<apteryx>it's the text mail equivalent to markdown ``` ``` ;-)
<mirai>what major-mode or package do I need for this?
<mirai>nvm, I figured it out
<podiki>howdy guix
<rekado>apteryx: when ibus stops working it’s most likely due to the cache(s)
<rekado>they embed store locations, which will eventually become outdated
<rekado>we’ve got a few bug reports about this
<apteryx>has someone looked into patching out the cache code paths of ibus?
<apteryx>(so that it doesn't create/use any cache)
<rekado>it’s been on my append-only TODO list since 2016 :)
<apteryx>*append-only*... haha
<apteryx>civodul: if we are to use any of the cross- something elsewhere than in inputs, it seems we'd have to do so in (gnu packages cross-base), right?
<apteryx>so all of the cross-compilers/toolchain procedures such as avr-toolchain (renamed make-avr-toolchain) would need to be moved to (gnu packages cross-base). Unless there are smart tricks we can do to use cross-* results in a package (inherit ) field.
<apteryx>in a different module
<apteryx>then these could no longer be used as packages, in the very narrow window where it worked, but at least it'd be reliable
<apteryx>bah no that wouldn't work... avr-gcc needs to inherit from the result of applying cross-gcc... which is at the top level. No matter where we put its definition it's going to be a problem.
<efraim>rekado: I've been thinking about mounting parts of .cache in a tmpfs so it gets cleared at each reboot. or maybe ibus has something like mesa where it can embed a hash of itself as part of the data?
<apteryx>mirai: if you want to reproduce my problem, you can simply add #:use-module (gnu packages avr) to the top of (gnu packages firmware), and try to build qmk
<apteryx>even if it's not use, the mere fact it's imported causes the bug
<rekado>efraim: we risk losing the configuration when resetting the cache
<apteryx>why is the configuration kept in a cache?
<apteryx>that sounds like a bug
<efraim>I would've thought it'd be in ~/.config
<rekado>let me check
<rekado>.cache contains a registry of input methods (with store locations). It looks like this is only a database of known input methods. It can be deleted and recreated with “ibus write-cache”.
<rekado>(“ibus read-cache” shows the contents in readable XML)
<rekado>ibus-libpinyin actually keeps a database of user phrases in .cache. It’s not exactly precious, but perhaps unexpected to lose it on every reboot.
<michzappa>I found that my .../gtk-3.0/whatever/immodules.cache file didn't contain any of my extra installed IME's (rime and anthy). So even after I set the GUIX_GTK_..._IMMODULE_FILE variable I wasn't terribly suprised nothing changed.
<michzappa>but 'ibus read-cache' does spit out stuff about anthy and rime
<michzappa>After running 'ibus write-cache' manually
<michzappa>And after restarting the ibus daemon I was able to add the IMEs from GNOME settings
<alethkit>Does anyone know of any channels that package tailscale?
<alethkit>I'm not too familiar with setting up system services
<podiki>alethkit: a quick search does show some packages on e.g. github, no idea if they work etc. but you can search there (as well as others like gitlab, codeburg and so on)
<mwette>Is there a comment to get a list of packages in my current profile? Sorry, but did not do a thorough scan (this time).
<mwette>I have guile installed but need to get to the includes and libs.
<RavenJoad>mwette: I think you can also go through the profile's symlink.
<RavenJoad>Is the only way to run (substitute ...) in a package phase?
<mwette>pwd
<mwette>di
<mirai>rekado: they should be going into XDG_STATE_HOME then
<alethkit>podiki: That makes sense! I wonder why they haven't been upstreamed yet.
<vivien>Hello Guix!
<vivien>I could not help but notice the gnome branch is failing because cogl can’t open GL as a gmodule
<vivien>(there may be other reason for failure!)
<viaken>mwette: guix package -I
<vivien>(reasons*)
<vivien>I’m trying with glib 2.77.3, maybe that one has a gmodule that can dlopen GL
<mirai>What happens when two packages provide the an executable with the same name? (i.e. which zcat will I get if I install zutils and xz)
<vivien>mirai, `ls -l $(guix shell zutils xz -- which zcat)` tells me that of zutils
<vivien>I’m not sure it helps ^.^'
<mwette>viaken: thanks much!
<charlotte1123581>anyone tried configuring samba4 DC before?
<podiki>mirai: I think it is position dependent, similar to a bug report....let me find
<podiki>mirai: https://issues.guix.gnu.org/58859 and linked issues
<RavenJoad>I assume Guix does not handle circular package output substitutions well? I seem to have blocked a system build that way.
<mwette>any idea why fibers does not show up here: https://paste.debian.net/1290936/
<vivien>Gnome update: well, no, glib 2.77.3 isn’t better.
<nckx>RavenJoad: Not entirely sure what ‘package output substitutions’ are, but no, circular anythings are something of an antigoal for Guix.
<nckx>mirai: It depends on order (‘guix shell zutils gzip’ != ‘guix shell gzip zutils’), but since Scheme lends itself perfectly to accidentally reversing lists, I bet there's no one consistent rule across all Guix command line/APIs/manifests/….
<civodul>oh, Matrix?
<Guest69> https://paste.debian.net/1290940/ Did I define my job correctly?
<bumble>I have basically stopped using matrix now that the bridge isn't engaged with irc
<bumble>seems like the activity level in matrix rooms is at the lowest point since I started using about 4 years go
<bumble>most of the interesting activity came from the irc side
<bumble>a bot of some kind started spamming long messages in the guix matrix channel and I left
<bumble>found this just now searching "matrix messaging decline" https://jakstys.lt/2023/my-declining-matrix-usage/
<vivien>I have to say, it’s a miracle when unmaintained applications that depend on unmaintained library still compile. It’s unfortunately not the case with abiword, that depends on cogl on gnome-team.
<vivien>libraries*
<vivien>Is it acceptable for guix to have an old version of glib for such cases?
<Guest69>Does that mean that there is no matrix bridge now anymore?
<vivien>Guest69, replace the "bash /root..." string literal with (system* #$(file-append bash "/bin/sh") "/root/guix-rpi/update-mirrors.sh")
<vivien>Or rather (system* #$(file-append bash "/bin/bash") "/root/guix-rpi/update-mirrors.sh") sorry
<vivien>You may need to import (gnu packages bash)
<nckx>Guest69: It does.
<Guest69>vivien: Thanks!  Why exactly is it required to give full path?
<vivien>There are 2 absolute file names in this examples: /gnu/store/xxx-bash-xxx/bin/bash and /root/guix-rpi/update-mirrors.sh
<vivien>The former is required so that your job may run whatever root has in its profile
<vivien>And the latter is because the script is executed from a cron job, whose current working directory we don’t know
<Guest69>system* is from Guile.  How do you know about that command?  Is it something that just requires working with it for some time?
<vivien>Usually you want to have the invoke function from (guix build utils), but in a cron job, for some reason, you can’t have it and have to use system*
<vivien>Oh they left
<nckx>Server shenanigans are afoot.
<nckx>civodul: Not Matrix. No more Matrix.
<nckx>Some services are being restarted, including the Web chat.
<vivien>Some would say liberachat is not more reliable than matrix
<vivien>(I would not because I’ve never had a chance to have or use a matrix account anyway)
<nckx>Matrix is famously unable to survive a reboot of the server you're using too 😉
<nckx>I think the 2604:bf00:561:2000:: wave might have been chat.sr.ht?
<Guest69>vivien: I read it through the logs.
<Guest69>webchat required reboot
<tylerm>Hello, im trying to figure out how to get PAM to require fprintd for fingerprint authentication, but i dont know how to to that in scheme. I was wondering if anyone here would know.
<vivien>If you want to use the more convenient functions from (guix build utils), you can always do (system* #$(program-file "my-program" (with-imported-modules (source-module-closure '((guix build utils))) #~(begin (use-modules (guix build utils) ...))))
<civodul>nckx: heh, i thought there were some unofficial Matrix bridges that people were using or something
<vivien>So that your program has access to the (guix build utils) module, but you make it a script and invoke that script as your cron job
<nckx>There *are* other protocols, as XMPP users will helpfully explain for half an hour if asked ☺ /j
<civodul>heheh
<formbi>hi
<formbi>I'm trying to package ch341eeprom tool, I currently have this: http://ix.io/4Fo9
<formbi>but it throws this error: «In execvp of gcc: No such file or directory»
<formbi>why doesn't it find GCC, what am I doing wrong?
<xelxebar>formbi: Build tools aren't automatically put on PATH. You'll need to call them directly from their path.
<formbi>ah, it needed «(lambda* (#:key inputs #:allow-other-keys)»
<xelxebar>Oh, wait, yeah. Lol.
<tylerm>formbi: do you setup your guix path environment in your shell profile?
<formbi>what do you mean?
<tylerm>first off where did you define your guix home envirnoment?
<tylerm>for me thats $HOME/.guix-home
<tylerm>in that folder there are two scripts, you want to source setup-environment, and call on-first-login in that folder
<tylerm>HOME_ENVIRONMENT=$HOME/.guix-home
<tylerm>. $HOME_ENVIRONMENT/setup-environment
<tylerm>$HOME_ENVIRONMENT/on-first-login
<formbi>I'm not setting up the home environment, I'm trying to write a package definition
<civodul>news and proposal re cached Git checkout growth: https://issues.guix.gnu.org/65720
<Guest69> https://paste.debian.net/1290948/ returns https://paste.debian.net/1290949/ but the path is correct
<Guest69>Okay, with a lambda function it reconfigured the system