IRC channel logs

2024-12-03.log

back to list of logs

<ddaglenn>Sorry, where would I go to ask these types of questions?
<Kolev>ddaglenn, PM me.
<vagrantc>ACTION and surely many others wonder about this awkward policy
<Kolev>vagrantc, nah, you gotta draw the line somewhere. I'm in favor of the policy.
<civodul>nutcase: no, that’s fine
<vagrantc>Kolev: i am in favor of the policy to be strict about non-free software, but the the "don't speak of it" part of it is astoundingly awkward. though i get why it is that way and do not see a definitely better way without lots of other problems.
<civodul>there is no “don’t speak about it” policy :-)
<civodul>ddaglenn: you can check #nonguix to get more info about the nonguix channel
<civodul>(which provides non-free software)
<vagrantc>civodul: so it is fine to post links or describe exactly where to get non-free software? :/
<x8dcc>Hello, how do I allow "non-forward updates" (temporarily)?
<vagrantc>ACTION never really understood where the line is drawn
<unmush>x8dcc: --allow-downgrades
<civodul>vagrantc: sure, it’s just that past a point, people should go to #nonguix or similar, because that’s a separate project
<unmush>(the name is a little misleading; it also includes allowing "sidegrades")
<x8dcc>unmush: Perfect, thanks!
<ddaglenn>Ahhhh, I see that now at the top.  I guess my question more broadly is, how would I handle installing "free software" that may not be in the guix repository but still use the guix software system?
<newguixuser>Hi, I am brand new to the Guix system. It was working fine for the first few days, but now, whenever I try to do anything like upgrading packages, I get a 504 error and a message saying "error: corrupt input while restoring archive from socket" is this a common issue? If so, how can it be fixed?
<vagrantc>yes, that is the clever way to rethink it :)
<unmush>ddaglenn: channels, see chapter 6 of the guix manual
<civodul>ddaglenn: you can pull from additional “channels”; see https://guix.gnu.org/manual/devel/en/html_node/Channels.html
<Kolev>civodul, vagrantc - I didn't know I was allowed to say nonguix here. Thanks for the clarification.
<vagrantc>there are also a few ways to do one-off packages, though channels are probably the most flexible
<ddaglenn>Years from now, when I've become a guix expert,  I will look back at my first interaction with the guix chat and laugh at how my first question blatantly violated the spirit of the group :]
<newguixuser>Do I need to switch what channel I'm using?
<unmush>newguixuser: sounds like your substitute servers are having issues, if you're okay building from source instead you could pass --fallback
<vagrantc>ddaglenn: it is a reasonable question :)
<Kolev>ddaglenn, you're welcome here! Glad to have you.
<newguixuser>unmush: thanks, I'll give it a try!
<newguixuser>I just now tried that, and I got the same error.
<newguixuser>I guess I need to change my substitute servers
<newguixuser>Is there a link that I can use to change my substitute servers?
<ieure>newguixuser, Substitute servers are part of your operating-system configuration. You need to edit the configuration and `sudo guix system reconfigure'. Do you know which server is giving you 504s?
<newguixuser>ieure: ci.guix.gnu.org
<vagrantc>you can also pass --substitute-urls="server1 server2 server3"
<vagrantc>if it is just a temporary thing
<ieure>newguixuser, https://guix.gnu.org/manual/devel/en/html_node/Using-the-Configuration-System.html has some examples of configuring substitute servers. %default-substitute-urls in (guix store) has the default list.
<ieure>newguixuser, Your specific issue isn't common, as far as I know, but Guix servers being flaky, in my experience, is.
<apteryx>so hm, I have a (guix build parsers rfc) module that I added to the #:imported-modules of python-build-system; I also have a #:use-module for it in (guix build python-build-system.scm), but for reasons that escape me, its public API is not available during a build (sanity-check phase fails with: Unbound variable: parse-rfc2822-file).
<sneek>apteryx, you have 1 message!
<sneek>apteryx, rekado says: Here's a few commits (v2) that let me build up to python-jupyterlab-server: https://issues.guix.gnu.org/74652
<apteryx>w.r.t the above unbound variable, is there some obvious I'm missing?
<apteryx>rekado: hi! great that you could fix it. I have lined up a few package upgrade/fixes on python-team trying to build my way to python-biopython (it now builds on my local branch).
<apteryx>infusing some more value in the Python build systems while I'm at it, should hit guix-patches soonish, after I've rebuilt lots of stuff.
<apteryx>re my silly unbound error; I had defined parse-rfc-2822-file but exported parse-rfc2822-file.
<apteryx>another bite from guile's gleefully exporting unbound symbols I guess ^^
<vagrantc>ACTION tries "guix build --dependents" for the first time
<wakyct>hi, this might not be a guix issue, but is anyone using a bluetooth headset with mic successfully? I'm at a loss because everything connects and my headset is showing headset and handsfree in the BT profile, and I even see the capture sink in pipewire (using Helvum). But I can't get the mic to work in any program.
<wakyct>the output goes to the headset OK though
<wakyct>is there some trick or plugin I'm missing?
<wdkrnls>Dear Guix, can I run guix system reconfigure from time-machine?
<raghavgururajan>In the past weeks, IceCat has been crashing out of the blue with the error "Exiting due to channel error." on Guix System.
<Rutherther>sneek: later tell wdkrnls: you can run everything from guix using time machine, including system reconfigure
<sneek>Okay.
<wakyct>wow it works. It's nice when that happens.
<mccd>Is there any kind of sandbox-capability built into guix? Something akin to enter repository and guix shell, but don't allow the shell to touch anything outside the repo?
<Rutherther>mccd guix shell has container flag that will do that. You can mount other paths if you needed as well
<mccd>Rutherther awesome
<mccd>ty
<fanquake>Hi. Trying to guix pull this morning, and it looks like the sha256sum for guile-bootstrap-hash might have been incorrectly changed in 4d9c5984fee481d74c2f504094b4797bbb4104d4 ?
<fanquake>ACTION https://gist.github.com/fanquake/fc552227a2443e79c707350dc6bdf18f
<fanquake>cc janneke
<janneke>fanquake: _oops_
<janneke>how did that get through?
<janneke>ACTION prepares a fixup
<janneke>fanquake: thanks, fix pushed
<fanquake>janneke: great, thanks
<janneke>wow, that was embarrassing
<civodul>Hello Guix!
<janneke>o/
<efraim>o/
<Lumie>Hello
<civodul>so! ‘guix publish’ keeps respawning on ci.guix, which makes things unreliable
<civodul> https://issues.guix.gnu.org/74632
<dariqq>hi, tried to reconfigure my system and it is failing to build the ath9k firmware with -Wimplicit-function-declaration erros. Also why is this using gcc-14?
<efraim>we've switched the cross-gcc to gcc-14
<csantosb>Hi Guix ! I'm trying to debug why guix image is not built in sr.ht, https://builds.sr.ht/~sircmpwn/job/1379955
<csantosb>I'm unable to reproduce in my laptop ...
<csantosb>Running `guix system image --verbosity=0 --image-type=qcow2 --image-size=16GB --save-provenance system.scm` gives
<csantosb>This (https://paste.sr.ht/~csantosb/3a17460fecbe078475a1bd83087d10eb40a40abd) error
<apteryx>I'm using (define-deprecated/public-alias %python-build-system-modules %default-python-imported-modules), but users of the deprecated %python-build-system-modules throw syntax-transform unexpected type, e.g.: In procedure filter: Wrong type argument in position 2: ((guix build cargo-build-system) (guix build cargo-utils) . #<syntax-transformer %pyproject-build-system-modules>). Ideas?
<csantosb>I guess this is related to hardware where they build images under sr.ht, right ?
<apteryx>anyone else using librewolf on gnome?
<apteryx>when I click on a link, I get some notification 'page is ready' or similar; I'd rather the browsers open directly instead
<apteryx>csantosb: are you trying to reproduce from the same exact guix commit that sr.ht is using?
<csantosb>apteryx: yes, same, guix 58a1342.
<Deltafire>apteryx: i get that with Gnome Files, if the directory is already open somewhere. It's quite annoying, it should bring the window to the front instead
<csantosb>I understand the problem comes from `guix build ath9k-htc-ar9271-firmware`
<dariqq>efraim: I see, this seems to break the %base-firmware packages
<apteryx>Deltafire: ah! some it's a global behavior in GNOME. Probably configurable, let's see
<Deltafire>links seem to open okay in IceCat, not really used Librewolf
<apteryx>I've read if one single application has the 'stay on top property', even if it's minimized, it would cause that behavior
<csantosb>For the record, in case anyone manages to reproduce, former system image is run in a virtual environment
<csantosb>With virtualization features: virtualization: AMD-V, Hypervisor vendor: KVM, Virtualization type: full
<apteryx>csantosb: interesting
<apteryx>Deltafire: I've asked in #gnome.
<janneke>ACTION works around SIZE_MAX problem for gcc-cross-boot0
<civodul>anyone looking into the guix.gnu.org certificates?
<civodul>ACTION tries
<janneke>ACTION pushes workaround to core-packages-team
<apteryx>that sounds like the next nice thing to investigate and fix for good on the CI... it's reoccuring at an annoying rate
<civodul>ok, done
<apteryx>what was needed?
<apteryx>restart nginx?
<civodul>yes
<z572>It appears that the elfutils cross-compilation has now failed
<civodul>it reflects poorly on the project
<civodul>Carlo submitted a patch for DNS-based challenges on guix-sysadmin
<civodul>but i’m not super qualified to apply/test it
<apteryx>I think none of us are, which explains the current situation, haha :-)
<apteryx>IIRC, it was working around the issue rather than really addressing it. Well, if it works, it's an improvement.
<apteryx>the real underlying issue (according to them, IIRC) being there are two competing machines trying to get the same cert, failing the http challenge or something something
<apteryx>like, bordeaux and berlin
<mjw>z572, I maintain elfutils upstream, but didn't know it could be cross compiled (but why not, might be useful)
<mjw>what doesn't work
<civodul>apteryx: yes, it addresses this particular issue
<civodul>and would allow us to have redundancy for the web site
<civodul>like guix.gnu.org pointing to both machines
<apteryx>is that not already the case?
<apteryx>the website keeps running when berlin goes down
<apteryx>changing topic, is it worthwhile throwing srfi-35 conditions instead of using 'error' in the build code? Does it get shown any nicer in the ouptut?
<apteryx>here what it currently looks like using basic 'error' + a formatted message: https://paste.debian.net/1338016/
<apteryx>ACTION tries with srfi-35 conditions
<z572>mjw: error: ‘__pread_alias’ writing 58 or more bytes into a region of size 10 overflows the destination
<z572>I think it's because our elfutils are a little old
<mjw>z572, which version, which src/line? It does sound slightly familiar
<efraim>it looks like
<mjw>If your version is more than 2 years old, it might be https://sourceware.org/cgit/elfutils/commit/?id=0873ae782d14e672e8344775e76b7fca0a8b41bf
<efraim>it looks like 'guix build syncthing --system=armhf-linux --tune=armv6' works
<z572> https://0x0.st/X7Tu.187.log
<z572>Thanks. I'll try
<mjw>ah, 0.187, yeah, that is old, latest is 0.192.
<mjw>If you want to just fix that error, the above commit probably will
<mjw>z572, btw. It is missing bzip2 and zstd support, you probably want the devel packages as inputs.
<mjw>elfutils Version 0.189 "Don't deflate!" will support zstd also for libelf
<mjw>But really, you want 0.192 :)
<janneke>ACTION reshuffels commits on core-packages-team and fies typos
<janneke>*fixes
<janneke>ACTION uses a new, nice and "fast" keyboard -- but typos... :(
<ray1729>janneke: fie typo, fie!
<apteryx>build code srfi-34 + 35 raise exception output: https://paste.debian.net/1338022/ It's a bit nicer than a misc-error thrown with plain 'error', as it gives more context.
<apteryx>(and the visible type has a meaning)
<apteryx>pretty verbose on the coding side though, which I'm not too big of a fan (yet?)
<janneke>ray1729: hehe :)
<apteryx>ACTION now to fix all the problems found with that new validate-version python phase ^^'
<z572> https://issues.guix.gnu.org/74669
<z572>mjw: i think this needs to be updated in core-packages-team
<mjw>z572, another advantage of upgrading the version, we stopped using ChangeLog files but put the ChangeLog entry in the commit message now :)
<janneke>ACTION sends long overdue mail for teams/hurd team
<civodul>janneke: yay for x86_64-pc-gnu!!
<civodul>i’ve seen reports about the failing ath9k-* package builds (cross-compilation)
<civodul>i suppose that’s related?
<civodul>also: uncaught throw to %exception: (#<&inferior-exception arguments: (match-error "match" "no matching pattern" "powerpc64le-linux") inferior: #<inferior pipe (0 1 1) 7f6bd777ad80> stack: ((#f ("ice-9/boot-9.scm" 1779 13)) (raise-exception ("ice-9/boot-9.scm" 1682 16)) (raise-exception ("ice-9/boot-9.scm" 1684 16)) (arguments ("gnu/packages/hurd.scm" 633 28))
<civodul>from https://ci.guix.gnu.org/jobset/master
<janneke>civodul: yeah -- and i managed again to do a brown-paper-bag commit -- accidental touch of aarch64 %bootstrap-guile-hash :(, oh well...
<janneke>civodul: ah, yeah that could very well be the switch to cross gcc-14
<janneke>err, no mathing pattern what?
<janneke>ACTION looks
<civodul>i’m on it
<janneke>ah great, thanks
<janneke>i guess we need some default for non-hurd arches?
<civodul>yeah
<civodul>for the firmware, it’s “implicit declarations” error
<civodul>not sure why we’d get them with gcc 14
<janneke>civodul: https://gcc.gnu.org/gcc-14/porting_to.html
<janneke>-Wno-implicit-function-declaration
<janneke>they even suggest creating a wrapper for it -- in the end that's what i did for gcc-final
<civodul>janneke: i tried -DCMAKE_C_STANDARD=99, which translates to -std=gnu99
<civodul>hopeing that “implicit declaration” wouldn’t be an error
<civodul>to no avail
<janneke>yeah, i believe you really need something like "CFLAGS=-g -O -Wno-implicit-function-declaration", (or even -fpermissive, but yeah)
<civodul>yes
<janneke>but thanks for trying nicer solutions, i've tried a number of things too
<janneke>patching configure might be another option -- assuming that configure fails
<civodul>pushed hot fixes!
<civodul>lemme know if anything’s wrong
<civodul>so we can have more hot fixes :-)
<janneke>civodul: <3
<civodul>looking forward to a qa.guix kind of tool that would truly guide us pre-merge
<janneke>that would be super nice
<fanquake>Seem to be running into a different issue running guix pull on aarch64
<fanquake> https://gist.github.com/fanquake/f02b8806ef3515cf9485fb45a6dbffff
<csantosb>civodul: please ! I think this is more than necessary
<csantosb>`guix time-machine --commit=58a134224e -- build ath9k-htc-ar9271-firmware` is broken
<csantosb>whereas `guix time-machine --commit=ef15b48379 -- build ath9k-htc-ar9271-firmware` is not
<civodul>csantosb: fixed in 6d34e751b545120fd4455cb6d718d02c9188d874
<civodul>csantosb: speaking of which, did yo look into fj.el? (i think we mentioned it the other day but i didn’t see your reply)
<civodul>uh
<civodul>lemme retry :-)
<civodul>csantosb: speaking of which, did yo look into fj.el? (i think we mentioned it the other day but i didn’t see your reply)
<csantosb>civodul: not by now, as I'm not planning to use codeberg more than this at this point
<csantosb>I'm an addict to sr.ht, I'm afraid; if someone knows of a good client, please !
<fanquake>cc janneke
<csantosb>We have sr.ht/~akagi/srht.el as a frontend to hut, but this is rather basic
<z572>security-updates: Last repository check: 27 hours ago.
<csantosb>civodul: I confirm that ath9k-htc-ar9271-firmware is fixed, thanks !
<z572>Why does cuirass always forget to check the repository?
<csantosb>And sr.ht guix image is back to life again 🥳: builds.sr.ht/~csantosb/job/1380288
<janneke>fanquake: thanks, fix pushed
<civodul>z572: it does not forget: this jobset is set up to be checked once per day
<civodul>though you can trigger it with a magic POST request
<civodul>csantosb: curious about your usage and feedback about sr.ht
<civodul>i looked at tickets at sr.ht and was unimpressed
<csantosb>civodul: I'll try to write something about
<noe>is glibc available to use in tests or does that break other platforms than x86?
<janneke>fanquake: and sorry about all the fallout
<civodul>noe: it is, but only in tests that require access to a full-blown install, like most in tests/pack.scm
<noe>great, just sent the fix for the AppImage tests
<civodul>coolio, thanks noe!
<noe>thank you for reviewing :)
<z572>Should we use (raise (condition (&package-unsupported-target-error (package pkg) (target target)))) instead "UNSUPPORTED_SYSTEM" ?
<civodul>ah, good question, i don’t know
<civodul>is that exception caught?
<civodul>if it’s caught from (gnu ci), then yes
<civodul>otherwise no
<fanquake>janneke: no worries. Thanks for the quick fixes
<csantosb>This is probably an old topic but, do we have any policy about releases ?
<csantosb>Releases eases users' life ... and are a good strategy to communicate
<janneke>fanquake: (y)
<civodul>csantosb: yes, there’s consensus that this should be done, but structurelessness causes each one of us to look at the others in the hope someone will step up
<civodul>we need to break out of that vicious circle :-)
<csantosb>(my use case is running tests on a guix image; would help a lot to do it on a release image instead)
<gabber`>would it make sense to tag a commit as a release around - let's say - every 10000 commits?
<gabber`>v1.3.0 was at 76971 commits, v1.4.0 was at 105808, we're now around 150k
<janneke>ACTION builds gcc-final on core-packages-team
<janneke>(reusing the fixes for hurd :)
<ray1729>gabber: is there a pre-release freeze period where stuff gets widely tested before a release is tagged?
<gabber`>ray1729: not the expert in this round but IIRC there were some "signals" (like big branches being merged when they finally built successfully) before tagging commits as releases
<csantosb>emacs's releases model seems good to me (semver looks like the path to go)
<gabber`>janneke: yesterday i failed installing GNU/Hurd due to netdde failing, i figure this has been taken care of?
<janneke>gabber`: i hope so!
<gabber`>nice! i'm gonna give the installer another try then!
<gabber`>ACTION super stoked to finally enjoy the HURD on hardware
<janneke>gabber`: no promises, but netdde was updated to latest
<janneke>what kind of problem did you encounter?
<dthompson>hope this isn't *too* off topic but spritely just launched its first supporter drive. we're big users/promoters of guix and we're also involved in projects to advance guix such as the goblins+shepherd initiative. if you like what we do, please consider a donation! https://spritely.institute/donate/
<gabber`>janneke: the installer was interrupted due to netdde failing to build
<janneke>gabber`: hmm, i haven't done anything in that area...
<gabber`>i think civodul did with commit d897f1b4
<gabber`>and ci.g.g.o reports a successful build from a couple of hours ago
<PotentialUser47>I'm currently getting this error when trying to build a guix shell https://paste.debian.net/plainh/ef00a5a2
<PotentialUser47>I assume this is just an issue with server and nothing I need to address on my end?
<ieure>PotentialUser47, Yes, substitute server fell over and died again. This happens often.
<PotentialUser47>Ok good to know
<efraim>rekado: I hope your hands didn't get tired typing ~2600 commit messages ;P
<trofi>I noticed that https://git.savannah.gnu.org/cgit/guix.git/commit/?id=8a7bd211d21f06c1234fbb82bb905d202d58f598 broke the build on systems with linkers defaulting to --as-needed. Specifically -lgcrypt is now specified before object files that use it and not after.
<trofi>sepeth: ^ . _LDFLAGS are usually for things like -L paths, and not library names themselves. i think _LIBS is a preferred way to add libraries to link list.
<rho`n>hello nutcase, did you figure out your home shepherd service issue?
<trofi>On top of that I still see `nix/local.mk: $(SQLITE3_LIBS) $(LIBGCRYPT_LIBS)` using `LIBGCRYPT_LIBS` without a corresponding AC_SUBST.
<janneke>gabber`: that would be nice, in that case we're still an effective team :)
<rekado>efraim: I actually did write a few hundreds of them.
<rekado>my hands are fine (thanks to Emacs registers), but it was rather tedious to hunt down build failures in ci.guix.gnu.org.
<efraim>ouch. that's actually why I ended up buying a split keyboard, was starting to feel some RSI pains from a standard keyboard and rapid typing
<Kimapr>i'm trying to run docker on Guix System, but `docker start` says `Error response from daemon: Unknown runtime specified runc`
<Kimapr>so i can't run any containers
<Kimapr>could the fact that i copied /var/lib/docker and /var/lib/containerd from a different distro be the problem
<nutcase>rho`n no, I didn't try again. And to be honest, I don't have an idea, how to debug the issue or what to try.
<Kimapr>Oh, yeah, that seems to be the issue. apparently every docker container remembers the runtime it was started with ?
<makx>anyone know about this guix pull problem here: https://paste.debian.net/1338079/
<Rutherther>makx: does it happen on retries as well? if so, I am not sure, but it could be corrupt store, so you can try guix gc to verify contents
<makx>i tried 3 times now
<makx>i'll try guix gc then
<makx>that didn't help
<Rutherther>what did you run?
<Rutherther>you need to run "guix gc --verify=contents" to verify contents, so you ran that?
<makx>i did not
<makx>i just ran guix gc
<makx>sorry i am not a massively experienced guix user (yet)
<rho`n>nutcase, on which OS are you running Guix?
<nutcase>rho`n guix system
<nckx> https://paste.debian.net/plainh/699e8f1f
<nckx>Upstream corruption.
<nckx>I think.
<rho`n>nutcase -- oh ok, running a guix system on a home server myself, but running user jobs through the main system config. I will configure some service in my home directory and see if I have the same issue
<lilyp>before I do something stupid: is there a way to leak files to a build container without interning them?
<Rutherther>lilyp: I am not sure what interning means. It's possible to leak files with --chroot-directory option to guix-daemon, but it shouldn't really be used as it destroys reproducibility. Is this what you meant?
<lilyp>"interning" as in "moving it to the store"
<Rutherther>okay. Then I think my answer covers it, and I don't think there is any other way, everything else has to come from the store as far as I know
<Kimapr>i'm trying to use i3lock but it doesn't accept my password..
<Rutherther>Kimapr: i3lock needs a pam service. Did you add a pam service for it?
<Kimapr>nope, i'll add one now. thanks!
<janneke>ACTION finds out about gexp->approximate-sexp, neat
<Kimapr>do i do it like in this cookbook page? https://guix.gnu.org/cookbook/en/html_node/Xorg.html
<lilyp>hmm, I guess the question is whether it's possible to specify an input that isn't a store path 🤔
<Rutherther>Kimapr: yes, but I would suggest to add (using-setuid? #f) to the configuration. Setuid shouldn't be necessary, and it's always additional security risk
<csantosb>Someone else experiencing networking issues upon `guix pull` or did I break my system ?
<ieure>csantosb, Yes, few others over the last couple days.
<csantosb>Uff, thaks.
<janneke>csantosb: could be related to guix-publish troubles https://lists.gnu.org/archive/html/bug-guix/2024-12/msg00000.html
<jlicht>ACTION is reading up on the latest and greatest contributing info pages
<homo>hi, does hugs depend on something that modern compilers don't have? gcc 4.9 fails to build
<jlicht>my pending patch series that will remove the old, insecure and probably never-should-have-been public Node.js@10 _technically_ seems to deserve a one month deprecation notice; should I send this notice to guix-devel explicitly?
<jlicht>Referring to https://issues.guix.gnu.org/74187 :)
<janneke>jlicht: that's how i read it; possibly cc your cover letter to guix-devel?
<jlicht>janneke: thanks for the sanity check: I'll be sending a v2 and CC guix-devel indeed :-)
<raghavgururajan>Hmm! IceCat crashes with "Exiting due to channel error." more frequently on Guix System. I think started at n-1 or n-2 update.
<trofi>Sent a possible fix for -lgcrypt underinking in guix-daemon as https://debbugs.gnu.org/cgi/bugreport.cgi?bug=74677
<csantosb>raghavgururajan: unrelated, but you made me wonder about the advantage of icecat-115 over librewolf-132 ?
<stochastic>When I do guix pull, a very long error pops up with meson
<stochastic>And a lot of \x0;
<csantosb>Apparently, pulling has to wait for a moment
<daviid>ll