IRC channel logs

2026-03-22.log

back to list of logs

<abbe__>hi
<ieure>Hello.
<abbe__>how to install a `computed-file' in system profile ?
<abbe__>apparently, i'm trying to test this https://codeberg.org/guix/guix/issues/5742#issuecomment-9990018 through a computed-file workaround
<abbe__>it's been a yak shaving last few hours :)
<ieure>abbe__, What are you trying to test, exactly?
<abbe__>geoclue doesn't work
<ieure>Correct, it doesn't. I filed the issue you linked.
<ieure>What are you doing with a computed-file?
<abbe__>to install the dbus permission stubs
<abbe__>it doesn't work because of missing dbus policy which dariqq has listed
<ieure>abbe__, It also doesn't work for other reasons, even with that patch.
<abbe__>that's geoclue's dbus service doesn't get activated
<abbe__>my computed-file hack https://www.irccloud.com/pastebin/cr2jmCPE/
<abbe__>breaks with adding to packages/profile-service-type https://www.irccloud.com/pastebin/nnV3LeRn/
<abbe__>so I wonder how to apply this without creating a fake package
<ieure>abbe__, You would need to use special-files-service-type to put those files into the profile.
<abbe__>oh
<ieure>One entry per file, with a file-like as its value.
<ieure>Which could be computed-file or local-file or whatever.
<abbe__>okay
<abbe__>thank you, let me try that
<ieure>abbe__, Are you hoping to figure out why it still doesn't work with that stuff in place?
<ieure>I tried the provided patch as-is, geoclue will activate, but it never returns a response with the location.
<ieure>I didn't see anything much about debugging the service itself to figure out why it wasn't reponding.
<abbe__>that I think because geoclue's location service has/had issues
<abbe__>the APIs it uses
<ieure>Okay. Well, I sure would like to have geoclue working again, so I hope you figure it out and send a patch.
<ieure>abbe__, It worked for me on Debian at the same time it was failing on Guix, so I'm not sure that's what was going on. Or everything that was going on.
<abbe__>#<&compound-exception components: (#<&external-error> #<&origin origin: "symlink"> #<&message message: "~A: ~S"> #<&irritants irritants: ("Read-only file system" "/run/current-system/profile/share/dbus-1/system.d/org.freedesktop.GeoClue2.conf.new")> #<&exception-with-kind-and-args kind: system-error args: ("symlink" "~A: ~S" ("Read-only file system" "/run/current-system/profile/share/dbus-1/system.d/org.freedesktop.GeoClue2.conf.new")
<abbe__>(30))>)>
<abbe__>genius
<abbe__>i guess have to add a fake dbus-service
<abbe__>ieure: that patch worked, at least no more dbus rejections :)
<abbe__> https://www.irccloud.com/pastebin/7AXFaw6W/
<donfelohke>guix package manager on openwrt? Setup
<abbe__>actually nevermind api works
<abbe__> https://www.irccloud.com/pastebin/PiEhxWsu/
<abbe__>the configuration is broken, it could use some more methods, and it just supports wifi :/
<abbe__> https://gitlab.freedesktop.org/geoclue/geoclue/-/blob/master/src/gclue-config.c#L673 supports geoip sources too
<abbe__>i'll look at that later. wasted too many hours on this yakshave
<ieure>donfelohke, Sorry, what
<donfelohke>ieure: guix package manager on openwrt?
<ieure>donfelohke, It's not supported and is IMO not a good idea.
<donfelohke>ieure: good idea for me
<ieure>donfelohke, I wish you all the best making it work.
<yarl>Hey ieure, could you take another look at https://codeberg.org/guix/guix/pulls/7148? I made the changes. The bot complains but I think it's a problem on cuirass and I don't know how to request a new evaluation.
<yarl>please*
<ieure>yarl, I agree it's a problem with the bot. I thought I left a comment on that earlier today, I guess I forgot to hit submit or something.
<ieure>yarl, You'll get a new evaluation if you rebase or push new commits. The comment I was going to leave is that you have long lines which need fixing, they have to be <80 chars. Push that change and you'll get a new eval, and I'll push your PR.
<yarl>ieure: ok, let me see.
<yarl>ieure: out of curiosity, how did you see that the lines are too wide?
<yarl>I mean, do you have an emacs configuration for this?
<yarl>Did you ran guix style? something else?
<yarl>ieure: Just pushed.
<yarl>Hmm, the evaluation failed again. I'll try to rebase.
<ente>d
<ente>oops
<yarl>ieure: evaluation success!
<ieure>yarl, I just noticed they seemed long when I was reviewing, when I checked out the PR to build it locally, I opened it in Emacs. There are packages that'll highlight long lines, I don't use one, I just put point on one I suspected was long and saw it was.
<ieure>yarl, Pushed.
<jakef>lilyp: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=80648
<jakef>seems like an emacs bug
<yarl>Thank you ieure.
<PotentialUser-91>I can't figure out how to define a working `shepherd-timer`. here's an example https://ctxt.io/2/AAD4K0slFA . it doesn't make the directory, it doesn't log anything, i just don't see any way to debug it. I do guix home reconfigure or guix home container on each change and it's becoming very taxing.
<PotentialUser-91>gosh it's so opaque
<PotentialUser-91>grepping for "test-timer-test" in .guix-home gives no results, apart from configuration.scm.
<czan>Does that service get listed in "herd status"? Is it listed as needing replacement in "herd status timer-test"? (I can't remember if timers get replaced like other services.)
<PotentialUser-91>yes it is. I'm triggering it manually with herd trigger timer-test
<PotentialUser-91>it just prints "Triggering timer." and nothing else happens
<czan>It is listed as needing replacement? Does "herd restart timer-test" do anything?
<graywolf>well, the code does not do anything though
<graywolf>?
<czan>"mkdir" is something, isn't it?
<graywolf>like, why should (list "mkdir" "/home/hello/test-timer-test") create a directory?
<graywolf>that is basically what you have in there
<graywolf>I think
<czan>The docs for shepherd-timer say: "Return a Shepherd service with the given PROVISION periodically running COMMAND, a list-valued gexp, according to SCHEDULE"
<czan>Seems right to me.
<graywolf>right, and command should be scheme
<graywolf>gimme a sec, will dig my btrfs check timers
<czan>Then I would expect that to give an error that it can't apply a string.
<czan>The example command in the manual is like this, though: #~(#$(file-append findutils "/bin/updatedb"))
<graywolf>hm, right, you are right and I am wrong
<graywolf>bad memory, bad
<graywolf>hm, turns out I am not actually using shepherd-timer anywhere, I always use shepherd-service with make-timer-constructor
<czan>Yeah, me too. And I'm pretty sure I have to "herd restart X" after each change to have the job actually change.
<graywolf>hm, yeah, that sounds right
<PotentialUser-91>ok, I had to herd restart the timer after having changed it from a previous, faulty one. I also learned that we can herd status <timer> to look into it.
<czan>Great! :)
<PotentialUser-91>🫶
<chris0ax>so i was looking into how to get Librewolf extensions packaged in Guix cause it looks really convenient and i realized the way its done in icecat is with a patch to respect an environment variable. Has anyone tried doing this for any of the firefox-based browsers besides icecat? This probably requires a similar patch to be put to make it work
<Rutherther>chris0ax: which one specifically? I believe all are using patches to support that dir
<chris0ax>Rutherther: i was trying to do for Librewolf personally. My bad! i was grepping MOZILLA_SYSTEM_DIR and only saw icecat, seems like the other browsers use ICECAT_SYSTEM_DIR
<Rutherther>yes
<Rutherther>but you shouldn't need to worry about the env var itself, as long as you output the extension to lib/icecat and install it to same profile as librewolf, it will work without any manual env var sets
<chris0ax>so i can package an extension for icecat and it should work librewolf as well, thats good to know!
<yarl>o/
<snamellit>Very very frequently when doing a `guix home reconfigure` hundreds (thousands?) of texlive packages are redownloaded. Is there so much churn there or is there some misconfiguration at my side?
<lilyp>we don't update texlive itself all that often, but it could well be a dependency
<Rutherther>snamellit: are you doing gcs? currently there are grafts so every time you gc you will lose them and they will need to be redownloaded
<lilyp>now obviously, anything that rebuilds texlive should probably be considered a world rebuild, but…
<lilyp>Rutherther: I had this even before the big graft
<snamellit>I do gcs from time to time. Usually when I am "done" with my updates and then pull one more time.
<snamellit>It does not rebuild on my machine, it downloads from a local substitute server from a machine with essentially the same config
<Rutherther>yes, it is normal with grafts you need to redownload everything that is grafted, because only grafted variants are kept, but you need the ungrafted ones
<Rutherther>s/kept/gc rooted
<snamellit>I'll look if I graft somewhere. I know I do with my nvidia machines but this is a boring intel i5.
<Rutherther>you don't graft yourself, glibc is grafted
<Rutherther>which means essentially every package is
<snamellit>so much to learn...
<a4censord>heya
<a4censord>i'm trying to package my blog, which is build with a ssg, but have been running into "permission denied (os error 13)" problems
<a4censord>The blog is with a single command that takes an --output-dir argument, i assume that i'm setting this incorrectly
<a4censord>I dont know what to set it as though
<a4censord>I've but the full source into https://paste.debian.net/hidden/816fad5c
<a4censord>hmm, maybe its related to `error (ignored): program `newgidmap' failed with exit code 1` during build setup?
<a4censord>but thats not happening when building other stuff (e.g. hello)
<lilyp>did we change code that goes into `guix system vm`? I'm having trouble spinning up my usual test vms
<kestrelwx>Could it have to do with the record change?
<a4censord>i made it work, but i'm still not sure what the problem was
<apteryx>ghc is "fun" to build
<ente>funfunfunfunfunfun
<lilyp>speaking of which, do we have ghc 9.6 or newer yet?
<lilyp>we really ought to update all our haskell stuff
<tschilp>Hi guix! I'm just testing `guix system image [...]` and start the created images with qemu. This is pretty cool. Right now I'm wondering, where the system config I used for the image creation ends up within the VM. Typically I get information about it by running `guix system describe`. If I do so within the Guix-VM there is no information about channels in use and no information about the configuration file. Where is this file (apart
<tschilp>from my the file on the host, that I used to create the image)?
<tschilp>s|my the file on the|the file on my|'
<Rutherther>tschilp: this is provided by the provenance-service-type. You need to supply --save-provenance to save it. Otherwise it won't be.
<tschilp>Rutherther: great, thanks!
<avigatori>o/
<redacted>If a PR of mine doesn't get reviewed after a while, should I follow up with someone? It's fine if it doesn't get merged, I just want to make sure I did my part.
<ieure>redacted, Did it get assigned to a team?
<ieure>redacted, There are more PRs to review than people to review them, so review is slow and uneven.
<redacted>Didn't get assigned to a team.
<redacted>How are reviewers chosen? I'd like to help review PRs if I can.
<ieure>redacted, CODEOWNERS is what auto-assigns, it's based on stuff in etc/teams.scm. Anyone is welcome to review any PR, whether you're a committer or not.
<Rutherther>redacted: feel free to submit a PR adding yourself to a team, pinging the team to review it
<redacted>Oh, I see. Thanks!
<lilyp>redacted: out of interest, which PR are you talking about?