IRC channel logs
2023-12-13.log
back to list of logs
<cmiller>If I have a channel, how do I add it with the -L flag? Is it the whole root directory or which directory do I need to add? <vagrantc>i would think that channels are a "guix pull" operation ... so you would have to pull to a guix version that had the channel configured ... but honestly guessing a bit here <podiki>cmiller: it would be the root directory of the channel; e.g. guix build -L. <some package> if you are in the root of your channel directory <podiki>and yes, not really a channel here, more just adding a directory to the load-path <podiki>(nearly?) the same as having the channel in your configuration for operations with -L I think <cmiller>Hmm, it still says unbound variable. Though it recognizes the correct path for the import. <isf>hello, I have a macbook 2,1 with GNU Boot, and trisquel installed, soon as possible I want re-install but my question is, is there a better free distro compatible with old hardware? <cmiller>Is something like "guix system build -L . system.scm" maybe wrong? <Kolev>isf, we love GNU Guix System here. <isf>Kolev: have Guix old kernel supported? <isf>they distribute GNU Linux-libre 2.7? <PotentialUser-79>Hello, I encounter a failure when I run `guix pull` after I install guix flowed by `Binary Installation` in the manual. <PotentialUser-79>Authenticating channel 'guix', commits 9edb3f6 to baeff7f (3 new commits)... <zamfofex>PotentialUser-79: Can you upload your logs somewhere? <cmiller>PotentialUser-79: Did you try to run the command again? Sometimes it breaks but running again will work. <cmiller>PotentialUser-79: Well, that probably means something is actually broken. In that case, I can't help, since I don't have the experience. But that store-connection error happens to me sometimes, too. Slow internet connection for example. <cnx>how do i make guix system reconfigure show more traces? <cnx>i'm now stuck with ``ice-9/eval.scm:159:9: In procedure scm_to_utf8_stringn: Wrong type argument in position 1 (expecting string): #f'' while hacking on a service <cnx>the only other line above is ``guix system: error: failed to load 'system.scm':'' <cnx>growing verbosity level to a million did not help <efraim>looks like you have a string-append or something similar that's recieving #f instead of a string <efraim>if port-number is a number you should probably use number->string on smartdns-server-serialze-boolean and in smartdns-serialze-port-number <efraim>I think I would also try to break it down into the smallest possible output, even if it doesn't produce an actually workable config, and to make sure some of the functions are working first. so to, say, start with only setting the IP and commenting out the rest of the config. If that produces what you're expecting in the config file then add back the next field <cnx>thanks, i'll try that, but is there any possible way to show more log? <efraim>you can try sprinkling 'pk' around the config. it'll print what the rest of the lambda would evaluate. not sure a bout making it output more log information though <xelxebar>The nix-daemon service never seems to start correctly from boot. Always have to manually restart it before using. <xelxebar>Before restart nix-env errors out with: opening pseudoterminal master: No such device <xelxebar>Oo! Just guix pulled. Congrats on the rust upgrade! <dariqq>hi, i have a local substitute server running with (advertise? #t) and a guix daemon with (discover? #t) but i think the the server is not being discovered as it only works if i add it to --substitute-urls. Any ideas how i can troubleshoot this? <dariqq>i did a guix archive --authorize with the key <dariqq>but maybe i am misunderstanding what the discover? argument should do? <attila_lendvai>i have added mine in a modify-services, guix-configuration, authorized-keys <dariqq>atilla_lendvai: i guess taht would also work. but i figured the whole point of substitute discovery is that is just works if it can find (authorized) servers and i dont have to manually specify them <attila_lendvai>dariqq, the servers are discovered automatically, but then silently ignored if they are not authorized on you client side <dariqq>if i invoke guix build xyz --substitute-urls="ip-address" it downloads the substitute from my local server but without the --substitute-urls it does not find the server. thats why i am a bit confused <attila_lendvai>oh, i see, then the discovery indeed doesn't work. maybe some avahi settings in your router/firewall? <dariqq>ah i think is something with my networking setup. (as both my substitute server and the device i am testing with are vms on the same host) and i the address of the substitute server is not reachable from outside of the host machine <attila_lendvai>when shepherd starts the services at boot, then /run/current-system/profile/bin/ is not part of PATH. is that intentional? <efraim>you either need to add it to the PATH or call packages by their full path <efraim>Of I suppose you could use the full path in /run/current-system/... <attila_lendvai>efraim, this involves sudo, so i had to add a PATH variable. (i'm trying to use guix shell as an inline simple-service to start up a daemon without writing a full-blown custom service for it) <attila_lendvai>civodul, in shepherd throw/catch unwinds the stack, and as such it makes debugging harder than necessary... i'm looking at a match-error thrown from inside service-controller, only with a useless backtrace. any recommendations how to proceed? ideally i should refactor throw/catch away, re-run my usecase, and fix the issue. <attila_lendvai>i have an idea: i'll just shadow catch with my own that uses #:unwind? #f until the refactor <civodul>attila_lendvai: match error in ‘service-controller’ sounds pretty serious; is that with current master? <civodul>i don’t have specific debugging tricks <civodul>in this case i’d check in the code all the messages that can ever be sent to the controller <attila_lendvai>civodul, it's with my heavily patched version, but i didn't (intentionally) mess with that part of the codebase <civodul>attila_lendvai: a simple trick is to ‘pk’ the output of ‘get-message’ in the controller <civodul>you can also (display-backtrace (make-stack #t) …) from the pre-unwind handler of ‘catch’ <gabber>in a $(guix time-machine -C channels.scm) context: aren't modules available form my channels automatically available in my load path in $(guix repl)? e.g. i do $(guix time-machine -C channels.scm -- repl -L local-path some-script.scm) and guix complains about some module (specified through one of the channels) being unavailable <attila_lendvai>i already have set up backtrace logging and error handling that can deal with nested errors. all i need now is a catch that does not hide the origin. <civodul>gabber: in ‘guix repl’ you first have to ,use(gnu packages) to make channel modules available <attila_lendvai>heh, it has finally dawned on me that changing #:unwind? is a change in semantics... :) i'll just log the backtrace before the unwind. <cbaines>I'm seeing files with these permissions in the store lrwx------ (this is when building on the hurd) <gabber>civodul: ,use is equivalent to (use-modules) and (define-module .. #:use-module (foo bar)), no? <civodul>gabber: ,use is equivalent to (use-modules …) indeed <civodul>as for reporting the bug maybe? or maybe it’s already filed? <efraim>oh good, guile-json isn't the optional dependency <vivien>efraim, I followed the rust-team merge from a distance, but I imagine that it represents a big chunk of work. Bravo! <efraim>I've started looking into implementing a cargo registry inside GUIX_ENVIRONMENT so it can be used as a substitute for crates.io <gabber>civodul: it is, indeed: #58841 and it seems to be addressed in #61343 <efraim>I assume I shouldn't use guile-json inside a build system <attila_lendvai>how does the guix daemon isolate the tmp directory for the build? is there a way to peek in there? this build is hung, and if i C-c it, then the directory is deleted instead of copied to /tmp/ <efraim>you can build it with -K and then it will stick around <efraim>if you just want to look inside you can use 'sudo ls' <attila_lendvai>hrm, it indeed does. last time it didn't (or i was blind). thanks! <vivien>attila_lendvai, in that situation I open htop and kill the process <civodul>cbaines: hello! i’ll reconfigure bayfront soon <fjl>civodul: Is there a some guidance about when stuff could be considered for upstreaming from Guix to Guile (for stuff that is general in nature, not Guix specific). It looks like guile-lib used to be a staging place for Guile but now it looks like it is mostly dead in that role and Guix is hoarding useful stuff. <civodul>fjl: hi! no specific guidance, but we can talk :-) <civodul>(guix records) for instance is generic; however it’s useful to be able to tweak it as we see fit <civodul>but then there are auxiliary modules that could probably move to Guile or to external projects <fjl>(guix records) is still evolving to some degree, as I look at the git history <fjl>but some of (guix combinators) or alist-let* (that is duplicated between Shepherd and Guix) are very stable <fjl>So, fold2 fold-tree fold-tree-leaves are good candidates <fjl>(ice-9 combinators) ? <fjl>OK. I'll prepare a patch <fjl>other candidates to upstream a <jonsger>civodul: uff, amazing how you improved cuirass recently. It's becoming a machine :) <fjl>are procs from doc/build <civodul>jonsger: aah, glad you like it, thank you :-) <civodul>i still leaves quite a bit to be desired <jonsger>the commit range feature when a build fails is awesome and saves a lot of "research" <jonsger>ACTION doesn't understand why he needs to build gcc-4.6.4, mes and all that stuff on current master... <fjl>civodul: for alist-let* I'm not sure where should it go <evilsetg> Would anybody here be interested in a Berlin guix meetup? <bienjensu>Is there any up to date guide or template on how to get a ruby on rails project working correctly with guix? <bienjensu>Is there an env I have to set for rails/bundle(?) for it to locate the propagated deps (as in the package def)? <attila_lendvai>civodul, why is there a separate <shepherd-service> type in guix, in (gnu services shepherd)? is this something that should be evened out, or will this stay like this? context: i need to add two fields to it (respawn-limit and respawn-delay), and i was baffled for a while until i relized that it's a thing. <attila_lendvai>mindlessly duplicating this feels wrong, especially that i need to define a default value in guix that overrides the default in shepherd <attila_lendvai>i kinda understand the reson for a separate type in guix, but the duplication is painful <tex_milan>Hello, I am using GuixOS, trying to use virtual-manager but it tells me, when I want to create a new virtual image, that it failed to find a suitable default network. Any idea what to do_ <f1refly>when using sway, I usually include /etc/sway/config.d/* on other distributions in my config file. What is the appropriate path to include when running sway on guix? <f1refly>I assume it's somewhere in the store, but I'm not sure where. $GUIX_PROFILE/etc/sway/ only has the default config file, no config.d <reedm>Realized I needed to add myself to the 'kvm' group