<fries>hi guix chat. so, lets say in this hypothetical situation, that a program has multiple outputs with different binaries that use the same install command. but, the first output has libraries that the other outputs depend on. do you know how you can make that binary see those libraries inside another output. kind of like depending on a library from a main output.
<ulfvonbe`>squeaktoy: haven't heard of desktop-packages, but there is %desktop-services
<ulfvonbe`>I guess you would add it to your packages field?
<ulfvonbe`>I don't know if that accepts general file-like objects or only packages
<ulfvonbe`>you might need to define a little package that puts it in (string-append #$output "/bin/startx") or something like that, not sure
<ulfvonbe`>or you could just extend special-files-service-type to put it in a well-known place
<ulfvonbe`>hmm, looking at the documentation for that field it does seem like it's just for packages
<squeaktoy>I'd like to know how to make use of this xorg-start-command thing
<squeaktoy>sx does work but my keyboard and mouse freeze up
<ulfvonbe`>Well, it's a file-like object that, when lowered, becomes a program in the store. You probably want it in your PATH, and the way to do that globally is to add it to thy system profile. But the system profile wants packages. So you have two options: either make a dummy package whose output contains it, or put it somewhere fixed in the filesystem like /usr/bin/startx with special-files-service-type
<ulfvonbe`>personally I think the former approach is better, though still ugly
<juliana[m]>hi, when running guix shell -CF guile I get the error message "ldconfig: /lib/libguile-3.0.so.1.6.0-gdb.scm is not an ELF file - it has the wrong magic bytes at the start." Is this a known bug? a cursory glance at https://issues.guix.gnu.org suggests not
<apteryx>stumbled on that too in the past; it's silly
<apteryx>ACTION can't reliably find the oldest commit 'guix-timemachine' ought to be able to travel to
<PotentialUser-27>hi, guix pull doesnt seem to want to update my installed applications like web browsers, email clients etc. Any ideas on what the issue might be? It doesnt give me any errors it just says nothing to be done
<PotentialUser-27>oh its just guix package -u? Im sorry i havent used guix in a while, thought it was pull but i guess that just refreshes the sources or whatever. And that message only showed once for me
<pret7>yep! I think it also updates the host guix itself.
<pret7>you'll have to update whatever profiles you have separately with guix package -p <PROFILE> -u
<pret7>I wonder if there's a way to update all profiles together
<atuin>I have made a custom channel for my packages which i keep in git, i have also a repo for the home / system configuration. The custom channel defines some functions to get patches for the packages i need. I have a home config that uses a packages from that channel that i want to add into the home configuration. What i want to achieve is basically to be able to test the config using the channel locally (passing -L) to guix home
<atuin>but seems that when loading the module that defines the package (a custom version of stow) i get a relative path for the module, so canonicalize-path fails telling me it does not exist
<atuin>seems the info for the module file is lost (current-filename) returns #f as well
<geri>i only know that (current-filename) returns #f when used in repl
<geri>worst case just replace your channels file temporarily with your channel and test, i think it shouldn't break unless you do home/system reconfigure
<nckx>When I use (current-filename), it turns every subsequent error into ‘ice-9/eval.scm:191:35: In procedure scm_to_utf8_stringn: Wrong type argument in position 1 (expecting string): #f’. Merely using it. Hard-coding its return value makes everything work again. It's wonky.
<geri>for my use-case canonicalize-path always betrays me as well
<nckx>So if (canonicalize-path) fails, doesn't that imply that (module-filename) is not relative to (getcwd)?
<bovid-19>I am running SBCL (Common Lisp) though Sly on Guix System. I assumed the correct way to load a system was `(require :system-name)'. This seemed to work so far, but now I tried to load `mito' without success (--> `Don't know how to REQUIRE MITO.'). `sbcl-mito' is installed. I also tried inside `guix shell sbcl sbcl-mito'.
<nckx>For reference, I downloaded a Gentoo VM live ISO, built QEMU to run, and tried running guix-install.sh on that VM, using ‘sendkey’ to send half of the keystroke because key handling is broken somehow. After all that trouble, I ended up with a (AFAICT) perfectly working /usr/local/bin/guix in $PATH.
<nckx>So it's rather frustrating when the response is ‘IDC, it's broken, fix it’.
<itd>There is no built-in authentication between cuirass and remote-workers. Anyone able to connect to cuirass's backend-port can pretend to be a remote-worker. Correct?
<itd>Also, I'm confused by cuirass-remote-worker-configuration's public-/private-key. It's not clear to me how to generate them. (But empty files seem to work as well.)
<itd>Looking at the Guix/cuirass docs, I failed to understand that without a reachable log-port the remote-workers are unable to send build logs to cuirass (despite being able to communicate via backend-port).
<itd>It would be nice not to see "Could not find machine information using Zabbix." on the remote-worker machine page, if the %zabbix-url parameter is #f (the default).
<apteryx>if someone is interested helping me to narrow down the oldest commit at which 'guix time-machine' can travel, I'm using this simple test: guix time-machine --commit=4c91332cced67bd7b9034035fb2b02c5728509a7 -- shell hello -- hello
<apteryx>currently the oldest commit which succeeded for me was 4c91332cced67bd7b9034035fb2b02c5728509a7, from Nov 13 21:43:45 2021
<nckx>squeaktoy: I got called away, but pushed two bugfixes (https://ttm.sh/Bfo.png made it quite obvious what was happening). Could you download the script again, remove any old Guix left-overs, and try again?
<ulfvonbelow>is there a hard one-package-per-commit rule, or could multiple packages be added in one if they're highly interrelated (for example, package A depends on package B, which inherits from A)
<nckx>The body is extremely flexible, in its own way, and I'd just write ‘gnu: Add A and B.’ as the title.
<takev[m]>Is there a recommended workflow for testing out new services? I have a system definition that that I have been running as a container, but it feels like there is probably a better way through the repl or something.
<squeaktoy>nckx: which file does the installer modify to customize the Bash
<squeaktoy>also how does guix modify /gnu/store from a user account?
<Kabouik_>https://youtu.be/9HgFUDR66pw sorry about it being a pic and not a pastebin, but the whole point of my guix pull is to upgrade foot to 1.15.1 which will support text selection by touch. Anyway, maybe my mistake is to run df -ih, h probably messes up when combined with i.
<zamfofex>Usually I’m good at working on things that matter to me more so than on things that don’t matter to me. 😄 My general interests (related to Guix) are (i) the Hurd, (ii) WebAssembly, and (iii) helping bring npm packages to Guix.
<zamfofex>At some point, I also investigated bringing C# and .NET packages to Guix, but that didn’t go very far.
<nckx>Kabouik_: I assume builds happen in /tmp on this system. Is /tmp a separate mount?
<nckx>Because of the truncated backtrace, there's not much more to go on.