IRC channel logs


back to list of logs

<lechner>itd_: the full log is a bit long, but maybe this is enough. thanks for looking!
<lechner>itd_: it says no Go files
<podiki[m]>check unpack path (or something like that) and module path of go packages, sometimes you have to set that so it points to the right directory of the package
<podiki[m]>itd_: ^
<miyulz>hi guix
<miyulz>Any idea why ./pre-inst-env guile -c '(use-modules (guix utils)) (pk (%current-system))' gives "WARNING: compilation of /home/miyulz.homedir/gits/other/guix/guix/diagnostics.scm failed: Unbound variable: trivial-format-string?" ?
<kaelyn>Hi #guix! I'm at it again with my ambitious graphics stack updates! :)
<kaelyn>I've sent in to update mesa on core-updates (and fix the vulkan layer manifests), and to update the vulkan packages and add vulkan-validationlayers
<Cairn>Awesome work kaelyn
<yewscion>Hello All, anyone have advice for building all of the source derivations for a package's dependencies? I'd like to download (for instance) all of the sources I would need to have to compile ghc and its dependencies from source, but the --source option just downloads the ghc source itself.
<Cairn>yewscion: Sorry, late response. Do you mean to build them right away? Because --no-substitutes will build everything locally
<Cairn>Sorry if that's obvious. I might just be misunderstanding.
<yewscion>Cairn: No worries. I mostly want to group all of the network activity together, and then proceed to the local builds.
<yewscion>So, download all needed sources now, and build them all later without needing a connection to the net.
<yewscion>The issue I've run into is that it will download the sources for the packages I explicitly request (ghc) but not for those that are implicitly depended on and also need to be built (like, for instance libffi or ncurses)
<yewscion>A workaround would be to manually request all of the packages needed during the call (probably using a fake call to discover all needed packages) but that is cumbersome and something I'd rather avoid.
<linj>Is nix-daemon still used?
<linj>A maintainer said it was in 2018
<kaelyn>Cairn: Thank you! :)
<podiki[m]>kaelyn: great! there's a failing test on i686 for mesa that can be worked around
<lilyp>linj: technically yes, but it's a forked one
<linj>lilyp: thanks
<podiki[m]>kaelyn: (I did it for another channel, I can send this on debbugs later too)
<linj>I want to change services of grub and luks, how the workflow will be?
<linj>* how the development workflow will
<linj>This seems only suitable for packages, but not for services
<brendyn>linj, depends to what extent you want to change it
<brendyn>looks like is offline
<unwox>does load for anyone? i get 'ERR_ADDRESS_UNREACHABLE'
<brendyn>i think also
<kaelyn>podiki[m]: thanks for that link! I'll add the workaround to my patch (in the morning).
<derpey_noise[m]>Is down?
<derpey_noise[m]>Trying to install the os and hanging at updating substitutes
<xd1le>derpey_noise[m], yes seems like it
<derpey_noise[m]>Any indication on when it'll be back up? It's been about an hour
<xd1le>I'm not sure
<xd1le>EU folks should be waking up soon. Maybe a few hours my guess.
<derpey_noise[m]>I don't suppose there is an external website anywhere that mentions these sorts of things?
<derpey_noise[m]>Been scouting a bit to no avail
<xd1le>Actually I was just looking for the same as well, I'm unaware of any.
<itd_>sneek: later tell lechner: As the error message said: There are no .go files in; you probably want one of the subdirectories instead (e.g., for siv: append -siv to package name & /siv to import path; I think).
<derpey_noise[m]>Just my luck the servers go down right after I download the image lol
<xd1le>yep rip :/
<derpey_noise[m]>Is it common that they go down?
<xd1le>not from my experience but I've noticed it a few times in the last few months
<xd1le>though the last one was a planned one I believe so maybe that doesn't count
<xd1le>so I wonder the issue for it being down this time
<derpey_noise[m]>Hopefully it isn't a recurring thing
<xd1le>derpey_noise[m], btw maybe you can try --substitute-urls=""
<roptat>ah wrong window :)
<brendyn>there is a recent patch that helps guix fallback to the other server
<Guest54>Hi, it seems that guix servers are unavailable today, with the exception of bordeaux, is this a problem from the hosting provider or from guix? If it is because of your hosting provider, maybe you should change it or use bordeaux by default because the blocking of users from other countries and this kind of unstable operation is bad for the whole
<Guest54>guix project.
<provoostenator>Hi there, I'm having difficulty with "guix pull" on a fresh Ubuntu machine with --no-substitutes.
<provoostenator>It insists on installing openssl-1.1.1l, probably because git-minimal 2.36 wants that, but I'm not sure how to analyse the dependencies.
<provoostenator>I mostly tried with commit=39dcbc7fa3c02ff5c9682f25e1c29667dbfe7827 which is where OpenSSL was bumped to 1.1.1q, but also on the latest master just in case.
<ennoausberlin>provoostenator: Can't you use git-minimal 2.38?
<provoostenator>ennoausberlin: that was my thought as well, but how I do override that in "guix pull"?
<provoostenator>I couldn't find any reference to 2.36 either.
<fps>guix pull failed for me again..:
<ennoausberlin>provoostenator: Where is your openssl1.1.1q coming? I don't see it here. Last is 1.1.1l on my guix
<ennoausberlin>Do you use an extra channel?
<provoostenator>But further down that file it still refers to l
<provoostenator>No extra channel
<provoostenator>AFC for a few hours, but will try suggestions later.
***Dynom_ is now known as Guest1852
<rekado>nckhexen apteryx I can’t see anything on the serial console on; are you going to reboot the server?
<pkill9>gnome-wayland decided to stop having problems for me, weird
<pkill9>no more blank screen on wakeup and having to switch TTYs to show the login screen, no more flickering of the screen
<Lumine>Good morning #guix
<zimoun>machines in Berlin are down, right?
<rekado>can’t connect to the serial console
<rekado>but they have power
<rekado>FWIW I’m not working on it
<zimoun>rekado: thanks. nckhexen, apteryx, more on the topic?
<Kabouik>We have a guix importer for elpa but not for melpa. I'm new to all this emacs-ecosystem but I think melpa is the cool thing now, is there a way to use the elpa importer for it maybe?
<Kabouik>I'd be interested in packaging the demap package (I know we have emacs-minimap already but I wanted to compare).
<efraim>sneek: later tell cbaines based on the history of gnu/packages/rust.scm, I assume rust on aarch64 stopped building when we switched from mrustc@0.10->rust@1.39 to mrustc@0.10->rust@1.54
<sneek>Got it.
<efraim>sneek: later tell cbaines I'm looking at different commits of mrustc that might work
<sneek>Will do.
<efraim>sneek: botsnack
<rekado>Kabouik: the importer accepts the ‘-a’ option to let it import from melpa
<Kabouik>Oh, nice. I checked its help but didn't make the link between --archive and melpa
<cbaines>efraim, awesome
<sneek>Welcome back cbaines, you have 2 messages!
<sneek>cbaines, efraim says: based on the history of gnu/packages/rust.scm, I assume rust on aarch64 stopped building when we switched from mrustc@0.10->rust@1.39 to mrustc@0.10->rust@1.54
<sneek>cbaines, efraim says: I'm looking at different commits of mrustc that might work
<civodul>rekado: hi! do you have clues regarding berlin.guix being unavailable?
<sneek>civodul, you have 1 message!
<sneek>civodul, apteryx says: in commit 78d55b703d1 (2018), it seems you forgot to update the docstring of (@ (guix scripts environment) launch-environment)
<civodul>looks like node 129 is also unavailable
<lizog>hi everyone, i'm new to guix and guile, would like to pack river, the wayland dynamic tiling compositor for guix, and it depends on zig to build; seems like it would be good if we have a zig-build-system. what could possibly be a good starting point for me to write it?
<Kabouik>(River looks interesting)
<unmatched-paren>lizog: Well, the best thing to do would be to look at one of the guix/build-system/*.scm modules and see how you could replicate that for Zig.
<cbaines>lizog, I'd first get some things packaged for Guix without a new build system, then go back and refactor based on those package definitions
<unmatched-paren>You'll need both gnu/build-system/zig.scm (for package-side stuff) and gnu/build/zig-build-system.scm (build-side stuff)
<rekado>civodul: no, I only know that the machine has power and serial isn’t responding.
<rekado>if you want I can force a reboot
<civodul>rekado: i there's no serial and it's not a more general network outage, i guess we have no other option but to reboot it?
<efraim>lizog: I'm about to go AFK but I have to ask, does building river with zig still require the sources to be available for each library? or can it used shared libraries?
*efraim goes afk
<civodul>good news: now 'guix substitute' says "retrying download of '/gnu/store/1gddq9hrlqsyr2pnbk9lqzm1ljbbl0vi-e2fsprogs-1.46.4' with other substitute URLs..."
<cbaines>it works for me too, is tried many times, maybe once for each thing you're trying to substitute, which does make it slow, but it does eventually work :D
<ennoausberlin>Hi. I have a question regarding emacs packages. Until yesterday I used doom-emacs under GUIX, but I like to get emacs packages managed by guix, too. Can someone tell me, how recreate the leader-key (space) functionality within GUIX?
<civodul>cbaines: yes, that's a bit ridiculous; there's should be some sort of caching for EHOSTUNREACH
<ennoausberlin>My fingers are used to Space f f now instead of C-x C-f. And there are a lot more bindings
<ennoausberlin>using space
<civodul>ennoausberlin: hi! never used it but i see there's an emacs-spaceleader package
<ennoausberlin>civodul: Yes, but I dont know what to put into my .emacs to get it working
<zimoun>ennoausberlin: are you looking for emacs-god-mode?
<lizog>unmatched-paren: thanks for the suggestion, it makes sense
<lizog>cbaines: you mean pack it with a existing build system and then rewrite it with a new one?
<ennoausberlin>zimoun: Never heard of it. Maybe. I used doom emacs daily for a long time. I still like it, but to have a reproducible dev environment I want to get the key functionality in my guix config and drop doom eventually.
<cbaines>lizog, yeah, probably use the gnu-build-system as a base and just replace the phases that need to be different
<unmatched-paren>lizog: I think civodul is suggesting to rewrite it once there are a considerable amount of Zig packages.
<unmatched-paren>sorry, cbaines is suggesting.
<cbaines>lizog, writing a build system will be easier once you've packaged a few things without it
<lizog>efraim: sorry i don't fully understand your question, but i have it successfully built in a guix shell, specifying all the dependencies mentioned here:
<unmatched-paren>lizog: efraim is asking whether you could build Zig dependencies into shared libraries or whether it requires you to provide the actual source at a specific location
***noah is now known as kyte
<zimoun>ennoausberlin: emacs-god-mode replaces modifier as C-x C-f by modal and x f.
<lizog>unmatched-paren: oh, at the moment zig doesn't have an official language package manager, and for river, it has some git submodules in it, but these are included in the release tarball, so i don't have to pack any external dependency
<unmatched-paren>Oh, yes, I forgot, Zig does lots of vendering.
<unmatched-paren>You'll probably want to unvendor the vendored dependencies.
<ennoausberlin>zimoun: I will have a look. Thank you
<lizog>unmatched-paren: seems like all these git submodules are just wrappers written by ifreund, the author of river: do i still need to unvendor these?
*rekado resets berlin
<unmatched-paren>lizog: If they're not useful elsewhere, then no.
<zimoun>ennoausberlin: for testing, you can run “guix shell emacs emacs-spaceleader -- emacs -q -l test-init.el” where test-init.el is inspired by the README explanations on
<lizog>unmatched-paren: that makes sense. i think it's better for us to consider managing dependencies of zig software once zig has its official package manager shipped
<unmatched-paren>Ugh, another language package manager.
<lizog>andrew kelley (the lead of zig software foundation) gave a talk a few days ago, talking about the mess between language package managers and system package managers: i think in the end he didn't give a convincing solution to the problem, but at least he is fully aware of the issue
<ennoausberlin>zimoun: That was helpful. Thank you
<rekado>something’s copying the kernel to /boot now
<rekado>is that by design?
<rekado>on boot?
<rekado>I thought this should only happen on reconfigure?
<antipode>Guest54: bordeaux is used by default, IIUC.
<antipode>(Both and bordeaux are)
*antipode checks
<morganw>ennoausberlin: I think modalka lets you construct your own modal system, and the page for it makes comparison to alternatives (including god-mode).
<antipode>After a quick search it appears whenever appears usually bordeaux is there too, with %content-addressed-mirrors as exception
<rekado> has booted
<ennoausberlin>morganw: Do you use it by yourself. Does it play well with evil-mode?
<rekado>but it’s very slow now
<xd1le>ennoausberlin, no it's supposed to be used instead of evil mode
<xd1le>like an alternative
<ennoausberlin>xd1le: That would be a greater step then. Hmm. I keep it bookmarked for now
<morganw>ennoausberlin: I just use the default emacs key-bindings, but if you want to try some alternatives, is packaged in Guix. I think Modalka itself is meant to be a clean implementation of how to implement a modal system, so it is up to you how you use it.
<xd1le>ennoausberlin, there's also meow mode and ryo-modal, which modalka readme doesn't mention
<xd1le>ahh morganw got you first
<ennoausberlin>morganw: xd1le: meow looks easier to grasp for me. I will give it a try
<xd1le>ennoausberlin, btw you can also just use general.el directly if you just want the space leader
<xd1le>along with a package like which-key
<xd1le>But I'm not sure about how to do things like hydra.. if that's what doom does. I've never used doom.
<xd1le>(spaceleader uses general.el)
<ennoausberlin>xd1le: which-key is very nice, if you introduce emacs to users of say vs-code. I already installed it using guix
<xd1le>yeah it is
<nckhexen>Morning, Guix.
<ennoausberlin>xd1le: I was using vanilla emacs, before switching to doom and got familiar with evil-mode. I was blown away how good doom integrated all tools. Now I want to go one step beyond and do a similiar setup by using guix and guix home
<Lumine>I just use evil and doom modeline in vanilla emacs
<ekaitz>civodul: could you give me a ride through the shepherd and some guidelines about how to run it and test it for dev purposes?
<ennoausberlin>Lumine: I got used to the space leader key and the nice mnenomics doom provided, e.g. space gg for magit-status
<linj>rekado: does it? the kernel and initrd are not copied to /boot last time I check ( two days ago )
<Lumine>ennoausberlin: I see
<Lumine>doomemacs is also optimized for speed iirc
<xd1le>rekado, do we know what happened and why it went down?
<nckhexen>linj: It was not a general question. Berlin is weird.
<Lumine>My emacs desktop is a bit sluggish to load, but that just might be because of my awful init.el
<nckhexen>I don't see anything indicating that it could be run only on reconfigure.
<ennoausberlin>Lumine: My emacs ignores init.el. I think it should not, but only .emacs in my home folder is evaluated on startup. I need to figure out this. I use emacs 28.2 from the normal guix channel
<Lumine>ennoausberlin: it loads it from .emacs.d/ for me, no other place seemed to work
<morganw>I just use .emacs.d/init.el too.
<unmatched-paren>ennoausberlin: I believe emacs ignores .emacs.d/init.el if .emacs exists
<unmatched-paren>they're both equivalent
<unmatched-paren>it also ignores ~/.config/emacs if you have ~/.emacs.d, which mystified me for a while :)
<ennoausberlin>Lumine: it did not for me. Strange. Now it works in .emacs.d too. But I was switching emacs versions a lot lately. Maybe the profile and env variables were messed up somehow
<ennoausberlin>unmatched-paren: I tried to follow Andrew Tropins guix for emacs package management and tried .config/emacs too. And I was mystified as well :)
<unmatched-paren>ennoausberlin: You basically need to cp ~/.emacs.d $XDG_CONFIG_HOME/emacs
<unmatched-paren>and mv ~/.emacs $XDG_CONFIG_HOME/emacs/init.el
<ennoausberlin>unmatched-paren: Thanks for the hint. Do you use guix home to manage these files too?
<Lumine>home is where the home is, or something
<ennoausberlin>unmatched-paren: Cool. I will probably ask more questions in the future. But for now I have to pickup my boy from school. Thanks to all of you
<unmatched-paren>ennoausberlin: here :)
<declantsien>How Emacs Finds Your Init File
<lechner>itd_: thanks!
<sneek>Welcome back lechner, you have 1 message!
<sneek>lechner, itd_ says: As the error message said: There are no .go files in; you probably want one of the subdirectories instead (e.g., for siv: append -siv to package name & /siv to import path; I think).
<unmatched-paren>lechner, itd_: that's not quite the right way
<unmatched-paren>if there's one subdirectory with .go files, that's fine
<unmatched-paren>lechner, itd_: see:
<unmatched-paren>this package has, like, 6 subdirectories, which would each need to be their own package
<unmatched-paren>instead we can use (delete 'build) and #:tests? #f to make the package "source-only"
<unmatched-paren>oh, btw, for the former way, you'd need to make #:unpack-path the same as the old #:import-path
<brendyn>is it possible to pause a build, like with SIGSTOP?
<brendyn>id love to be able to pause and resume chromium builds
<derpey_noise[m]>Is it common to have guix's install hang at xf86-video-nouveau?
***wielaard is now known as mjw
<mroh>brendyn: maybe try ludo's tip:
<nckhexen>brendyn: Sure.
<nckhexen>It won't hurt anything if that's what you mean. Do it all the time.
<brendyn>nckhexen, which process do you pick to stop?
<brendyn>i see
<brendyn>guix process
<brendyn>i wonder if it would be possible to breakup the chromium package in to a set of several intermediate derivations that finally gets linked in to the final build
<civodul>rekado: the thing that copies the kernel is an activation snippet, so it runs both on reconfigure and on boot
<civodul>rekado: thanks for restarting it!
<civodul>looks like we experienced a shepherd crash!
<civodul>this is exciting
<civodul>bug reported
<fanta1>would not have happened with init
<sektor[m]>What is an error 100?
<sektor[m]>I saw it this mornign when I was pulling. I had to switch off substitutions.
<sektor[m]>But it failed with an error 100 while compiling openssl.
<unmatched-paren>well, ENETDOWN is apparently 100 on linux
<unmatched-paren>but that's an errno, not an exit code...
<abralek>Hi Guix!
<f3n1x>hi guixers ! newbie here... I'm willing to add my first service (syncthing, for the matter) to init.scm , by following the doc. here :
<f3n1x>it is requiring me to define $XDG_CONFIG_DIR/shepherd/init.d/synchting.scm .. but i cannot find $XDG_CONFIG_DIR . I'm puzzled... what am i missing ?
<f3n1x>ah... ok, seems that i need to run it as a sudoer . Fine now. Does setting shepherd user services needs to be done with 'sudo' ?
<gabber>IIRC $XDG_CONFIG_DIR usually is ~/.config
<f3n1x>nice to know... may i edit those files as a regular user ?
<gabber>f3n1x: are you trying to install a system-service in a home-configuration?
<f3n1x>yes gabber, exactly
<gabber>i don't think that's possible
<gabber>home-services have a "home-" prefix
<f3n1x>umh... well, i just humbly want to run a service (syncthing ,for the matter ) at boot time. how to ?
<gabber>as a system-wide service defined in your system-config.scm
<gabber>you are running Guix System, right?
<f3n1x>yes, i do !
<gabber>home-services only start on login (i think)
<f3n1x>that makes sense . For my use case (single user of an old laptop), there would be nobig differences..though
<provoostenator>Update: I was able to work around the issue by manipulating the timestamp, as suggested by emzy in #bitcoin-core-builds:
<gabber>well, from Guix's perspective the big difference is that there is no unprivileged/home-service defined for syncthing -- hence this simply does not work
<gabber>but feel free to write one and contribute!
<f3n1x>... why not ! thanks, thanks, thanks
<gabber>f3n1x: ... because! pretty sure that there's differences underneath ;)
*f3n1x has so much to learn ...
<pkill9>hello guixers
<pkill9>I am running guix worldwide
<f3n1x>hi pkill9 ... worldwide ?!
<pkill9>yes, I'm out in the world on my laptop running guix system
<nckhexen>In public??
<f3n1x>Ah. Enjoy ! From my side ... i see a ~/.config/guix/channels.scm file ... in order to add a new service, may i define it in a new ~/.config/guix/system-config.scm ?
<mirai>Hi, I'm trying to write a guix service manifest, what does the '(service-extension activation-service-type ...' do?
<mirai>I've looked at some of the existing services and I see that they usually have a 'NAME-service-activation' function and a NAME-shepherd-service
<gabber>f3n1x: you can define your system config wherever you like (version control might make your life easier), copy the currently running configuration (there's usually a copy at /etc/config.scm, but to be sure use the path mentioned in the last line of `guix system describe`) and add your own configuration there and finally `sudo guix system reconfigure path/to/your/systemconfig.scm`
<f3n1x>thanks, thanks, thanks gabber
<gabber>f3n1x: no problemo, muchach@
<Kabouik>75% of my keystrokes are doubled in Google Docs (I know, colleagues didn't give me a choice) when using Icecat (and I think I got the same issue with Nyxt). I never observed that on any other Linux machine. Anyone else with dirty hands in a Google Doc have observed that on Guix?
<Kabouik>I can't see how it could be related to the OS and not the browser, but since I observed the same in Nyxt, which I think doesn't use the same engine as Icecat, I don't know…
<gabber>i've been looking into the creation of a (proper) home-emacs-service but i've stumbled upon an issue. creating config files through guix is no issue; but (iirc) emacs uses ~/.emacs as a malleable(!) configuration entry-point with (custom-set-variables) and such on top. so i'm left to believe that ~/.emacs shall remain malleable (by the user and by emacs) so my current path is to create a service which generates an emacs-configuration
<gabber>and checks ~/.emacs whether there's a (load "/gnu/store/foo-emacs-config") line in the file. could this be a senseful way for my mission?
<gabber>unmatched-paren: you seemed interested last time, WDYT?
<unmatched-paren>gabber: you can change the file used with custom
<unmatched-paren>to something like ~/.config/emacs/custom.el...
<unmatched-paren>and then load that file, if it exists
<unmatched-paren>actually, that's what i do
<unmatched-paren>you could use --eval to run that code when emacs --fg-daemon is started
<gabber>i get that; but when we store config files in /gnu/store they're read-only -- and emacs uses (needs?) one config file to store stuff interactively, no?
<gabber>(and also thanks for commenting :)
<unmatched-paren>what do you mean by that?
<unmatched-paren>you could just make emacs load the custom.el file
<gabber>we have option 1: -- a simple shepherd service that doesn't care about configuration files and such.
<gabber>or we have option 2: (only the lower part; doesn't work yet) where we allow the user to add their own configuration in file-like objects in an <emacs-configuration> object. the crux is that in this latter way of realizing the service we would have to both: leave the .emacs file (for emacs to add it's font face and other configs) but also check whether the service's file-like objects are being loaded by emacs
<gabber>i'm just not sure whether checking file contents in an emacs-activation g-exp is the Guix way to solve the problem at hand
*f3n1x current 's system description says '/gnu/store/LOONGHAASH**78p7k-configuration.scm ' file
<f3n1x>umh... , extending that file ^^ , may i add my services there and reconfigure my system à-la '$ guix system reconfigure /gnu/store/LOONGHAASH**78p7k-my-configuration.scm ' ?
<f3n1x>I have some doubts... i might be missing some basics here. umh,... may i otherwise put it like this '~/.config/guix/my-configuration.scm' ?
<nckhexen>You can put it wherever you want. IMO it makes more sense to put it in, e.g., /etc/guix/system.scm as it's system-wide, but that's just, like, my opinion, man. You can't edit files under /gnu/store, though.
<f3n1x>ah... let's try. If something breaks, may a guix ... 'rollback' be possible ?
*f3n1x fingers crossed lol
<nckhexen>If it breaks so bad you can't even boot, do so after selecting an older generation at the GRUB menu.
<f3n1x>Nice to know. just to double check... i might directly append my (for the matter) 'synchting' service to /etc/guix/system.scm ?
<f3n1x>and then 'guix system reconfigure /etc/guix/system.scm' right ?
<nckhexen>Not literally append as in >>, but you can append it to the (services (list …)).
<nckhexen>Just in case ☺
<f3n1x>Crystal clear.ThX for the reminder.
<nckhexen>If you do it wrong, Guix will just complain, or at worst you'll put it somewhere where it has no effect. It's unlikely to break boot unless the service itself is broken.
<f3n1x>if i'm adding a 'syncthing service', may i add it to the packages list section of config.scm too ?
*f3n1x getting inspiration from taiju 's ?
<unmatched-paren>gabber: i think option 2 is better, so long as it uses file-likes, and not something like the rde emacs service which basically uses a list of sexps as an init.el
<nckhexen>f3n1x: Unlikely, although I don't know what this ‘service’ looks like. (packages…) just makes, well, packages available in the system profile, it cannot start services on its own.
*nckhexen foodz.
<f3n1x> /me ^^ got "guix system: error: symlink: Permission denied: "/var/guix/profiles/"
<unmatched-paren>f3n1x: did you ``sudo guix system reconfigure''?
<f3n1x>ah... sudo is required ! ThX
<f3n1x>sudo guix system reconfigure /etc/config.scm
<f3n1x>why does guix say 'guix system: warning: cannot determine provenance for current system' ?
<unmatched-paren>if you've never guix reconfigured before, or you've reconfigured from the source checkout previously, that will display
<unmatched-paren>nothing to worry about
<f3n1x>nice to know.
***justache is now known as justHaunted
<lechner>Hi, is it possible with 'guix deploy' to separate the operating-system declaration that can also be used with 'guix system reconfigure' from the machine specification that is used with 'deploy'? I asked in #scheme and was told that 'include' might work, but it hasn't so far. Thanks!
<civodul>zimoun: congrats for the Capitole du Libre talk! :-)
<civodul>lechner: exactly: 'include' is what you're looking for
<civodul>it amounts to pasting the contents of the file in the place where it's used, just like #include in C
<lechner>civodul: thanks! i removed all top-level 'define's for some custom packages but still see the presumably setuid-related operating-system.scm:264:14: error: program: unbound variable here
<lechner>civodul: btw, i think you ought to win some kind of prize on Guix's behalf
<lechner>civodul: that's with this deploy machine list
<yewscion>Hello All, anyone have advice for building all of the source derivations for a package's dependencies? I'd like to download (for instance) all of the sources I would need to have to compile ghc and its dependencies from source, but the --source option just downloads the ghc source itself.
<yewscion>I mostly want to group all of the network activity together, and then proceed to the local builds. That is: I'd like to download all needed sources now, and build them all later without needing a connection to the net. The issue I've run into is that it will download the sources for the packages I explicitly request (ghc) but not for those that are implicitly depended on and also need to be built (like, for instance libffi or ncurses).
<yewscion>A workaround would be to manually request all of the packages needed during the call (probably using a fake call to discover all needed packages) but that is cumbersome and something I'd rather avoid.
<civodul>lechner: oh right, you need to (use-modules (gnu system)) explicitly at the top of operating-system.scm
<civodul>this is done for you when you pass it to "guix system", but not when you pass it to "guix deploy"
<civodul>a weird discrepancy
<lechner>civodul: alas, that did not solve it. also, i do (use-modules (gnu system setuid)) already. why isn't this picked up here?
<civodul>lechner: what if you do (load ...) instead of (include ...)?
<civodul>yewscion: hi! did you try "guix build --sources=transitive PKG"?
<lechner>nckhexen: Hi, how does this look for a draft bug filing about the guix daemon deleting valid user roots, please?
<nckhexen>lechner: Looks good? Only gripe: the full-paragraph ad for gocryptfs was jarring. I don't see the point of it myself. It's not a gocryptfs-specific issue.
<lechner>nckhexen: no problem, i'll drop it. any tags in debbugs?
<nckhexen>I'd wait (with tagging).
<nckhexen>Thanks for this. Reporting bugs can be tedious.
<tex_milan>Hello Guixers! I am trying to use `name-service-switch` in my system config, but it complains that `hosts` is unbounded. I do use-modules with (gnu system nss). What is wrong? BTW I am trying to use mdns4 instead of mdns to fix my CUPS troubles (printer is found, but after few minutes CUPS doesn't see it anymore, after restart of CUPS everything is OK).
<tex_milan> (name-service-switch
<tex_milan> (hosts (list
<tex_milan> ; First, check /etc/hosts
<tex_milan> (name-service (name "files"))
<tex_milan> ;; If the above did not succeed, try with 'mdns_minimal'.
<tex_milan> (name-service
<tex_milan> (name "mdns4_minimal")
<unmatched-paren>tex_milan: please use a paste service such as ;)
<lechner>nckhexen: thanks for your help! here is the first bug
<tex_milan>Sorry, next time will do. So how is it possible that name-service-switch is recognized, but not hosts? (The former is type the latter is item in the type).
<bricewge>civodul: For the issue #58290 about guix deploy's error channel-open-session, a workaround is to increase ssh's MaxSessions
<nckhexen>tex_milan: Can you share the full configuration? It looks like you might be missing a surrounding ‘(name-service-switch …)’ here but it's impossible to tell.
<nckhexen>lechner: \o/
<nckhexen>tex_milan: I.e., (name-service-switch (name-service-switch (hosts …) …)) ; the first is the operating-system field name; the second constructs a name-service-switch record as its value. They happen to share a name.
<tex_milan>oh, I see!
<nckhexen>Right. Ding :)
<tex_milan>thanks nckhexen! Sometimes, seeing things twice is expected :D
<nckhexen>It is October after all.
<nckhexen>(My ‘happen to’ was a bit disingenous: it's a common pattern in Guix, and done on purpose. You get used to it.)
<tex_milan>Yep, but hey, it is more readable than Nix DSL (I am refugee form Nix land :D)
<nckhexen>I agree. (Same, if very very long ago.)
<nckhexen>I wouldn't say I'm a refugee, more like I fell in love in a foreign land.
<unmatched-paren>Odd. I've never even used Nix, and I find its DSL perfectly readable.
<nckhexen>Hardly odd.
<nckhexen>It is still allowed.
<tex_milan>:) Feel same. Well, Nix is OK if all you want is already there and you just configure your system. Once you need to override package, add package, do something non-trivial it gets very weird. I can speak C, C++, Erlang, Java, JavaScript, bash, xml, lisp, I even hacked pandoc (its in Haskell)... but I found my limit, it is Nix DSL :D
<unmatched-paren>I guess I've just not seen a "RAZE IT TO THE GROUND"-worthy .nix file...?
<tex_milan>I never knew what and why exactly I am doing what I was doing (copy paste from web in hope it does what I need). With Guix, I seem to understand what the lisp does.
<tex_milan>It may be my problem, I am not compatible with Nix DSL.
<nckhexen>unmatched-paren: I have, but I don't think the language was to blame. One could write bountiful horrors in Scheme, in any language.
<antipode>lechner: "guix build --sources=transitive" currently does not get all the transitive sources, see
<antipode>* lechner -> yewscion
<itd_>unmatched-paren: Thanks for the golang correction. :) (I wasn't aware that 'source-only' packages are permitted.)
<lechner>Hi, why does the example for guix deploy here include (allow-empty-passwords? #t) please? is that necessary (or advisable)?
<nckhexen>Speaking of golang (OK, not really, but a segue's a segue): is this considered sane there, or is there a better pattern? <> (note the ‘recursion’ of sorts).
<unmatched-paren>nckhexen: that's pretty much exactly what itd_ was talking about :)
<unmatched-paren>with the two go-filippo-io-cmd-age* packages
<nckhexen>And what was the ‘golang correction’ that I didn't read? 😛
<unmatched-paren>you can combine go "subdir" packages into one by making a "source-only" package, which you do by deleting 'build and disabling tests
<unmatched-paren>though i'm really surprised that age package doesn't work without the go-age packages...
<nckhexen>I haven't tried.
<nckhexen>I was just cleaning up the commit messages & lint warnings so far, but it caught my attention.
<unmatched-paren> <- in here, there's go files in the root, so it should work without splitting it into several "subdir" packages
<nckhexen>Also because I'd like to move it to password-utils.scm.
<unmatched-paren>you should only need source-onlyies if the package is basically several packages, split into several directories
<unmatched-paren>this one simply is one package with a bunch of subdirectories for tidyness
<unmatched-paren>this one, for example, warrants source-onlyness
<unmatched-paren>note the absence of any .go files in the root
<unmatched-paren>so, yeah, try removing those -cmd-age packages
<unmatched-paren>i think it should work?
<nckhexen>We'll see!
<nckhexen>I was already trying it TBQH, but it's good to have some confirmation.
<nckhexen>I mean of course I know what I am doing wat.
<lechner>unmatched-paren: thanks for the golang packaging hint! should i make all propagated inputs "source-only" packages?
<nckhexen>I love Go and know its basics haha.
<unmatched-paren>lechner: No, only the ones that feature multiple source directories, each with their own go.mod.
<unmatched-paren>lechner: Like this one:
<unmatched-paren>Oh, wait. Those don't have their own go.mod.
<unmatched-paren>Okay, multiple source directories, with a go.mod in the root, but no .go files in the root.
<lizog>hi guix, i was trying to pack river, as i've mentioned before, using gnu-build-system, but i encountered a problem. i tried substitute the build phase for a new one, as written in this paste link: however, when i was trying to build this package, it shows error it the output log file. it failed at the build phase, and the log is in the end of the pastezone file
<unmatched-paren>lizog: hmm, does that "install" at the end of the command mean to install the program?
<unmatched-paren>lizog: why are you deleting patch-source-shebangs and patch-generated-file-shebangs?
<unmatched-paren>they do no harm, even if they don't do anything here
<unmatched-paren>also, check comes before install :)
<unmatched-paren>and i'm pretty sure it's ``zig test'', not ``zig build test''
*unmatched-paren guix shell zig
<unmatched-paren>yup, ``zig test''
<unmatched-paren>looks like there is no ``install'' target though
<lechner>unmatched-paren: how about this one, please? it builds but the tests fail
<nckhexen><check comes before install> It's OK if it doesn't for good reasons.
<patched[m]>If there are recommended way to fix unicode characters not rendering? Seems to be a problem for me on the many programs
<unmatched-paren>nckhexen: They didn't rearrange it, they only put their (replace 'check) after (replace 'install) :)
<unmatched-paren>lechner: source-only packages should have tests disabled (if that is what you're talking about :))
<lizog>unmatched-paren: yes, the install there means build and install, and as suggested in river's packaging guide(, here i just firstly install the binaries in the build phase and copy them in the install phase. does that make sense?
<unmatched-paren>lizog: Ah, so you are installing.
<lizog>unmatched-paren: it's actually ``zig build test'', as i've tested it in a guix shell
<unmatched-paren>I think you should delete 'build and just put that command in 'install.
<lizog>unmatched-paren: here it just installs the files inside the source directory
<unmatched-paren>And then just use #$output as the prefix.
<unmatched-paren>Oh, yes, actually, lemme just gexpify that package for you...
<lizog>unmatched-paren: oh that seems more reasonable, i'll just give it a try
<unmatched-paren>One moment.
<unmatched-paren>lizog: Here you go.
<unmatched-paren>Make sure to import (guix gexp).
<unmatched-paren>I think this should work?
*unmatched-paren goes to test it
<lechner>unmatched-paren: is not a source-only package in this recipe but the test ffail due to missing sources. do i have to propagate the missing input, ogletest ?
<unmatched-paren>lechner: It should be a native-input.
<lechner>unmatched-paren: it isn't either!
<unmatched-paren>lizog: IMO building from git is cleaner:
<unmatched-paren>of course you need to import (guix git-download) there :)
<kaelyn>following up on Cairn's pointer from last night, I've updated my mesa patch to include the i686 test fix:
<lizog>unmatched-paren: does the git-fetch also pulls all the submodules?
<unmatched-paren>lizog: If you add (recursive? #t) to the git-reference, it will.
<podiki[m]>kaelyn: great! (twas me)
<unmatched-paren>Intriguing. I still get AccessDenied.
<unmatched-paren>Aha, you need to set XDG_CACHE_HOME.
<podiki[m]>kaelyn: I should note that I did not test extensively, but the failure only happened on i686 not x86_64, though I don't upstream mentioned architecture
<podiki[m]>kaelyn: thanks also for the vulkan updates, glad to see it!
<lizog>unmatched-paren: in the 'check phase?
<unmatched-paren>lizog: Oh sequence of random non-alphanumeric symbols.
<kaelyn>podiki[m]: whoops! Sorry about that misattribution. I got two conversations conflated ;)
<unmatched-paren>You need wlroots 0.15.
<unmatched-paren>Looks like you'll need to wait. A while.
<podiki[m]>no worries!
<podiki[m]>I have a 6th sense for mesa stuff being mentioned (it was my first real patch attempt in the previous core-updates cycle)
<Cairn>Hehe, yep; I just praised the work you were doing.
<unmatched-paren>wlroots 0.15 requires some updates to core graphics packages, which are in core-updates, which will be merged in, what, a couple of months...?
<unmatched-paren>So. Sad. :(
<podiki[m]>kaelyn: have you looked at libglvnd stuff at all by any chance? there were discussions last year about building mesa with that option and how to handle things like foreign distros (that provide other gl drivers)
<unmatched-paren>"So" there is not intended as an amplifier, by the way.
<lizog>unmatched-paren: oops, i think andrew tropin's repo has wlroots 0.15. so maybe i'll just use it locally and wait for the merge of core-updates and then compose a patch
<unmatched-paren>lizog: <- This is as far as I got before I got the "requires wlroots 0.15" error.
<lechner>unmatched-paren: i already have (use-modules (guix gexp)). is this on the build side?
<unmatched-paren>lechner: looks like you might be nesting gexps somewhere?
<unmatched-paren>can you show the package definition?
<lizog>unmatched-paren: may i ask another noob question: what is this ``cc-for-target''?
<unmatched-paren>lizog: Basically, it returns either "gcc" if natively compiling or "TARGET-gcc" if cross-compiling.
<kaelyn>podiki[m]: No, I haven't looked at the libglvnd stuff and am not familiar with the discussions that have happened so far. I originally updated the vulkan packages so I could package vulkan-validationlayers for use with a vulkan tutorial; I also use mesa and Guix System pretty much exclusively so haven't (yet) encountered the use cases for libglvnd
<unmatched-paren>In a normal situation where you weren't ripping out most of the phases, it would allow you to refer to the name of GCC without breaking cross-compilation.
<unmatched-paren>In this case, you're using Zig, so you'd probably need to do some extra stuff to make cross work anyway, but it's good practise to use it.
<lechner>unmatched-paren: should all the propagated-inputs in gocryptfs be native-inputs?
<unmatched-paren>lechner: Um, that's a link to the build log.
<unmatched-paren>lechner: Basically, the ones that are used only during the build should be native.
<unmatched-paren>So, anything used for e.g. a build script, a test, etc.
<unmatched-paren>This applies for anything, not just Go, by the way.
<lechner>unmatched-paren: sorry,
<lizog>unmatched-paren: thanks for all the helps, i learned a lot!
<apteryx>where is hosted? seems the cert is expired
<unmatched-paren>apteryx: i think rekado hosts it, but i'm not certain
<unmatched-paren>lizog: no problem!
<apteryx>or just doesn't have any cert support
<unmatched-paren>lechner: You should probably use new-style inputs there :0
<unmatched-paren>apteryx: wrks for me.
<unmatched-paren>Maybe a cache issue?
<lechner>unmatched-paren: yeah, i know. i am so new i figured i would work with the importer first.
<apteryx>unmatched-paren: that's from icecat
<lechner>unmatched-paren: will that fix my issue, though?
<unmatched-paren>apteryx: I'm on qutebrowser
<apteryx>it works but it says 'not secure'; from ungoogled-chromium also
<unmatched-paren>lechner: Aha, you see in go-oglemock, you use a gexp in a `(...)?
<lechner>apteryx unmatched-paren: also serves plain HTTP, but HTTPS works here, as well
<unmatched-paren>In there, you're basically quoting your call to (gexp ...), causing it to be run on the build side
<unmatched-paren>so, use (list ...) instead
<efraim>cbaines: I got good progress on a much newer mrustc commit, but then my machine went OOM. Trying again with qemu emulation, and also testing that commit on riscv64-linux
<unmatched-paren>But also do the other things. They won't fix it, but you should still do them.
<efraim>if it works out then I'll push it to staging I guess for the rust rebuild
<podiki[m]>bootrapable works over https for me too, says certificate valid from oct 11 (recent)
<lechner>unmatched-paren: in line 106?
<unmatched-paren>lechner: Yes.
<lechner>unmatched-paren: tricky! i did that in the other one, but only because my eyesight is so poor. i habitually replace the quotes with their longform when no one is watching
<yewscion>sneek: later tell civodul guix build --source=transitive PKG did the trick! Thanks, I dunno why I missed that option.
<sneek>Got it.
<yewscion>sneek: botsnack
<nckhexen>I do the opposite with lambda → λ in emacs (since Guix outlawed upstream λs :), perhaps you could let your editor take care of it? It would help reading all code, not just your own.
<nckhexen>lechner: ☝
<mirai>Where should config files be created? What's the difference between 'NAME-service-activation' and NAME-shepherd-service?
<unmatched-paren>mirai: An "activation service" is basically a service that runs code when you do ``guix system reconfigure''
<unmatched-paren>so you might use one to create a directory that the program expects to exist, for example
<unmatched-paren>whereas a shepherd service allows you to specify daemons for Shepherd to handle and monitor
<unmatched-paren>so you can write a shepherd-service that provides 'foo, so when you ``herd stop foo'', the service will be killed based on the ``stop'' field of the shepherd-service record
<tricon>nckhexen: you are a good person for that recommendation.
<tricon>and now i want to play Hexen.
<nckhexen>…I just realised my nick is FSDG-problematic.
<podiki[m]>after being shunned for lack of wayland, how the tables have turned!
<tricon>nckhexen: if there's a FLOSS engine that can load the "WADs", it wouldn't be.
<nckhexen>(Or is it? I used free software to play it. /me recently finished playing it, after >20 years.)
<podiki[m]>though didn't hexen use the doom engine that got open sourced?
<nckhexen>The ‘functional data’ stuff still confuses me.
<mirai>unmatched-paren: thanks
<tricon>nckhexen: it's a bit of a blurry, grey area, eh?
<podiki[m]>oh the hexen source is out there too, was released
<nckhexen>I don't think it was open-sourced then, but I didn't check.
<nckhexen>It's a heavily modified Doom engine yes.
<nckhexen>*was, now it's just gzdoom :)
<tricon>i used to leave a modded Quake server running on a tiny VPS in Chicago. i should start up some old-shool FPS servers.
<podiki[m]>not sure the license on hexen source, but the dos version exists, dunno if it is easily usable
*nckhexen unshuns podiki[m]. I'm sorry, I forgot that was still on.
<nckhexen>I did play Hexen on Wayland.
*podiki[m] dances as the unshunned, loud and proud
<vivien>Hi, I’m trying to make a relocatable guile executable. There is add-to-load-path, but not add-to-load-compiled-path, should I modify %load-compiled-path directly to add something to the compiled path?
<podiki[m]>....didn't last long :)
<tricon>whenever i see the word "shun", i always think of The Office.
<unmatched-paren>vivien: I think so.
*unmatched-paren wishes ``guild pack-executable'' was a thing that existed
***mark__ is now known as mjw
<civodul>rekado: you around? we'll need another berlin reboot :-/
<sneek>civodul, you have 1 message!
<sneek>civodul, yewscion says: guix build --source=transitive PKG did the trick! Thanks, I dunno why I missed that option.
<nckhexen>civodul: I am. Why?
<civodul>nckhexen: i messed up while debugging
<civodul>this is embarrassing
<nckhexen>It's not.
<nckhexen>Just so you know, reboots take about 20 minutes.
<civodul>ooh right
<nckhexen>Much SAN, very btrfs, etc.
<lilyp>quick question: how do I substitute* a file that uses idiot encoding
<unmatched-paren>lilyp: maybe you could ``uconv'' it into utf-8, then ``substitute*'', then ``uconv'' back?
<nckhexen>Kernel booted. Now we wait. I mean, in addition to the waiting we just did. Slightly different muzak plays.
<unmatched-paren>s/uconv/iconv/ i guess
<lilyp>(with-fluids ((%default-port-encoding #f)) ...) was the solution
<civodul>nckhexen: dankon!
<civodul>actually maybe it was already rebooting?
<civodul>thing is, gdb passes SIGINT straight to the inferior
<civodul>SIGINT on shepherd means ctrl-alt-del...
<nckhexen>There was no console output, but that's not hard proof.
<nckhexen>That console is flakey, or has rules we don't quite understand.
<nckhexen>civodul: We could test that hypothesis with this console session open, but it would take another ~20min.
<nckhexen>Up to you?
<apteryx>there'll be a new Btrfs option in 6.1 that should make our boot zippy again
<nckhexen>Tell me more.
<rekado>have I been summoned?
<apteryx>we could also drop 'compress-force' and use 'compress' alone instead; non-compressible files then store in 128 MiB chunks instead of 512 KiB
<nckhexen>Yes, but berlin is rebooting already.
<nckhexen>In fact it is copying everything to /boot as we speak.
<rekado>about that console: it’s weird because we configure more than one console
<rekado>I remember that I was able to fix it (in one of apteryx and my debug sessions some months ago) by changing the kernel arguments
<rekado>worth playing with those args on one of the build nodes
*nckhexen …your boot is important to us, please stand by, a rootfs will be with you shortly…
<nckhexen>civodul: SSH away.
<apteryx>nckhexen: then search for block-group-tree on that page
<civodul>nckhexen: no no, i won't cause another reboot
<civodul>chances are shepherd will cause it within a few days/weeks anyway :-)
<civodul>(i haven't been able to reproduce the problem locally yet)
<nckhexen>apteryx: That sounds perfect.
<lamp140>i set up the httpd service and now get this message when the access: "Forbidden
<lamp140>You don't have permission to access this resource."
<lamp140>i also set up the mariadb service but i cannot access the database
<nckhexen>I'm heading to bed, but to help debug this the channel wil need a lot more info: starting with your Guix system configuration, as well as any relevant httpd.conf, mariadb commands you tried, etc. You can use for all that.
<nckhexen>Night #guix o/
<csepp>folks, anybody's got some magic incantations to make using git with the guix codebase not suck immensely?
<gabber>unmatched-paren: thanks! i'll give it a go then!
<unmatched-paren>csepp: How does it suck immensely? :)
<csepp>i'm trying to commit something piece by piece and the formatting style makes things like splitting diffs impossible, or at least way too involved.
<unmatched-paren>Usually, I write one package at a time, committing each one, instead of writing them all at once then committing them all.
<unmatched-paren>Magit's diff splitting features are really nice, but I'm betting you aren't using Emacs?
<csepp>also resolving even small conflicts when popping from the stash is a damn nightmare, git's diff is so wrong.
***stikonas_ is now known as stikonas
<csepp>unmatched-paren: i am, but somehow using emacs has no magically fixed my workflow. :)
<csepp>i'm sure i only have to install another dozen major and minor modes.
<lamp140>nckhexen: thank's
<lamp140>my config.scm:
<unmatched-paren>csepp: If you use the Magit staging window, you can select code to commit line-by-line.
<lamp140>my intention is to climb the nextcloud
<csepp>unmatched-paren: oh, that's useful. thx. on that note, could you or someone pleeease write these down in the manual/cookbook? or i guess i can try transferring my notes but it would be good to know how people who are productive at hacking guix do it.
<unmatched-paren>csepp: Maybe we could add a bit to "The Perfect Setup" in "Contributing" about using Magit with Guix...
<csepp>Yupp. The perfect setup definitely does not cover everything IMHO. (It should also have something about email probably.)
<csepp>inb4 snark about "everything", i mean everything that is needed for contributing large patches without giving up after a month.
<lechner>Hi, is the 'botsnack' command required when using sneek?
<unmatched-paren>lechner: no, it's just a bit of fun :)
<unmatched-paren>sneek: botsnack
<lechner>unmatched-paren: thanks!
*vagrantc notices sneek being more excited to see people who give botsnacks
<lechner>sneek: later tell nckhexen Will you please share your visual Emacs lambda mod? Thanks!
<lechner>sneek: botsnack
***jonsger1 is now known as jonsger
<lechner>Hi, does guix build catch circular, native prerequisites?
<unmatched-paren>lechner: It will do an infinite loop before crashing due to an overly large heap if you have recursive inputs
<lechner>unmatched-paren: i do not how to resolve a loop between and
<unmatched-paren>lechner: try disabling tests for oglematchers
<unmatched-paren>and removing the ogletest dependency from its inputs
<lechner>unmatched-paren: thanks! that worked
<lechner>unmatched-paren: why would this error not resolve with x/net being a native input, please?
<lechner>unmatched-paren: this is my latest spec
<unmatched-paren>lechner: x/net should not be a native input
<unmatched-paren>because it's used by the actual program/library, not the tests or build scripts
<lechner>unmatched-paren: i thought all golang module serve source
<lechner>maybe our version is too old
<unmatched-paren>lechner: you need to propagate libraries used by other libraries
<unmatched-paren>but you can just put libraries used by go programs in inputs
<unmatched-paren>s/libraries/go libraries/
<lechner>unmatched-paren: what kind of a library is x-net, please?
<gnou_liber>Hello, I recently installed the Guix system with i3 on my PC, what happened is that I had a problem when I wanted to change the i3status configuration, I could not find the configuration file anywhere, could someone help me?
<unmatched-paren>lechner: it's just a regular old go library...
<lechner>unmatched-paren: last time i checked, they did not have any
<lechner>in the traditional sense
<lechner>i thought golang always produces a static executable using only syscalls