IRC channel logs


back to list of logs

<TakeV>Hey there! I am somewhat new to submitting patches. If I have something I would like to send in (an update/fix for remmina), do I also CC the mentors team?
<apteryx>TakeV: hi! It seems a good idea to CC the mentors if you'd like to get an introduction to the process
<TakeV>apteryx: Thanks!
<TakeV>Hmm, how odd. I keep getting an error when trying to send the email via git-send.
<mirai>apteryx: re shared-mime-info, I think that patch is broken / needs to be redone
<mirai>I think it doesn't build anymore / hash changed
<mirai>though it did build when I sent it (?!)
<mirai>it's more profitable to review #61015 instead, since that one helps immensely to get xdg-shared-mime updated
<helpful-friend>"[PATCH core-updates 0/9] Modernize and fix docbook-xml"
<mirai>if I fix the hash in the shared-mime-info and merge the docbook-xml changes then the package should build straight away
<mirai>otherwise it will fail somewhere
<mirai>(the modify-phases would also be unnecessary with docbook-xml changes)
<mirai>TakeV: the manual has a Contributing section
<mirai>and some recommended values to configure git send-email, etc.
<TakeV>mirai: That is the part I am getting confused about. I am using that documentation, and there is a failure. Hmm.
<mirai>what are you seeing?
<apteryx>mirai: I see
<apteryx>TakeV: are you using "git send-email $(etc/teams.scm ...)" ?
<apteryx>this is known to have a problem
<helpful-friend>"can't substitute etc/teams.scm command as doc suggests"
<TakeV>apteryx: Yeah. Oof.
<mirai>does it make a difference to cc teams or not?
<mirai>don't they all receive the torrent of mails from ML anyway?
<mirai>or read through a NNTP mirror of the MLs?
<apteryx>mirai: fstrim reviewed!
<apteryx>mirai: the idea is that they don't want/can't pay attention to the torrent that is guix-patches anymore
<apteryx>and can focus on changes they care most about
<apteryx>(I don't blame anyone from not following guix-patches directly, I don't either)
<apteryx>I do use emacs-debbugs though.
<mirai>apteryx: thanks! I'll revise it and CC you after addressing the points from the feedback
<apteryx>sounds good
<bumble[m]>"git:send-email" is added to my packages list processed with specifications->packages and reconfigure home succeeded, but using git send-email command shows this,
<bumble[m]>`git: 'send-email' is not a git command.`
<sneek>Welcome back bumble[m], you have 1 message!
<sneek>bumble[m], lechner says: / it was not anything you said. mondays are usually quieter than other days. also nckx, who was the heart and soul of this channel, somehow disengaged. it probably did not help that i got into an argument here yesterday. people avoid mean places and instead spend time with friends
<bumble[m]>does anyone know what I am doing wrong?
<bumble[m]>ah thanks :)
<bumble[m]>I see
<bumble[m]>referencing home config files from unwox and unmatched-paren shows that they specify send-email using this expression `(list git "send-email")` but I'm not sure how to format this to work with "specifications->packages", not used by those config files
<bumble[m]>i see there is not a nice solution installing git:send-email with specifications->packages
<helpful-friend>"Re: How to select git:send-email output in config.scm?"
<bumble[m]>wishing specifications->packages would handle the colon itself
<spacecadet[m]>is there an easy way to drop a file in /etc? just testing and breaking things
<unmatched-paren>bumble[m]: try specifications->packages+outputs
<unmatched-paren>(not sure what the exact name is)
<ArneBab>lechner: kein Problem — Namenverrutscher passieren :-)
<sneek>Welcome back ArneBab, you have 1 message!
<sneek>ArneBab, lechner says: / Bitte entschuldige, dass ich Deinen Namen im Email etwas verkuerzt habe. Es war ein Versehen, und aus Eile
<bumble[m]>@unmatched-paren thank you for your response
<jpoiret>bumble[m]: I don't think it works with specifications->packages, so what I do is use specification->package for most packages but then use ((specification->package spec) output)
<jpoiret>now i try to import the modules that have the packages and use the package symbols
<abrenon>hi guix
<gabber>i wanted to use the redshift home service on my Sway/wayland desktop but refrained - since randr and redshift don't work on Wayland. now i found gammastep which seems to be a drop-in replacement for redshift (this isn't mentioned anywhere, but the help text of both applications are almost identical). going the bit-too-tired-half-step forward yesterday night i set the (redshift) attribute of the redshift service to gammastep -- which
<gabber>of course didn't work because in the service the package-path gets "/bin/redshift" appended. so i've come here to ask what the easiest way is to provide the service for sway/wlroots users: a) duplicate and alter the code for the gammastep service b) change the current definition of the redshift service so that it doesn't depend on a hard-coded "/bin/redshift" string c) something else
<abrenon>gabber: are the options, configuration, etc. the same between both ? as in, is it only a matter of replacing the name of the binary ?
<mirai>gabber: you can be daunting and generalise the services instead
<mirai>a night-mode-service-type
<mirai>which can intelligently choose the backend it uses or leave it up to the user
<gabber>abrenon: i think they are!
<gabber>mirai: yes, that was exactly what i had in mind
<gabber>how would such a service find out whether Xorg or Wayland was in use?
<mirai>perhaps that should be manually done
<abrenon>(long shot) …in the config of the DE service ?
<mirai>via a parameter wayland? #t / #f
<abrenon>that's more realistic : )
<mirai>or via sorcery in the service constructor
<gabber>but currently the reshift service is a home service whereas GDM is a system service
<mirai>what's the problem with that?
<gabber>no problem, i'm just wondering how my home service can figure out a specific property of a system-service
<mirai>services don't get to see each other
<mirai>not without resorting to lurid horrible runtime things
<gabber>that's what i figured (:
<gabber>so my main concern is that if i create a generalized night-mode-service-type i would omit/clear/remove the (package) field which lets a user specify their own version of a package. isn't this against some (maybe unwritten) golden Guix rule?
<efraim>gabber: you might be able to reuse something from the list `env | grep wayland` (like XDG_SESSION_TYPE) in a getenv to check for wayland
<mirai>^^within the service constructor
<gabber>that would work! but then we would be unable to tell (at reconfigure time) which package would actually be used (and might need to provide both redshift and gammastep), right?
<gabber>this somehow feels a bit un-guix-y
<mirai>you can define 2 record types here
<mirai>the generalised service config
<mirai>and a night-mode-backends, whose fields are: { redshift: <package for redshift>, ... }
<evilsetg[m]>Would it be possible to do this with the extension mechanism?
<evilsetg[m]>But just adding the package to a general night-mode-service does seem easier.
<dcunit3d>for a project with guix.scm or manifest.scm, where's the best place for development dependencies like lsp servers and linters to go?
<dcunit3d>i'm using buffer-env.el to load a project-local environment, but i'm not sure which of these files should include such tools. i don't really want them in my path all the time.
<gabber>mirai: what's that night-mode-backends supposed to be? a separate configuration? do you have any services in mind that do what you're referring to?
<dcunit3d>i've thought about concatenating manifests from other files, but this seems a little heavy-handed
<dcunit3d> has things like clojure-lsp and rde features in the guix.scm
<helpful-friend>"~abcdw/guix-clojure: guix.scm - sourcehut git"
<gabber>ok, there's this one difference: in gammastep the section is called [general] (it's the section that defines color temperatures and such), in redshift this is called [redshift]... hrmpf
<mirai>iyzsong[m]: oops, my bad, I didn't see your work at #58086 and ended up reimplementing it with #61964
<helpful-friend>"[PATCH] gnu: Add fstrim-service-type."
<helpful-friend>"[PATCH] services: Add fstrim-service-type."
<mirai>gabber: let me dig up an example
<jpoiret>dcunit3d: I would recommend using both guix.scm and manifest.scm, with manifest.scm containing the dev dependencies
<jpoiret>although there's no handy shortcut to get a development manifest in scheme (i think?)
<jpoiret>maybe that could be done :)
<dcunit3d>i was thinking i might have a second guix-develop.scm package that inherits from the first and adds development dependencies
<dcunit3d>will guix shell load both of these?
<dcunit3d>both of guix.scm and manifest.scm?
<jpoiret>no, you need to concatenate the inputs of the guix.scm package to manifest.scm
<jpoiret>something like (define the-package (load "guix.scm")) then using (native-inputs the-package) and the like
<jpoiret>not clean, but less boilerplatey than copying everything
<dcunit3d>i see, thanks for the tip
<jpoiret>i think manifest has precedence, although I'm not sure
<mrvdb>Trying to use 'guix install --without-tests=kio kio' , but tests still run. What am I overlooking?
<jlicht>hey guix
<sneek>Welcome back jlicht, you have 1 message!
<sneek>jlicht, lechner says: / thanks for the pointer about i get tripped up because the address for bugs is, and I still use the old interface from time to time
<mirai>jpoiret: what address are you using when you submit new patches/bugs
<mirai>sneek's comment might be the key to the “X via Guix-patches via <guix-patches....>” mystery
<sarg>hey guix, has somebody packaged doom emacs already?
<jlicht>sarg: I don't think so, but I have trouble seeing how one would package doom in the first place
<jlicht>probably a shortcoming of my imagination though :-)
<jlicht>is it a 'normal' elisp package these days?
<sarg>not, but hlissner is working on extracting core out of it
<jpoiret>mirai: the same as my usual one :)
<jpoiret>i think the initial mail sent to guix-patches gets its header rewritten but I'm not sure
<mirai>jpoiret: with or without the [debbugs.] part
<jlicht>(it bounces without, afaik)
<mirai>no idea then
<jlicht>You get a response that your mail could not be delivered from your friendly neighbourhood mail server
<sarg>jpoiret jfyi, your mails end up in spam on gmail. Maybe something wrong with the setup of your mail server
<mirai>surely it's not gmail / the largest mail providers throwing their weight around instead?
<robin>gmail's getting quite...selective about what email it accepts (better have dmarc+spf+dkim configured correctly, although that doesn't seem to be much of an obstacle for spammers)
<robin>forwarding to gmail via smtp also appears to be a Bad Idea that can, apparently, get your mailserver marked as a spammer even if you filter out obvious spam first
<robin>i thought that was maybe just a problem for tiny mail servers (like the one hcoop runs for ~100 members), but:
<helpful-friend>"Chris's Wiki :: blog/spam/DKIMSigningMostlyMandatory"
<gabber>mirai: i think i won't have much more time today to work on that but feel free to sneek me a reference to an example :) TIA!
<jpoiret>I have dmarc+dkim+spf on my mailserver though :)
<jpoiret>sarg: so was your mail in response to mine?
<sarg>jpoiret you mean the last mail in gpt ticket? No, that was me following your recommendation on documenting local testing to help maintainers
<jpoiret>ah, thanks! I was a bit confused when I read it this morning
<sarg>jpoiret but should the image from `bare-bones.tmpl` be bootable? It has `/dev/sdX` in targets
<sarg>I assume the template has to be edited manually
<robin>jpoiret, ah, that's unfortunate. hcoop only has spf so i hoped full dkim and dmarc support would fix things...
<mirai>guix system: error: /home/ika/src/guix/gnu/packages/gd.scm:100:2: package `perl-gd@2.73' has an invalid input: ("_" #<syntax-transformer perl-extutils-pkgconfig>)
<mirai>something broke in latest guix
<robin>file under "things google does that you can't actually do anything about" i guess
<mirai>nvm, nothing broke, I had stale .go files
<efraim>mirai: try `git grep -E '\(gnu packages gd\)' | cut -f1 -d':' | sed -e 's/scm/go/' | xargs -n1 rm` and then run make again
<mirai>false alarm
<mirai>my bad
<apteryx>jpoiret: do you have some config guide somewhere for the email server stuff? I've been punting forever on setting up one myself.
<jpoiret>efraim: I use the simpler `grep --include '*.go' -l perl-extutils-pkgconfig -R ./gnu | xargs rm`
<jpoiret>apteryx: i'm using a docker container that configures everything for you. I plan on moving to guix for my server at some point
<efraim>jpoiret: I suppose that would also ignore ones that had already been recompiled without the offending package, or that never had it to begin with
<apteryx>jpoiret: ah!
<apteryx>mirai: do you think the fstrim service would be good to add to the base-services?
<apteryx>seems most distribution do so now
<mirai>apteryx: I don't know if its linux specific though
<mirai>it uses util-linux package for one
<apteryx>ah yes, so it should be added conditionally if the kernel is Linux
<apteryx>if that's doable
<mirai>another concern is that some might prefer discards in fstab?
<mirai>I think I had to manually enable fstrim service in fedora
<mirai>at least N releases ago
<mirai>not sure what's the status nowadays
<apteryx>supposed to have been enabled by default in fedora 32
<mirai>do you want me to send a v3 with a separate commit that adds it?
<apteryx>it can be kept separate
<apteryx>I'll push the fstrim service itself shortly
<nom[m]>anyone know how to disable pulseaudio and use pipewire instead?
<mirai>apteryx: ok
<jpoiret>I think util-linux is compatible with Hurd
<mirai>but will it trim?
<mirai>or will it: a. obliterate b. do nothing?
<jpoiret>isn't fstrim done using ioctls? if so, this won't work on the hurd
<bumble[m]>is there a command that "cleans up" or removes the "guix-home-legacy-configs-backup" directories in the home directory? is it recommended to simply rm -rf these once in a while?
<jpoiret>but yeah, it probably won't work on the hurd, a good old (if (target-linux? ...)) will work
<sarg>bumble[m]: why do you keep getting these? It should be a one-time thing
<bumble[m]>The guix commands I use basically are "sudo guix system reconfigure guix.system.scm" and "guix home reconfigure guix.home.scm" --I don't know where these directories come from
<bumble[m]> this link indicates they are created each time home reconfigure is used
<sarg>what's inside? these are created by guix-home symlinking script
<bumble[m]>@sarg they are completely empty
<bumble[m]>I had not looked in them before...
<sarg>sure? `find` inside, there must be dotfiles in it
<bumble[m]>$ ls -a 1673427559-guix-home-legacy-configs-backup/
<bumble[m]>./ ../ .bash_profile .bashrc
<sarg>yup, so it seems that these two are being overwritten by something else and guix-home has to restore the files each time
<bumble[m]>hm ok thank you
<apteryx>jpoiret: for btrfs at least, it seems discard=async is on by default with Linux 6.2
<apteryx>not sure about other file systems
<apteryx>if it's on by default for all file systems, then we wouldn't need to add the fstrim service to %base-services
<apteryx>to be researched
<jlicht>all that filesystem-related stuff still feels like cargo culting to me
<jlicht>(when I do it, at least)
<jlicht>I have a USB keyboard connected over USB-c dongle to my laptop; is it possible to use that keyboard to type my FDE encryption password in grub?
<winter>Is there an effective way to, with Guix System, use a different version of a library for every package, including those e.g. referenced by services that I might not necessarily have easy control over? (I've asked this in the past, but I'm curious if there's any method -- even if roundabout -- to do it right now.)
<mirai>apteryx: good news! regarding #60788, I managed to make a qemu VM see traffic
<helpful-friend>"[PATCH] services: Add vnstat-service-type."
<mirai>bad news: it takes approx 5 minutes for it to start picking up
<mirai>so a test suite is possible
<mirai>but very slow
<mirai>I never thought I'd find a use for the 'discard' service in inetd
<jlicht>winter: you can use package-input-rewriting, but this does imply you won't have substitutes
<apteryx>mirai: is it just because of its default config (5 minutes window)
<helpful-friend>"Defining Package Variants (GNU Guix Reference Manual)"
<apteryx>if so, we could make it very short (e.g., 10 or 20 s)
<winter>jlicht: sure, but wouldn't that require me having (in)direct access to the package used in a service?
<winter>like, some way to specify what package a given service uses for <whatever>
<mirai>I can try turning some knobs
<jlicht>In most cases, not having control over the package used by a service would be a "bug"
<jlicht>well, rather open invitation to contribute rather :-)
<winter>fair, but i'd say littering package-input-rewriting everywhere isn't ideal as well ;)
<winter>but i digress
<winter>hmm, i do wonder
<winter>i could just rewrite the inputs to a graft if i just care about runtime, right?
<winter>therefore, no unnecessary building
<jpoiret>adding a graft requires you to have access to the original package definition
<jpoiret>so it's as effective as just changing the original package directly
<jlicht>add to that, grafts are supposed to be compatible for some definition of the term with the original package
<winter>in this case, it is
<mirai>apteryx: ah, no, it's because 5min is the smallest unit
<winter>jpoiret: right, but i'll avoid having to rebuild every node of the tree, since it'll just graft, right?
<mirai>that it can record
<winter>(i don't see how your message relates to what i said, was i saying something wrong jpoiret?)
<jpoiret>winter: what I meant was that if changing the original package definition is already an option for you, package transforms are useless, right?
<apteryx>I think the default fail2ban sshd jail only looks at /var/log/secure
<apteryx>while sshd spam attempts appear in /var/log/messages
<apteryx>is anyone else using the fail2ban service?
<apteryx>mirai: OK! perhaps it could be left out then, 5 minutes is a bit too long to my taste; but it's still very useful to know if the service starts correctly and can be restarted
<mirai>I managed to narrow down to ~2 mins?
<apteryx>ah, that's probably OK!
<mirai>it looks like its faster if I request --xml/--json or oneline output (??)
<mirai>bizarre stuff
<apteryx>we have tests which waits for the system to graphically boot and analyze screenshots with ocr, these are probably up there too
<civodul>apteryx: hi! i think you inadvertently pushed the 'base-for-issue-62196' tag
<civodul>which reminds me i never remember which remote to use to get branches for issues handled by qa.guix
<mirai>is there an analogue of void-port but for "/dev/zero" ?
<civodul>jpoiret: hi! do you know what the deal is with 'core-updates-frozen'?
<civodul>or anyone else :-)
<apteryx>civodul: oh, pushed to guix's repo?
<ekaitz>hi! how can I use mysql in a guix shell? it doesn't let me connect!
<ekaitz>and also, where is it going to store data? in my home folder?
<cbaines>civodul, it's
<abrenon>ekaitz: but the service is running on your system ?
<ekaitz>abrenon: that's what I want to know: guix shell mysql and then what?
<ekaitz>I need to run mysqld with a --datadir and a --socket... and what else? anyone has a good recipe of this?
<minima>sarg: i was able to build an image that uses my guix home configuration; i did it from a local guix checkout where i manually applied abcdw's changes (this home service ); thanks for helping with this, you put me on the right track
<abrenon>ekaitz: I'm afraid the best recipe is… the service itself
<abrenon>I've done pretty much the same this weekend trying to run nginx from a guix shell: I ended up having to contradict the whole thing changing all the carefully chosen values in the config in order to get one that would just run from my shell
<abrenon>eventually I spawned a VM with the service, had it running in a couple minutes and lived happily ever after
<jpoiret>civodul: wdym what's the deal with c-u-f? I just don't think it's been used yet
<sarg>minima, that's great. Though the solution is a bit rough still. These profiles are not registered as gc roots and not symlinked in /var/guix properly
<sarg>btw, you don't need a local checkout, you could just make an `image.scm` file with the Andrew's code, which should return an <image>
<minima>sarg: oh yes, true, i'd even seen david wilson doing that here - i don't know why i got fixated in using a modified checkout...
<ekaitz>abrenon: yeah that makes sense... I'll give that a go, thanks a lot
<mirai>ah, of course my test never worked
<mirai>hostfwd=tcp:: hardcoded for port-forwardings
<mirai>apteryx: IMO the spacing between the fields should have been kept
<mirai>it makes things easier to read
<mirai>it's a nightmare if the records grow very large
<mirai>the substitute servers are glacial
<mirai>apteryx: no dice, spawn doesn't work in guix (yet?), the test will have to be sent later
<civodul>jpoiret: ah ok; i just noticed that branch and wasn't sure what the status was
<oriansj>mirai: well building everything takes time
<jpoiret>i think it's from the last c-u merge
<jpoiret>does it even still exist though? I don't see it on cgit
<jpoiret>civodul: wrt. the freestanding bug, I think it also affects native compilation
<jpoiret>oh, but I guess if the other configure flags are under a (if (target-hurd? target) ...) it should be fine
<jpoiret>I'm still quite new at autoconf hacking :p would not know how to bypass these tests myself
<civodul>jpoiret: yeah, making minimal changes to the build system to achieve some effect is sort of an art :-)
<civodul>the libstd++ one is harder though, i don't have a good idea
<rekado>mirai: the substitute servers are fast. Connection is good. Peering is not.
<jpoiret>heh, I don't think there is a clean solution to this
<poselyqualityles>For configure flags, how difficult is it to search for a specific includes path in the inputs list and reference it in the #:configure-flags argument? e.g. `(#:configure-flags (list (string-append "--with-some-lib-includespath=" (my/find-in-input <some way to reference inputs> <input> <path>))))
<civodul>jpoiret: yeah, that's unfortunate; it's a case where equivalence classes among same hash-modulo-refs outputs would help a lot
<jpoiret>or ways to let origins be thunked! and get memoized per-target maybe
<jpoiret>but I'm not too sure how compilation helps, if it turns out compiled origins vs. thunked origins really matters performance-wise
<civodul>thunking 'patches' would lead to a stat storm
<civodul>so we'd need some extra work elsewhere to counter that...
<jpoiret>why stat storm?
<civodul>because we might repeat those search-patch calls many times
<jpoiret>not if they're memoized per-target!
<civodul>well, OTOH origin lowering is cached, so maybe that'd worked
<civodul>they're already memoized per-target, in the object cache in (guix gexp)
<civodul>so maybe that's okay after all
<jpoiret>oh, then that's great :)
<civodul>it needs careful checks though
<jpoiret>also, isn't search-patches compilation-time?
<jpoiret>it could be
<civodul>it's not compilation-time
<civodul>Maxime and i played with a bunch of ideas on this theme a while back
<jpoiret>by the way, I don't know if you've seen the new import cycle issue on master (that's also on core-updates now). I'm not sure what to do here, my best guess would be to move pre-commit to python-xyz, or maybe autoload it
<jpoiret>`guix show git` will trigger it :)
<civodul>argh, that should trigger the emergency alert
<civodul>ACTION goes afk for a bit
<apteryx>mirai: I was on the fence about it, but keeping the fields together in the same record appears to be the more conventional style in the code base
<apteryx>(together as in without blank lines in-between)
<apteryx>musescore wants to download 'sound plugins' from its hub; is this free software?
<jpoiret>musescore seems to be GPLv2 so yes they have to be GPL-compatible
<apteryx>perhaps they are the copyright owner, in that case they can do whatever
<apteryx>and they do:
<apteryx>I mean, they are
<platoxia>'guix system reconfigure' fails to boot for me for 6.2.x series kernels. It seems to be failing when the video driver loads.
<platoxia>^This is on Guix System without any modifications on my part.
<jpoiret>apteryx: wdym by that? I don't think the musescore CLA influences the license restrictions imposed upon plugins?
<jpoiret>platoxia: do you have any relevant kernel messages? /var/log/messages should contain prior boot messages as well
<jlicht>jpoiret: why would a GPLv2 condition apply to them, if they don't have to license the core under GPLv2 themselves?
<jlicht>(to themselves through the CLA, not to us mortals)
<apteryx>jpoiret: I meant that the plugins could be non-free although they link to musescore, as the Muse Group owns the whole copyright (via their CLA) on the plugins they hosts
<apteryx>the issue is the musescore we distribute points users to it
<apteryx>and after some research what they call 'Muse Sounds' is indeed non-free and a plugin
<jpoiret>IANAL but doesn't the GPL force things linking to a GPL program to also be GPL-compatible?
<jpoiret>isn't that the whole point of the LGPL?
<jpoiret>to avoid the above "issue"
<apteryx>jpoiret: yes, but if you own the GPL code you can relicense it as proprietary or whatever
<apteryx>you can't sue yourself
<apteryx>I've opened #62391 to remember
<jpoiret>right, but as long as the code is GPL, all addons should also be GPL-compatible
<jpoiret>does that mean we should also not package any project with a CLA that would allow it to relicense?
<apteryx>no, that's not the point of #62391, the issue is an FSDG one
<jpoiret>that was a genuine question though
<jpoiret>i'm not sure I have any opinion on it
<apteryx>jpoiret: I don't think it matters, GPL + CLA remains free software
<platoxia>jpoiret: Yes, it seems like an elogind problem with gdm: elogind[386]: New session c1 of user gdm. gdm-session-worker: GLib-GObject: g_object_unref: assertion 'G_IS_OBJECT (object)' failed. gdm: Gdm: GdmDisplay: Session never registered, failing. elogind[386]: Removed session c1...It loops through this until it says: gdm: Gdm: GdmLocalDisplayFactory: maximum number of X display failures reached: check X server log for errors.
<platoxia>jpoiret: To be clear, it loops through attempting to create new sessions; c1, c2, ... until the final error about maximum number of X display failures reached.
<lilyp>jpoiret: the GPL does not apply to the author – that's a legal loophole in any system
<lilyp>so if you are the author of a program and it uses no other GPL parts, you can add proprietary stuff on top and no one can stop you
<lilyp>if it does use something like readline for example, then you are liable for violating the license of readline, but not the license of the thing you yourself authored
<lilyp>this is in any case something we colloquially refer to as "a dick move", but again a legal one
<lilyp>as is making a hitherto GPL'd program proprietary: again, nothing to stop you from doing so, but you still may still have to provide sources for older versions that were GPL and you better bet that if it's a popular program people will fork it and you won't see that sweet profit
<jpoiret>lilyp: never thought of that. doesn't seem avoidable though, unless the author relinquishes their copyright (which might not even be possible in some countries)
<jpoiret>platoxia: eh, looks like the GDM display is failing to start
<jpoiret>do you have any kernel messages indicating an error for some device?
<jpoiret>could be in /var/log/debug
<ArneBab>I’m currently testing --tune and enabling that for many more packages — focussing on those where Clear Linux has an advantage over Ubuntu in
<ArneBab>I created a guile-3.0-latest/tunable package, but the r7rs benchmarks to not tell a clear story yet … I may have to hack them to actually give errorbounds.
<mirai>apteryx: I'm planning to gradually shift the define-configurations to have a space between fields
<mirai>by gradually = as I go over them or add new services
<platoxia>jpoiret: nothing of note in the debug log. Here is the gdm greeter log, ... Also, Xorg.0.log shows no errors or warnings.
<mirai>platoxia: looks like a graphics card problem?
<mirai>what device is it
<platoxia>mirai: it works fine when I boot Guix System with a previous kernel <6.2
<platoxia>mirai: it is an GeForce GTX 960
<sarg>after half an hour of yak shaving I now have autologin to EXWM on tty1 and the first command in .xsession is to lock the screen. With this setup I don't have to wait 5 seconds while EXWM boots after entering the password
<sarg>(to be fair, the long startup is not exwm's fault, it's doom and its batteries, 320 packages or so)
<apteryx>mirai: it should be discussed first to guix-devel :-)
<ichernyshovvv[m]>hi,guys, is there a way in grub-bootloader to specify kernel parameters for automatically generated guix generations entries? i want to specify resume=/dev/sda8 to be able to load last session I saved via hibernation
<lilyp>ichernyshovvv[m]: I use the following:
<lilyp>it may or may not work for you – I don't make any guarantees that it will work if you reconfigure in between
<ichernyshovvv[m]>lilyp: thank you!
<jpoiret>sarg: can't do `guix system image -t efi-raw gnu/system/examples/bare-bones.tmpl` with your new patch, still failing with "/gnu/store/70v5bf0ns2sd5z477dcai6205y36d36w-grub-2.06/sbin/grub-bios-setup: error: cannot find a device for . (is /dev mounted?)."
<jpoiret>platoxia: there were some changes in nouveau for 6.2 I've read somewhere
<sarg>jpoiret but shouldn't efi-raw imply `grub-efi-bootloader`?
<jpoiret>oops, that's right :)
<apteryx>platoxia: hi! have you seen #59361? we may be on the same boat
<apteryx>lechner: where is helpful-friend? I already miss them.
<platoxia>apteryx: jpoiret: Yes, it seems to be the same problem as bug #59361. I'll add the latest kernel 5.1.x to my config and see how it goes.
<mekeor[m]>hello. how do i locally check out the "node-18-updates" branch of guix? "git branch -r" only lists the master branch for me.
<mekeor[m]>i don't understand what's wrong with my local clone of the guix-repository.
<sarg>`syncthing-shepherd-service` could be used with slight modifications in `home-environment`. I wonder if guix already has examples of shepherd services suitable both for system and for home usage.
<apteryx>platoxia: was it working with any linux 6.* ?
<sarg>mekeor[m] do you have the right origin?
<apteryx>for me it fails straight from a commit between 5.19 and 6.0, as shown in this git bisect here:
<mekeor[m]>sarg: "git remote -v" lists as "origin"
<mekeor[m]>ACTION tries to clone guix into a new local repository
<sarg>mekeor[m] dunno. If it helps you - here is my .git/config
<unmatched-paren>hello, guix :)
<mekeor[m]>hi ( :)
<unmatched-paren>feedback wanted for :)
<mekeor[m]>sarg: thanks! it seems to help. i replaced "master" with "*" in my line "fetch = +refs/heads/master:refs/remotes/origin/master"
<mekeor[m]>sarg: does "git branch -r" list all remote branches for you? or only those you have ever fetched?
<sarg>all of them
<platoxia>apteryx: Yes, it works fine prior to the 6.2.x kernels.
<ArneBab>rekado: I sent two patches as guix bugs to add tunable? . #t to more packages. I thought you might be interested.
<jimtaylor>Well, that's odd. A minute ago I was getting SSL errors trying to run a "guix pull" on any of my machines, but it's fine as of about a minute ago. Any idea what happened?
<jimtaylor>Erm, swap that first "minute ago" with "a few minutes ago"
<morganw>jimtaylor: it can happen if the system time is wrong, and then the ntp service fixes the time in the background (i.e. your CMOS battery might be flat)
<sarg>jimtaylor: it looks that is down
<jimtaylor>So it does. Hope it's nothing too serious.
<mekeor[m]>i was thinking i screwed up my networking. i always "ping -c 3" to check my networking :D
<jimtaylor>Seems to be mostly back-in-business now.
<evilsetg[m]>I am getting a strange error when trying to use guix package --export-manifest. The file gnu/packages/python-xyz.scm throws undefined variable error over the pre-commit package. However this package is correctly imported from gnu/packages/version-control.csm if I check in my '~/.config/guix/current/share/guile/site/3.0/ directory'. Running guix pull does not fix this. Does anybody else have this problem or knows how I could fix it?