<nij>Hello! I have a manifest in the repl. How do I install it in a specific profile (in that repl)?
<nij>Or I should always save that to a file, and invoke `$ guix install -f ...` from the cli?
<nckx>nij: Untested, but you can try importing (gnu scripts package) and using build-and-use-profile. Or some other procedure if that's not the right one. But, honestly, I know I'd go for a quick & lazy (system ...) instead.
<jgart[m]>Does anybody happen to know a particular example where a dependency is needed as both a propagated-input and a native-input? What would you do in that case? Duplicate it?
<jgart[m]>The case where I need this is python-abjad. It requires lilypond for the tests to pass but it also needs lilypond as a propagated input since the abjad api directly calls the lilypond binary from a method in order to compile a pdf
<nipcat>This does not work, though. I removed the gdm-service-type but it still starts up. I then tried to specify my own %desktop-services, but when I try to use the avahi-service-type in my own config it says "guix system: error: service 'avahi-daemon' requires 'networking', which is not provided by any service"
<lfam>nipcat: Regarding the "requires networking" thing, I would add (service dhcp-client-service-type), which will provide networking for you. Plus, you probably want to use a DHCP client to get IP addresses assigned automatically
<nipcat>lfam: error about something missing ... I basically simply copy pasted all the services defined in desctop.scm as %desktop-services ... I don't understand why that shouldnt work
<lfam>Hm, that's not what your paste showed. Was that in another revision of your config.scm?
<nipcat>(service ntp-service-type) does not work either
<lfam>What do you mean it doesn't work? Can you be more specific?
<iyzsong-w>re 'xinit', there is a 'xorg-server-service-type', which should work out-of-box with 'sx'
<nipcat>if I add (service ntp-service-type) I get "guix system: error: service 'ntpd' requires 'networking', which is not provided by any service"
<jgart[m]>lfam: The issue is that I would have to duplicate lilypond in the propagated inputs as well as the native inputs since lilypond is also required for tests to pass
<nipcat>lfam: of course i changed that trying to minimize my config to hunt for that bug
<jgart[m]>lfam: Should I just put lilypond in both, native-inputs and propagated-inputs?
<nipcat>lfam: but trust me, I did try that first .. simply defined my own %desktop-services exactly like in desktop.scm
<lfam>jgart[m]: What I'm saying is that you don't have to do that. Just use propagated-inputs, and the right thing will happen with respect to native compilation when cross-compiling (which is what native-inputs is about)
<lfam>jgart[m]: Propagated inputs are available in the build environment
<nipcat>ok adding dhcp-client-service fixed that networking dependancy .. although I still don't understand how .. because it's not defined by the original %desktop-services either
<lfam>nipcat: If you simply copied and pasted %desktop-services from it is defined, without any other code, I would not expect it to work. You would also need to import the variety of modules, and perhaps procedures defined in that file
<lfam>Those modules are imported at the top of the file, with #:use-module
<nipcat>lfam: of course, I imported the modules accordingly
<nipcat>lfam: otherwise scheme would had complained about unbound variables, not about servicse that are not provided :)
<lfam>I suppose that %desktop-services handle DHCP via network-manager-service-type
<lfam>By the way, if you install xdot (from the xdot package), you can do this: `guix system shepherd-graph config.scm | xdot -`
<lfam>It shows the services of your config.scm in a graph
<nipcat>yeah i just thought about that .. it makes sense now!
<lfam>Well, let us know if you have more questions. I wish I could answer your original question (how to use xinit)
<lfam>I really like the startx / xinit workflow: boot to console, start X by hand if wanted
<nipcat>just one more mystery for me would be why the remove gdm-service-type call wont work
<lfam>That's unexpected to me, but the graph tool might help unravel the mystery
<lfam>You could even use it before reconfiguring, just to be sure it's how you want it
<nipcat>yeah i used to do startx with arch .. i dont really mind gdm, but I kinda wanna define all my emacs packages in a manifest file for my user profile .. which is not possible when i use exwm via display manager
<nipcat>it seems to expect all the emacs packages in the system profile
<lfam>Hm, there must be a way to make it work :) Hopefully something with experience with exwm can chime in
<jgart[m]>lfam: Thank you! I rebuilt successfully with your suggestion. I'll send a patch soon with the package update.
<Oxddc>I am torn with using Guix any longer. I dislike how it has become political. I dislike how they dont want to be apart of the FSF GNU and just using the GNU name and brand as "Gathering under a New Umbrella." I dislike how they say its for inclusivity where FSF GNU shows no signs that they are not inclusive. Why am I typing this here? Because its the GNU head devs how are always at the tops of these lists. What are their motives? It
<Oxddc>seems pretty irrational and ridiculous. I wish I could use Gathering under a New Umbrella Guix without having to worry about what the head devs politcal motives are, but it is very hard not to. Especially when contributing to packages and other ways. It is all very sad. I like the Guix ecosystem, but these actions do not sit well with me to want to support the project furthur. It seems like its all half truths.
<Frosku>Yes, which is concerning. If this is i.e. a 'union' for gnu developers, that's a good thing in my view, if this is splitting gnu between 'Stallman GNU' and 'Not Stallman GNU' it seems like a horrible idea.
<rekado>but I really suggest taking the discussion off #guix; I don’t like to dominate a channel with a discussion that is really at best a tangent for Guix
<rekado>I heard that before, but the truth of the matter is: it’s a decision the assembly will have to make if the assembly members have a problem with the CoC as is. That’s perhaps a little circular, but … you know, such is bootstrapping ¯\_(ツ)_/¯
<manumanumanu>Hello! I just installed guix using the script, but now I try to run guix as my user, I get access denied to my profile.lock: guix install: error: open-file: Access denied: "/var/guix/profiles/per-user/manumanumanu/guix-profile.lock"
<raghavgururajan>nckx: Hey o/ I was wondering if you could setup .authorization on c-u as well?
<mdevos>of inheems voer, dat is wat korter dan inboorlingenvoer
<nckx>I feel like we need to find a metaphor for cross-compilation (‘kruiscompilatie’ zegt het Web) that fits better. All I can find are articles talking about het hostsysteem and de targetcompiler :-/
<mdevos>'herkauwd voer' klinkt als iets dat je niet wil. Dat is de juiste connotatie voor propagated-inputs denk ik
<mdevos>hash -> hachee, of controlecijfer? Tweede is eigenlijk niet correct
<nckx>mdevos: I've punted on ‘store’ so far (there's an untranslated ‘store’ somewhere that needs to be fixed). I think I've used ‘magazijn’ once. I considered depot but thought it was too Belgian (you'd know).
<mdevos> In onderzoek van het Centrum voor Leesonderzoek uit 2013 werd "depot" herkend door:
<mdevos>Volgens Frans: ‘projecteer <un péripherique> met ...’
<mdevos>projecteren klinkt passend, maar toch ook weer niet duidelijk ...
<nckx>Inderdaad wiskundig accuraat doch eerder obscuur.
<PotentialUser-65>Hi all. An executable is missing libstdc++.so.6 and having installed firstname.lastname@example.org I don't see it either under .guix-profile/lib but `locate` does show it somwhere in /gnu/store ... what am I missing?
<pkill9> PotentialUser-65 is this an executable you downloaded?
<pkill9>because if so guix doens'tby defualt support running those executables
<nckx>So on x86, --target=aarch64-linux-gnu will use an x86 GCC binary that creates an aarch64 output. --system=aarch64-linux will run an aarch64 GCC binary, using transparent qemu-binfmt emulation, that creates an aarch64 output.
<nij>Hello! I'm setting up a foreign input method (chinese, japanese, korean) on my guix system. Currently, it works well for my terminal and qutebrowser, but it while I type through them in emacs and icecat, it appears that they are not existing at all.. namely what show up are plain english characters. This could be a tough problem to resolve, but I hope I can get some hint here.
<nij>Hmm.. I set them in a terminal, and launched emacs in that terminal.. but it still doesn't lemme type in CJK.
<raghavgururajan>nckx: Let's say my local copy of c-u is 9 commits ahead of remote c-u. If I `git rebase HEAD~9 -i` and edit on 5th commit, and then if I push; only the first 5 commits will be pushed right?
<roptat>nij, I think you need that in the whole environment, not just a terminal
<roptat>but, even though I have those variables, it doesn't work for me either (well, ibus is detected, I can see it change the behavior of typing dead keys like ^, but it doesn't find anthy, which might be due to stray cache refering to non existing store paths)
<nckx>raghavgururajan: I have to leave again but in general I'd say: in such a case, be completely explicit in what you push (git push origin HEAD:core-updates, git push origin HEAD~5:core-updates, or ever git push origin <commit>:core-updates), don't rely on what you think should happened if you kept track of all state perfectly. Git is stupid; don't be too clever.
<nckx>mdevos: If you find one let me know, I've used ‘commit’ so far. Mainly because it's extremely common in all Dutch git tutorials/docs.
<nckx>mdevos, jlicht: Suggestions for ‘install the package that the code within FILE evaluates to’ (holy split infinitives, Batgirl)? ‘waartoe … evalueert’ is hella meh, and ‘beschreven door’ feels like cutting one corner too many.
<nckx>efraim: It just depends on what you push, at the risk of being tautological.
<nckx>Pushing HEAD:core-updates or core-updates:core-updates differ.
<nojr>another question: do I need to specify JAVA_HOME when installing openjdk? which would be the path for it via guix package -i ?
<mdevos>En dan is nog de vraag als je verschillende versies van een pakket als htzelfde pakket beschouwd ... Is "hello" op core-updates hetzelfde pakket als "hello" op master? Kunnen we wel van een "hello" pakket spreken op core-updates wanneer core-updates weer eens kapot is?
<nckx>I really really recommend being explicit, not relying on ‘so this should do that by default, right?’.
<nckx>mdevos: There's a string in English somwhere like ‘can' find package X for version Y’ which seems to acknowledge your point, but this is all quite esoteric.
<efraim>raghavgururajan: nckx suggested using 'git push origin HEAD:core-updates', I would either create a temporary branch to push just the commits I wanted or create patches of the ones I didn't want to send, reset HARD to the commit I want to push, push those and then reapply the patches
<efraim>(or find the previous HEAD using git reflog)
<nckx>raghavgururajan: HEAD will be where you stopped, HEAD~N will go back N commits *from there*. Again, danger, just put away the tightrope.
<efraim>so I guess 'git checkout -b cu-to-push; git reset --hard HEAD~6; git push origin HEAD:core-updates; git checkout core-updates' is what I'd probably do
<nckx>raghavgururajan: You can always ‘git show HEAD’ and ‘git show core-updates’ to see the facts, no need to guess.
<nckx>It will make it clear what's going on more than we can
<nckx>Well, ‘sounds good’ is always relative in these cases.
<nckx>‘Sounds utterly alien because computers speak English’.
<apteryx>is the operating system PATH set correctly in a marionette test OS? (system* "pgrep" ...) suggests not
<apteryx>ah, peeking at (gnu tests ganeti) I see (setenv "PATH" "/run/current-system/profile/bin") was used
<solstag>Ni! Hey folks, I've successfully installed guix in a vm, now I'm trying to generate an image from within it, but `guix system image install.scm` yields `guix system: error: failed to load 'install.scm': Permission denied`. Any ideas? The file is there and its mode is 644. Sorry to bother.
<darth-cheney>Hello all, guix newbie here. I have a question about environments
<darth-cheney>I regularly use python venv and nodeenv as part of my dev workflow, but I use this liberally
<mdevos>when I hack on the project, I do "guix environment -l guix.scm"
<mothacehe>solstag: hey! what's the origin of the install.scm file you are trying to use?
<mdevos>the first time I run it (or when packages are updated), it takes some time to compile everything
<mdevos>but the second, third ... times things are fast enough for me
<pkill9>you can also use direnv to automatically enter a guix environment when you enter the directory
<darth-cheney>I guess the part that is confusing for me is what is the result of the guix environment? Is it a local dir that I can source like venv? How can I flip back and forth between environments?
<mdevos>guix environment starts a shell that has environment variables set
<mdevos>the environment variables are such that the packages you requested are visible
<pkill9>and it saves the guix environment, so it will always be quick. i.e. won't build/download updates if you update guix
<solstag>mothacehe: well, I copied it from /gnu/store/.../install.scm
<darth-cheney>mdevos is there a way to quickly source the env myself? I'm asking because I use ansi-term in emacs and I have custom setup stuff that happens there. I like to have only one term buffer open
<mdevos>darth-cheney: see the "-p" option of "guix install"
<mdevos>(that's not a use case I often have though, so I can't help you much there)
<darth-cheney>ok thanks everyone, that's probably enough info to play around for a bit and see what I can do
<nckx>The common phrase ‘externe host’ is also meh.
<gr0n>hi. i'm running guix on top of trisquel and am trying to install texlive with it, but am hitting an issue with latexindent (Can't locate LatexIndent/Document.pm in @INC (you may need to install the LatexIndent::Document module))
<solstag>mothacehe: sorry, it was my fault all along; my modifications were causing the error when ran; there was never a problem accessing install.scm; thanks a lot and sorry for the bother; at least your suggestions did lead me to figure it out. cheers!
<mdevos>maar het is wel in de xtc-koffieshop ... ik zal andere bronnen moeten raadplegen
<cdegroot>Question on the erlang package - it has "glu" and "mesa" as propagated-inputs, which means it pulls in everything and the kitchen sink (I even saw cups pop by...). I'm not even sure it is needed to run wxWindows (which probably caused this), but given that not many people actually use wxWindows, is it an idea to have an "erlang-wx" package that just mixes erlang and these two and do the regular package without these deps?
<cdegroot>Moet ik mijn zus vragen? Die is lerares Nederlands :)
<lispmacs[work]>I'm living pretty much paycheck-to-paycheck, though, so I can hardly afford to keep my 6 year old Gigabyte motherboard running (replacing the fans and such)
<cdegroot>```Engelse ww worden opgenomen in het Nederlandse klanksysteem. Dus roetsjte, geroetsjt, [kesjte], [gekesjt]. De ik-vorm blijft echter gehandhaafd: cache``` zegt ze.
<lle-bout[m]>I suggest OpenPOWER for desktop/workstation and RISC-V for laptop when it becomes available, though chips with less power consumption based on OpenPOWER might also appear though not the focus of IBM now it seems (they're the main vendor of chips for now)
<terpri>as they say, lots and lots of cash is the price of liberty (one reason i'm not a huge fan of the hardline anti-blob policy though i understand the reasoning behind it)
<lle-bout[m]>nij: lispmacs[work] : I think obsessing on freedom with hardware when you can't afford it is not worthwhile, better get some refurbished hardware with proprietary parts and reasonable performance and wait for an affordable option, that's my personal opinion, I have a Dell XPS 13 2020, I just wish better options freedom-wise appear soon, but I can't take in so much frustration, I also need the performance-per-watt for my
<cdegroot>lle-bout[m]: for pretty much the same reasons I got a T5810 a new lease on live last year :)
<lle-bout[m]>There's so much proprietary hardware available second+ hand that works and it would be unreasonable for me to just trash that for freedom
<mdevos>Now, if only $VENDORS would always publish the firmware as free software, and give everyone a mechanism for replacing the firmware with your own copy!
<lle-bout[m]><mdevos "Now, if only $VENDORS would alwa"> Yes, so the electronical waste would be even lower since users could make communities to maintain their own hardware software-wise forever if they needed to.
<terpri>there's some world wherein the anti-blob policy would increase overall software freedom (pressuring vendors, motivating reverse-engineering etc.) but i don't think we live in such a world atm
<lfam>It's a situation where "the perfect is the enemy of the good", IMO
<gr0n>even though i run fully free trisquel, i'd rather see someone use ubuntu on their system than see them use windows or macos
<lle-bout[m]><lfam "We don't want to push away peopl"> I wouldnt want to, but I also don't want to spend time *supporting* proprietary stuff, but that work's mostly done with Linux vanilla these days so..
<lfam>Another thing to consider is that the foundations of GNU were created on hardware that would never meet the RYF standards
<nij>:-((( after three hours I'm still trying to set up a minimal working personal channel.. `guix pull` kept failing.. i wonder if someone could help out :-)
<gr0n>lle-bout[m]: it is pretty common to see it happen, but it's the cracks in "convenience" start to show when you realize that the thing you want to do on some system doesn't work because the company doesn't want you to do it
<lle-bout[m]>Let's be honest since I got my x86 laptop, I barely use my OpenPOWER desktop, because transportation is actually really important for me.
<gr0n>so said convenience quickly turns to frustration.
<gr0n>even though i have the option to use a perfectly functional MBP, i still prefer my x200 because the MBP just doesn't let me do things i want to do
<gr0n>btw, do you guys know if guix works on non-GNU/Linux things, such as the BSDs? we have a very large deployment of FreeBSD for research purposes at a university and we currently build everything ourselves when it comes to packages using poudriere (typical FreeBSD way to do it). guix would be really helpful to use when it comes to maintaining different versions of things on user accounts, and not requiring a
<cdegroot>I just installed Guix on Debian/arm64, did I get the "this is the version we want you to test with" installer?
<cdegroot>Ik heb me altijd laten vertellen dat onze zuiderburen meer gesteld zijn op de Nederlandse taal en beter hun best doen het ding in stand te houden dan de Hollanders, die met plezier Duits met Engels mengen en dat dan Nederlands noemen ;-)
<jlicht>De laatste 'vergadering' die ik had staat me niet meer bij, maar 'meetings' heb ik elke dag
<mdevos>Here's an untranslatable fluff word: Provisioning
<mdevos>‘Provisioning for machines that are accessible over SSH and have a ....’
<cdegroot>(rent naar zijn boekenplank met woordenboeken...)
<mdevos>ik heb een vertaalwoordenboek Engels->Frans en Frans->Nederlands
<gr0n>i wonder if this is because of the directory structure guix sets up, specifically ~/.guix-profile/share/texmf-dist/scripts/latexindent, so we maybe need to set an additional -I in the latexindent invocation?
<roptat>gr0n, maybe... I'd like to understand the lib part in the code above, to understand what it does and why it's supposed to ensure @INC contains latexindent
<Noisytoot>sneek: tell roptat because lfam ignored me, so if I don't use sneek, lfam won't see the message (and it's unrelated to the reason that he ignored me)
<sneek>roptat, Noisytoot says: because lfam ignored me, so if I don't use sneek, lfam won't see the message (and it's unrelated to the reason that he ignored me)
<gr0n>roptat: yeah, i haven't looked into it in too much detail. i assume it uses Document later on, assuming it's in the @INC path instead of trying to guess where it should be based on the texlive installation
<gr0n>but i'm not sure if that's the actual reason.
<jeko>Broad question : how do you manage your emacs dotfiles for your Guix System ? I mean, do you have it somewhere defined into your system definition ? or do you git pull it when your system is ready to operate?
<roptat>gr0n, I'll need time to download the substitute :p
<gr0n>roptat: in that case, i can try to look at each include path and try to see if the right thing is there
<roptat>/gnu/store/jr9czrjx7slnda355xfmbv6hkkdpl5qi-texlive-bin-20190410/share/texmf-dist/scripts/latexindent sounds the most promising
<gr0n>roptat: the only thing in there is latexindent.pl (the actual script). however, there is no directory
<roptat>so where does your ~/.guix-profile/share/texmf-dist/scripts/latexindent come from?
<cdegroot>is it me or is the substitute coverage for ARM lacking? My poor RPi4 is running hot compiling all sorts of stuff. Not a biggie, if it melts I'll return it as DOA, it's new :P, but wanted to know whether that's expected.
<lfam>cdegroot: I'm not surprised it's lacking. We recognize a lack of build capacity for 64-bit ARM as a problem, and we are taking steps now to improve it
<lfam>It's hard because the market for capable hardware is still in its infancy
<lfam>But, there are some options, and we are trying to acquire them
<Telc[m]>next license missing :) " Apache-2.0 WITH LLVM-exception"
<cdegroot>lfam: ok, if it's expected, i'm good. I wish I could contribute. Want to have my spare RPi B? :P
<lfam>I think the Pis won't make great build farm machines. We need hardware that is designed for high-performance (server chips), rather than low power consumption (mobile chips)
<cbaines>and the "experiment" I'm running with guix.cbaines.net for armhf-linux and aarch64-linux builds is going pretty well, armhf-linux builds are still happening, so I'm waiting to see how far it gets, but I'll email guix-devel when I have a little more data
<mihi>Hello :) My quest on Guix Hurd continues. Last time "guix pull" inside died at some point (after building openssl) as there were no CA certificates bundled, so this time I'd try to give it some CA certificates first. I could copy some from my (Debian) Linux system, but was wondering what the canonical way is to get them on a Guix system. Some package to install?
<lfam>Then, I would do `guix build nss-certs`, and then manually set up the environment variables mentioned in that manual section that I linked to
<mihi>but what works is to pass a VISO to the Hurd VM as /cdrom, mount it, drop into kernel debugger (apparently there is a breakpoint in cdrom driver somewhere), exit kernel debugger with "cont" and copy the files from there into the system :)
<nckx>raghavgururajan: They've been around in #guix a lot recently, and I didn't expect this behaviour from them either. But it happens.
<nckx>‘authenticated to services’, to use the formal term :)
<lfam>Guix development is based on mutual respect. I set my client to ignore that user, and asked them to stop trying to contact me. Instead of dropping the issue, they used sneek to speak to me anyways
<lfam>They were warned to stop doing that, and they did not stop
<lfam>mihi: We are planning a new release (1.2.1) on Sunday. Hopefully it will be easier to start with that. We could even run some tests now, before the release, if you could give us advice about what to try
<lle-bout>lfam: hey, you saw xorg security releases recently?
<mihi>somebody even forgot to link the MAKEDEV script somewhere outside of the store, so my way to make devices for my CDROM is 'find the MAKEDEV script in store and symlink it manually'
<lfam>This assumes you are working from the 1.2.0 release. If you are working from the current master branch (ie with `guix pull`), then your tests are the same as what we'd do
<lfam>You can find that out by clicking the "Evaluation" number (21564), where it shows which Git commit the build is based on. Then, use `git log` or other tools to check which branch the commit is from
<civodul>apteryx: hi! if you're giving "make release" a try, you can ping me in the coming days if you notice anything wrong
<civodul>or if there are offloading issues or anything
<apteryx>civodul: alright, yeah I test this weekend. Also I've been trying to finishing up some tests for a jami service, I think it's almost complete. We should be able to use this a live bridge soon. One issue I had is attempting to import the json module inside the gexp of a computed file, like so: https://paste.debian.net/1194049/
<nckx>‘Machine’ or ‘computer’ or so; it's not waard puns.
<mihi>mdevos, I saw in the logs that you also tried to join the adventure of Guix Hurd. How far did you get? :)
<mdevos>hier is een lang woord: initrdmoduleveiligheidscontroles
<mdevos>ga er veiligheidscontroles van X en Y van maken
<nij>lfam: I'm installing a fresh guix on my machine using the ISO you posted a few hours ago. So far so good, except it doesn't detect my wired internet (yes I'm using graphical).. which is weird cuz my machine is always on ethernet and it has been working for days.