<ScaredyS1uirrel>I am using "guix system init /mnt/etc/config.scm.test /mnt" from the install iso for x86_64 <ScaredyS1uirrel>and my host name and users and the mount points and swap devices keyboard layout locale and users <leoprikler>then copy /etc/configuration/desktop.scm to /mnt/etc/config.scm first and edit that <leoprikler>that should be simpler than writing out everything <ScaredyS1uirrel>and that's my prompt and i am running the 1.0.1 amd64 gnu guix distribution install iso thing <leoprikler>okay, then paste /run/current-system/profile/share/guile/site/2.2/gnu.scm <civodul>nckx: note that access logs are scattered across several files <nckx>civodul: I used all *access* files. <civodul>ScaredyS1uirrel: hey, this may be frustrating but this tone is uncalled for <nckx>ScaredyS1uirrel: Would you kindly be kind. <leoprikler>okay, this file indeed has no error as far as I can see <leoprikler>so, let's get back to config.scm, what does that look like? <leoprikler>and your claim is that (use-service-modules <whatever>) fails? <ScaredyS1uirrel>but I could install the wgetpaste package so guix is partially working but not to install the thing <apteryx>maybe the error reporting is just plain bad when attempting to pass your config.scm.test which doesn't return an operating-system record. <apteryx>error reporting is still pretty much an area needing polishing in that area of Guix, so I wouldn't go believing everything it spits out. Better try an actual config that is supposed to work (and complete) at first. <leoprikler>okay, now try adding (use-service-modules <whatever>) to that config.scm <ScaredyS1uirrel>I use that one and: root@gnu ~# guix system init /mnt/etc/config.scm /mnt/ <leoprikler>I'd put service-modules below package-modules, but other than that it seems fine <leoprikler>try uncommenting your services and see whether the parentheses match up <ScaredyS1uirrel>I got that working and now it says that all the packages list has a list of wrong length <leoprikler>I have no idea, why you need those packages (other than openbox and nss-certs) at system level, but so be it. ***Blackbeard[m] is now known as Blackbeard[m]1
***Blackbeard[m]1 is now known as Blackbeard[m]2
***Blackbeard[m]2 is now known as Blackbeard[m]
<leoprikler>okay, that was my fail for not trying to build the system before pasting it <ScaredyS1uirrel>erm what about the (map specification->package+output (list "openbox" "neovim" "irssi")) thing? <leoprikler>I took a wild guess that this was the cause of your issue. <ScaredyS1uirrel>when I took it out the final error is gnome-desktop-service-type is an unbound variable <leoprikler>as far as your packages are concerned, you do need nss-certs at system level, and if you want to use openbox, you need that at system level too <leoprikler>the other packages (neovim etc.) look like they would fit better into a user profile ***Guest93493 is now known as daviid
<ScaredyS1uirrel>the gdm service I'd just use the service removal thingy to remove my display manager <leoprikler>perhaps spawn a terminal at vt7 and use the xinit packageß <leoprikler>perhaps if you install it systemwide like openbox? <nckx>Guix won't even try to startx as a regular user. There's a way to do it as root, but it's been years since I've tested it. <ScaredyS1uirrel>nowadays it just for the xinit command as both about some /gnu/store/yyyy-xinit-1.4.1/bin/X not being found as the X server <nckx>Specifically, (service special-files-service-type `(("/bin/startx" ,(xorg-start-command))) ; used to work 🤷 <nckx>‘Other fine display managers are available.’ <leoprikler>the manual has an example with slim on VTs 7 and 8 <ScaredyS1uirrel>ok I have (use-modules (gnu) (gnu services desktop) (gnu services xorg) (srfi srfi-1)) <leoprikler>you only need (gnu) (srfi srfi-1), because you already have (use-package-modules) and (use-service-modules) <nckx>Oh, they got it working? That was an unexpected happy end 🙂 <ScaredySquirrel>well well it says it is applying 3 grafsts for /gnu/store/p63yyyy-guix-1.0.1.drv and then it says guix system: error: cannot close compressed log file (BZip2 error = -6) <nckx>Why do you get all the weird bugs no one has ever seen before o_o <nckx>ScaredySquirrel: Yep, it's BZ_IO_ERROR. <nckx>Your disc isn't (almost) full or anything fun like that? <nckx>Then I honestly don't know what's wrong or why you keep hitting all these ‘special’ problems :-/ <ScaredySquirrel>I told you that shadow and tails might be touching my computer right now <nckx>I've used Guix for years, never in my life has it shouted ‘-6!’ at me and eaten my slippers. ***jonsger1 is now known as jonsger
***ScaredySquirrel is now known as eyeseyes
***eyeseyes is now known as ScaredySquirrel
***ScaredySquirrel is now known as randomrandom
***randomrandom is now known as ScaredySquirrel
<lispmacs>hi, I was trying to run a certain Java app, but it complain that it cannot load SWT library because there is no "swt-pi-gtk" in the path. Does anybody know what I need to install? <xelxebar>looking at gnu/system/mapped-devices.scm, I see a few form feed characters sitting about on their own line. Is this a convention of some sort? <lfam>xelxebar: Yeah, it is some sort of Scheme convention for delimiting sections. I don't know the details, try asking in #guile or waiting a while here <nckx>lispmacs: Should be simply ‘java-swt’. It provides lib/libswt-pi-gtk-4758.so and more. <nckx>xelxebar: Yep, it's just that — a metaphorical ‘new page’. <lispmacs>nckx: okay, thanks. I thought I did a regex search for something like that but must have missed it <bandali>hey folks, anyone able to add a network connection from nm-applet? <bandali>i get an insufficient privilege error <lispmacs>nckx: hmm, still seem to be getting the same error. maybe some path issue <nckx>It's always been a privileged operation here (‘sudo nmtui’), we're probably missing some polkit fanciness. <nckx>lispmacs: Possible; not a Java fellow, <bandali>i think we should do that; perhaps just use netdev instead of network <nckx>Also not an fd.o fan or fellow, I'm afraid. The thing uses JavaScript configuration files for heaven's sake. <nckx>bandali: Do you use GNOME? <bandali>nckx, no, i'm currently on xfce, soon switching to just exwm. how so? <nckx>bandali: I suspect GNOME users don't have to sudo to connect to new networks or we wouldv'e got bug reports long ago 🙂 I wonder what their /etc/polkit-1/rules.d contains, if anything. <bandali>nckx, right... fwiw, i can connect from nm-applet with no issues, i just can't create a new one <jackhill>nckx: actually, I'm a gnome user who does. I get the gnome polkit authorization dialog, but instead of my password (or not putting in the dialog), I have to enter root's password. <nckx>bandali: Same with nmtui (for me that's a feature, but it's probably not intentional). <bandali>yeah so right now as a work around i just log into root and do that, but it'd be nice to have a proper polkit rule <jackhill>it only happend when adding new networks, not when (auto)connecting, so I guess it didn't annoy me enough <bandali>i'd send a patch if i knew how to add one <jackhill>agreed that fixing the polkit rules would be nice <nckx>jackhill: Not at all, whether you consider it a bug is up to you, I'm just suprised GNOME users haven't complained yet, seems like the kind of thing of which they would not approve 😛 *nckx genuinely surprised, also off to bed o/ *jackhill is happy he was here to pipe up ☺ <apteryx>what do people do to have a Guix VM with a install-enabled guix in it? <apteryx>(e.g., which I can use to test 'guix install', 'guix remove' etc.?) <apteryx>the one provided by 'guix system vm' shares its store with the host, which cripples such use <lispmacs>are there any fics chess clients currently in guix? <lfam>apteryx: You could use vm-image or disk-image with QEMU ***daviid is now known as Guest78798
***Guest78798 is now known as daviid
<g_bor>Any idea why the zabbix-frontend did not start on berlin? <sneek>civodul, you have 1 message. <sneek>civodul, nckx says: The blog post it at the top of Planet GNU and scheme.dk now, in case you were worrying (you weren't). *civodul still debugging http pipelining over TLS <g_bor>do you know why the zabbix -frontend did not start? <g_bor>I did not wnat to touch it yesterday, maybe there is something valuable in the log... <g_bor>Is there something that still needs to be done to get it running? <civodul>rekado_: ah right, i'm running nginx "manually" currently, for debugging purposes <g_bor>civodul: did you find anything out about this tls thing? <g_bor>What I found is that the 408 is intentionally not sent back to the client, so that looks like expected. <leoprikler>sneek later tell ScaredySquirrel: I've experienced your -6 error before, but IIRC it was not Guix' fault. I think I forgot to `herd start cow-store` or my hard drive was already dying. Not sure which of those it was. <roptat>I tried updating OCaml yesterday, but this update is going to be harder than usual <roptat>camlp4 is dead and its latest version doesn't build with ocaml 4.09, so I removed it, and fixed every dependent. I also removed ocaml-bisect, which is obsolete and depended on camlp4 <roptat>and now, after updating ocaml (which was really easy now that they properly use the autotools), a lot of things are failing to build, so I'm updating everything, with new packages, ... <roptat>so I'll do it, but it'll take a bit more than the 10 minutes I expected ^^' <g_bor>civodul: a decrypted dump of a failing connection might shed some light. <g_bor>All in all this seems to be a simple timeout, looks like the server is expecting more client traffic to come in, which never arrives... <g_bor>I have to go now, see you later! <roptat>also, the JRES (a French conference where I'll present Guix) is coming soon, and my train back was canceled :/ <roptat>but I'm ready, so I'm going anyway. I'll find another solution for going back home :) <civodul>JRES is next week, with the Dec. 5th strike right in the middle <roptat>right, so I'll be able to go, but not come back ^^' <civodul>some of us will be in Marrakesh for the R-B summit around that time :-) <roptat>I guess other participants are in the same situation, so I'll try to hitchhike or something :) <civodul>"guix pull" (~= git fetch) doesn't work for me <janneke>i fetch from: git+ssh://git.sv.gnu.org/srv/git/guix <civodul>ah ha! "guix pull" can't do that, though *janneke did not see the `u' in guix pull <roptat>I tend to do guix pull in git repositories, and git build in random locations ^^' ***Francima- is now known as Franciman
<civodul>half a victory is better than a defeat <civodul>ping that bug if you still have problems while retrieving substitute info! <zimoun>roptat: nice this conference. Do you a have bigger slot? And nice associated article! Maybe, it could be converted to an entry for linuxfr.org or something? <zimoun>roptat: if you are blocked in Paris when traveling back, drop me an email :-) <zimoun>ok, cool! it will have a lot of time to present the awesome Guix features ;-) <roptat>an article on linuxfr would be nice, but it'll have to wait after the conference and after my thesis defense on the 11th :p <roptat>yes, if you're in Rennes, you're welcome, although I need to know if you come so they let you in <zimoun>thank you but unfortunately, I cannot be in Rennes the 11th. Good luck! Even if you have already done the hardest part (manuscript), I guess. <civodul>roptat: oh, congrats for the upcoming defense! <civodul>well, usually one says "congrats" afterwards, but we know you've done that hard part already ;-) <civodul>roptat: if you want we can post a paragraph on the blog before your talk, with a link to the article <anon987321>i was trying to test out emacs's xwidget support, but the emacs-xwidget package still returns that emacs hasn't been compiled with xwidget support <stikonas>how long do I usually have to wait till patches on guix-patches are reviewed? <roptat>usually less than a week, but sometimes some patches slip through, so don't hesitate to send a reminder :) <rekado_>anon987321: what did you try? I have built an xwidget Emacs in the past. It needs a few more inputs and an additional configure flag. <Franciman>Hi. Would you suggest using guix on a foreign distro? <roptat>not ideal, but perfectly usable :) <rekado_>Franciman: it’s not ideal (ideal would be Guix System) but I still suggest using it. It depends a bit on what you need from it. <stikonas>rekado_: are you maintaining java packages? <Franciman>hmm I see, thing is, that at this stage, probably it would be better for me to just use guix to create envs in which develop things <stikonas>oh ok... I just had some patches that make OpenJDK bootstrappable from C++... <rekado_>can you send them to the mailing list please? <zimoun>stikonas: which patches are we talking about? <Franciman>one thing I am a bit confused about guix system <Franciman>but when I do guix pull, it's only for my local system <Franciman>but if I update things, they can break if they depend on the global profile <rekado_>‘guix pull’ gets you a fresh version of Guix. <zig>for the current user? <rekado_>you can (and should) use that very same version of Guix to upgrade your packages and reconfigure your system. <rekado_>‘guix pull’ is only a mechanism to get a new version of Guix. <rekado_>it’s independent of your user’s default profile <Franciman>because guix is a global thing. It is known systemwide <rekado_>‘guix pull’ installs to ~/.config/guix/current <rekado_>that’s not global unless you make it this way <Franciman>ok, here is the thing. I have a global system <Franciman>this local guix should be used to update the global system <Franciman>now, the one upgrading the system should be the admin <stikonas>at the moment some precompiled java classes are used in java bootstrap path, presumably by accident <rekado_>stikonas: thanks for this patch. This looks good to me. I’ll apply them later today when my baby lets me. <Mrtn[m]1>civodul: You mean on the IRC-side of things? <Mrtn[m]1>I don't know how it looks .. I don't even know if I can log in as my puppet, without breaking anything? <bandali>(so e.g. netdev users could add network connections through nm-applet or nm-connection-editor) <alextee>does guix accept alpha/beta software upstream? or only stable stuff? <ng0_>Mrtn[m]1: think netsplit, but smaller. <sneek>Welcome back ng0_, you have 3 messages. <sneek>ng0_, ArneBab_ says: however this many times slower might be changed by 2.9 <sneek>ng0_, ArneBab_ says: as far as I understood, (read) got much slower with 2.0 — so much slower that some people couldn’t work with it. <sneek>ng0_, ArneBab_ says: in addition some non-documented tricks used by lilypond stopped to work on 2.0 and some errors appeared multiple times, which cause frustration that led to reduced investment into getting to 2.0. My opinion is that this problem was caused by mistakes on Guiles side (you shouldn’t change things that break the one tool using your language that achieved total domination in its domain). But that’s done now, and I hope it ***ng0_ is now known as ng0
<efraim>sneek! Don't leave us hanging! Tell us Arne's hopes and dreams! <sneek>us, efraim says: Arne's hopes and dreams! <ng0>the bot is really simple <efraim>i should've seen what I walked into there <zig>is there a package definition for guile-fibers that run on guile-3.0 (aka 2.9.5) <civodul>bandali: someone just submitted a patch to do something similar for the "wheel" group, on guix-devel <civodul>so i guess you could take the same approach <bandali>civodul, ah cool, i'll have a look thanks <apteryx>sneek: later tell lfam thanks for the tip about VMs with a working Guix! <anadon>What am I getting wrong? Google has been of no use with this error. <efraim>anadon: you have to import (gnu packages gcc) to use gcc, but gcc is already in the build environment so you can remove that input <anadon>`(gnu packages gcc)`? From the examples, I thought it was `(guix packages gcc)` for referencing packages into a `guix` module. <rockandska>zimoun: Hi, i modified my docker environment to force USER=root and now there is no symlink change but the final size is still ~1.3Gb even after guix gc. I will summarize this on ML too. <nckx>anadon: The package definitions are in (gnu packages …). (guix packages) is [part of] your Scheme toolbox for creating & working with them. <anadon>Huh. I would not have guessed that. <nckx>If you see (guix packages anything) in an example it's a typo, please notify the author. <anadon>I read that as the pattern, then my mind just filled that in as soon as I saw the 'g' on the other lines for dependencies. This was user error again. <nckx>anadon: There are two ways to do what I think you're trying to do above. If the default GCC is too old, you can pass ("gcc" ,gcc-8) as an input (don't forget to import (gnu packages gcc)). <anadon>Does this also hold for `gcc-9`? <nckx>anadon: Because bumping GCC is a big change (lots may break) and even if nothing breaks it rebuilds the world, so it goes onto a special branch that in practice is only merged every few months. <anadon>when I replace all instances of `gcc` with `gcc-8` it can't find the package. <nckx>anadon: The second approach, if GCC is in fact new enough (it often is in practice with the right options) is to add something like #:make-flags '("CXXFLAGS=-std=gnu++17"). <nckx>anadon: Why ‘replace all instances’? Only add ("gcc" ,gcc-9) to native-inputs [sorry, I typo'd ‘inputs’ earlier]. Don't search-replace all gccs with gcc-9s… <nckx>How you pass the -std= flag depends on the build system, no way around that, so see ‘grep -r -- 'std=c' gnu/packages/*scm’ for different possibilites. <anadon>Huh, curious. It looks like the cstdlib is being used from gcc 7.4 rather than 9.2. <anadon>It is getting really close to working. <anadon>OK, `gcc-9` is used in the inputs, but cmake is grabbing `gcc` 7.2. <anadon>`#:build-type` doesn't seem to be the wanted flag. It looks like `#:make-flags` _may_ be right. <nckx>anadon: #:build-type is a very CMake-specific flag that takes a single word (often ‘Release’, ‘RelWithDebInfo’). Which doesn't sound related to what you're doing. <anadon>The existing packages do this sort of thing in `#:configure-flags`, but that almost immediately causes an error and still doesn't fix the use of gcc 7.4. Let me paste where I am. <zimoun>rockandska: ah interesting! So yes please send an update to the mailing list :-) <rockandska>zimoun: i just realized too, that during my staps, it is always the old daemon who is running since i don't restart between staps, so i take a look if restart it change something too and will upgrade tests cases on ML <rockandska>didn't try to reproduce this behavior on a VM too and would be interesting if what i see is global or directly related to docker <nckx>anadon: Could you paste the error? Also, please just use ‘gcc’ instead of ‘C++17+’ and try again. Something might depend on that name. When I said I sometime use a different string than the exact package name, I didn't mean like this — some code may depend on a specific key. <nckx>anadon: That's just the progress, all the interesting things will be here: View build log at '/var/log/guix/drvs/9f/pfc2alj9bwmgymw75qa8fbh4xkjsv8-magic-enum-0.6.3.drv.bz2'. <nckx>(The file name differs for every attempt after every change, don't keep re-opening the same one!) <nckx>anadon: You can use ‘bzless <log>.bz2’. <anadon>So the `#:make-args` (or similar) affected that? I don't understand how those are related. <nckx>‘CMake Error: The source directory "/tmp/guix-build-magic-enum-0.6.3.drv-0/build/CXXFLAGS=-std=gnu17" does not exist.’ *nckx looks for the package definition again. <rekado_>that’s because cmake doesn’t accept CXXFLAGS= as a configure flag. <nckx>anadon: Doesn't CMake expect -D… flags? Otherwise: setenv in a phase, the breakfast of homeless champions. <nckx>-DC(XX)FLAGS is apparently not a thing. <anadon>Claims to be CMAKE_CXX_FLAGS, but still no dice <nckx>anadon: So you tried CMAKE_CXXFLAGS (my definition of ‘normal flag’) as well? <rekado_>that should be -DCMAKE_CXXFLAGS, no? <nckx>rekado_: Yeah, #:configure-flags '("-DCMAKE_…") ; right. <anadon>Nada. `c++: error: unrecognized command line option ‘-std=gnu17’; did you mean ‘-std=gnu11’?` So the compiler mixup needs to be addressed first. <anadon>So it looks like it is getting almost ingsted. <Franciman>does anybody have a custom package for firefox based off firefox precompiled binarie? <Parra>you can set CXXFLAGS as an env var or.. <nckx>anadon: Hm. Surely gcc-9 is new enough. Personally, I'd add ‘--version’ to CXXFLAGS just to sanity check that version 9 is actually being used. <Parra>another option is to set a random variable.like -DCXX_EXTRA_FLAGS and then setting them inside the cmakelists.txt <anadon>nckx The file paths for libraries strongly implies 7.4, not 9.2. Let me test that however. <nckx>Yeah, I somehow suspected that. I don't understand why that is, though. <anadon>It looks like it is explicitly using 9.2, but is still at 7.4. <anadon>Is this a ML question at this point? <nckx>I've never used CMake + custom gcc. I'm pretty user it just works with the GNU build system :-/ <nckx>anadon: Nobody here's piping up, so it would seem so. <nckx>‘no pending requests.’ 🤷 <nckx>anadon: I was looking for messages, not requests. I've approved someone! <anadon>For "joshua.r.mrshall.1991@gmail.com"? <anadon>Am I somehow being filtered out? I'm not getting any ack. <nckx>Maybe I'm looking in the wrong place (‘Membership Management’ & ‘Tend to pending moderator requests’). <nckx>I am not a qualified mailman professional. <anadon>I've gotten the email drafted for when you approve me. however that works. <nckx>Your first mail will need to be approved, but it won't get lost. <nckx>If it doesn't show up for me it will show up somewhere, because *someone's* been approving new senders 🙂 <nckx>guix-devel@ but guix-help@. guix-patches@ but bug-guix@. Simples! <anadon>Both "help-guix.gnu.org" and "guix-help@gnu.org" are not resolving. <nckx>anadon: What error are you getting, exactly? <lispmacs>hi, I've got this problem where, in GnuCash, the Enter key is not doing what it is documented to do (it works fine in Debian). It is supposed to move to next split line in transaction, but it does nothing. I am wondering if this is happening for me only <anadon>lispmacs I think you're in the wrong channel. I think you want either a channel specific to GnuCash, or Gnu in general. In is for Guix. <lispmacs>anadon: well, the #gnucash folks are going to say, we programmed it to do a split line move, and it does it fine our distros, so it must be a problem with your distro or DE settings <nckx>Well, it is apparently specific to GnuCash on Guix. <Parra>is it possible to rebuild all dependencies of a package changing the the glibc or with different compilation flags? <Parra>without refactoring all dependencies one by one <lispmacs>anyway, if somebody could install gnucash and try it, that would answer my question <lispmacs>ideally somebody running Gnome, but other DE results might be of interes <anadon>I can't find out how to actually fix the locale warnings. I've installed them, rebuilt them, configured for en_US.utf8. What could I be missing here? It is spamming my console. <nckx>elais[m]: It is very compared to GNOME on a Core 2 Duo. <nckx>Not everyone can go full tiling wm just because they have an old machine. XFCE is great for those users. <Parra>anadon: I had the same bug with locales in a guix running inside alpine and I only removed the LC_ALL env var and worked <nckx>anadon: Are you running Guix System? <Parra>I'm running guix inside an alpine <anadon>No, Ubuntu and trying to hybridize them. Later on, the only way for me to have guix at work is if I get it co-installed with CentOS7. <Parra>oh sorry I mismatched the username <Parra>both of you are green nckx and anadon <nckx>anadon: Is GUIX_LOCPATH set to $HOME/.guix-profile/lib/locale? <KE0VVT>nckx: LXQt or MATE would be my choice for old computers. <nckx>anadon: And is there something there? <rockandska>Parra: just for my personal information, don't you have a problem with the size of your image ? <nckx>KE0VVT: Interesting, I wasn't expecting Mate there. I've themed XFCE to emulate GNOME 3, which is nice. <anadon>anadon@goodadvicemallard:~/Documents/code/guix-packages$ ls $GUIX_LOCPATH/2.29/en_US.utf8LC_ADDRESS LC_IDENTIFICATION LC_MONETARY LC_PAPERLC_COLLATE LC_MEASUREMENT LC_NAME LC_TELEPHONELC_CTYPE LC_MESSAGES LC_NUMERIC LC_TIME <Parra>rockandska: I don't really care anyway, meanwhile it isn't like windows server which with msvc is around 30GB <Parra>for example, the debian image for building metacall is around 1GB <nckx>anadon: And are all your packages up-to-date (guix package -u .)? They need to link against glibc@2.29. <Parra>but it's the build image, the output is around 100mb <rockandska>Parra: 300MB after having made guix pull without guix gc ? <Parra>rockandska: I didn't check in detail, maybe more <Parra>I think docker compresses the layers <Parra>we can check it out on Docker hub I think <rockandska>Parra: huge for a docker images in my opinion :) <Parra>it's built through buildkit so maybe it improves the layering too <Parra>rockandska: it is, but it's a builder image <rockandska>Parra: i try to build a lighter guix docker image too and have some problem with the garbage collector, this is the reason whi i asked <Parra>from metacall/guix as builder <Parra>#then build your tarball here <Parra>copy --from=builder tarball.tar.gz <Parra>the output image will be very thin although the builder image is 400mb <Parra>rockandska: I asked in the mailing list and the best solution is to use guix to create the docker image by itself <Parra>my next approach will be to use the alpine image to bootstrap guix and then build the docker image inside docker <rockandska>Parra: i'm aware of multi stage building with docker <Parra>then publish it to dockerhub <Parra>this will generate the minimum size possible <Parra>but you may remove 100mb at most <rockandska>guix pack -f docker seems to add some unused thing in the result image but didn't had the time to post on ML about it <Parra>what do you mean by unused thing? <rockandska>test with guix -f docker bash for example, it includes some dependencies not needed in final and the result is ~80mb compress vs 5Mb for the bash docker image (based on alpine) <Parra>Ludo said this: guix system docker-image config.scm <nckx>anadon: Does happen when running any Guix-installed software, or only when you run ‘guix’ itself? Maybe the daemon needs to updated, or the Environment= line in /etc/systemd/system/guix-system.service is off. <nckx>If it only happens when running guix. <Parra>maybe you can investigate with that <Parra>rockandska: but I'm building a docker image with guix inside, are tou talking about the same? <rockandska>Parra: yeah, but i did some different things than your <Parra>if you find a safe way to make it smaller tell me, I'm interested <rockandska>Parra: the bare config seem interesting but don't know at this time how to use it with a docker context. I'm looking for a way to do CI on some profile by using docker too <nckx>anadon: If ExecStart=/var/guix/profiles/per-user/root, try running guix pull as root: sudo -i guix pull <nckx>That's the extent of my non-Guix-System knowledge 🙂 <nckx>(And then systemctl daemon-reload & restart guix-daemon.) <anadon>That is actually very critical for me to take note. It means I'll still have to deploy guix with puppet. <KE0VVT>nckx: But does it emulate the functionality of dynamic workspaces? <nckx>KE0VVT: My mother doesn't care. 😃 <nckx>GNOME if great if all you're running is GNOME. <KE0VVT>nckx: Is your mom logging into a Guix System and using it? :o <nckx>Once you tire of the cool animations and fire up IceCat, things get bad. <nckx>KE0VVT: She doesn't know it but she is. <KE0VVT>nckx: Wow. Although I'm addicted to the GNOME workflow. <KE0VVT>I wish there was something that functioned like GNOME but without the fancy animations. <nckx>and compositing, which IIRC stopped being possible long ago. <rockandska>guix team: just curious, but the documentation states that after having made 'guix pull' the user need to expand $PATH / $INFOPATH. <rockandska>- why not using .config/guix/current/etc/profile ( who expand $PATH ) ? <rockandska>- why INFOPATH is not in expand in .config/guix/current/etc/profile ? <pkill9>what's wrong with gnome? the resource usage? ***dftxbs3e_ is now known as dftxbs3e
<nckx>anadon: Mail approved, just took a while. <nckx>pkill9: It took significant resources from IceCat. <nckx>rockandska: I don't understand. What does ‘expand’ mean? <rockandska>nckx: cf guix pull doc "Thus, make sure to add it to the beginning of your search path so that you use the latest version, and similarly for the Info manual" <rockandska>nckx: i wondering why just using ~/.config/guix/current/etc/profile and why in ~/.config/guix/current/etc/profile $INFOPATH is not set too <nckx>rockandska: INFOPATH is not included in the profile's profile file (ehm) because texinfo is not part of that profile. <rockandska>nckx: we talk about guix itself, and the doc ask to do it, so ~/.config/guix/current/etc/profile could include "export INFOPATH="$HOME/.config/guix/current/share/info:$INFOPATH"" too <nckx>Search paths are only set if a package in the same profile can make use of them. In this case, the texinfo that would consume the INFOPATH is in a different castle^Wprofile. <nckx>rockandska: No, it can't, not without changing how search paths are handled. <nckx>PATH is a different matter. I don't know why the docs ask you to set it explicitly; looks like a non-Guix-System thing TBH. <rockandska>if you set PATH after guix pull it seems that the user will not use the last guix version <nckx>rockandska: That is correct. But: I think these two export commands only make sense on other distributions, and shouldn't we explicitly ask users to add both lines to their ~/.bash_whatever instead? <nckx>I haven't read the whole chapter. Maybe that's suggested elsewhere. <nckx>rockandska: These lines can't (simply) be added to ~/.config/guix/current/etc/profile for reasons, but I agree that they should be saved to ~/.bash_profile or similar. <rockandska>nckx: yeah, but since for package we source .guix_profile/etc/profile, i foudn more elegant to only source .config/guix/current/etc/profile, and in case, someday, guix need other env var to be set/update, the user will not have to be awore of that addition if all of them are in .config/guix/current/etc/profile <nckx>rockandska: Sure, I get it. <rockandska>nckx: i don't get why INFOPATH can't be set as PATH in .config/guix/current/etc/profile since it only expand it for guix doc, but since i'm not a guix expert i will trust you ^^ <nckx>The short story is that ~/.config/guix/current/etc/profile isn't generated in a way that allows adding random lines like that 🙂 Doing so would be slightly more convenient for users *in this specific case*, but it would make profiles *in general* less pure. I think… <nckx>I think, instead, the docs should clearly explain to users if and how to create a custom ~/.bash_profile or ~/.bashrc on other distributions. Of course I can't do that. <nckx>That clearly separates the things that make this a special case from the general code that creates ~/.config/guix/current/etc/profile, and gives the user a clearly supported place to make their own customisations. <rockandska>Parra: uncompress your image size is 1.17GB where the one i just build seems to be 574MB for the latest guix pull <Parra>it's still acceptable, my metacall/core:dev image has a similar size <Parra>rockandska: how did you achieve that image? is it working properly? <rockandska>Parra : ./hooks/build will build guix:1.0.1 / guix:latest <Parra>rockandska: you don't do pull and update <rockandska>Parra: no, too recent, you need to restart the container to have the last guix-daemon running after guix pull <rockandska>Parra: don't stick to the Dockerfile, i do guix pull ( but not update ) <Parra>but the pull isn't cached in a laywr <Parra>it is done after doing run, right? <rockandska>Parra: take the time to take a look at ./hooks/build <Parra>when you commit you make a layer of the current running image, right? <Parra>it's easier and more portable than my solution <Parra>it may help to reduce the image size <rockandska>Parra: it will not work in your case, because you will always have the old daemon running <rockandska>Parra: guix pull update the daemon too and guix gc doesn't clean thinks who are always in use from what i've observed <Parra>if I do guix pull it gets patched <Parra>and on the entry point I launch the new daemon in a separated thread <Parra>with --priviledged to allow the clone as you <Parra>or I can restart it during build time or whatever <nckx>anadon: Again, include: View build log at '/var/log/guix/drvs/zw/dm194p8n8mi4kfbw83z945j05g03dv-guix-package-cache.drv.bz2'. <anadon>(repl-version 0 0)Generating package cache for '/gnu/store/n9rwj3jlrf262jgj0ynwf5ihf7kdxvi7-profile'...(exception misc-error (value #f) (value "~A ~S ~S ~S") (value ("No variable named" uglify-js in #<interface (gnu packages lisp) 7ffff0d17be0>)) (value #f)) <nckx>anadon: Could you paste your package expression? <anadon>That was just from a pull. No arguments. <nckx>mbakke was right, uglify-js was moved to (gnu packages javascript) and in your error log we can see that something is still looking for it in (gnu packages lisp). <nckx>anadon: So the problem is with the current state of the Guix repository, nothing you can do about it. I understand that it can be frustrating and I'm sorry about that, but these things happen. <mbakke>ideally we'd have some git hook to prevent this from happening :/ <anadon>Even a total rebuild of all packages from source? <mbakke>anadon: you can pull the commit before, i.e. `guix pull --commit=183fd2070c7c50675df419e456d6bc036b4aa490`. <mbakke>or even --branch=staging if you're feeling adventurous... <mbakke>nckx: are you able to look into it? my git checkout is...uh, busy. <anadon>I'll try for the older. I want to test and fix my setup. I'm not prepared to meaningfully contribute to upstream. <nckx>I can put work aside to look into it but plan A is Pierre speedily replying to the mail I'm about to send. <nckx>anadon: Did --commit=$older help? <rekado_>lets revert abf43a67d5929bf279edce3e790ef1348dda32a5 <rekado_>lets fix this with a little less urgency soon <Parra>rockandska: thanks for the Dockerfile, I really learnt with it <Parra>and also I'm glad to see that other people uses guix in docker too <Minall>Right now I'm on a guix live USB, I can't install guix properly <mbakke>Minall: what error are you getting? <Minall>mbakke: the installation partitions my device badly, I don't get an EFI directory, <Minall>So at the end of my installation I get an error that grub can't find the efi directory <Minall>Is there a way to install grub without needing to run everything again? since I have to download and install everything again <Minall>Is there a way to make like an 'arch-chroot' into my new system? <Minall>The problem is basically that I have to download everything that I already downloaded again, and build it <mbakke>Minall: if you run `guix system init` again (without rebooting), it will continue where it left off <Minall>mbakke: It will copy everything again to /mnt :( <mbakke>if you reboot, it will have to download all substitutes again <mbakke>that reminds me, there are a couple of other EFI bug reports with the installer :/ <mbakke>Minall: does /sys/firmware/efi exist in the live environment? <Minall>So I can't install grub without installing everything again? <mbakke>Minall: is /mnt/boot/efi mounted? <nckx>sneek: later tell Minall: Is copying everything to /mnt a deal breaker? Running grub-install --boot-directory /mnt/boot --bootloader-id=Guix --efi-directory /mnt/boot/efi might work fine, it also might result in an unbootable system if not all configuration was installed… <nckx>mbakke: They've left :-/ <nckx>anadon: Glad to hear it. It took so long here that I had to cancel it, so couldn't verify. It's been ‘fixed’ on master now. <anadon>Go pulls are stable, but still no fix on the setlocale warnings. <nckx>It will have something to do with the daemon's environment, not your user's, but I don't know the rest by heart. The warnings are ugly but harmless though. <rockandska>Parra: feel free to ask in private if you have some specific questions not directly related to guix ***poet` is now known as poet
<Franciman>I still need to understand: Do I need to set GUIX_LOCPATH="$HOME/.guix-profile/locale" or not? <Franciman>so I find it more sensible to make guix use the locales in my profile <nckx>Franciman: There is no ‘globally installed’, just a root profile and yours. They are equals. So whose $HOME are we talking about here? <Franciman>nckx, hmm I don't understand, they are not equal <nckx>Setting it to root is fine, setting it to your user's is fine if you run guix-daemon from your user profile (I used to). What matters is that GUIX_LOCPATH is set in the daemon environment, setting it on the command line will not do anything. <nckx>Franciman: They are two equal Guix users; ‘root’ is not global. <Franciman>but I saw that locales are in /run/current/locale <nckx>Ah, you're on Guix System? <nckx>So many locale problems today, I lost track 🙂 <nckx>Thought you were someone else. <Mrtn[m]1>OK ... why does guix ask me to set GUIX_PROFILE to the exact same thing it is already set to?! <nckx>Mrtn[m]1: If it's what I think it is: because it's part of a two-line message and GUIX_PROFILE is used in line 2…? <Mrtn[m]1>Which pastebin do we use here? I know it is looks like feces if I paste on a channel bridged to IRC ... <nckx>Mrtn[m]1: paste.debian.net, thanks for asking. <nckx>Mrtn[m]1: Looks like my answer above was exactly correct. <nckx>There's simply 0 value in splitting it into 2 lines with a (when …) to check whether your GUIX_PROFILE happens to match. <nckx>It's the 2nd line that matters anyway. <Mrtn[m]1>But, as you can tell, the variable is already set, so why warn about it? <Mrtn[m]1>nckx: I get what you're saying ... but let me show you this: <nckx>A good counter-argument to my opinion would be ‘I don't use a mouse to paste everything I see’, that would be fair 🙂 In that case a patch to (when …) the first line would probably be accepted. <nckx>Mrtn[m]1: And GUIX_LOCPATH is unchanged, I presume? <Mrtn[m]1>This is from .bashrc ... I have those exact lines in there, which means the commands are run when I log in ... <nckx>Yeah… If you can write a clever but bullet-proof check for ‘all variables are identical sod off Guix’, it might get accepted. <nckx>It just has to be fool-proof. <Mrtn[m]1>nckx: Ahh, so what you're telling me, is that it doesn't care if I already did what it asks me for, it just warns always? <nckx>Well, no, it does *a* check, I don't know off-hand what it is and you seem to be showing me that it's not perfect. <nckx>I certainly don't see that message every time I modify my profile. <nckx>At least I don't think so… <Mrtn[m]1>nckx: I see it about every time I upgrade <nckx>I'm sure that gets old fast. <g_bor[m]>Mrtn: the main thing here, is that the file pointed to by guix profile is changed. The second line sources the file. <Mrtn[m]1>nckx: Which locale are you using? I have a suspicion it might be related to that .... <Mrtn[m]1>g_bor: Oh ok ... so it appears each time the profile changes! That makes sense. <nckx>Hum, I got dragged into looking at the code (I wasn't going to get dragged into looking at the code 😛), and I don't see a check. Now I'm starting to doubt my own memory. Surely I can't have got hint-blind by running guix too much… <nckx>Nope, I apologise for spewing feces. I guess I just really ignore whatever Guix complains about nowadays. That's not good. *nckx adds ANSI blink codes to all warnings. <Mrtn[m]1>Well, the most important thing for me is, that it doesn't mean there is something broken, it's just something that Guix really likes to tell me. :) <nckx>Now I don't know what to think. <nckx>Mrtn[m]1: What did you mean by ‘every time you upgrade’ above? <g_bor[m]>nckx: I believe it is triggered when packages with a search path are involved. <Mrtn[m]1>nckx: I might have gotten that wrong. I can't check at the moment ...