<HappyEnt>Or is there some global package override like in nix? <bavier1>HappyEnt: from a git clone, use pre-inst-env <janneke>joshuaBPMan: so, the breakage seems to be shepherd related? <joshuaBPMan>janneke: I'm not certain, I just know that your command fixed my issue. <joshuaBPMan> guix time-machine --commit=d39885a8a9e0e03c2bf6277d475d384168bba642 -- system reconfigure <joshuaBPMan>all kernel modules loaded fine. (including the touchpad). <janneke>which means you're on 5.4.20, right? <HappyEnt>bavier1: perfect that should work thank you :) ***daviid is now known as Guest78499
<janneke>joshuaBPMan: it was HappyEnt who said that commit was the last that worked for them ***Guest78499 is now known as daviid
<janneke>i did not bisect any further, but for now suspect shepherd-0.7.0 <nojr>Hi, quick question: Why is the Django package stuck at v 1.25? ***jonsger1 is now known as jonsger
<apteryx>nckx: thank you, I will check it a bit later <apteryx>has anyone else noticed how running QEMU on Guix resets the terminal and then you loose the scroll-back buffer? Especially annoying in tests, where something crashes before QEMU starts, yet it cannot be inspected. <apteryx>someone from #coreboot told me it must be seaBIOS sending an escape code; someone from #qemu told me my terminal is misbehaving (could reproduce the problem in xterm but not in gnome-terminal and others). <apteryx>I could reproduce the issue with sakura & xfce4-terminal too. <apteryx>you can try "C-l" in your terminal to see if you have the issue. <apteryx>apparently QEMU does more than "C-l", because in xfce4-terminal I can scroll back after pressing C-l, but running QEMU, the scroll back is still deleted. Humph. <roptat>it happened to me outside of qemu, I couldn't scroll back on a terminal anymore, but that was on fedora <roptat>I don't know what can cause that... <guix-vits>apteryx: echo -e '\n'`date` >> log.txt; quemu ... &>>log.txt as WA? *guix-vits too tied to favorite terminal emulator <apteryx>it really is SeaBIOS sending the ansi escapes: reset + resetwrap + clearscreen + modesoff ***apteryx is now known as Guest32737
***apteryx_ is now known as apteryx
<efraim>Has anyone used shepherd as a user and run guix services? <civodul>efraim: i run shepherd as a user but with a hand-written config file <civodul>roptat's Guix Home can handle user services IIRC <efraim>I'm experimenting with running guix services as a user with sudo privileges on a foreign distro ***gour_ is now known as gour
<efraim>I might need to refactor the code a bit, right now I declare the service in an os-config file. not sure it'll be happy if I also export the service <efraim>and two warnings from using guile3.0-guix: <efraim>WARNING: Use of `load' in declarative module (guix ui). Add #:declarative? #f to your define-module invocation. <efraim>WARNING: (guix build emacs-build-system): imported module (guix build utils) overrides core binding `delete' <efraim>ok, i see the note in (guix build utils) about delete <efraim>service internals of the shepherd manual is empty <zimoun>janneke: have you built successfully with the time-machine? bug#39575 <efraim>I just noticed that one of the gexp examples in the manual symlinks emacs to vi <civodul>janneke: you should now be able to log in on the overdrive! <janneke>civodul: hmm, overdrive.guix.info asks for a password/does not seem to accept my key *janneke goes afk for lunch ;-) <civodul>janneke: that's overdrive1.guixsd.org, not the same one (see PM) ***zalox_ is now known as zalox
<janneke>civodul: time-machine, bug#39575 not real success, i was composing a reply; let me finish that <raingloom>i'm gonna try fixing the Desmume package. found some patches from Arch. should I merge them into one patch or is it better to separate the gcc6 fixes from the gcc7 fixes? <guix-vits>raingloom: first test at least one of them ;) <nckx>apteryx: Ditto. I always assumed it was QEMU using a secondary buffer or something for more accurate VTE-in-VTEing. <raingloom>guix-vits: i checked them both separately and it looks like both are needed *guix-vits prepares to look smart <guix-vits>raingloom: If the fix for GCCs versions 6 and 7 is logically the same, unite them in one patch. If the fixes are different, separate them (not an dev myself). *janneke finally validates a non-intel-linux boostrap fix on core-updates :-) <guix-vits>ME please unite, please separate, i was meaned ;) <zzappie>efraim: whoa :) I was expecting vim user to have done it. <nckx>raingloom, guix-vits: Actually if the arch patches work without modification I prefer to keep them intact, each with a header noting ‘copied verbatim from Arch at <link>’. Makes them easier to track compared to creating another subtly different variation. <nckx>See gnu/packages/patches/xf86-video-voodoo-pcitag.patch for how I like to do it (habit from another project), but choose whatever format you like. <reepca>tor button in icecat doesn't seem to be working... it shows that I'm connected to the proxy but https://check.torproject.org/ indicates that traffic isn't being routed through it. Any idea what's up with that? <raingloom>nckx: in this case Arch also just copied it and the header has some info on where it's from. that should be enough, right? <nckx>reepca: Might want to cross-ask in #icecat. I don't remember it ever working. <kmicu>(Guix maililing lists also have some chit‑chat about the issue.) *jonsger resumes work on Nextcloud <guix-vits>kmicu: But it does run all of your HTTP traffic through the tor network, WHICH IS PRETTY COOL <zzappie>I got a question: I sometimes need to quickly run and test a random project that says something like just sudo pip/npm install loads of stuff that is not in guix packages. Do manifests provide means to do arbitrary scripting after profile with core packages has been initialized? <zzappie>or better put, what do you think is the best way to handle situations when you dont have time to write package definitons and need quickly spawn isolated invironmmanen containing bunch of stuff installed by particular language's package manager on a guix system and make it persist between invovations *zzappie guix-vits: yep thanks. Typos is my weeek spt *guix-vits wonders if spt means spit XD <guix-vits>zzappie: What about make an "fixed" environment and install with Pip and such on top of it? <guix-vits>zzappie: "Sometimes it is desirable to isolate the environment as much as possible, for maximal purity and reproducibility." --container at the link above? "Fixed" is wrong wording. <zzappie>Im doing something like "guix environment --manifest=manifest.scm --user=bcdb-js --container --ad-hoc coreutils --link-profile". Than I could run some commands on top. But as if I get it right changes made won't persist between guix environment invocations <roptat>There's a user-service like feature, but only for configuration, it won't run anything on the aser's behalf <sirgazil>zappie: I use "guix package --manifest ... --profile ..." to create environments for my projects instead of using "guix environment". <roptat>Sorry, responding to a message from 4 hours ago <sirgazil>zappie: And for Python projects whose python package dependencies are not yet in Guix, I'm using python environments for the Python packages plus Guix profiles for other packages. <guix-vits>zzappie: i'd done guix environment --container bash; and i'd `echo abc > xyz.txt`; after i'd quit the environment, x is steel there. idk for bigger things, though. <sirgazil>zzappie: And gradually packaging python packages for Guix so that I don't depend on pip anymore. <zzappie>sirgazil: Oh i've missed --profile option for "guix package". Yeah I know that it is beter to "guix import" and package things for reproducibility reasons. But Im not yet proficient enough with guile and guix to do it quickly enough. So I'm mixing two worlds here. But im gradually incorporating guix into my projects <sirgazil>zzappie: Yeah, when I wanted to create development environments, I thought "guix environment" was the thing to do. *zzappie sirgazil reading just now :) <zimoun>civodul, janneke: the bug#39575 about time-machine is weird. janneke and I, we are not failing for the same reason when building (python vs guile ;-)). Moreover I have noticed other commits around the offending one failing too (build error of Guile 2.2.4 or Python 3.7.2). <zzappie>Speaking of docs... I've came accros a nice article about documentation. https://www.divio.com/blog/documentation/ Is there any roadmap for documenting things? It seems like guix grows quickly and and using guix has an effect on many development activities thus might benefit from different more interactive types of docs. <nckx>zzappie: No roadmap I know of. ‘Just volunteer’ 🙂 That article is part of what led to the Cookbook. *guix-vits hates Cloudflare and its captchas. Some people just use tor to solve bad routing issues. <zzappie>nckx: Nice :) meanwhile ROADMAP was last updated in 2015. What about guix blogging contest? *nckx doesn't know what that is. <nckx>And yes, we need a roadmap to update the roadmap. <zzappie>Everyone blogs about how they use guix in their development and good usecases are revised and merged to docs <nckx>That would suit the cookbook. <zzappie>this way we also popularizing guix in different communities and platforms <nckx>zzappie: Do you see this as an actual contest (with announcements, winners, maybe prizes) or just encouraging more people to blog more about more Guix? <guix-vits>Funniest thing is, that anyone can popularize or promote [gi:ks]... But critique in speeches is somewhat dangerous without a careful specifying (OS, err.. SD.. khm... we mean that exact distro from gnu.org!), or else a lot of [gi:ks] may be angered. <zzappie>nckx: I duno I just made it up :) Could be both <nckx>guix-vits: OMFG ITS GUIX SYSTEM <nckx>We are not that quick to anger. <guix-vits>nckx: Are the <strike>D'Artangnans of Free Software</strike> Geeks really not that quick to anger? <nckx>guix-vits: Apart from the package manager/system split (which isn't really a split IMO, but some people see it that way), what do you feel one needs to be ‘careful’ about? There is only one GNU Guix and no forks that I know of. <nckx>guix-vits: I don't know how much you're joking, but if this is really your impression (even in jest) I'm genuinely surprised. <nckx>I regard the Guix community as pretty damn chillaxed. <zzappie>Yeah there is lots more arguing about guix on linux reddit :) <nckx>Until you try to call $PAGER like a sane person because omg systemd does. /s *nckx goes back to hacking on Guix Linux. o/ *guix-vits performs JOKE (vocal): ... yes, just there is two real [gi:ks] distros (which is really only one). The <strike>Guardians of Cardinal</strike> Arch, Debian, Fedora, Gentoo ... Puppy ... are cool, but aren't [gi:ks] distributions, of course (m'sew)! *zzappie guix-vits minds are going to be blown when someone will port guix to bsd. *ngz wonders why everyone is using /me <roptat>Apart from my using of tabs, arg <guix-vits>roptat: Not much from me: Why not to just use wget? Something like: "(guix (bad-connection #t)) ;; use good old wget"? Typo: "Performa partial". <roptat>Sure, but it's more a workaround than a real fix :) <roptat>And you have to know the expected filename too <joshuaBPMan>that's weird. guix system reconfigure is pulling in a lot of haskell stuff. ghc-* <guix-vits>roptat: "(system (append filename_to_download, ("wget")))" -- i'm sorry for pseudo-Lisp of mine. <roptat>Oh, but then there's a dependency on wget <roptat>And we'll even need a binary of it, since you can't wget wget before you build it <roptat>joshuaBPMan: I thought guix graph could help, but apparently it only works on packages, not on the current-system <roptat>It works with derivations though… so maybe try this: guix system reconfigure /etc/config.scm -n <roptat>This will give you a derivation named system.drv <roptat>Then try guix graph -t derivation /gnu/store/…system.drv > system.dot <guix-vits>roptat: first install wget, then set (bad-connection #t), then wget any amount of wgets? Should work... <roptat>guix-vits: but how do you refer to wget from the build side? <guix-vits>Like in bash (rough): `x=`mktemp`; wget $x`, but in Lisp. Should be possible. <joshuaBPMan>hey guix, maybe I'm just being really silly, but I do not understand how to turn the tls-auth parameter on the openvpn client service. <reepca>joshuaBPMan: methinks the 'tls-auth-client' part of the documentation is describing the type of the object <joshuaBPMan>reepca: hmmm let me try to read that documentation again then. <reepca>unfortunately I can't find any documentation on how to get an object of that type... <bavier1>joshuaBPMan: I think it needs to be a string <bavier1>joshuaBPMan: quick glance at the code suggests acceptable values are #f or a string <joshuaBPMan>bavier1: thanks. I was looking at the source code and still didn't understand it. <joshuaBPMan>so (tls-auth "tls-auth-client") seems to let guix reconfigure...we'll see what happens. <roptat>joshuaBPMan: you can use -n and find the openvpn.conf.drv, then use guix build to build that anl review, it should be faster :) <roptat>You'll have to look at the openvpn manual though, it corresponds to the tls-auth configuration <joshuaBPMan>roptat: sorry, but I'm not certain what I am supposed to put after the "-n"...let me look at the manual. <joshuaBPMan>roptat: I think I need to use "tls-client". That's what my working config file says. <roptat>What is the value for tls-client? <roptat>(Redacting any private info of course) <joshuaBPMan>Julien was kind enough to supply a auth-user-pass option for me. <joshuaBPMan>roptat: oh. Nice to meet you sir. Thanks for the help. I'm still having trouble connecting, but that's ok. I'm getting closer. <roptat>Ok I chink I see where the issue is <joshuaBPMan>I'm still getting a TLS Error: TLS handshake failed in /var/log/messages. <roptat>We don't have a tls-client option, and we always provide the client option instead <roptat>tls-auth will not do what you want <roptat>We simply have no way to generate your config yet <roptat>I'll give it another good look tonight to try anl generate your config <joshuaBPMan>roptat: ok. let me try to find my config file that is being generated. then I can compare and perhaps push a patch. <joshuaBPMan>Ok fair enough. I'll try to race you to the correct patch. :) Fair warning, I'm a newbie. *guix-vits lists that process lists that process lists... <roptat>If you're faster than me I'll be glad to push any patch you'll provide :) *guix-vits hopes that someone will care about maintainability in this race... <joshuaBPMan>hmmm, I'm not even certain how to find the openvpn conf file that guix generates. <guix-vits>joshuaBPMan: By `find /gnu/store -name if_known.conf`? <roptat>When you change your config, you can add -n to your guix system invocation, which will show the list of derivations to build, including openvpn.conf.drv <roptat>You can pass that drv to guix build without having to build the whole system <apteryx>using DDG as a search engine in icecat, when I press back, it shows be a blank page, which is quite infuriating (it should go back to the main query page, with my previous search results). <roptat>It's faster to check a change you make, anl you can test the whole thing when you have male all the changes you want <bandali>apteryx, works fine here, though i use LibreJS and/or uMatrix, which blocks their JS; that may be why it's not broken for me <joshuaBPMan>that worked. I feel like guix ought have a command like guix system show config file for openvpn... <joshuaBPMan>guix build /gnu/store/stasn55uh5an55h4una4thu34snth/shepherd-vpnclient.scm worked, but it was still a bit awkward to find the config file. <apteryx>bandali: you nailed it. Re-enabling GNU LibreJS fixed it. <bandali>apteryx, great, cheers :-) if you wish, feel free to write to bug-gnuzilla@gnu.org about it, but i have a hunch that we may not be able to do much about it <bandali>also, anyone try building libreoffice on an x200? i'm curious how long it'd roughly take <guix-vits>sneek later tell joshuaBPMan: x=`mktemp`, `sudo guix system reconfigure -n /etc/lightweight-desktop.scm &> ${x}`, then `grep openvpn ${x}`. It should highlight location. <guix-vits>sneek later tell joshuaBPMan: or `grep 'drv$'`, then grep ... khm :) <guix-vits>roptat: 2 is descriptor for errors. it direct them to 1, which piped. At least it isn't >> here. Thanks! <guix-vits>sneek: later tell joshuaBPMan: better do as roptat: no file, `... 2>&1 | grep ...` *guix-vits fall asleep... <shader>how do I find the configuration file for nginx? <shader>I'm trying to set up nginx and certbot <shader>but certbot doesn't seem to be working, and I can't tell why <shader>never mind, I found the conf line on in the reconfigure output <shader>it seems to have the code inserted by certbot, but I don't have any evidence that certbot has or will run <shader>it did make the /var/www directory <shader>or... do I have to separately install the certbot package? <shader>I don't think I had to for openssh or nginx... <Ashjkaell>Hi everyone o7! Has anyone else had problems at boot with Linux-Libre-5.4.20 or have I singlehandedly borked my reconfigure? :-) <bavier1>Ashjkaell: there were a few here yesterday who were reporting difficulties with 5.4.20 <Ashjkaell>bavier1: Aha, many thanks, it gives me solace in the pain of others ;p <Ashjkaell>(I must say guix has been invaluable in the ability to roll back to another generation) <bavier1>Ashjkaell: I believe there may have been some bug reports opened too, so you might check those out to see if there are details and/or workarounds <shader>anyone know how to get certbot working? <shader>hmm. maybe it will just run eventually on it's own <roptat>shader, you can look at herd schedule mcron <roptat>Also make sure you don't have a server block that references a key you don't have yet, as nginx will fail to load <shader>I managed to find the command by editing the service, discovering mcron, editing mcron, discovering the jobs file, and running find on the store to look for that <shader>yeah, I ran into that issue first <roptat>And certbot needs nginx running to get a certificate <atw>shader: I did but it was confusing. When I put a certbot service and nginx service in my operating-system, but then nginx refused to start because the private key hadn't been created yet <roptat>So one of tge jobs is certbot-command. I think it's supposed to run once, but you can run it by hand as root <shader>do I need to manually configure the cert locations? <atw>I should have written some notes... <roptat>No, it's all part of your configuration already, no manual conhiquration is required <roptat>The cert will be created in /etc/letsencrypt/live/your.domain/fullchain.pem <shader>I meant, do I need to manually add those paths to the main nginx configuration? <roptat>Once you have the key, you'll be able to change nginx configuration and reconfigure, restart nginx and you'll be done <roptat>Once you have the certs, not before <atw>^ yeah, that's where I got tripped up <shader>I'm getting a lot of "incompatible bytecode kind" messages during system reconfigure... any ideas? <roptat>Could it be a guile 2.x vs 3.0 issue? <vagrantc>janneke: so, my debian package builds of guix build for aarch64/arm64 just fine ... as i think it's using guile 2.2.x <vagrantc>janneke: so i have a theoretical workflow to get the pinebook pro installed from debian :) *vagrantc hopes there's some hope for guix pull getting fixed <apteryx>I want to add a 'test' or 'run' action to herd, to force executing a job for test purposes. <apteryx>but that's a bit far on my todo stack. <apteryx>shader: I've seen those as well. Don't seem critical. <janneke>vagrantc: ah, that's another good path <vagrantc>janneke: my debian packaging branch builds ... i haven't tried applying your pinebook pro patches on top, but in theory it might work :) <janneke>vagrantc: great, yes danny has some good new pointers too *vagrantc has been meaning to reply to the bug thread <janneke>too bad my pinebook does not boot at all -- filed a support request :-( <janneke>good, we can probably get pretty decent support without too much effort <janneke>vagrantc: yesterday shtwzrd said they'd attempt my guix recipe and look at accelerated X *apteryx works around qemu resetting the scroll back buffer with 'make system-check TESTS=btrfs-on-root |& strings' *janneke pushes build fixes for arm and hopefully powerpc to core-updates <shader>so how do I find out what configuration file is being used by a service? <shader>ok, it looks like nginx is only getting the server block from certbot, and not the other configs I added in my system file <shader>am I supposed to compose them somehow? <shader>weird; if I restart it, it uses a different config that has all three... <roptat>shader services are not restarted after a reconfigure, you hade to do it by hand <roptat>Examining my system, I found that libevent has a dependency on python, but that's only for a script. Putcing it in a separate output should reduce the closure size from 180 MB to about 80 maybe. I'll try to do that tonight :) <roptat>Since it's required by ntp, it's about 80 MB that can go away from my system's (and everyone else's system) closure <shader>where do I find the list of services available to shepherd, and their source? <shader>I mean, I know I can get them from my system config file, but I was wondering if there's a way to check what's actually running <shader>and how would I add a service withough reconfiguring the whole system? <shader>I've been generating a ton of generations since the only way I have to test a service is reconfiguring my system... <bricewge>shader: Read the output of `herd doc root action load`. <bricewge>The load action can use a specific file in shepherd, `sudo herd load root $myfile` <roptat>herd status will give you the list of services <roptat>The ones installed on your system that is, not the ones available in guix <shader>how would y'all suggest automatically deploying a static website stored in git to guix? <roptat>Any idea why the shepherd is using a completely different guile than the rest of my system? <shader>I see the static-web-site-service-type, etc., but what is Haunt? <roptat>It's a static website framework written in guile <roptat>generator maybe is a better word <shader>so how does the build / deploy process work for the guix website then? <roptat>Ah right, I was looking at the service definition, not the package one… <civodul>a post-processed "guix graph --references"? <civodul>you'll need to share tips & tricks :-) <roptat>With tge markpath script from two years ago, I remembered I had it saved at framagit <roptat>framagit.org/tyreunom/guix/snippets/3601 for my saved version <roptat>We could add some options to guix graph though <roptat>Also it's a bit sad that it doesn't read symlinks by itself "/rur/current-system is not a package name" is a bit annoying :) <raghav-gururajan>New Linux-Libre version breaking GDM session. When I boot, I see "New gdm session for c1" followed by "Session c1 removed". <Ashjkaell>raghav-gururajan: It definitely looks like it! <jackhill>believe I found a typo in the manual, in section 4.8 "unmodified to the 'guix' command if the specified revision", I think the if should be of. <jackhill>should I propose a patch to fix this, or report it, and leave it as a first issue for the interns? <Ashjkaell>No problem! By curiosity, do you have a thinkpad T400? (It's my machine and the one in the issue too.) <janneke>civodul: just a quick doc headsup, thanks for your comments!! i'll respond in with more detail after a rewrite; this kind of feedback was exactly why i created a new bug for it :) <civodul>tsmish: cool, feel free to post to the bug! <raghav-gururajan>Ashjkaell Last known commit that works for me is "41884bf". Anything after that breaks my system. Hbu? <civodul>so it says 205c1e04e04b9a9338c7219ff82bd13f000fb8c8 is the culprit <civodul>though i'd have expected the next one <Ashjkaell>Yes, I can definitely reproduce the last steps of the bisect posted by tsmish <janneke>that's what shtwzrd asserted; i haven't bisected it to one yet <Ashjkaell>(Stupid question.. What does shepherd have to do with kernel modules?) <civodul>i thought 8b9cad01e9619f53dc5a65892ca6a09ca5de3447 was a more likely culprit <Ashjkaell>Hmm.. On a borked system /proc/sys/kernel/modprobe points to the same modprobe as a working system (commit d39885), but the modprobe it points to is not in the store. <jackhill>Ashjkaell: interesting, I hadn't noticed that. I've posted the contents of my file to the bug <jackhill>roptat: for me it's /run/current-system/profile/bin/modprobe, so yes <Ashjkaell>jackhill: Do you mean `which modprobe' or `cat /proc/sys/kernel/modprobe'? *kmicu is happy to see some numbers “Over 3000 GB were served [by Nix’s Cachix] in December 2019” <jackhill>Ashjkaell: I mean what in /gnu/store/daq5zs7ni529zh3xxgyhidna52wa17js-modprobe <kmicu>(That like 600 Icecats… so not much I guess 🤷) <Ashjkaell>(A funny side effect of the bug is that the annoying vterminal beep my t400 does normally is disabled) <janneke>so, what if activation-service does not get run at the proper time; could that happen? <civodul>janneke: it runs before shepherd is started so that should be fine <janneke>yeah, i get only so much time before the system freeze or the keyboard starts looping -- not very tempting to boot into that to gather information <jonsger>Nextcloud 18 running on Guix system. Still very hacky (no package), but it lets me stay on Guix for my server OS :) <nckx>jonsger: Woot! Does it use Node? <shader>jonsger: I was almost really excited; was just talking about setting up nextcloud, but I guess I'll have to keep waiting for a package... <jonsger>shader: the package is still a long way. But I'll share my configuration as soon it's polished up. Then it's like extracting a zip, running two nextcloud commands and thats it <bdju>Maybe if it made it to the top 100 more people would try it out. 