IRC channel logs

2022-05-02.log

back to list of logs

***stikonas_ is now known as stikonas
***rekado_ is now known as rekado
<Ox151>hi, what is the correct place to put the GUIX_PROFILE="" . $GUIX_PROFILE/etc/profile prompt that shows when installing a package? i copy and pasted the 2 lines that is outputed into my bash_profile, but i still get the message and GUIX_PROFILE is not added to my env variables. am i mssing something?
<mekeor[m]>0x151: did you try to use a login bash? try to run `bash -l` and then `echo $GUIX_PROFILE`
<Ox151>mekeor[m]: yeah that worked, but shouldnt it be set by default once i log in once for all bash instances?
<Ox151>i just saw in the guix best practices to include a 'for i in $GUIX_EXTRA_PROFILES/*' block, ill try that.
<mekeor[m]>you need to log out and re log in
<mekeor[m]>e.g. using your login manager
<mekeor[m]>or using the vty-login-prompt thing
<Ox151>yeah thats why i am confused, I had it set previously and have logged in multiple times. oh well, try it again.
<Ox151>on a side not, does anyone know what is writing all these nil#-Guix~ files to my root directory?
<mekeor[m]>you can also make your terminal emulator spawn bash -l instead of just bash
<mekeor[m]>Ox151: just a wild guess: maybe some kind of bash-alias or -function or a script of yours?
<Ox151>yeah could be related to emacs. sorry, I saw the Guix at the end of the file an thought it was related to Guix OS, but forgot my hostname is Guix. please disregard. Ill ask in #emacs
***rekado_ is now known as rekado
<lechner>Hi, is there a way to provide Kerberos kdc and admin services from a Guix System?
<Rampoina>is there any plan to support appimages for guix pack?
<apteryx>Rampoina: nobody is working on it that I know, but that'd be a welcome addition
<apteryx>I'm planning to add RPM at some point, if nobody beats me to it
<Rampoina>apteryx: nice!
<apteryx>lilyp: is renpy not python 3 compatible?
<apteryx>it built fine with python 3, great
<lilyp>apteryx: not yet afaik
<lilyp>the python2 renpy internals are already futurized, but all the game logic stuff appears to still depend on python 2 last I checked
<lilyp>full python 3 migration is planned for renpy 8, which will be out whenever
<civodul>Hello Guix!
<mothacehe>hey civodul!
<AceTheMercenary[>Yo
<littlebobeep>civodul: What is WKD??
<littlebobeep>What eactly is wrong with pgp.mit.edu because vagrantc seemed to indicate this is for some reason not sufficient for web of trust?
<rekado>apteryx: /dev/sd{r,q,t,s} are four paths to the 100TB volume on the SAN; would you like to move anything there? Install the system there? Copy the store?
<attila_lendvai>i wish https://issues.guix.gnu.org/54674#11 could get some love. i can't publish my channel until it gets merged (or gets rejected, and i will have reworked my service code).
***dustfing` is now known as dustfinger
<civodul>ah, i missed mothacehe
<civodul>littlebobeep: WKD is "web key directory"
<wingo>hey guix ppl
<wingo>i have a dumb question
<littlebobeep>civodul: Thanks, what's the difference between WKD and browsing an PGP keyserver with a web browser? You called gpg --locate-keys something like WKD, but if just on the CLI how does that involve the web?
<wingo>so i need to build v8 from git on this guix machine. for that i need to run the whole depot_tools thing (weird google dependency management system)
<wingo>probably the easiest path to victory on this laptop is to run debian in docker on guix
<wingo>how do i do that? do i need to reconfigure my system to run a docker service or something?
<wingo>otherwise if people know how to get depot_tools working on guix that would be fine too
<singpolyma>wingo: do you need docker or can guix support chroot?
<wingo>i don't necessarily need docker, probably other things can work
<wingo>oddly i tried gnome boxes to install a debian dvd and it failed in mysterious ways
<littlebobeep>wingo: https://commondatastorage.googleapis.com/chrome-infra-docs/flat/depot_tools/docs/html/depot_tools_tutorial.html#_setting_up
<wingo>yeah but when i run "gclient" having just checked out depot_tools it fails with a strange (to me) python3 error
<allana>wingo: I'm probably just a fly on the wall on this, but I can say that the docker-service-type works well for me on my guix system.
*wingo tries docker-service-type
<allana>I think a few years ago I had issues running that service and I had to add "overlay" to the operating-system's initrd-modules.
<civodul>wingo: yeah, docker-service-type should let you do that sort of thing i guess
<civodul>littlebobeep: if i'm not mistaken, --locate-keys implements the WKD mechanism
<littlebobeep>civodul: Okay but how is that "web"?
<civodul>that's what it's called, but yeah, it'd debatable
<civodul>(maybe i'd be more accurately called "DNS key directory", dunno)
<civodul>*it'd
<littlebobeep>I see, okay
<littlebobeep>civodul: Do you know about the web of trust issues with pgp.mit.edu that vagrantc mentioned?
<civodul>littlebobeep: i think we were referring to https://dkg.fifthhorseman.net/blog/openpgp-certificate-flooding.html no?
*attila_lendvai has debootstrap-ped a debian chroot successfully not so long ago
<sneek>wb yewscion!
<yewscion>sneek: botsnack.
<attila_lendvai>wingo, maybe this will help: https://linuxconfig.org/install-debian-server-in-a-linux-chroot-environment
<sneek>:)
<attila_lendvai>wingo, actually, here's all i have: https://paste.debian.net/1239681/
<wingo>attila_lendvai: neat trick! i should try something like that one day :)
<wingo>managed to get v8 compiling from git in docker btw. seems to work ok
<attila_lendvai>are services and home services interchangable? i can't seem to simply add a syncthing service in my home config... i get: guix home: error: no target of type 'boot' for service 'shepherd-root'
*attila_lendvai is trying out home services for the first time
<abrenon>hello guix
<gnoo>attila_lendvai: no they're not. it's better to create your own home services.
<gnoo>also they can't communicate with each other.
<attila_lendvai>gnoo, i'm sure i'm still missing some context here, but it seems to be a reasonable assumption that they should be interchangable. is there so much difference between shepherd running as init, and as a user process, from the perspective of a service implementation?
<gnoo>service syncthing probably depends on network which since the shepherds don't communicate, the user shepherd can't know when that happens.
<attila_lendvai>gnoo, hrm... that's an interesting design issue... doesn't sound like something that will be resolved in the near future.
<jpoiret>attila_lendvai: in any case, most system shepherd services are tied to things like being able to start as a specific user, have access to things under /var, etc...
<jpoiret>and yeah, there's no general abstraction unfortunately
<attila_lendvai>jpoiret, interesting tidbit here: i have implemented 3 system services in my channel, and i have already started to abstract out the serice related part of their config into something uniform. i'll propose something on the mailing list if i reach a good enough understanding of this.
<civodul>oh i'm a pronk, i mixed up shepherd destructors: service started by make-systemd-constructor and stopped by make-forkexec-constructor
<civodul>that shows the limitations of that interface, i guess
<attila_lendvai>home-environment-variables-service-type generates shell expressions that don't quote the value strings... is there a reason for that? or is that just an inconvenient omission?
<podiki[m]>civodul: is mesa suitable for staging? I heard in the past it was, but most recently we did in core-updates; if so I can get an updated patch done this week (unless 22.0.2 needs lots of work, otherwise I'll do the last 21.3.x)
<attila_lendvai>anyone from the home service devs around? why isn't there automatic quoting of the value in environment-variables->setup-environment-script ?
<yarl>Hello.
<yarl>What is the proper way to rebuild guix after "git pull"? I did "guix shell --pure -D guix" then "./bootstrap" then "./configure --localstatedir=/var" then "make" but got many  "Please consider running po4a-updatepo to refresh it." messages.
<yarl>And many "source file newer than compiled" but I assume that then those are being recompiled.
<yarl>right?
<abrenon>yarl: I think they are but in anycase you can still "make clean-go" if the message remains (will trigger a recompile of all involved module next time you "make", of course)
<abrenon>as for the po4a message, I don't really know, I thought it had to do with special operations from the translation maintainers who have to update certain files from weblate but I'm not sure
<abrenon>don't you have the message on a clean repo ?
<abrenon>(in any case, with the occasional ABI incompatibility, a full make clean is required once in a while, so it can't hurt if you can spare the CPU time)
<yarl>abrenon Thank you, yeah, the .go file are recompiled and the .po too, probably. another make does not warn.
<unmatched-paren>it's (guix gexp) that exports (gexp) et al., right?
<sleepydog>unmatched-paren: yes
<unmatched-paren>sleepydog: ok, then i'm confused (i added gexps to a package [in guixrus], and i have `(guix gexp)` imported, but guile says that `gexp` is unbound...)
<unmatched-paren>OOH i haven't unquoted before gexping :P
<unmatched-paren>is there a gexp equivalent to %build-inputs? is it #$inputs?
<sleepydog>are you inside of a builder?
<unmatched-paren>i'm inside `(arguments #:make-flags (list ...))`
<unmatched-paren>sorry `(arguments #:make-flags ,#~(list ...))
<sleepydog>i think you can do something like #$(package-inputs this-package)
<sleepydog>gexp's are kind of mind-bending
<unmatched-paren>ok
<sleepydog>i'm not sure if that include implicit inputs, if it matters
<lechner>Hi there, what's hydra.guix.org, please? I only see SSL_ERROR_BAD_CERT_DOMAIN
<drakonis>it was previously an instance of nix's hydra, before cuirass was ready
<podiki[m]>do we know at all what hardware guixers are using? Mesa has removed old DRI drivers in v22 (some we already didn't build on non-x86 platforms anyway)
<podiki[m]> https://www.phoronix.com/scan.php?page=news_item&px=Mesa-Classic-Retired
<podiki[m]>I locally updated libdrm, wayland-protocols, and mesa to the latest, which all builds fine
<podiki[m]>maybe it is too much for staging (perhaps a separate branch could be built so people can test) but once a year mesa updates in guix is rather slow for anyone on recent hardware
<unmatched-paren>is there something like `--keep-failed` except it keeps even successful builds?
<unmatched-paren>a package i'm making builds successfully but doesn't produce the right output
<unmatched-paren>so i want to inspect the build environment
<roptat>add (throw 'error) to a build phase :p
<unmatched-paren>roptat: ok, fair enough :P
<unmatched-paren>apparently the build directory is empty? how is that possible?
<unmatched-paren>(this is a go package)
<unmatched-paren>it's supposed to spit out a binary but the binary isn't being installed to $out/bin
<dustfinger>I have never used guix. I just purchased an mnt-reform (https://mntre.com/media/reform_md/2020-05-08-the-much-more-personal-computer.html) and was thinking of installing guix + exwm on it. The system only has 4GB. Someone recently mentioned that I might find guix a bit slow on an underpowered machine. They also mentioned that NixOS seemed faster than guix. Will the mnt-reform struggle with guix?
<sneek>Welcome back yewscion!
<yewscion>sneek: botsnack.
<sneek>:)
***rekado_ is now known as rekado
<unmatched-paren>ah, i have it working now :)
<lechner>drakonis: thanks!
<vagrantc>dustfinger: substitutes for aarch64 are sometimes a bit slow or never show up, so you'll likely end up building some hefty packages yourself
<vagrantc>dustfinger: no idea hjow much tinkering you'll need to get it to work. i might try it sometime soonish
<dustfinger>vagrantc: Thank you for the feedback.
<vagrantc>dustfinger: you'll definitely need to "bring your own bootloader" as that won't be supportable
<civodul>podiki[m]: hi! mesa would be for core-updates, according to https://guix.gnu.org/manual/devel/en/html_node/Submitting-Patches.html
<civodul>(sorry i saw your question just now)
<podiki[m]>civodul: no worries! I heard rumors that in the past mesa would be updated in staging
<brettgilio>If anybody is interested, there is a new general PLT group #plt - Thanks for attending my ted talk. Long live GUIX
<brettgilio>(Also hi, long time no see)
<podiki[m]>what would you think of a separate branch to build mesa+related updates so people can try? because I'm wondering if anyone with much older hardware will lose support in mesa 22
<dustfinger>vagrantc: Okay. Sounds like fun :-)
<civodul>podiki[m]: that said, you could ask for advice on guix-devel, maybe others will agree to have it done in staging
<civodul>brettgilio: howdy, wb! which ted talk?
<bjc>finally got out of tree kernel modules into the initrd
<brettgilio>civodul just poking fun at my advertisement announcement for the plt group :)
<bjc>it was easier and harder than i expected. mostly because of all the false starts on getting it working
<civodul>brettgilio: heh alright :-)
<bjc>but the actual code is pretty straight forward
<bjc>how long has mumi had the encircled "!"? and what does it mean?
<podiki[m]>civodul: yeah, might be worth a discussion there, and more generally I want to bring up things like feature branches (which we do have some and seems we have capacity to increase, easing the burden on e.g. big core-updates merges)
<civodul>podiki[m]: yes, i'm all for feature branches; that makes debugging easier
<civodul>bjc: it's for issues marked as "important" or "serious" i think
<bjc>ah, thanks
<podiki[m]>civodul: Yup! And easier for someone to test just the set of changes in the branch easily, just a quick change in their guix pull. Let's use all that good technology we have
<tricon>dustfinger: I am curious to see how you like the laptop. I have a PineBook Pro that I really like and have a general interest in "eccentric computers".
<civodul>speaking of which, i lost track of the emacs-28 branch :-/
<civodul>is it on ci.guix yet?
<jackhill>civodul: I think I've run into an edge case with the net inetd-style sshd. On a host using network manager with a wireguard interface, port 22 doesn't seem to accept connection on the wireguard address.
<civodul>jackhill: oh? but what does "netstat -tupla |grep ssh" say?
<civodul>prolly worth a mail on bug-guix or we're going to forget :-)
<lechner>Hi, what is the most concise way to instantiate a (gss-service-type) in config.scm, please? https://guix.gnu.org/manual/devel/en/html_node/Network-File-System.html
<lechner>with all defaults
<atka>hey guix, is there a technical reason docker is still on 19.03?
<drakonis>i think nobody bothered updating it recently
<drakonis>although i find you might want to use podman nowadays
<atka>yes, podman is still on an old version as well, and I don't think rootless mode works
<atka>ideally I would use guix containers but I'm not there yet
<atka>I can do the work to get things updated but I'm busy at the moment
<civodul>lechner: (service gss-service-type)
<civodul>that gives you the default config
<lechner>civodul: i get this, though http://paste.debian.net/1239731
<atka>I'm using guix for servers primarily so having up to date container tools and runtimes is needed
<civodul>lechner: did you write: (append %desktop-services (list (service gss-service-type))) ?
<civodul>remember that 'append' expects its arguments to be lists
<zamfofex>Hello, Guix! I hate to ask this so bluntly like this, but it has been a while since I opened this: <https://issues.guix.gnu.org/54832> I know people are busy, but I was just wondering whether it’s still not forgotten. This feels important to me because I still want to be able to use that Glibc version to update Mig and see once again if I can run the Hurd with networking.
<lechner>civodul: thanks! i was off by a parenthesis. i need to find a lightweight code formatter
<lechner>unfortunately, now i get http://paste.debian.net/1239733
<lechner>also, this statement about rpcbind-service-type does not seem to be true for gss-service-type "Many NFS related services use this facility. Hence it is automatically started when a dependent service starts."
<lechner>hi, why does emacs stumble over the read-only flag on /etc/config.scm, while zile does not, please?
<jackhill>civodul: it does show it listening: "tcp 0 0 0.0.0.0:ssh 0.0.0.0:* LISTEN - ". I'll write up a bug later. If there's any other useful information to gather, I'm happy to do so.
<lechner>Here is my formatted config http://paste.debian.net/1239734
<jackhill>hmm, looking at that output, the wirguard IP is v6, and I wonder, which the host doesn't have on any non-wireguard device, so I wonder if that's actually the issue
<lechner>jackhill: i have recently seen ssh not listen on ipv6
<bjc>you can configure it to v4/v6 only, but by default it listens on both
<civodul>lechner: looks like pipefs-service-type is broken or something
<lechner>civodul: thanks! i'll try to give the defaults explicitly
<civodul>lechner: oh i see; you can work around it by commenting out the pipefs-service-type line in your config
<lechner>civodul: i wish! guix system: error: service 'gss-daemon' requires 'rpc-pipefs', which is not provided by any service
<civodul>ah hmm
<lechner>is that the right place to look? https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/services/nfs.scm#n108
<maximed>sneek: later tell attila_lendvai: For quoting, see https://issues.guix.gnu.org/54469
<sneek>Will do.
<maximed>attisneek: botsnack
<maximed>sneek: botsnack
<sneek>:)
<podiki[m]>catching up...
<podiki[m]>atka: there is some WIP patches for docker update, it is a bigger update and need to figure out docker-compose (now part of docker); I haven't tried it
<podiki[m]>civodul: I heard of the emacs-28 branch and that it was built, but also never actually saw it on the CI