IRC channel logs

2020-12-20.log

back to list of logs

<profmakx>is there no samba service for guix?
<vagrantc>sneek: later tell ZombieChicken the wip-pinebook-pro branch had a working kernel and bootloader, although how to actually install it a bit more of an adventure
<sneek>Will do.
<vagrantc>sneek: botsnack
<sneek>:)
<pushcx>Is there a terminfo package somewhere? I did a barebones takeover of a Debian vps with Guix and after a successful reboot my terminal (alacritty) is unhappy. Backspace inserts a space, odd linebreaks, etc. (Maybe it's not terminfo?)
<pushcx>git.savannah.gnu.org is down - when I run 'guix pull' I get 502s, or a very long pause and then 'guix pull: error: Git error: SSL error: syscall failure: Connection reset by peer'. Is there a status page somewhere so I could get a feel for how large the outage will be?
<sss2>looks like overloaded
<sss2>same here
<ryanprior>there's a fairly up to date mirror at https://github.com/guix-mirror/guix
<ryanprior>3 hours out of date
<ryanprior>commits are signed, so you can verify whether it's a legit mirror
<pushcx>Can a user run their own services, or only root?
<marusich>A user can, but I can't remember how. If you search the guix-devel email list for keywords like "shepherd user services" you'll find discussions and examples.
<marusich> https://lists.gnu.org/archive/cgi-bin/namazu.cgi?query=shepherd+user+services&submit=Search%21&idxname=guix-devel&max=20&result=normal&sort=score
<pushcx>marusich: thanks!
<guix-vits>pushcx: also there was a blogpost on guix.gnu.org/blog
***pushcx_ is now known as pushcx
<ride>Hello! How do I fix messed up icons on gui applications installed via guix? I am on a foreign distro and the messed up applications are both qt and gtk.
<ride>I will send screenshots in a second, but I will have to blur some information out.
<ryanprior>ride: does this happen in all gui applications?
<ryanprior>Or I guess more specifically, which ones have you tried?
<ride>ryanprior: I just found the same issue on an application installed with another package manager, so it does not seem to be a problem with guix. Thanks anyways!
<ryanprior>Okay, hope you can get it sorted out!
<bandali>regarding the savannah issues earlier, there was a network outage affecting all fsf/gnu servers earlier tonight
<bandali>things are back to normal now, though
<ryanprior>Thank you for the update bandali :)
<bandali>cheers ryanprior :-)
<guix-vits>do some list of 3party substitute servers exists? i have found none.
<guix-vits>remember it was mentioned there, searched logs to no avail.
<marusich>I haven't heard anything about it. The only third-party substitute servers I've heard about are mirrors of the official build farm, rather than build farms of their own.
<marusich>There is one for China, I think, since the download speed is not so great between China and Europe. You might find information like that if you search the guix-devel email list for keywords like "substitute mirror"
<marusich> https://lists.gnu.org/archive/cgi-bin/namazu.cgi?query=substitute+mirror&submit=Search%21&idxname=guix-devel&max=20&result=normal&sort=score
<marusich>I don't think there's a list, but I might not know.
<Aurora_v_kosmose>How would one generally bootstrap guix?
<Aurora_v_kosmose>oh, manual has it.
***apteryx is now known as Guest47059
***apteryx_ is now known as apteryx
<raghavgururajan>Hello Guix!!!!
<guix-vits>hello there man.
<guix-vits>how are you been? how gnomes, how cmake?
<raghavgururajan>I am not bad.
<raghavgururajan>How about you?
<raghavgururajan>gnomes are stuck.
*raghavgururajan took zopliclone and he will be out like a light in some time
<olivuser>hello guix o/
<olivuser>when you go about to tweak your system configuration and run into errors, how do you iron them out? I mean what procedure do you personally follow to identify and erase those errors? I'm new both to programming and programming in guile, so I wonder if a utility such as the guix repl is of any use in such case and how it would have to be employed
<guix-vits>raghavgururajan: thank you, i trying to guess what distro will tolerate my ass-growing hands to install it on sbc. then i'll boot into x86-64 guix, and udate it also.
<guix-vits>oliveruser: adding one feat at a time.
<guix-vits>*hello
<olivuser>guix-vits: thank you for your reply. Still, how to proceed when adding this one feat doesnt work?
<olivuser>I mean currently I am trying to configure mcron jobs, and I had been trying to replace GDM with lightdm. If any of such "one feats" fails, how do you proceed?
<guix-vits>olivuser: guix system search servicename
***chrislck_ is now known as chrislck
<guix-vits>guix.gnu.org/manual/devel/ on single page, to search for things (or M-x info in emacs)
<guix-vits>also post to this channel the config that doesn't work
<guix-vits>people can dry-run it, to see the error
<guix-vits>those that cannot this or another way, may appretiate the error message/pasted backtrace.
<guix-vits>sometimes people use strace, if thing builds, but the result is abscent.
<guix-vits>i think if problem is bigger, it may well worth to write to mailing list.
<guix-vits>olivuser: idk mcron, buit why you cannot swap gdm? manual has example for 'modify-services.
<guix-vits>no.. something with 'remove.
<guix-vits>slim-service-type has example how to swap gdm
<guix-vits>olivuser: ^
<guix-vits>he, i'd forget someone added boatloads of anchors to html manual
<guix-vits> https://guix.gnu.org/manual/en/guix.html#index-slim_002dservice_002dtype
<olivuser>guix-vits: thanks for the details, much appreciated. So far, I have noting that needs to be reported, because I can slowly start to figure out the easy things by myself :)
<guix-vits>civodul: hello. any way for anchors in html manual, on single page version, to point to nodes?
<guix-vits>copy pasting the link, like this: https://guix.gnu.org/manual/en/guix.html#index-slim_002dservice_002dtype
<guix-vits>loads the whole manual.
<mroh>guix-vits: This link jumps to the anchor. Browser has to load the page to do this, no?
<guix-vits>mroh: single page a whole manual != node a manual's page
<guix-vits>i want to easily share small pages
<guix-vits>can the single page be generated this way?
<cbaines>guix-vits, on the subject of third party substitute servers, both I and nckx run one. They're at https://guix.cbaines.net and https://guix.tobias.gr respectively
<olivuser>is anyone here using emms+mpd?
<olivuser>if so, is the configuration any different from a "usual" linux machine? I mean I just configured the mpd-service-type, but I am wondering if emms will simply work or if supplementary steps are required
<guix-vits>cbaines: thank you.
<guix-vits>sneek: later tell marusich: thank you.
<sneek>Got it.
<guix-vits>sneek seen hector?
<sneek>Nope.
<guix-vits>B)
<profmakx>mhhm. wonder whether i should start runniing one or two subsitiute servers for aarch64
<dftxbs3e>sneek later tell lfam, removing that third-party channel fixes that gcrypt bug
<sneek>Okay.
<dftxbs3e>profmakx, would really love to run a substitute server too
<dftxbs3e>Does GNU Guix accept contributions in the form of pull requests from a remote?
<profmakx>i suppose there are two issues with running a aarch64 substitute server: finding aarch64 hardware with enough oomph and (for me) having the upstream bandwith for those packages to be useful for anyone else
<guix-vits>profmakx: consult dannym?
<dftxbs3e>profmakx, Ampere Altra seems powerful
<profmakx>I am still a bit surprised by the fact that there isn't a sambas service yet; is it somehow complicated to make one? is ther a solution that i didn't find yert?
<guix-vits>as i understood, much of aarch64 is semi-proprietary at best.. idk, tho
<dftxbs3e>Don't know about Ampere Altra's proprietary-ness but it has OpenBMC and EDK2 UEFI BIOS (both Free Software)
<guix-vits>gnu approved distros seem to focus on those armv7 ? idk :)
<dftxbs3e>What hardware does current GNU Guix CI run on?
<guix-vits>anyway, cannot recommend my board for Free endevours (rockpro64). love it so far, tho
<dftxbs3e>I doubt it's RYF-certified
<guix-vits>heh, guess if GNU Guix CI runs on virtual aarch-es? joke
<profmakx>I am extremely happy with my rockpros
<profmakx>they are not really suitable as build machines though
<guix-vits>on guix?
<guix-vits>hdmi?
<profmakx>I have guix on one of them, but not guix-sd
<guix-vits>arr
<profmakx>haven't gotten that far yet
<guix-vits>my worked, then pew
<profmakx>hdmi does work with a vanilla kernel; no idea about a libre kernel, havent tried, wouldn't know what would be in the way potentially
<guix-vits>tried last working commit, no
<guix-vits>hdmi, when it was, worked on libre kernel
<guix-vits>but only place where hdmi really worked, in right definition, was armbian legacy (buster?)
<guix-vits>so, definitely there are frendlier boards (bsd has no hdmi at all on rockpro :p)
<profmakx>¯\_(ツ)_/¯ freebsd
<guix-vits>open, netbsd
<profmakx>anyone using guix system containers with some sort of networking setup? I was under the impression that I could meybe use them in the same vein as linux containers?
<GewaltDisney>if I want to install packages from one user to another without downloading them again, what is the best way to do this?
<GewaltDisney>user account*
<profmakx>wouldbn't they be in the global store?
<guix-vits>try symlinking. user who has a symlink may be unable to pull, tho
<guix-vits>~/.guix-profile --> ~other/.guix-profile, or same for ~/.config/guix/current
<guix-vits>or so
<GewaltDisney>word
<guix-vits>?
<GewaltDisney>sorry, word is slang for "got it, thanks" in the states
<guix-vits>cool
*guix-vits galaxy is in danger https://github.com/NixOS/nixpkgs/issues/107250
<Noisytoot>GewaltDisney: All of the files in ~/.guix-profile/bin and ~/.guix-profile/lib are symlinks to the global store
<Noisytoot>GewaltDisney: so just install them normally
<GewaltDisney>cool, thanks for the heads up, i just wanted to make sure that wouldn't initiate a ton of unneccessary downloads
<guix-vits>tbh, .guix-profile is the installed packages, .config/guix/current is "package list" + `guix` that was fetched by guix pull.
<ani>Hi guix.
<guix-vits>sneek: later tell any hello
<sneek>Got it.
<jonsger>roptat: it seems that JOSM doesn't run on wayland, I'm not sure if this is a guix bug or an upstream one
<asterope>I was wondering, is there a way to prevent an user from filling up /gnu/store by installing too much packages?
<rekado_>dftxbs3e: we don’t do pull requests, but you can link to your repository and we’ll check it out
<guix-vits>asterope: plug out the eth cord
<amirouche>is the person working on guix + gnunet around ?
<dftxbs3e>rekado_, what are the cons of pull requests?
<cbaines>I've never used git request-pull, but the downsides I can think of is that it makes reviewing and sending comments over email slightly trickier, as you have to find the commits and copy the content you want to comment on in to an email
<amirouche>dftxbs3e: there is https://www.gnu.org/software/repo-criteria-evaluation.html => no PR based system has A also PR review is mess.
<cbaines>but generally I'm not too familiar with it
<mbakke>o_O wip-arm-bootstrap, exciting!
<amirouche>github (and prolly gitlab) as email and reply to email for PR but reading the a whole review even via github gui when they are multiple reviewer and push force et al.
<amirouche>I like PR but it is far from perfect. The happy path when the PR is perfect and just need to click merge is neat. but that is rare.
<cbaines>amirouche, my guess is that this conversation is about Git style pull requests, not GitHub style pull requests
<amirouche>oh
<cbaines>I've never used Git pull requests, but this is the docs https://git-scm.com/docs/git-request-pull
<cbaines>I think it's quite common for Linux development, especially when you're requesting someone merge lots and lots of patches
<amirouche>so far my prefered workflow was git pull the branch that must be reviews, then comment code inline, commit and ask the OP to pull my review
<amirouche>compared to patches, it allows to comment on code that is outside the patch perimeter.
<mroh>jonsger: I guess, the eog and eog-plugins bump/patch also belong to gnome, right?
<jonsger>mroh: yes
<mroh>ok, ty!
<jonsger>updating gnome all together was at least how we did in the past
<mroh>what exactly is part of gnome?
<jonsger>good question :)
<mroh>I think, this "rule" results in total stagnation for gnome related things. (and far to much bureaucracy (for guix)).
<mroh>How about updating kde and/or qt related stuff all together -> no more obs or kdenlive bumping inbetween...
<rekado_>mroh: we update epiphany whenever there’s a new version and that’s officially part of Gnome.
<rekado_>the point is to upgrade things together that need to be compatible
<rekado_>that’s the case for some Gnome things but not for all of them.
<rekado_>the current branch where Gnome upgrades happen is wip-desktop, not wip-gnome-3.36
<aecepoglu[m]>brettgilio: No builder for dotnetcore. Got it.
<aecepoglu[m]>I guess my question for the moment is if I am porting a package (jellyfin) to guix should I first create a builder for it or build it on top of trivial-build
<cbaines>aecepoglu[m], I'd use the gnu-build-system, and modify the phases where useful
<cbaines>the trivial build system is only for things with a trivial build process
<aecepoglu[m]>okay, it's not so much about "trivial" vs "gnu" but "trivial | gnu" vs "creating a builder". I wonder if people build dotnet packages on guix
<cbaines>I don't know much about C# or .NET but it doesn't look like jellyfin really uses a build system
<aecepoglu[m]>except "dotnet build", no
<cbaines>right, so I'd just start simple
<cbaines>build systems are to generalise similar processes across multiple packages
<cbaines>so it's best to write them when you have several similar package definitions to generalise across
<aecepoglu[m]>:) thanks
<aecepoglu[m]>but, how de we know when there are "several" packages?
<aecepoglu[m]>does anyone ever do a refactor, realizing "oh we have X number of opam packages all of which have implemented their own build/install etc codes. Lets create an opam builder for those"
<cbaines>I don't have a specific example
<cbaines>In the case of C#, maybe you can find similar packages by looking at which packages use mono
<alextee[m]>anyone using chromium have this issue? my logins get reset when i restart the browser
<alextee[m]>i need to re-login to every web app i use
<mbakke>alextee: toggle the flag in settings that clears cookies on quit
<leoprikler>I get a hash mismatch on p11-kit, is this expected?
<leoprikler>oh, it's unstable github yet again
<helaoban>hey guix: After a fresh 'guix pull' I can no longer reconfigure my system with 'guix system reconfigure <config-file>'.
<helaoban>I'm getting a hash mismatch for p11-kit.
<leoprikler>helaoban: just got the same, working on it
<helaoban>leoprikler: ah, didn't see your message till after I was done typing!
<helaoban>leoprikler: so this was just a bad sha256 pushed upstream?
<leoprikler>I believe it's Github tarballs being shaky
<alextee[m]>mbakke: ah found it thanks!
<helaoban>leoprikler: If run guix pull --commit=<commit-before-update-to-p11-kit>, guix will allow me to downgrade?
<helaoban>ah, there's the '--allow-downgrades' flag.
<mbakke>ah, the TeX Live 'oberdiek' package has been split up into many smaller ones
<mbakke>hm, but these smaller ones are generated
<profmakx>does anyone actually *use* guix system container? it somehow escapes me how one'd setup such a containe with networking etc (and it not rebuilding ssh host keys every time i start it)
<rekado_>mbakke: splitting up the oberdiek package is a good move.
<mbakke>rekado_: it was forced upon me by the TeX gods!
<rekado_>mbakke: hail to the TeX gods!
<mbakke>currently I'm adding all the pre-generated files, which feels kind of wrong
<rekado_>you asked about kpathsea earlier: I was under the impression that this is part of texlive-bin; at least it should be, in my opinion.
<rekado_>it’s one of the core scripts to compute locations
<rekado_>uhm, if the source is really still the oberdiek package we could create new packages that inherit from oberdiek but only build part of it.
<mbakke>rekado_: yes indeed, there is likely something odd with texlive-base or texlive-union, I get an error message from kpathsea when building mit-scheme, but adding kpathsea to its union works, heh
<mbakke>rekado_: the oberdiek packages seem to have their own .dtx in source/latex/<pkg>, but I'm not sure where the "generic" variants come from
<mbakke>oh, they are also in source/latex
<civodul>profmakx: as you note it's still quite rudimentary; sharing your use cases and needs would help improve it
***grumble is now known as SeasonsBeatings
<profmakx>civodul: so what I would like to do is define what's essentially a "system" (so, packages/services, user accounts) and then run them like with lxc;
<profmakx>civodul: I can hack up network config with namespaces (i did it by hand just now)
<civodul>profmakx: what does lxc "consume" as input? would "guix system docker-image" help?
<profmakx>but that should somehow be part of the container script
***jess is now known as sandy-claws
<profmakx>lxc uses a rootfs and some network config that creates interfaces
<profmakx>(and i don't know what a "docker image" entails, so I can't really answer that question)
<profmakx>if its not a huge multi-month project i'll happily hack up patches
<civodul>profmakx: does it take like an ext2 root file system?
<mbakke>phew, now I need atveryend.sty, I hope the name is symbolic
<dissoc>im trying to make a package for suricata but during the build it runs a command in the Makefile using cargo to grab dependencies for the rust part of the package. im not sure how to handle this.
<dissoc>maybe use the cargo build system and then add the required gnu phases and strip out the Makefile part??
<dissoc>are network requests not allowed in package definitions?
<dissoc>it seems like i cant do a wget or a curl via (invoke...)
<davidl>pushcx: for problems like "backspace inserts a space, ..." you can try just running: stty sane
<cbaines>dissoc, package builds can't make network requests, as that would go against declaring inputs to a build
<cbaines>dissoc, dealing with cargo is pretty complicated, but yeah, maybe you can adapt what the cargo build system does
<pushcx>davidl: Unfortunately that does not change the behavior.
<mbakke>dissoc: there are examples on using cargo outside of cargo-build-system in 'icecat', 'mozjs-78', and 'librsvg'
<dissoc>cbaines: that makes sense.
<dissoc>mbakke: thanks. i will look at those
<davidl>pushcx: :-/
<pushcx>I added 'export TERM=xterm' to my .bashrc and that's close enough that I haven't yet seen any issues with alacritty, but it would still be nice to have its proper terminfo. Installing the 'alacritty' package seems like way overkill as this is a tiny vps, not a desktop.
<mbakke>brettgilio: is it possible to make straw-viewer use an absolute reference to youtube-dl instead of propagating?
<pushcx>Is anyone aware of an example of packaging a rails app for guix?
<cbaines>pushcx, I did some work a while back on packaging Rails itself, I also played around with using Rails without bundler https://git.cbaines.net/rails/guix-demo/
<pushcx>cbaines: Thank you! Do you happen to know if this is currently working?
<cbaines>I don't
<cbaines>are you looking to use Guix to develop a Rails app, or package an existing one?
<pushcx>Package an existing rails app to run on a guix-based vps.
<pushcx>(and if that works, it'd be natural for it to take over my dev env, really)
<cbaines>rubygems isn't particulary necessary, but it's not a problem either. Bundler might be a problem though, and I'm assuming this app uses bundler
<cbaines>also, Rails as a framework still lives in a world where a reasonable deployment pattern is running the app from the Rails root
<pushcx>It does use bundler. And... uh, I've been in Rails for a dozen years so it's my normal, could you be more explicit about what's unreasonable there?
<cbaines>well, assuming you package the Rails app for Guix, the files are going to be within /gnu/store
<cbaines>which is readonly as far as the app is concerned
<pushcx>Ah, right. I mean, I'd live with a checkout of the rails repo sitting around r/w, I just don't see bundler (esp native extensions) playing well with a guix distro.
<cbaines>if you're wanting to switch to using Guix, you could consider just dropping bundler
<cbaines>these changes worked for me https://git.cbaines.net/rails/guix-demo/commit/?id=27d364f337cc1e1b9838e0a1451e229d86097c7a
<pushcx>Huh. Where's the seam between this and guix? I see the 'gem' calls in config/boot, but I can't tell what wires that into guix?
<cbaines>It's all about search paths, Guix will manage the GEM_PATH for the Guix ruby packages (which include the rubygems gem)
<cbaines>so you'll get GEM_PATH set appropriately in profiles
<cbaines>Rubygems can search the GEM_PATH for gems, and adjust ruby's LOAD_PATH accordingly
<pushcx>Given that you successfully packaged rails - the git submodule shouldn't be needed here, right?
<cbaines>indeed
<cbaines>I should probably update the example, and/or maybe move it to the cookbook https://guix.gnu.org/cookbook/en/html_node/
<pushcx>Thanks for confirming. I'm getting the hang of thinking about things in the Guix way, so I really appreciate you talking through this. I'm going to see if I can get this demo app running in a prod-shaped config (nginx, worker pool, etc).
<cbaines>you're welcome :)
<rekado_>does anyone know if there’s a way to use PAM authentication with a Guix application on top of a foreign distribution?
<rekado_>pam is implemented as a library, but perhaps there’s a service I could install on the foreign distro to provide a different interface?
<rekado_>I need this for running RStudio Server from Guix on a foreign distro.
<rekado_>in the case of RStudio Server I can override the PAM helper binary that talks to the foreign distro’s PAM. That would be a binary that’s compiled on the target system.
<rekado_>but I wonder if there’s a general recipe for dealing with PAM authentication across the divide.