IRC channel logs

2020-08-30.log

back to list of logs

<joshuaBPMan>nckx: thanks. I guess I missed that. It shouldn't be too hard to fix though.
<nckx>Train those Guix Sysadmin skillz.
<joshuaBPMan>nckx: haha. Yeah, I've been looking through the berlin configuration file. There's lots of good stuff there!
<nckx>👍!
<joshuaBPMan>Also, if someone wants me to host a static website, please let me know.
<str1ngs>nly: Hello I have created an experimental webrtc branch. if you pull nomad ans switch to the webrtc branch. then guix environment nomad-git . you can test with make run. substitutes are also provided on http://gx.bufio.org you'll probably want those to avoid large rebuilds. you can test with this snippet in ~/.nomad. http://paste.debian.net/1161792 . please only use this for testing I need to do somewhere on the 'permission-request
<str1ngs>signal before you can use this for real. let me know if you have any issues. also test with jitsi if you can please.
<str1ngs>nly: s/somewhere/some work/
<str1ngs>nly: if you have an actually web cam you can drop the #:enable-mock-capture-devices #t for jitsi testing
***Guest46951 is now known as KE0VVT
<brendyyn>How can change a file-system so that the system still boots if it fails to mount?
<joshuaBPMan>brendynn: I think there is a way to say that this filesystem is not needed for boot...
<brendyyn>joshuaBPMan: needed-for-boot? is #f by default. i think that just means it wont be available before / is mounted
<brendyyn>my issue was my second drive was not recognised for some reason, so everything failed after that and i was stuck in shepherd
<joshuaBPMan>needed for boot is #f by default? hmmmm....DId you mark it as needed for boot?
<brendyyn>no
<mroh>try autofs and these boot problems are gone.
<brendyyn>mroh: how does one set that up with guix?
<mroh>brendyyn: try http://www.rohleder.de/2020/07/04/autofs/ as a starting point.
<mroh>for the conf files see `man autofs.{conf,master}`
<brendyyn>rather advanced
<peanutbutterandc>Hey there, I am working on a package that uses the cmake build system, and at the end of configure phase I am being warned "Manually specified variables were not used by the project: CMAKE_INSTALL_LIBDIR"
<sneek>peanutbutterandc, you have 2 messages!
<sneek>peanutbutterandc, leoprikler says: you give me too much credit, origin inputs are used in all of guix
<sneek>peanutbutterandc, leoprikler says: you give me too much credit, origin inputs are used in all of guix
<peanutbutterandc>leoprikler, Well, thank you, nevertheless for teaching me the guix-fu technique of doing that, sensei
<peanutbutterandc>...now, I did add another phase before 'configure that (setenv "CMAKE_INSTALL_LIBDIR" out/lib) and it still doesn't quite seem to work...
<peanutbutterandc>any ideas please?
<nckx>peanutbutterandc: The warning is warning you that ‘you’ manually specified a variable on the CMake command line, which the project didn't use. I.e.: it had no effect, was pointless, but harmless.
<nckx>I say ‘you’ because Guix passes some default variables like CMAKE_INSTALL_LIBDIR, not you. However, manually passing it *again* won't help 😛
<peanutbutterandc>nckx, But `guix environment -L . --ad-hoc the-package-i-just-made` && cd $GUIX_ENVIRONMENT && tree shows that there is no lib directory. And the package was acting kinda' buggy (whereas my manual non-guix build works well, kinda) so I thought this might be an issue
<nckx>Pretty sure that setting it as an environment variable has no effect at all.
<peanutbutterandc>nckx, I see.... would you like to take a look at it, too? https://termbin.com/jes0
<peanutbutterandc>It builds well... and works, but can't seem to be able to import videos
<nckx>Why should there be a /lib directory? The project not using CMAKE_INSTALL_LIBDIR is an indication that there are no libraries.
*nckx is building the thing.
<peanutbutterandc>When I try to create a new project and load a video (File > New Project > Input Video), it says "Number of video streams -1"
<peanutbutterandc>I wouldn't know, really. I'm just shooting with my eyes closed, hoping it'll hit the bull's eye. :D
<peanutbutterandc>that sounds awfully bad.... I'm trying to fix this issue the best I can with my limited knowledge
<peanutbutterandc>much better :D
<nckx>slowmoUI?
<peanutbutterandc>yes sir
<nckx>o7
<peanutbutterandc>Also, I noticed that build is a lot faster when it is being done by guix than when it is done outside manually
<nckx>Number of video streams: 1
<peanutbutterandc>What!?
<nckx>This is for a random .mkv in my home directory.
<peanutbutterandc>That is not so very cash money (reproducible) of the package at all
<peanutbutterandc>Can you load that video into slowmoui?
<peanutbutterandc>but perhaps I should build using guix time machine. That probably will fix the reproducibility issue
*nckx extracts thumbnails.
<peanutbutterandc>Oh wow
<peanutbutterandc>I'll throw a time machine up then. What commit are you in, please?
<peanutbutterandc>`guix describe`, I mean
<peanutbutterandc>Wow, never once in my life did I ever think I'll say "I'll throw up a time machine" and mean it.
<peanutbutterandc>...jumpstart
<nckx>I'm a few local commits ahead of 73a3085e346f7e6286acdc3e7b0ea3194ff0c317 that ‘should not’ affect this.
<peanutbutterandc>Thank you
<peanutbutterandc>Lo and behold 73a3085 is you, too. (Author, I mean)
<nckx>I used guix environment --ad-hoc ffmpeg -- `guix build -f foo.scm`/bin/slowmoUI .
<peanutbutterandc>Hmmm.... that is different. Let me try that first then before using a time machine
<nckx>So slowmoUI is not actually in that profile; I like testing things that way.
<nckx>It teases out bugs.
<peanutbutterandc>nckx, I do the same thing. guix environment is super cool.... ah! I thought i had ffmpeg in my normal profile (I was so sure!) but turns out I have not. Well, time to (wrap-program, then!
<peanutbutterandc>Thank you for your help
<nckx>Happy to.
<peanutbutterandc>Turns out to be a stupid mistake after all, like always
<nckx>Don't be hard on yourself. Be happy it's easy to fix.
<peanutbutterandc>I am glad all of you here are so kind and willing to help. Thank you very much for that. I really appreciate everyone here who has taken me under their wings one time or the other. (:
<peanutbutterandc>and you out-number others for the most part so thank you very much
<nckx>:) I'm proud of the community we have here.
<peanutbutterandc>Yes, it is a great one. And what I like most is that it doesn't quite seem forced. But everyone is just very chill. And willing to help out. So, maybe this is by virtue of a lot of chill people hanging out rather than a code of conduct.
<jgart[m]>I packaged this a few weeks ago and it builds successfully: https://notabug.org/jgart/guix-packages/src/master/working/python-pyjokes.scm
<jgart[m]>Would it be appropriate for upstream?
<jgart[m]>jokes as a service from the command line
<jgart[m]> https://pyjok.es/
<jgart[m]>Or should I just make an independent guix channel offering it?
<jgart[m]>I think it's important that we get a good laugh from the terminal every now and then
<nckx>I mean, an entire Guix channel (with introductory signature and of course home page) offering a single pyjokes package would be legitimately *funnier*...
<nckx>As the person who added ‘sl’ I'm probably not the best one to ask, but I think it's OK as long as the package is good.
<brettgilio>Is magit causing a stack size issue for anybody else using 27.1 from guix?
<brettgilio>The issue to be specific is Lisp nesting exceeds 'max-lisp-eval-depth' in a Warning buffer
<nckx>jgart[m]: That means writing a proper multi-line description &c.
<nckx>Go for it.
<mroh>brettgilio: check, if you have emacs-seq loaded (maybe as a dependency).
<nckx>brettgilio: emacs 27 was merged earlier than expected, which broke some packages. It's unlikely to be you, if that's what you're asking. That said I don't use magit.
<brettgilio>nckx: I definitely do.
<brettgilio>Use seq I mean
<nckx>That was mroh but I was about to ask the same thing so yay.
<nckx>Efficiency.
<brendyyn>jgart[m]: yes, there is no rule that says packages need to be "serious" or anything. if its free software and is desired or needed by anywone then it can be packaged
<jgart[m]>brettgilio: does magit have good support for revising history? I'm looking for a tool like this: https://github.com/pflanze/cj-git-patchtool
<brettgilio>jgart[m]: magit has everything
<brendyyn>you just press l to visit the history, grep through it and so on
<mroh>brettgilio: I think, if you use any of those `guix refresh -l emacs-seq` packages, your emacs27 is toast, not only for magit.
<brettgilio>mroh: thanks for the tip. I am going to pin emacs26 in the meantime
<nckx>If the breakage is this bad, should there be an emacs-26 package for now?
<brettgilio>nckx: I am fine with inferiors. I think we may be too far in to revert emacs now
<nckx>Oh, not revert, for sure.
<jgart[m]>brettgilio: I'm still meaning to get back to your email. Thanks for sending the info about debugging the guix channel for your emacs config package. I haven't found the time to test that again. I'm hoping to reply soon
<nckx>But OK, even adding a separate emacs-26 back is probably work that could be better spent on fixing the breakage.
<brettgilio>jgart[m]: oh that was you! No worries. I am messing with it right now
<brettgilio>nckx: it's probably just my philosophy but I agree. We should have started with a wip branch, we should just fix it and advise people use inferiors until we finish fixing it
<brendyyn>Hmm. Doom breaks my emacs's guix extension. When i run guix-edit i get No geiser repl for this buffer. if i run run-geiser, then running guix-edit says starting repl and errors with no prompt found
<jgart[m]>cool! Ideally it is how I would like to have my emacs configs managed (i.e. as a guix package definition)
<jgart[m]>brettgilio: brendyyn It would be cool to write something like this managed by guile/guix https://github.com/plexus/chemacs
<jgart[m]>so that one can boot into different "emacs distros" (i.e. init.el/~/.emacs) with the help of guix/guile
<jgart[m]>or maybe chemacs just needs to be packaged for guix... ha I haven't tried it
<brendyyn>there is more ascii art than actual code
<jgart[m]>ha
<jgart[m]>something that behaves like guix profiles and manifests but for switching between different emacs configs could be nice. Might be useful for those emacs users that want to quickly jump between doom and no doom
<mroh>jgart[m]: doom means: no guix emacs packages, yes?
<jgart[m]>yes
<jgart[m]>switching between managing you emacs packages with guix and using melpa, marmalade, etc...
<str1ngs>jgart[m]: guix has profiles :)
<mroh>maybe you could have 3 guix profiles: one "root", and the other two that inherit from "root" with guix-emacs and doom-emacs.
<jgart[m]>if a user wants to build packages on guix system in the traditional way for testing, experimentation without guix should they ```guix install debootstrap lxc``` and build them using one of the above?
<brendyyn> i dont have any special config appart from installing Doom emacs. with emacs -Q i can run guix-edit just fine
<str1ngs>jgart[m]: you can build it manually. say you want build bash locally. you can do guix environment bash then configure and make ... etc
<str1ngs>what I call the GNU three finger salute.
<jgart[m]>mroh: that sounds good. I know there was some talk on the mailing list of shepherd managing configs as a service. Has there been any progress on that?
<jgart[m]><brendyyn " i dont have any special config "> but if one wants to quickly switch between doom and guix managing emacs then -Q might not work?
<brendyyn>i dont care about how its managed, it was me discussing that. The only problem i have is that the emacs-guix extension is broken with . it wont start the repl
<brendyyn>it wasn't*
<raghavgururajan>jgart[m], Hey o/
<jgart[m]><brendyyn "i dont care about how its manage"> Are you using guix-el on a foreign distro?
<brendyyn>im on guixsd
<jgart[m]>raghavgururajan: Hey!
<jgart[m]>brendyyn: Not sure what it might be then. Maybe someone else here knows. It worked for me last time I used it on guix system but I haven't adopted guix-el fully into my workflow yet.
<jgart[m]>I've mostly been using guix from the command line with my shell aliases
<jgart[m]>it would be nice to make a port of this to vis at some point: https://guix.gnu.org/packages/vim-guix-vim-0.1/
<jgart[m]> http://guix.gnu.org/en/packages/vis-0.6/
*raghavgururajan sings "You don't know you're beautiful, that's what makes you beautiful" (1D), at Guix's logo
<jgart[m]>Now if I can only figure out the subset of sam structured regex I need to mimic a la carte from paredit in vis...
<peanutbutterandc>How do I get a list of all files in a certain directory inside of a phase?
<brendyyn>i think with (find-files dir ".")
<brendyyn>that may be recursive though
<peanutbutterandc>brendyyn, Ooooh! Thank you very much!
<brendyyn>this actually instructs people to put guix in their profile which im pretty is is always wrong https://github.com/alezost/guix.el
<peanutbutterandc>Okay... so, I just packaged a program and while running it, it says "QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-usernamehere'"
<peanutbutterandc>Now, my normal XDG_RUNTIME_DIR would be /run/user/uidhere
<peanutbutterandc>How can I change the package to use that? Not that it would do anything, still... just curious
<peanutbutterandc>should I put it in (wrap-program too?
<brendyyn>yeah you can wrap variables
<peanutbutterandc>brendyyn, thank you once again, cap'n!
<brendyyn>im not sure if that variable should be wrapped though
<brendyyn>mine is set to XDG_RUNTIME_DIR=/run/user/1000
<brendyyn>but that was set automatically somehow
<raghavgururajan>peanutbutterandc, brendyyn: DG_RUNTIME_DIR=/run/user/1000 is set by elogind-service
<raghavgururajan>peanutbutterandc, If you ran the program in a pure env, then you can ignore that warning. elogind is something that user will have to setup. The program installed and running with elogind-service, will not show that warning.
<raghavgururajan>> brendyyn‎: im not sure if that variable should be wrapped though
<raghavgururajan>It should not be. Because, if the program is wrapped with "XDG_RUNTIME_DIR=/run/user/1000", then the program will look for that dir/file, which will not exist in pure-env or in system without elogind.
<peanutbutterandc>raghavgururajan, I am on a foreign distro.... And yes, I did ran it in a pure env... so perhaps that is why it's saying that....
<peanutbutterandc>raghavgururajan, Yes... impure environment and it works just fine
<peanutbutterandc>Thank you!
<zjgkkn>Hi! Have this error when trying compile xmonad "collect2: error: ld returned 1 exit status 'gcc' failed in phase 'Linker'"
<zjgkkn>related packages in config scm: xmonad, ghc-xmonad-contrib, gcc-toolchain, ghc
<raghavgururajan>peanutbutterandc, Cool!
<peanutbutterandc>Okay.... I have a question: what is wrong with this .desktop file: https://termbin.com/4a4ul ?
<peanutbutterandc>It doesn't show the icon in the menu
<peanutbutterandc>but then again, the icon is being referred to using absolute path
<guix-vits>zjgkkn: Do You compile xmonad with `guix build xmonad`?
<guix-vits>*Hello
<raghavgururajan>guix-vits, 0/
<zjgkkn>guix-vits: no, but just try it -- doesn't work. May be `guix build ghc ghc-xmonad-contrib xmonad` will help?
<peanutbutterandc>raghavgururajan, Hey there, a package I packaged put it's icon in GUIX_PROFILE/share/icons (outside any other dir)... it is just a single 500x500 png file. Is share/pixbuf a better place to put it? o.O
<peanutbutterandc>s/pixbuf/pixmaps/
<raghavgururajan>peanutbutterandc, Is the icon GTK or Qt?
<peanutbutterandc>Qt
<peanutbutterandc>it's png but the program is in Qt
<peanutbutterandc>also where can I find information regarding such thing? XDG? Freedesktop?
<peanutbutterandc>and the best practices regarding that
<raghavgururajan>Just a sec
<raghavgururajan>peanutbutterandc, So the that icon is only used by that program right?
<peanutbutterandc>raghavgururajan, Yes, sir
<peanutbutterandc>strange thing is: while I was running from `guix environment --ad-hoc prog`, I can see the icon in the panel, but now that I've installed it into my profile that I can't see the icon at all
<raghavgururajan>peanutbutterandc, LoL why sir. So $prefix/share/icons is the default location for installing icons. Any sub-dirs following that, depends wupon use cases.
<raghavgururajan>peanutbutterandc, Wait a sec, you mean panel inside the program right?
<peanutbutterandc>raghavgururajan, No I mean the DE's panel
<peanutbutterandc>but I can't see it now, again. curious
<peanutbutterandc>I guess it has to do with the .desktop file then.
<raghavgururajan>I see. So it goes under $prefix/share/icons/hicolor/500x500/your-file.png
<peanutbutterandc>raghavgururajan, could you please take a look at this desktop file please: https://termbin.com/4a4ul ?
<raghavgururajan>hicolor is the XDG default. Honored by both Gtk and Qt
<raghavgururajan> https://specifications.freedesktop.org/icon-theme-spec/icon-theme-spec-latest.html#install_icons
<raghavgururajan>Sure, just a sec.
<peanutbutterandc>Hmm... so if I am to move the icon to that directory, I should move it manually after 'install, and change the .desktop file
<raghavgururajan>Ah! So it is coded to look at $prefix/share/icons. Therefore, the file location should be $prefix/share/icons/Your-File
<peanutbutterandc>raghavgururajan, And it is there. I wonder if it is because of the lack of .png at the end
<peanutbutterandc>but it refers to that file using absolute pathname
<peanutbutterandc>and still the icon isn't showing up
<raghavgururajan>Yes, you are correct. The .desktop file should be patched with .png
<peanutbutterandc>anyways, is there a procedure to move/rename files inside build phases?
<raghavgururajan>Or you could rename the icon file to just AppIcon
<guix-vits>zjgkkn: Plese paste the commands You ran, and the error. Someone should help (maybe even me, not a dev myself).
<jgart[m]>It would be great to have a website that aggregates working public guix channels. Kind of like those sites that list public instances of online instances i.e. https://instances.invidio.us/
<raghavgururajan>peanutbutterandc, Which path you wanna take? Patch the .desktop file or to move/rename the icon file?
<raghavgururajan>I suppose the former is less work.
<peanutbutterandc>raghavgururajan, I am kinda' tempted to do both: move the icon file to a better place and refer to that
<peanutbutterandc>I'm learning anyways
<peanutbutterandc>prefix/etc/icons/Appicon.png doesn't look all that great
<peanutbutterandc>what is the procedure to move/rename a file inside a build?
<raghavgururajan>peanutbutterandc, Just a sec
<peanutbutterandc>okay
*jgart[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/OFxiADkvDVjVAUuhNoFBcZNu/message.txt >
<raghavgururajan>So you are going to move+rename the file to $prefix/share/icons/hicolor/500x500/apps/AppIcon.png and patch the .desktop file respectively, correct?
<guix-vits>jgart[m]: Lenin1917. Ленин жив.
<guix-vits>"Even Lenin use Guix"tm
<raghavgururajan>peanutbutterandc: Here, https://bin.disroot.org/?187d7ffff218d8eb#8oJFBSdxKUPk4ahARv6i8zoYELgyMZmQv3syPSH3B333
<peanutbutterandc>raghavgururajan, (rename-file) moves files too? o.O
<peanutbutterandc>didn't know that... although mv
<raghavgururajan>peanutbutterandc, Yeppie!
<peanutbutterandc>raghavgururajan, oh wow! It does (just tested it in the repl). Hmm... the documentation should be updated (in the source code)
<peanutbutterandc>Thank you very much, BTW (:
<raghavgururajan>Anytime!
<brendyyn>to rename a file means to move it to the new path. whats the difference?
<brendyyn>... except the only the file name part of the path changes, nothing beftore the last /
<peanutbutterandc>brendyyn, this rename-file procedure works like mv; so perhaps it should be documented accordingly? "this works like `mv`"
<peanutbutterandc>rename kinda felt like what you just said
<peanutbutterandc>but it really works like (mv-file)
<raghavgururajan>> ‎jgart[m]‎: It would be great to have a website that aggregates working public guix channels. Kind of like those sites that list public instances of online instances i.e. https://instances.invidio.us/
<raghavgururajan>+1
<raghavgururajan>jgart[m], One issue though that any non-official instances cannot be constantly verfied for FSDG compliance. As it will be under control of the custom channel's owner and not guix maintainers.
<raghavgururajan>Any other channel can silently start distributing non-free software.
<jgart[m]>raghavgururajan: right. It would have to be a use at your own risk list/website
<raghavgururajan>jgart[m]: Now I wonder, did you mean to channel or substitute-server?
<jgart[m]>probably not endorsable by guix.gnu.org because of FSDG
<raghavgururajan>*mean to say
<jgart[m]>raghavgururajan: a website for each would be nice :)
<jgart[m]>since they are different things even though they are related
<brendyyn>Its an interesting proposition. I've thought about it too. What kind of channels should exist where the packages wont be put into guix proper, besides proprietary software?
<jgart[m]>I think it would be good to make a distinction between a substitute server and a channel not offering substitutes
<raghavgururajan>jgart[m]: Yeah! the substitute server should not be an issue. Because, when a user uses official channel but tries to download from custom substitute-server, then guix won't download due to mismatch of hashes.
<raghavgururajan>IIRC, channel = repo/source
<jgart[m]>brendyyn: custom software, overlays, experiments
<brendyyn>suppose for example guix deleted python-2, so someone created a python2 channel for that old stuff. but in that case i think it would be good for guix proper to include an updated /purpose/ based list of such channels
<raghavgururajan> https://guix.gnu.org/manual/en/html_node/Channels.html#Channels
<jgart[m]>brendyyn: suckless software with custom patches in a custom guix channel
<jgart[m]>that can be one thing
<jgart[m]>but a channel doesn't have to package software per se
<brendyyn>i need to explain what i mean. With debian/ubuntu, i often add a new Source for something specific, but over time that repo might get abandoned and there will be another one somewhere to replace it
<jgart[m]>this can be a guix package: https://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-001-structure-and-interpretation-of-computer-programs-spring-2005/video-lectures/
<jgart[m]>all the sicp videos as a guix package
<jgart[m]>there can be two versions of that package
<jgart[m]>the audio only version and the full video version
<brendyyn>so for example there could be a suckless channel, but guix proper will include a list of free software channels that have names like "suckless", which will point to which ever channel is the go to for latest suckless software. if in the future that channel gets abandoned but someone else forks it, guix can just change that path and things wont break for everyone
<jgart[m]>kind of like the way sicp is packaged as an info manual
<raghavgururajan>jgart[m], We can list channel mirrors though. All mirrors will be signed same.
<brendyyn>yeah, one of the things i want to do is create packages for all creative commons dictionaries i can find
<raghavgururajan>For example, I created a mirror for all guix's repositories, at https://git.disroot.org/guix
<jgart[m]>raghavgururajan: very nice!
<jgart[m]>It would be nice to have a guix channel that packages free tech books and/or books from project gutenberg
<jgart[m]>I'd rather search and download free books with guix than with my browser
<brendyyn>so guix's end game is to become a web browser? im not sure this idea would pass with management
<jgart[m]>Project Gutenberg has over 60,000 free eBooks
<jgart[m]>brendyyn: we have nomad for that https://o-nly.github.io/articles/nomad-web-browser/
<jgart[m]>what we need is `guix import gutenberg` as a keybinding from within nomad
<jgart[m]>so that I can add guix ebook package definitions as I browse Gutenberg's collection of 60,000
<jgart[m]>by add I mean append to gutenberg.scm
<jgart[m]>it seems like most of the books on gutenberg.org are available as Plain Text UTF-8 :)
<raghavgururajan>> jgart[m]‎: I'd rather search and download free books with guix than with my browser
<raghavgururajan>`guix search --web`? :-D
<jgart[m]>in nomad would be cool
<jgart[m]>nyxt has a shortcut that calls youtube-dl for example
<jgart[m]>nyxt also has a shortcut to call git clone on the current website
<jgart[m]>nomad could call `guix import gutenberg` followed by guix build or install
*raghavgururajan dreams GDE - Guix Desktop Environment
<peanutbutterandc>raghavgururajan, So... I did all of that and guess what? Still no icons
<jgart[m]> https://nyxt.atlas.engineer/article/the-nyxt-thesis.org
<jgart[m]> https://nyxt.atlas.engineer/article/emacs-hacks.org
<raghavgururajan>peanutbutterandc, What? No icons in the prog or no icons on panel?
<peanutbutterandc>raghavgururajan, In the panel and the menus
<peanutbutterandc>...of my DE
<jgart[m]> https://nyxt.atlas.engineer/article/command-line-programs.org
<raghavgururajan>peanutbutterandc, Did you double-check if the .desktop was patched sucessfully? You can use -K option to keep the build-dir
<jgart[m]>> * <@freenode_raghavgururajan:matrix.org> dreams GDE - Guix Desktop Environment
<jgart[m]> https://github.com/mwitmer/guile-wm
<peanutbutterandc>Yes sir
<raghavgururajan>peanutbutterandc, Ah I think I found out what the issue. According to XDG standards, there *should* be a icon at 48x48, to be used by DEs in panels and menus.
<peanutbutterandc>raghavgururajan, i cri evritiem it's only a 500x500 thingy. Oh well. Let me look at the `source` if there is something else there. Thank you very much for your help
<raghavgururajan>peanutbutterandc, Quoting FreeDesktop, "Minimally you should install a 48x48 icon in the hicolor theme. This means installing a PNG file in $prefix/share/icons/hicolor/48x48/apps"
<jgart[m]> https://invidious.tube/watch?v=sIN_-sbJ6QY
<peanutbutterandc>isn't that "minimally"? o.O
<raghavgururajan>🤷‍♂️
<guix-vits>jgart[m]: #nomad-browser
*guix-vits submerges
<raghavgururajan>peanutbutterandc, I checked linphoneqt. It has /share/icons/hicolor/scalable/apps/linphone.svg. But no other icon files. My bet is, an application *should* have a icon in either *48x48* non-scalable format (png) or in scalable format (svg).
<peanutbutterandc>raghavgururajan, I see... that makes sense... I'll try to find some another way to fix it. Thank you very much once again.
<raghavgururajan>peanutbutterandc, I think that because, scalable icons can be resized, where non-scalalable icons cannot be resized. So anything larger than 48x48 may not be compatible to be fit/viewed in menus.
<peanutbutterandc>Hmm... yes, probably
*raghavgururajan checks other GUI applications
<raghavgururajan>peanutbutterandc: Twinkle has /share/icons/hicolor/{16x16:24x24:32x32:48x48:scalable}. So there is 48x48
<peanutbutterandc>Hmmm...
<jgart[m]>just wanted to share something cool I just found out. If you want to use the guile repl from the terminal with vi keybindings and readline you can use rlwrap with this in your $HOME/.inputrc https://paste.gnome.org/pdimj63vu
<jgart[m]>no need to edit ~/.guile
<jgart[m]>`guix install rlwrap`
<str1ngs>guile has (ice-9 readline) already
<str1ngs>no need for rlwrap
<str1ngs>vi bindings also unholy :P
<jgart[m]>does it have vi support?
<peanutbutterandc>yes, (activate-readline) is required though
*str1ngs sprinkles holy water
<peanutbutterandc>lol
<jgart[m]>rlwrap has readline baked in
<jgart[m]>I've used active-readline before
<jgart[m]>this is just another way to achieve that
<str1ngs>jgart[m]: read always has vi bindins if you set -o vi and also add set editing-mode vi in ~/.inputrc
<jgart[m]>*(activate-readline)
<str1ngs>readline*
<jgart[m]>str1ngs: ohh ok good to know. thanks!
<str1ngs>no worries
<jgart[m]>str1ngs: do you happen to know how I can add custom meta commands in the repl?
<jgart[m]>I want `,t`
<jgart[m]>for going to top-repl
<str1ngs>I did at one point now I forget.
<jgart[m]>instead of having to press ctrl + d multiple times
<jgart[m]>to exit repls
<str1ngs>I don't think you can just back out that way
<jgart[m]>I back out with ctrl + d
<jgart[m]>or ,q
<str1ngs>that probably works the best
<str1ngs>I use geiser and I eval expressions files
<str1ngs>only use REPL under certain conditions
<jgart[m]>I also want to be able to reload my global environment and $HOME/.guile without exiting the repl
<str1ngs>err eval expression in file I meant.
<str1ngs>,l "/home/<user>/foo"
<jgart[m]>and without using emacs
<jgart[m]>I tried that with no luck before
<str1ngs>or (load (string-append (getenv "HOME") "/.guile")
<jgart[m]>I tried that too with no luck
<jgart[m]>I get errors
<str1ngs>should be ,l "/home/<user>/.guile"
<jgart[m]>path is correct
<jgart[m]>will that reload the global environment afresh?
<str1ngs>it will load ~/.guile that is all
<str1ngs>,l uses premitive-load IIRC
<jgart[m]>I put this in my .guile
<jgart[m]>(define ll (load (string-append (getenv "HOME") "/" "test.scm")))
<jgart[m]>but it doesn't work the same as ,l "/home/jgart/.guile"
<jgart[m]>I prefer not to create a global variable but I don't want to type the long command above. Ideally a custom meta command would be good here
<str1ngs>jgart[m]: change load to primitive-load
<str1ngs>check your bindings with ,b
<str1ngs>odds are it just in the scope of (guile-user)
<str1ngs>plus you can excursion into modules with ,m
<raghavgururajan>peanutbutterandc: Btw, how did you know that the size of the icon is 500x500?
<jgart[m]> https://paste.gnome.org/pidfgshi5
<jgart[m]>str1ngs: that's what I have in my .guile
<jgart[m]>but ll still fails to load test.scm
<jgart[m]>and update the global environment
<jgart[m]>I checked with ,b
<jgart[m]>I realize the two lines there might be redundant but I
<jgart[m]>str1ngs: any suggestions as to what I might be doing wrong?
<raghavgururajan>peanutbutterandc, may be just place file under 48x48 :-P (might work)
<str1ngs>jgart[m]: this will make a ,re meta command http://paste.debian.net/1161818
<str1ngs>jgart[m]: but you need to do more to get it to work.
<str1ngs>throws nknown file name for module #<directory (guile-user) 55b015f44140>
<str1ngs>jgart[m]: I need to get some sleep ping me if you get the ,re macro sorted out.
<jgart[m]>str1ngs: ok will do. thanks for all the help today
<peanutbutterandc>raghavgururajan, I opened it up with gimp
<raghavgururajan>I see.
<peanutbutterandc>well... I'll have to see. For now I'm pretty tired of this thing. :[
<raghavgururajan>500 is too big for an icon. I doubt if gimp auto-resized or something.
<raghavgururajan>peanutbutterandc, In your previous patch, replace 500x500 with 48x48, as a hail-mary.
<raghavgururajan>Cool!
<peanutbutterandc>raghavgururajan, Okay. I'll try to do it then
<peanutbutterandc>one last try
<peanutbutterandc>raghavgururajan, Nope, it does not work. All right I'll have to request icons upstream then, I guess
<peanutbutterandc>Thank you for your help, nevertheless
<nefix>good morning! Is it possible to build an operating system definition for a RPI3 and be ready to dd it?
<raghavgururajan>peanutbutterandc, Oh :( Sorry!
<nefix>e.g. guix system rpi-image system.scm
<peanutbutterandc>raghavgururajan, Its okay. Thank you!
***sneek_ is now known as sneek
<raghavgururajan>peanutbutterandc: When you are charged-up, try this. I found out something new. In .desktop, make icon line just as `Icon=AppIcon`. No png, no /gnu/[..]. Let the rename stuff be /share/icons/hicolor/48x48/apps/AppIcon.png
<raghavgururajan>That is because, `Icon=` attribute already defaults to $XDG_DATA_DIR/icons/hicolor/48x48/apps. If there is /gnu/[...] in `Icon=` line, then DE's will look for $XDG_DATA_DIR/icons/hicolor/48x48/apps/gnu/store/.../share/icons/hicolor/48x48/apps, which doesn't exist.
<raghavgururajan>I checked with other GUI apps. They have `Icon=prog-name` in .desktop file and `prog-name.png` file under /share/icons/hicolor/48x48/apps
<raghavgururajan>peanutbutterandc, ^
<peanutbutterandc>raghavgururajan, I was doing just that so far
<peanutbutterandc>:[
<raghavgururajan>Hmm.
<user12344556>Hi anyone ever tried to run guix package manager on rasbian/raspberry 3b+?
<raghavgururajan>peanutbutterandc, May be `Icon=AppIcon` in .desktop and location at `/share/icons/AppIcon.png`?
<user12344556>I just installed it but at guix pull it fails to build derivations with a segmentation fault on guix-packages-base
*raghavgururajan is also tired. 😴
<raghavgururajan>Later folks!
<peanutbutterandc>raghavgururajan, https://termbin.com/h6cp
<peanutbutterandc>just in case you feel like trying
<peanutbutterandc>thank you very much for your help
<raghavgururajan>peanutbutterandc, You said https://termbin.com/h6cp did not work right? May be try my last message ^
<peanutbutterandc>raghavgururajan, that is how it stands right now.... i've tried basically all those variations
<peanutbutterandc>anyways you and i both need some rest from this
<raghavgururajan>> peanutbutterandc, May be `Icon=AppIcon` in .desktop and location at `/share/icons/AppIcon.png`?
<raghavgururajan>I meant this. Your link contains moving to different location
<peanutbutterandc>oh okay i'll try it later
<peanutbutterandc>thanks
<peanutbutterandc>(i am too sick of it for now)
*raghavgururajan is somehow postive about this
<raghavgururajan>Cool!
*raghavgururajan back to sleepy night-shift
<peanutbutterandc>:)
<fnstudio>hi all! i'm following what reported here https://guix.gnu.org/cookbook/en/html_node/Basic-setup-with-manifests.html
<fnstudio>at some, in particular, this bit is reported: `guix package --manifest=/path/to/guix-my-project-manifest.scm --profile="$GUIX_EXTRA_PROFILES"/my-project/my-project`
<fnstudio>my understanding is that `--manifest` is the option to pass the instructions that i've to have ready in a .scm file
<fnstudio>on the other hand, the path passed via `--profile` doesn't have to be an existing file, but it just tells guix where to save the profile
<fnstudio>is my understanding correct?
<fnstudio>in other words it's not completely clear whether a profile is a name, a url to a file, or what
<fnstudio>*clear to me :)
<leoprikler>for this usage of --profile, yes, it's where to put it
<leoprikler>for other uses of --profile [e.g. --list-installed] it is the path to an already existing profile
<leoprikler>in either case --profile is always the location of the profile, not a name or something
<fnstudio>leoprikler: cool, thanks; and how about the `manifest.scm` that i pass via `--manifest`? is it good practice to save it in the profile folder? or can it be discarded straightaway after its use?
<leoprikler>well, you can save it right next to the profile, but in the profile would be a bit difficult
<fnstudio>right, difficult as in "it'd be overwritten"?
<leoprikler>where you put it is your preference. I personally use ~/.config/guix/manifest*.scm
<fnstudio>cool
<fnstudio>i'll try and follow that, thanks
<leoprikler>btw. has someone tried using non-package gexps with manifests?
<fnstudio>(cool, it worked, i now have my first (extra) profile, yay)
<Kimapr[m]>fnstudio: note that you can install things into profiles other than your main one without a manifest, just use the -i (install) and -r (remove) options with -p
<Kimapr[m]>e.g. 'guix package -p .games -i xmoto'
<fnstudio>Kimapr[m]: excellent, that's helpful, thanks
***catonano_ is now known as catonano
***modula is now known as defaultxr
<simendsjo>Hi. I tried `guix pack guile3.0-chickadee` to see how large the baseline would be if I should distribute a small project using it. Some of the dependencies are quite large, and I'm not sure why everyone is there. Is everything here necessary? Both llvm and gcc? https://gist.github.com/simendsjo/c9a6fca44f7b31680da90303eeb87bc2
<roptat>simendsjo, I think you can obtain the same result with "guix size guile3.0-chickadee" ;)
<roptat>now "guix graph" might give you more insight as to why all of these are included: guix graph -t references guile3.0-chickadee | dot -Tpng > graph.png
<roptat>(but it doesn't include propagated inputs, so it won't show everything)
<simendsjo>roptat: guix size reports 194MiB, while the pack output is 969M.
<roptat>yeah, I think it's the same reason: guix size doesn't show propagated inputs
<roptat>so for instance I don't see mesa in the graph
<simendsjo>No llvm either.
<roptat>so that's probably because guile3.0-chickadee propagates guile3.0-opengl and guile3.0-sdl2 (llvm is a dependency of mesa)
<roptat>you can see it with guix graph guile3.0-chickadee --path llvm for instance
<roptat>I don't think we have a type of graph that shows you exactly what's in the pack: the references type is a subset of them (only references, but not propagated inputs), the default one is a superset of them (it shows all the inputs, even when they're not used at runtime)
<simendsjo>Thanks, that's quite useful. Not really feasible to give someone a 1GB program for a really small thing though.
<roptat>well, people are used to huge games, no? :p
<simendsjo>Haha, but then they get more art than computer generated squares :)
<leoprikler>computer-generated triangles ;)
<brendyyn>simendsjo: it would be interesting to explore making portable apps with guix. i think you would need to define custom packages that stripped out many things. the thing about guix is that it includes literally every dependency
<roptat>splitting llvm into bin / lib / devel would be a win of 37 MB already
<brendyyn>but many of them could be deleted and integrated with the system
<brendyyn>for example what of there was a phase with guix pack that could delete the bundled terminal and replace it with a link to /bin/sh
<leoprikler>well, yeah, but how do you ensure that /bin/sh behaves how you want it to
<simendsjo>Yeah, I expect far from 1GB is needed to run this. I have no idea how it should be done, but it sounds time-consuming and that you'd need experience with guix.
<brendyyn>there are no guarantees for anything, its about compromising for practicality.
<brendyyn>well my idea actually has almost nothing to do with guix. for example you make a guix pack, extract it, then delete things that are not needed, and a strict that replaces them with links to their location on the host system
<brendyyn>you could delete documentation too
<raghavgururajan>sneak, later tell peanutbutterc: Let me if it works :-)
<raghavgururajan>sneek, later tell peanutbutterandc: Let me if it works :-)
<sneek>Okay.
*raghavgururajan +zopiclone = zzZ
<janneke>phew...seems i somewhat recovered from the 27.1 upgrade
<rndd>hi everyone! is there way to make my own directory for patches in my guix channel? any guides?
<civodul>janneke: i guess that'll be my task for tomorrow :-)
<civodul>rndd: hi!
<civodul>you can store patches anywhere in your channel, and then use 'local-file' to refer to them
<civodul>local-file allows you to refer to a patch relative to the source file where 'local-file' is used
<leoprikler>rndd: define your own search-patches procedure mimicking the behaviour of (@ (gnu packages) search-patches)
<janneke>civodul: hi! yeah, took me a quite a while to find that using (set-frame-font "ubuntu mono-20" nil t) makes debbugs-gnu segfault ...
<janneke>mail setup, starttls, 'cl deprecation -- well, great to see some progress i guess
<bandali>there's a lot of goodies in 27.1; including a couple of (annoying) bugs fixed in ERC :-)
<leoprikler>and sadly a bunch of baddies as well
<leoprikler>I'll miss you, XIM 😢️
<janneke>bandali: oh great, i should look into that -- i have been using ZNC.el to make erc usable with a znc server
<mjsb>Good afternoon everybody :) ! I was wondering, do you have any timeout trying to reaching out guix.gnu.org ? I was exploring the manual but it suddenly stopped answering
<janneke>mjsb: works for me...
<janneke>any url in particular?
<bandali>janneke, ha! i, too, use ZNC.el for now, but ERC should be usable with ZNC without ZNC.el, save for some minor annoyances addressed by ZNC.el
<bandali>hopefully we can improve that
<mjsb>janneke: I was reading https://guix.gnu.org/manual/en/html_node/Build-Systems.html
<janneke>bandali: ah, okay good -- yes, that would be great
<bandali>:-)
<janneke>changing (require 'cl) to (require 'cl-lib) in ZNC.el seems to work and help get rid of the deprecation warning
<bandali>ha, i'd wrapped it in an (eval-when-compile ...) for myself
<bandali>leoprikler, but that's (nearly?) all software :-) also, was XIM removed in 27.1?
<janneke>oh, and doing defun* => cl-defun -- that took me a bit to figure out
<bandali>right
<leoprikler>i'm not quite sure tbh
<janneke>not sure how to help "the internet" here -- everyone is running their stale copy of znc.el %-/
<bandali>yeah, i use my own fork
<bandali>would be cool if the author accepted patches
<leoprikler>the relevant code was changed back with the X11 update and it continued to work after my patch was applied
<janneke>any chance you know of a still-working jabber.el fork?
<leoprikler>the upstream patch shortcuts a lot of logic that would create an otherwise sane session
<bandali>janneke, i do not, unfortunately :-/ there's https://github.com/legoscia/emacs-jabber but i don't know what shape it's in
<civodul>janneke: segfault?! uh
<bandali>for jabber/xmpp, i'd probably look into something like bitlbee so i could use it from erc
<leoprikler>i'm not quite sure what the correct thing here is
<ngz>Hello. Is ci.gnu.guix.org down? I cannot retrieve substitutes at the moment.
<civodul>hi ngz!
<civodul>it's not down
<civodul>what error do you get?
<janneke>bandali: yeah, i've been using that but with 27.1 it seems broken now, even when substituting cl-lib
<bandali>janneke, i see..
<leoprikler>after investigating a bit, the libx11 update was related to xim and I think adding some extra code just to address this case is probably wrong
<ngz>civodul: I get (pardon my french ;) "guix substitute: error: connect: Connexion terminée par expiration du délai d'attente". I cannot ping ci.gnu.guix.org either.
<leoprikler>shouldn't it be guix.gnu.org?
<civodul>it's not pingable
<leoprikler>either way i'm not sure if CI replies to ping
<civodul>lemme see
<civodul>ngz: i tried "guix weather ..." and "guix environment --ad-hoc inkscape -- inkscape" and both work for me
<civodul>weird
<civodul>could it be something on the way to berlin? :-)
<ngz>civodul: could you try retrieving "cppunit"? It may be related to this package.
<Noclip>That's not normal, is it?:
<Noclip>"drwxrwxr-t 1101 root 30000 5,8M Aug 28 13:21 store"
<guix-vits>Noclip: I have the same in /gnu
<guix-vits>drwxrwxr-t 1 root guixbuild 11M Aug 30 13:36 store/
<civodul>normally 30000 resolves to "guixbuild"
<civodul>(on Guix System)
<civodul>ngz: cppunit works too
<Noclip>guix-vits: Interesting. A fresh unpacked guix tarball shows the following:
<Noclip>"drwxrwxr-x 48 root root 960 Jan 1 1970 store"
<civodul>did you try "telnet ci.guix.gnu.org https"?
<guix-vits>Noclip: What did They done with guixbuild!?
<ngz>Something eludes me then. I freeze at "substituting /gnu/store/yb6szgh4b8dwrf0n1wm489kmj9xy547x-cppunit-1.14.0...". I was able to substitute clucene a few seconds ago...
<ngz>civodul: telnet: Unable to connect to remote host: Connection timed out
<civodul>something's broken!
<Noclip>guix-vits: Wait, maybe that's wrong.
<civodul>ngz: perhaps you can try "traceroute" or something?
<guix-vits>ngz: i have ~200 KiB/s connecting to ci.guix.gnu.org now (sometimes drops to 90 KiB/s). At morning it was better (2-5 MiB/s).
<civodul>or access it over Tor until brokenness has vanished: https://guix.gnu.org/cookbook/en/html_node/Getting-substitutes-from-Tor.html#Getting-substitutes-from-Tor
<civodul>guix-vits: is free.fr your ISP?
<civodul>we've noticed occasional problems in the past
<ngz>Odd. Traceroute doesn't complete in 30 hops.
<guix-vits>civodul: No, GoodLine (Kemerovo).
<bandali>civodul, what was the reason again that guix discourages package mirrors like traditional distros?
<civodul>bandali: it's not discouraged, it's just a bit harder to set up
<civodul>well you need an nginx caching proxy
<bandali>ah, i see
<civodul>like: https://git.savannah.gnu.org/cgit/guix/maintenance.git/tree/hydra/nginx/mirror.conf
<ngz>It doesn't complete in 50 hops. 49th one is at ae1.0.ebr2-new.Amsterdam1.level3.net (4.69.136.110)
<civodul>but it only helps if there's enough users
<ngz>Trying 100 hops.
<bandali>civodul, gotcha! do you have any numbers on roughly the disk usage needed to run one?
<bandali>s/disk usage/disk space/
<civodul>it depends on what its users do with it
<civodul>if you can on demand and people all use x86_64-linux, you need less space than if people use all 4 arches
<bandali>i see
<civodul>likewise, if users only ask for emacs and guile, you need less space than if they also ask for libreoffice and texlive :-)
<bandali>hehe, true :-)
<civodul>ngz: actually it doesn't complete either here, but that's probably because ICMP is blocked there?
*civodul is a network noob
<ngz>civodul: does it complete with traceroute -T on your side?
<mjsb>nqz: civodul: myself i'm encountering a similar issue, since i'm having delays on page loads from guix.gnu.org or any related material (manual, cookbook), but after a while it sometimes succeeds
<civodul>ngz: hmm there's no -T in traceroute from inetutils
<civodul>mjsb: "sometimes", woow
<civodul>sounds terrible
<ngz>Oh? OK. I'm using Debian's.
<Noclip>guix-vits: No it wasn't wrong, the fresh store belongs to user and group root!
***boeuf is now known as bdjux
<mjsb>i wonder if there could be any network issue from my side, but i can reach without problems any other website, anyways now it seems back to normal :) !
<ngz>OK the issue seems to be gone by now
<joshuaBPMan>hey #guixoids!
<ngz>For the record, traceroute ci.guix.gnu.org doesn't complete, as expected, but traceroute -T ci.guix.gnu.org aka tpctraceroute does (not ICMP based).
<Noclip>guix-vitscivodul: On my foreign distro 999 seems to be the guixbuilder group and not 30000.
***guix-vits is now known as guix-vitscivodul
<guix-vitscivodul>Yessir
<guix-vitscivodul>civodul: ^
***guix-vitscivodul is now known as guix-vits
<guix-vits>sneek: later tell vagrantc: (aarch64) Do You have any pointers: how to make sddm, slim, or anything like that work? I can use sway, and xterm (for example). But the display managers do not work..
<sneek>Got it.
<guix-vits>sneek: botsnack
<sneek>:)
<joshuaBPMan>guix-vits: I personally do not use a display manager for sway. I just set up sway to automatically start when I log in from a virtual console.
<joshuaBPMan> https://wiki.archlinux.org/index.php/Sway#Autostart_on_login
<guix-vits>joshuaBPMan: Thanks, me too.
<guix-vits>I just want to try some X-based WMs.
<guix-vits>Maybe i can set some higher resolution with X (my aarch64 thingy glitches in relation to graphics currently (linux 5.8, arm64-generic).
<guix-vits>
<guix-vits>And just now i get seg-fault with xinit (guix environment --ad-hoc openbox xinit xorg-server). IDK why.
<pkill9>there's a TUI login manager I want to package
<pkill9>this is the one https://github.com/nullgemm/ly
<joshuaBPMan>also, question time. I've got gnucode.me running again (finally). It is being served by a guix system on a linode. I've set up certbot and nginx per the manual and civodul helped me to do it....I'd like to set the site to redirect to https...I've done that before. But will that eventually cause certbot to fail, if it needs to re-issue a certificate? because when it tries to go to
<joshuaBPMan> http://gnucode.me./well-known/some-werid-path-to-verify-I-own-the-server/ it will get redirected to https://gnucode.me/.well-known/some/weird/path/to/verify/that/i/own/the/server/... Will that re-direct to https 'cause certbot to fail eventually?
<joshuaBPMan>pkill9: that's cool!
<ngz>pkill9: very nice finding.
<rekado_>Noclip: having store items with 1970 timestamp is by design
<guix-vits>Hello rekado_.
<Noclip>rekado_: I know, I was just confused because it differed from the fresh install tarball.
<Noclip>rekado_: But what's really weird is that there is a directory in my store that has a recent timestamp and not the one from 1970.
***guix-vits is now known as vits-no-x-aarch6
<vits-no-x-aarch6>;P
<Noclip>"drwxr-xr-x 3 root root 4,0K Aug 28 05:15 6qyfv6psgb89ndq1w3qb3bwgvcbshhwp-ghc-pandoc-2.7.3-static
<Noclip>drwxr-xr-x 2 root root 14M Aug 28 13:21 .links"
<rekado_>that’s also okay
<rekado_>the .links directory contains hardlinks to deduplicate files in the store
<roptat>I also have some "délai d'attente dépassé" issues with ci (but also with my own substitute server)
<roptat>just now: substitute: guix substitute: warning: ci.guix.gnu.org : la connexion à échouée : Aucun chemin d'accès pour atteindre l'hôte cible
<Noclip>rekado_: But shouldn't the pandoc directory also have a 1970 timestamp?
<str1ngs>wierd curl: (7) Failed to connect to ci.guix.gnu.org port 443: No route to host
<roptat>yep, similar error here
<str1ngs>dig works
<roptat>no route to host means the name was resolved, but there's no way to go to that address from where you are
<Noclip>Firefox also fails on connecting to http://ci.guix.gnu.org/.
<str1ngs>probably running guix system reconfigure! :P
<rekado_>I also can’t connect
<rekado_>there’s no maintanenance that I’m aware of
<vits-no-x-aarch6>"Чебурнет пришол" :)
<Noclip>ci.guix normally has a web interface, right?
<str1ngs>I thought it might be maintenance with it being Sunday
<rekado_>no
***vits-no-x-aarch6 is now known as guix-vits
<rekado_>we would announce disruptive maintenance
<str1ngs>emergency maintenance! :P
<rekado_>and no: “guix system reconfigure” does not lead to outages either
<rekado_>again, no emergency maintenance; I’d be involved in that
<str1ngs>:P usually implies j/k
<roptat>civodul, you can access ci? what does curl tell you?
<rekado_>I don’t joke when ci.guix.gnu.org is involved
<rekado_>roptat: unlikely
<rekado_>I’ll connect to the serial interface
<Noclip>Is https://guix.gnu.org/ now down, too?
<rekado_>it’s the same system
<str1ngs>rekado_: so that means I should'nt?
<rekado_>you’re free to joke
<str1ngs>good
<ngz>traceroute reaches vpn.gwinatm.mdc-berlin.de (141.80.238.2) tho
<ngz>which is apparently the destination.
<guix-vits>Noclip: for me, yes.
<ngz>so (incredible guess) something is happening at mdc-berlin.de
<roptat>it reaches 141.80.238.11 from here, which is on the same network as ci
<rekado_>my guess is that the server may have run out of root fs space again, because we didn’t do anything about it last time
<msavoritias[m]1>Maybe its related to this
<msavoritias[m]1>blackmobile: https://news.ycombinator.com/item?id=24322513
<roptat>msavoritias[m]1, I don't think so, we can reach the destination network, but not the specific server
<civodul>roptat: i can no longer access ci.guix.gnu.org
<rekado_>I can’t get on the serial thing due to reasons
<nckx>Hi all! Is b-- oh.
<rekado_>gotta wait for my colleague then
<civodul>can't access it over tor either
<rekado_>I’m also on parental leave again
<civodul>ah! that you should take care of :-)
<rekado_>someone shut off my workstation in the office, which I’d use as a gateway
<rekado_>it’s a bit ridiculous
<civodul>uh
<rekado_>and my pubkey’s been removed from the alternative gateway server
<rekado_>probably just an error while … *checks notes* reinstalling it
<bandali>*_*
<rekado_>“reinstalling” servers is such an anachronistic thing
*rekado_ shudders
<nckx>So this won't be fixed today?
<rekado_>well, I’m still trying
<civodul>are other things on that network reachable?
<Guest49>Hi everyone. I'm a new user of Guix. Is IceCat (browser) missing digits for you? For instance if you look at a version number you only see the dot.
<bdjux>that problem sounds familiar, but I haven't had it in a long time
<Guest49>Any advices or solution ?
<bdjux>maybe missing fonts
<rekado_>civodul: without seeing any details I’m pretty sure that the server froze up because it ran out of space.
<rekado_>I’m in a betting mood
<ngz>rekado: but it has been flaky for a while now. Is it compatible with your bet?
<rekado_>the flakeyness has no good explanation. It is separate from this issue, I’m certain
*rekado_ is on the management interface now
<rekado_>it doesn’t respond on the serial console.
<rekado_>I can force a reboot
<bdjux>test
<rekado_>test back
<bdjux>nice. was having some issues with the xmpp bridge
<rekado_>ci.guix.gnu.org is booting now
*rekado_ presses the any-key
<rekado_>it says: loading '/gnu/store/12l3z6vbnyn06jpd86hjb1pxwgk7lmkx-system/boot'...
<rekado_>and that’s it
<rekado_>may have to boot an older generation
<rekado_>reboots
<joshuaBPMan>rekado_: bummer. Thanks for working to fix it though!
<simendsjo>Is ci.guix.gnu.org down for everyone? https://downforeveryoneorjustme.com seems to think so. But there has also been a global outage today..
<rekado_>simendsjo: it’s down
<rekado_>it’s rebooting
<rekado_>but it’s also down
<simendsjo>Ok, so nothing I can do on my part then :) Thanks. I'll just wait a bit.
<rekado_>same problem with an older generation: loading '/var/guix/profiles/system-367-link/boot'...
<rekado_>then nothing
<rekado_>this is a generation from end of july
<joshuaBPMan>rekado_: geez! That's odd. :(
<rekado_>the root raid is degraded
<joshuaBPMan>oh no. bummer. :(
<zjgkkn>oh :(
<rekado_>it’s rebuilding the raid
<rekado_>could be that this is why the boot is frozen
<efraim>pkill9: I have a package for ly and the start of a service which needs more debugging. Want me to send it to you?
<pkill9>efraim: sure
***caleb_ is now known as KE0VVT
<Kimapr[m]>are guix.gnu.org and ci.guix.gnu.org ded again?
<pkill9>apparently
<pkill9>someone else said they are for them
<pkill9>and downforeveryoneorjustme.com
<rekado_>so, the raid is now rebuilding
<rekado_>at 6%
<rekado_>I’ll wait for the rebuild op to be complete before continuing to boot
<rekado_>because previous boot attempts simply got stuck
<rekado_>I’d much rather not take care of this now
<rekado_>7%
<rekado_>yay.
***caleb_ is now known as KE0VVT
<rekado_>11%
<rekado_>you’d think that this should go on in the background without impact on booting
<rekado_>and perhaps that’s totally unrelated to the booting problems we have
<rekado_>I’m just a lowly bit shifter
<rekado_>14%
<rekado_>send “ugh” to 330330 to get even more of these exciting progress updates
<ArneBab_>does gpodder work for you? I get errors and nothing shows :-( TypeError: gobject `gpodder+gtkui+app+gPodderApplication' doesn't support property `application_id'
<ArneBab_> File "/gnu/store/pchwq79ixamc2chpm9iiyx7x0z1maagj-gpodder-3.10.15/lib/python3.8/site-packages/gpodder/gtkui/app.py", line 52, in __init__
<ArneBab_> Gtk.Application.__init__(self, application_id='org.gpodder.gpodder',
<str1ngs>ArneBab_: looks like a gobject introspection issue.
<str1ngs>Gtk.Application.__init__(self, application_id='org.gpodder.gpodder' looks like a legitimate way to create a GtkApplication with a application-id property
<str1ngs>ArneBab_: is gpodder package in guix?
<str1ngs>sorry missed the store hash
<str1ngs>ArneBab_: try staring a new terminal, does it still do it?
<str1ngs>starting*
<str1ngs>though this wraps GI_TYPELIB_PATH so it should work regardless
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix ⚠️ CI server is being repaired ⚠️ | get Guix at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs & patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: http://logs.guix.gnu.org | 1.1.0 is out! https://guix.gnu.org/blog/2020/gnu-guix-1.1.0-rele'
***nckx changes topic to 'GNU Guix ⚠️ the CI is being repaired ⚠️ | get Guix at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs & patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: http://logs.guix.gnu.org | 1.1.0 is out! https://guix.gnu.org/blog/2020/gnu-guix-1.1.0-release'
<nckx>Damn it. 2 more characters.
***nckx changes topic to 'GNU Guix ⚠️ the CI is being repaired | get Guix at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs & patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: http://logs.guix.gnu.org | 1.1.0 is out! https://guix.gnu.org/blog/2020/gnu-guix-1.1.0-release'
***nckx changes topic to 'GNU Guix ⚠️ the CI is being repaired | get Guix at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs & patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: http://logs.guix.gnu.org | 1.1.0 is out! https://guix.gnu.org/blog/2020/gnu-guix-1.1.0-released/'
***ChanServ sets mode: -o nckx
<str1ngs>ArneBab_: I'll see if I can replicate this.
<nckx>roptat: Could you remove _IMAGE from the ISO label? I don't think it adds value and versions can get long.
<rekado_>str1ngs: it wraps GI_TYPELIB_PATH, but it doesn’t override it, does it?
<str1ngs>rekado_: I'd have to check if it concatenates an existing GI_TYPE_LIB_PATH but offhand I'd say no it doesn't. other wise it would loss functional packaging.
<alextee[m]>is meson updated yet?
<str1ngs>also would be good to check gobject-introspection no how it handles this for propagation.
<alextee[m]>nope, core-updates seems to be on 0.53.2
<str1ngs>gpoppler does rely on that propagation hence the wrap. which is nice, but could be problematic.
<str1ngs>does *not*
<rekado_>str1ngs: how does overwriting GI_TYPELIB_PATH result in a loss of functional packaging?
<str1ngs>rekado_: over writing is fine. its concatenation in the wrap that could loss functional packaging
<str1ngs>let me recheck this declaration I just did a quick glance.
<str1ngs>err gpodder :(
<str1ngs>(gi-typelib-path (getenv "GI_TYPELIB_PATH")) this is problem matic
<str1ngs>depending on context though
<str1ngs>my guess in the build environment it is safe to use this.
<str1ngs>technically speaking a typelibs are propagations .. err run time dependencies. but for guix that does not translate well. so wrapping works nicely here.
<str1ngs>all typellib*
<str1ngs>rekado_: sorry for the rant. to answer your question better if the wrap did this it would loss functional package. export GI_TYPELIB_PATH="$GI_TYPLIB_PATH:/gnu/store/output...:/gnu/store/output" pseudo code.
<rekado_>we *do* do that by default
<rekado_>wrap-program is most often used in prefix mode, not in = mode.
<rekado_>so it’s something like GI_TYPELIB_PATH="/gnu/store/output...:/gnu/store/output:$GI_TYPLIB_PATH", really
<rekado_>rekado: ugh
<rekado_>rekado: 62%
<lfam>Is that the status of the array rebuilding, rekado_?
<str1ngs>without know interworkings of wrap I'll assume it does the right thing :)
<str1ngs>thoug regards to GI on guix I'm still hunting down an order precence issue that effect g-golf and guile-gi.
<rekado_>lfam: yes
<rekado_>str1ngs: it’s not always right. I think we’re too often picking “prefix” when “=” would have been better. We’ve been suffering with PYTHONPATH wrapping since forever, because the order of directories on that variable has not the obvious semantics that you’d think.
<pkill9>efraim: you never sent the package
<str1ngs>rekado_: aye I hear ya.
<jonsger>hm datefudge 1.23 can't be found on the origin server as well as on software heritage
<sneek>jonsger, you have 1 message!
<sneek>jonsger, civodul says: guile-ssh segfault reproduced!
<jonsger>oh nice or bad :P
<jonsger>civodul: should I still send my findings on the libssh bug to the tracker?
<jonsger>`guix copy` could help when CI is down, but sadly it's a bit broken due to this guile-libssh thing :(
<civodul>jonsger: i pushed a fix yesterday, in https://issues.guix.gnu.org/42740
<civodul>note that you can use 'guix publish' to work around the lack of ci.guix.gnu.org
<roptat>nckx, I'll do it, thanks :)
<jonsger>civodul: so guix publish is not hit by this guix copy issue?
<nckx>roptat: Thanks! I don't think anyone actually enforces the 16-character limit but good riddance.
<joshuaBPMan>Sweet. The guix website is back up again. Thanks whoever did that.
<nckx>👏rekado.
<lfam>The Unifont "clapping" emoji is not the best one
<lfam>Lol I had to copy and paste it into a browser to see what it was
<lfam>I guess the options are limited for Unifont
<rekado_>at 93%, but it booted
<joshuaBPMan>rekado_: thanks. May I ask what you suppose caused the problem ?
<rekado_>haven’t checked yet, baby needs to sleep
<joshuaBPMan>rekado_ hahaha
<ngz>hedgewars seems to fail building. Build log reports a few entries like "ld: cannot find -lHSregex-tdfa-1.2.3.2-24BpKmE04Wg88SsVhgqjyZ". Could this be related to recent Haskell changes?
*nckx thinks Unifont's emojoes are cute, but can't get HexChat to use them to display that one.
<nckx>ngz: There's a bug report about that (#43112) but it hasn't hit issues. yet.
<lfam> <https://bugs.gnu.org/43112>
<ngz>nckx: Yes, I'm current looking at this bug report and the log posted there.
<nckx>Good, I was about to post it here with a warning.
<rekado_>I accidentally booted a pretty old system; I’ll reboot a newer one and see if that works. Otherwise I’ll boot the same old system again to investigate.
<rekado_>(this is generation 372)
<rekado_>but first: why did it fail?
<rekado_>I see only 4G free on the root file system, but this may still be enough
<civodul>hey rekado_, thanks a lot for taking care of this
<civodul>rekado_: 4G should be enough, unless some things were not being offloaded
<civodul>like the image builds (the #:local-build? change that apteryx made recently could be the reason)
<str1ngs>ArneBab_: guix environment --ad-hoc gpodder -- gpodder works for me now that ci.guix.gnu.org is up. can't replicate the application-id property issue
<rekado_>hint: “Aug 30 16:35:10 localhost postgres[85719]: [29-1] 2020-08-30 14:35:10.232 GMT [85719] LOG: could not close temporary statistics file "pg_stat_tmp/global.tmp": No space left on device”
<rekado_>a bunch more like that in the following minutes
<joshuaBPMan>rekado_ that would explain something
<rekado_>I’ll compress some of the log files
<rekado_>the 81G of nginx logs are obvious candidates
<rekado_>we also have lots of SSH log spam
<joshuaBPMan>rekado_ have you ever heard of endlessh? https://nullprogram.com/blog/2019/03/22/
<joshuaBPMan>I guess it's a way to help "deal" with people trying to shh into you server. I really want to try it at some point.
<nckx>Is there a Guix system service for that?
<nckx>I know we have the package.
<lfam_>Seems like rotating the logs would be more helpful
<joshuaBPMan>lfam_ most likely.
<joshuaBPMan>nckx: I hope there is a service.
<rekado_>lfam_: we do rotate logs already; but it started a little late
<nckx>Cleaning up /var/log/guix/drvs might help too.
<rekado_>so we’ve got a whole bunch of old logs
<rekado_>I can’t allocate enough time to figure out which of them are still useful for us, so I’ll just make them smaller and move on
<nckx>But I can't think of a plausible place for 1 TB to hide.
<rekado_>joshuaBPMan: we have a thing to ban SSH script kiddies
<joshuaBPMan>rekado_ I'm not certain what that means... I'm guessing something like fail2ban?
<nckx>Worse. Hard-coded # of connections per time period.
<rekado_>joshuaBPMan: https://git.savannah.gnu.org/cgit/guix/maintenance.git/tree/hydra/modules/sysadmin/services.scm#n279
<joshuaBPMan>thanks!
<nckx>All successful? Doesn't matter, on the list you go.
<rekado_>I think mhw once added this
<rekado_>nckx: you’re welcome to propose something better
<nckx>I won't. I don't understand the problem.
<nckx>Did this crash SSH?
<rekado_>did what crash ssh?
<rekado_>nothing crashed ssh afaict
*rekado_ is confused
<nckx>The kids. Before the iptables hack.
<rndd>hi everyone! i run "guix gc -d 2d" but notes in grub menu are still exists. am i doing something wrong&
<rndd>?
<rekado_>rndd: remove system generations first
<rndd>rekado_: how?
<rekado_>nckx: no, this was inherited from old hydra
<rekado_>rndd: “guix system delete-generations”
<rndd>oh
<nckx>rekado_: But which problem(s) did they cause?
<rekado_>nckx: we still have the … situation of getting lots of SSH connection attempts, but I don’t think it’s a real problem
<rndd>rekado_: maybe there need to be a note about it in "invoking guix gc"?
<efraim>pkill9: sorry, i wandered away from my computer for a while and forgot about it :/
<rekado_>rndd: no, because GC will only collect unused things. If you never declare that you don’t use older system generations they will never be collected.
<rekado_>same with old profile generations
<rndd>0_0
<rekado_>nckx: we don’t have any problem with connection attempts; it doesn’t bog down the server if that’s what you mean to ask
<nckx>rekado_: If/when you have time: what's /store's purpose?
<rndd>rekado_: how i should delete old profile generations?
<rekado_>nckx: I just noted that SSH connection attempts still appear in the logs, causing a lot of useless text to skip over
<nckx>My question was whether they crashed SSH, which you answered above.
<efraim>pkill9: https://dpaste.com/34VT6AU4D for the service, https://git.sr.ht/~efraim/my-guix/tree/master/wip/ly.scm for the package
*rekado_ is still confused, but it may be a problem with this brain
<rekado_>:)
<rekado_>rndd: “guix package -d” will delete all but the current generation of your default profile.
<rekado_>with an argument it will delete matching generations
<rekado_>I recommend skimming or reading the manual.
<rndd>rekado_: i know, i should read it
<rndd>one day
<rndd>i promise
<rekado_>no need to read it all, not in one sitting anyway; it’s too long for that
<rekado_>but it introduces the concepts that Guix uses and implements, and it explains them better than I could here in the chat
<rndd>bytheway, what about gradle in guix? is there way to build java project that use gradle without java?
<rekado_>do you mean to write a package definition for an application using gradle? This would likely not work right now.
<rndd>=(
<nckx>I see what /store is for now. du -hs just took suspiciously long (still running) for what it contains.
<rekado_>gradle has bootstrapping problems
<rekado_>nckx: sorry, missed your question about /store
<rekado_>it holds the kernel and initrd for all installed system generations;
<rekado_>we need to have it on the root disk because of a booting quirk
<rekado_>when we boot the server the built systems are all on the external storage array
<rekado_>but the storage array is not visible before linux has booted
<rekado_>very frustrating
<rekado_>so at least linux and the initrd must be on the local root file system
<rekado_>why is it called /store, though?
<rekado_>I don’t know!
<nckx>Yep, it turns out to be very similar to what I do to boot bcachefs. Thanks! I was just thrown off by how slow ls was (I had to ^C it after a minute or so) for what turned out to be ‘few’ items.
<rekado_>it just kept looking for things in /store, so I said: whatever, you want it there? I’ll give it to you!
<nckx>🙂
<rekado_>it’s still slow because the RAID is still being rebuilt
<nckx>Debugging by capitulation.
<rekado_>last few percentage points
<rekado_>still at 95%
<rekado_>it’s probably much slower now that we’re using the system
<nckx>I've suspended all my jobs.
<rekado_>after some time and removing some logs I’ll reboot it again
<rekado_>well… I’m guilty of slowing things down. I’m compressing a rotated log file.
<rekado_>though I guess I could just delete it.
<rekado_>we don’t seem to be the log-reading types
<lafrenierejm>Does anyone have a simple manifest file they'd be willing to share? I'm getting "error: specification->manifest: unbound variable" and I'm not sure how to fix it.
<nckx>Actually, which mount(s) is/are on the degraded array?
<rekado_>lafrenierejm: what command gives you that error?
<rekado_>lafrenierejm: by the way, it’s specificationS->manifest.
<lafrenierejm>rekado_: The missing "s" fixed the immediate error. Thank you!
<rekado_>nckx: /mnt/root-fs (which is a remounted /)
<rekado_>nckx: /mnt/root-fs exists so that we can access the shadowed /gnu
<nckx>Thanks. Linux doesn't make it easy to track complex bind mounts, or I don't know where to look.
<nckx>Well, complex == new to me 🙂
<rekado_>I think that we can safely erase /mnt/root-fs/gnu
*nckx as duck: ...so where would the corresponding db for that be? (or have been?)
<nckx>Judging by the age I suspect you're right.
<rekado_>originally, it was very messy and I wasn’t sure which parts of it were visible by the time linux took over
<rekado_>but it looks like only /store is needed, then linux does its thing, bind-mounts /gnu/store from the external storage, and loads the system from there.
*nckx AFK.
<nckx>o/
<hendursaga>I'm trying to load Drakma using Quicklisp but it's complaing it can't find libssl.so - what do I install/configure?
<rekado_>perhaps you could set LD_LIBRARY_PATH=$(guix build openssl)/lib and try again?
<ArneBab_>str1ngs: I still get the error; could I have some version mismatch?
<ArneBab_>str1ngs: thank you for trying to replicate!
<rekado_>ArneBab_: does it still happen when you unset GI_TYPELIB_PATH before running the application?
<ArneBab_>rekado_: ah, no, it doesn’t! nice! GI_TYPELIB_PATH= gpodder
<ArneBab_>^ this works
<ArneBab_>… 57 podcast episodes to download … thank you!
<ArneBab_>how can I fix that in general?
<ArneBab_>rekado_: I think I know how … I have GI_TYPELIB_PATH in my ~/.bashrc — likely a leftover from when Guix did not have its own profile
<ArneBab_>in my ~/.bash_profile
<ArneBab_>… no, I take it back, it’s commented out
<ArneBab_>what I have is this:
<ArneBab_>hash guix
<ArneBab_>export GUIX_PROFILE="/home/arne/.guix-profile"
<ArneBab_>source "$GUIX_PROFILE/etc/profile"
<rekado_>does that file set GI_TYPELIB_PATH?
<rekado_>it might be due to other packages you have installed
<ArneBab_>yes, it does:
<ArneBab_>export GI_TYPELIB_PATH="${GUIX_PROFILE:-/gnu/store/0i2bvr70g44h8b1kkzs5f7l9ci0kv403-profile}/lib/girepository-1.0${GI_TYPELIB_PATH:+:}$GI_TYPELIB_PATH"
<rekado_>this might be from gobject-introspection (or a package propagating it) or from nomad
<rekado_>both have a GI_TYPELIB_PATH search path specification, which may be dangerous
<rekado_>(I think we should generally prefix more variables with GUIX_ to avoid conflicts)
<str1ngs>ArneBab_: is GI_TYPELIB_PATH set in your profile. echo $GI_TYPELIB_PATH to be sure
<str1ngs>rekado_: nomad has a known limitation with GI_TYPELIB_PATH as I mentioned before the progation is work around.
<str1ngs>though I have a semi stable declaration that doe'snt use propagation
<rekado_>str1ngs: the variable is set by Guix
<rekado_>so if nomad is not in the profile then it’s due to the gobject-introspection package, which may be propagated by another package.
<rekado_>ArneBab_: if you’re curious why it appears you can view the generated “manifest” file, which is part of every profile
<str1ngs>rekado_: right but if nomad is in the profile it progates gi dependacies and provides a GI_TYPELIB_PATH
<str1ngs>just because you mentioned nomad I'm clarifying what happens
<ArneBab_>str1ngs: yes, it is, and it’s long.
<ArneBab_>python-pygobject
<ArneBab_>gobject-introspection
<ArneBab_>… it is set by quite a few packages
<str1ngs>ArneBab_: try this unset GI_TYPELIB_PATH; gpodder . you might need to restart your term after. GI_TYPELIB_PATH= gobber . migh work too
<rekado_>str1ngs: see above.
<ArneBab_>str1ngs: that’s what rekado suggested and it works
<ArneBab_>(still, thank you!)
<str1ngs>well rekado_ mentioned nomad, is that in your profile?
<str1ngs>because that could exhibit this isse.
<str1ngs>err issue*
<ArneBab_>not right now
<ArneBab_>(had to remove it because it failed to build)
<str1ngs>kk so not nomad related? just to make sure because I have a declaration that does not use propagation.
<ArneBab_>I don’t have nomad right now, so not nomad related
<str1ngs>good stuff sorry for the confusion.
<ArneBab_>should I upload my manifest somewhere?
<ArneBab_>(is there anything sensitive in it?)
<str1ngs>just past a guix pacakge -l I would think.
<str1ngs>is gobject-introspection in your profile btw?
<ArneBab_>doesn’t look like it
<ArneBab_> https://gist.github.com/ArneBab/50f51eed3c06aa53b6cfaaa00f22b6a2
<str1ngs>how large is your manifest ?
<ArneBab_>(paste.debian thinks that this is spam)
<str1ngs>lol
<ArneBab_>my manifest is around 20k lines
<str1ngs>ArneBab_: have you logged out and logged into your DE lately. it will effect environment varibles
<str1ngs>known guix limitation
<ArneBab_>yes, this morning
<rekado_>ArneBab_: you can find gobject-introspection in the manifest file and then walk up to the package responsible for it.
<ArneBab_>python-matplotlib
<ArneBab_>(if I got that right)
<ArneBab_>can’t remove taht …
<ArneBab_>that
<str1ngs>ArneBab_: I would bisect your manifest. until GI_TYPLIB_PATH is removed from .guix-profile/etc/profile
<str1ngs>you can use -p to create a temp profile to bisect
<ngz>ArneBab_: it may be related to http://issues.guix.info/35308 I reported a while back.
<rekado_>str1ngs: no need for that
<rekado_>str1ngs: just looking at the generated “manifest” file is enough to understand where things come from
<rekado_>ArneBab_: if you only use it for certain activities it may be worth moving your Python things to a separate profile and “activating” it when that activity comes up
<str1ngs>ArneBab_: definately python-matplotlib
<ArneBab_>can we fix this somehow?
<str1ngs>remove python-matplotlib
<str1ngs>test then
<ArneBab_>I meant fix python-matplotlib
<ArneBab_>I’ll test that
<ArneBab_>to be sure that it really is the reason
<rekado_>python-matplotlib is much too big; we have added a lot of inputs to support all features, and many are propagated
<rekado_>“fixing” it would mean to cut down on that
<ArneBab_>python-matplotlib actually *needs* all that
<rekado_>yes
<str1ngs>ArneBab_: right test first though
<rekado_>so, “fix” is not applicable here
<str1ngs>we can try and fix after :)
*rekado_ —> zzzZ
<ArneBab_>I still remember installing it from scratch on a university server a few years ago
<ArneBab_>rekado_: good night!
<ArneBab_>took 3 months …
<str1ngs>rekado_: g'nighth
<ArneBab_>I’ll test tomorrow
<str1ngs>ArneBab_: okay, let me know how it works out.
<NieDzejkob>argh, big grafts take so much memory that every time one happens in the background, my computer locks up for a few seconds
<roptat>NieDzejkob, not memory I think, IO