<drakonis>is gtk/qt getting isolated from your settings? <meena>i'm not sure i know what that means <drakonis>well, i didnt know if it was seeing your cursor settings <cdegroot>What's the issue exactly? If you want to make a Guix package for the project, I'd say go with the flow and include these libraries statically? <Aurora_v_kosmose>cdegroot: Well, vendoring its own versions that are incompatible with upstream means duplication of packages. Guix also seems to promote the use of dynamic linking for most things. <cdegroot>Yes, but I guess pragmaticism will actually get that package included, battling upstream to reverse what smells like a done deal, not so much :) <cdegroot>Statically linked libraries for a single package won't conflict with DLL versions of the same elsewhere, so apart from some extra disk space, I don't think it should be a big concern. <cdegroot>(I'm assuming that these upstream libraries have licenses that make this all ok, of course) <Aurora_v_kosmose>Licenses seem okay. The upstream project just seems to have gone full cathedral and barely even replies to issues on their own forum. <cdegroot>I'd see these forked/static libraries as just code that belongs to the package now. <jts>Hey y'all. I need some help. I've been running GuixSD for a couple months now, and I've really fallen in love with it. But it feels like any time I want to use a programming language that relies on remote package repositories, I inevitably end up not being able to get some package or other that I need, and that isn't in Guix. I could theoretically package all of these myself, but that's really annoying and time consuming when I want to <jts>work on other stuff. Am I missing something? Or do I just need to use a different distro if I want to work in these languages without manually packaging dependencies? (The specific languages I've had this experience with are Julia, Rust, and OCaml.) <singpolyma>Does cargo not work on guix system for some reason? <jts>cargo seems to work in the general case, but there are instances where it fails. For example, I tried running `cargo install racer` and got an error suggesting I use rustup to add a component. <jts>I'm very familiar with `guix import` and have used it to package several OCaml libraries (part of a patchset I haven't completed) but it's still annoying to have to package dependencies just to play around with stuff <cdegroot>I don't know about the language you mention but my main programming language I just use the built-in package mamagement system (until the day I have the time to write a huge mirror scripts for all packages - to me it should be all or nothing, I have better things to do ;-)) <zacchae[m]>Yeah, for python, I use pip unless I can manage to get every package in guix. Similarly for my breif usage of octave. <jts>maybe I'm just trying to use some specially cursed libraries XD ***iyzsong-www is now known as iyzsong-w
<allana>Hi #guix! General question, does "guix pack" also bring native-inputs along for the ride, or just the runtime dependencies? <civodul>allana: hi! it brings just run-time dependencies, those that can be seen with, say, "guix graph -t references PKG" ***iyzsong-www is now known as iyzsong-w
<nouun>Hey, was wondering if there's a way to somehow edit file permission for /etc/light/targets/sysfs/backlight/auto/minimum through my config so I don't manually have to change it? <nouun>Sorry, meant the "/sys/class/backlight/intel_backlight/brightnes" file ^ <civodul>nouun: hi! currently there's no service that does that <civodul>you could write a service, using 'simple-service', that does (call-with-output-file "/sys/class/..." (lambda (port) (display "1\n" port))) <civodul>i'm looking for an example of a daemon that would typically be "socket-activated" under systemd <ekaitz>civodul: the thing I asked you the other day... any idea? <AwesomeAdam54321>civodul: I don't think there are any, services use their sockets anyways when they're fully up and ready <AwesomeAdam54321>With socket activation, preparing a socket that the daemon listens on before the daemon is even started doesn't make it any faster <civodul>jpoiret: but that one is started automagically on demand, per user; i don't have a shepherd service for it <civodul>ekaitz: hey, what was the question again? <jpoiret>hey :) i think on systemd-based distros it uses socket activation but I might be wrong. at least the arch wiki mentions that ***julm is now known as Guest5469
<ekaitz>civodul: i wanted to make a gcc-toolchain that contains a custom gcc, but it's unable to find the libraries <civodul>do you have an example paste somewhere? <ekaitz>civodul: wait, not that branch but riscv <civodul>ekaitz: looks like gcc-from-source overrides the phases and flags found in gcc-4.8 <civodul>gcc's a complex beast, so you'd rather reuse as much as possible of the existing package definitions <ekaitz>and how can I keep the flags and add what I'm adding there without overriding everything? <civodul>ekaitz: using substitute-keyword-arguments <civodul>there are examples in commencement.scm and elsewhere <ekaitz>also civodul I tried to compile gcc by hand and the normal gcc-toolchain says it doesn't find gnu/stub-32.h <leinad>Is `guix shell totem gst-plugins-ugly` supposed to work in the sense that the GNOME Video player (Totem) can make use of the ugly GStreamer plugins? <ekaitz>but the guix package as I defined there is able to compile <leinad>I am trying to play a H.264 video but `guix shell totem gst-plugins-ugly -- totem foo.mp4` cannot play it saying: Missing plugin: gstreamer|1.0|totem|H.264 (Main Profile)-Decoder|decoder-video/x-h264, level=(string)4, profile=(string)main, colorimetry=(string)bt709 (H.264 (Main Profile)-Decoder) <ekaitz>leinad: if you install it in the system works? <ekaitz>leinad: you can always make a `rollback` later <civodul>leinad: you may need to add gstreamer so that the GST_PLUGIN_PATH variable is set <abrenon>wouldn't it be (yet) another case of path variable, and a "main" gst package would be needed to have that variable populated ? (just a wild guess) <civodul>ekaitz: hmm not sure; perhaps you can start by overriding nothing but 'source'? <ekaitz>civodul: I'll give a try to the substitute-whatever but also my system searching for gnu/stub-32.h is weird... I'll check that too and maybe ping you back withe results :) <civodul>ekaitz: re stub-32.h, in general you cannot use -m32 with the compilers of Guix <civodul>that's because we don't enable multilib support <leinad>ekaitz: by installing in the system, you mean in the system profile /run/current-system/profile? <leinad>civodul: adding gstreamer also did not work :/ <abrenon>leinad: in the shell you opened with both gstreamer and gst-plugins-ugly, is the variable GST_PLUGIN_SYSTEM_PATH defined ? <leinad>weirdly it is set, but to some path under ~/.guix-profile <leinad>no sorry... GST_PLUGIN_PATH is set to ~/.guix-profile/lib/gstreamer-1.0 <leinad>GST_PLUGIN_PATH is set to /run/current-system/profile/lib/gstreamer-1.0 <leinad>aah sorry, I need coffee, GST_PLUGIN_SYSTEM_PATH=/run/current-system/profile/lib/gstreamer-1.0 <jpoiret>are you trying to do `guix shell gstreamer gst-plugins-ugly totem`? <jpoiret>it works for me, GST_PLUGIN_SYSTEM_PATH ends up pointing to the fresh guix shell profile <leinad>Ok, so building a profile with: guix shell --pure gstreamer gst-plugins-ugly totem grep coreutils <jpoiret>(i don't claim that totem works with it though) <leinad>and running `env | grep GST` yields GST_PLUGIN_SYSTEM_PATH=/gnu/store/50c1gxa1xk9i9lcvfdnh36b7830ncz7s-profile/lib/gstreamer-1.0 <leinad>so this works... but then running `totem foo.mp4` does not work <civodul>weird; could it be that some other plugin is missing? <leinad>it explicitly says H.264 plugin is missing and also AAC <jpoiret>at least that's what the internet tells me <jpoiret>well, for H264, maybe you'll need the plugins in gst-plugins-ugly <leinad>it is working with `guix shell gst-libav gstreamer totem` <leinad>I also thought I needed the ugly plugins by looking at the inputs (specifically libx264) of that package. <abrenon>great ! so not a matter of guix, but a matter of which plugins are needed and provided by which gst-plugins set ? <leinad>abrenon: yes, i think so, sorry for the noise <abrenon>no, don't worry, it's useful knowledge in any case, and I'm was more reassured than anything else to find that it wasn't an issue with the packages or how to use them <ekaitz>civodul: that helped! I'll check the compilation process to see if I can fix that <sneek>unmatched-paren: Greetings!! <unmatched-paren>hello guix! how could i get the version of a package? i've tried (package-version PKG), but that procedure doesn't seem to exist <unmatched-paren>sneek: later tell jts: "For example, I tried running `cargo install racer`" <- isn't racer obsoleted by rust-analyzer? that is in guix <unmatched-paren>cdegroot: actually, a full package repo mirror script seems like a fun idea... it'd be really tedious fixing up the problems manually, though ;) <katco>hey all! `guix pull` is failing for me, and asking that i report a bug. i commented out the channels i'm subscribed to in an attempt to see if it's something with my setup, but the issue persists. any ideas? https://paste.debian.net/1235866/ <katco>this seems suspicious, but i ran emacs's `check-parens` on the file and there's no problem: <katco>> ice-9/read.scm:126:4: In procedure lp: gnu/packages/databases.scm:4628:1: unexpected end of input while searching for: ) <unmatched-paren>looks like there's an unmatched-paren in gnu/packages/databases.scm :P <katco>that's what i thought :p perfect user to respond ;) <nouun>Is there a psuedo package where I can create a package from a sh script stored in a string? <cdegroot>~unmatched-paren yeah, it's not very high on my todo list :P <nouun>The source isn't hosted anywhere and I just used to have it in my NixOS config and packaged it with writeScriptBin. <katco>unmatched-paren: on master, which i assumed was that commit. hm. lemme check <katco>let me confirm what's going on first <unmatched-paren>ok, try looking through `/var/guix/profiles/per-user/...` for an old-looking (but not too old) `current-guix-N-link` directory <unmatched-paren>then, try to run `/var/guix/profiles/per-user/.../current-guix-N-link/bin/guix pull` to pull with the old guix maybe? <unmatched-paren>katco: i'll try pulling my guix checkout and searching through the log for something that relates to `gnu/packages/databases.scm` <katco>no commits have touched this `databases.scm`: `git diff ff4585d..8e994ea512 -- ./gnu/packages/databases.scm` (empty result) <katco>`guix describe` tells me that `ff4585d` is the commit i'm on; `8e994ea512` is tip <unmatched-paren>...have you possibly accidentally edited your local copy of the file? <katco>ok, i'll try that now. fair warning: there is a non-zero chance this is an issue with my ram <katco>maybe my last pull wrote `databases.scm` to disk corrupted? <katco>pulling with the previous version seems to be working *unmatched-paren is currently imagining a tiny solar ray hitting katco's ram, flipping a bit in the databases.scm file stored in it, changing one unicode character from a parenthesis to a space <katco>there is a whole story behind this. i found out i was inadvertently overclocking my ram. after i stopped doing that a few months ago, i still have not regained confidence that one of my dimms is not bad <abrenon>how does one overclock ram by accident ? (I wouldn't even know how to do it on purpose ^^) *unmatched-paren is now imagining another solar ray hitting katco's ROM, flipping the ENABLE_RAM_OVERCLOCK bit to true <katco>abrenon: i was using a clock-speed in my bios that i thought was supported by my motherboard/cpu, but it isn't. <katco>i build a new computer like once every ten years, so i forget everything and then hope i can learn it all again to build something that works :p i probably made a mistake during the burn-in period after this build <katco>unmatched-paren: the pull using the prior `guix` worked, yes! i uncommented my channels and i'm running a standard `guix pull` from path now <unmatched-paren>i wonder out of curiousity if we could dig into that same directory and see what the problem was... pretty sure you should be able to locate the faulty guix there <katco>i have some free time if you have suggestions. do you suggest i find the `databases.scm` in that profile? <unmatched-paren>depends on whether the problem was in the guix you attempted to pull, or the guix you were pulling with. <katco>regular `guix pull`, with channels, works now <katco>tysm unmatched-paren! i'm never completely sure whether these issues are because i fiddle a lot or a proper issue <katco>yeah i usually run `guix gc` on some cadence, so it's good to know that i should hold off on doing that for a bit <katco>i knew how to use old versions of guix, i think i'm just always unsure when i see an error if it's an actual error or because of something in my setup <unmatched-paren>but you should wait to do `guix gc --delete-generations`; it frees up TONS of space, but will stop you ever doing that (or rolling back system/home configs) <linus>hey, when i try to boot from the usb (with the latest iso of guix system), it stops after a few lines, one of them being <linus> Error: Driver 'pcspkr' is already registered, aborting... <apteryx>linus: yes, it's harmless, although it'd be nice to fix it <apteryx>I think jpoiret had some idea about how to go around fixing that pcspkr module loading error? <unmatched-paren>what's pcspkr actually used for, anyway? i assumed it was for audio, but my audio still works with it disabled... <jpoiret>did I? I have some faint recollection of this <jpoiret>although this shouldn't be an issue wrt booting <jpoiret>the issue might come from something else, eg. graphics driver, do you happen to have a graphics card that would require non-free blobs? *unmatched-paren remembers that they still get annoying syslog messages all over their login prompt <linus>okay, thank you for the information, that's the only line containing "Error". <linus>Any idea how i could find out why it won't boot? (it just stops printing new lines to the terminal) Could this be a problem with the linux-libre kernel ? (i have AMD Ryzen 5 Pro 5650). <linus> would think that even tho they may not be fully supported it would atleast boot/ print out a more meaningfull error? <jpoiret>unfortunately, oftentimes it does not <jpoiret>you could go to grub, and on the Guix line you could press `e` to edit, and add nomodeset to the kernel command line <linus>okay, ty for the information, will try <jpoiret>there are other wayland backends though <linus>unmatched-paren my priority would mainly be to get it working to a usable state, but ty for the warning <jpoiret>oh, seems i'm confusing things, it's the HW accel vs software rendering i was thinking about *unmatched-paren qutebrowser -> :open does wayland require kms <jpoiret>although you could run wayland under X11 :))) <unmatched-paren>although apparently someone once patched it to run under the framebuffer... of course, there would be no hardware acceleration at all ;) <linus>ty for your help, esp. (jpoiret). I understand it may be not optional now but i installed guix on my laptop successfully thanks to you, ty <dlowe>huh, I thought the amd gpus were all libre drivers now <unmatched-paren>well, i heard AMDGPU isn't working in linux-libre yet, even though it's free-source <jpoiret>iirc, amd gpus still have non-free blobs <dlowe>nVidia didn't help with nouveau though <unmatched-paren>i mean... there ARE plenty of people who find our walls of parentheses scary... <nouun>They're just hugs for the code! <dlowe>people are trying to make javascript guix be a thing via guile's js support <unmatched-paren>dlowe: even here, in this isolated corner of the free software community... we cannot hide from JS <nouun>dlowe: They're doing what now? <unmatched-paren>because i don't think js has a way to refer to names that aren't valid js identifiers? <dlowe>singpolyma was giving a talk about it <singpolyma>unmatched-paren: yes, js can do it no problem. lua can't that I know of but would love to hear more I don't know lua as well <singpolyma>ah, if I just do ["blah"] that's a toplevel? cool. that will be useful <singpolyma>I'm using a procedure to work around that in my lua-guix experiment right now, but this is better <apteryx>uh, 'guix package --search-paths -m my-manifest.scm' returns export PYTHONPATH="/home/maxim/.guix-profile/lib/python3.9/site-packages", among other things <unmatched-paren>singpolyma: wait, actually it's the same as js... foo.bar['baz'](quux) <singpolyma>unmatched-paren: but what about top level? in js there is always a "global object" (in browser it is window, in node global, in guile we just use the top level this) so I can this["blah-blah"] <apteryx>I think the PYTHONPATH search-pah comes from ansible <unmatched-paren>so singpolyma for the above you can do `_G["some-scheme-procedure"](foo, bar)` <unmatched-paren>as `lua modules == lua tables`, you can just do `module_from_guix['foo']` <efraim>has anyone managed to cross compile gfortran? <phodina[m]>does somebody know what is the syntax for dependency download in inputs using new syntax? <phodina[m]>unmatched-paren: Ok, but how do I refer to the package itself later? <unmatched-paren>phodina[m]: hm, not sure; i ran into that problem myself once; i gave up and just used the new syntax... <civodul>phodina[m]: yeah, you can use the old syntax if you want a label on your origin, or you can refer to the origin right from a phase <lilyp>It helps being the one who committed both <phodina[m]> lilyp civodul unmatched-paren thanks, that's exactly the example I was looking for :-) <apteryx>has anyone else tried to convert markdown (.md) to HTML with pandoc and have a table of content (--table-of-contents) ? <apteryx>It produces the HTML file in my case, but *without* a ToC <apteryx>ah, perhaps this "Hint: If you want a table of contents in html output, you also need to supply -s which generates a standalone document." <attila_lendvai>did anyone manage to scan using an older HP printer/scanner (1522nf)? any hints? <unmatched-paren>there's no neovim@0.0 package as my attempt to `guix edit` it shows... <unmatched-paren>the only `neovim` objects are 0.4 from `gnu` and 0.6 from my channel <attila_lendvai>i see" openat(AT_FDCWD, "/gnu/store/[hash]-hplip-minimal-3.21.10/lib/libm.so", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory) <attila_lendvai>and: openat(AT_FDCWD, "/var/lib/hp/hplip.state", O_RDONLY) = -1 ENOENT (No such file or directory) <unmatched-paren>i suppose hplip needs to be patched to look for libm in the right place? <apteryx>attila_lendvai: that libm.so error is probably not fatal; it's just the way shared libraries are scanned for, I think *attila_lendvai has realized his user is not part of the group 'scanners' *attila_lendvai needs to relogin <attila_lendvai>yep, the libm.so thing seems to be non-fatal. and the hplip.state, too. no idea why it fails... ***the-porcupirate is now known as porcupirate
*attila_lendvai gives up in frustration <civodul>i think the Shepherd is ready for a release candidate in the coming days! <singpolyma>Concurrent yeah. I think parallel is reserved for multiprocessing usually? <drakonis>by virtue of no longer doing things in a linear fashion <civodul>so startup might be faster, but above all, that gives more flexibility and robustness