IRC channel logs
2026-03-25.log
back to list of logs
<gabber>can you try to "manually" log into the machine through ssh? <gabber>there is the (multiple, IIRC) -v options for `ssh` calls <gabber>using the same credentials/mechanism as defined in your (presumably named) machine.scm? <czan>Does "guix deploy -x 'echo hello'" work? <redacted>Huh, "guix deploy -x 'echo hello' ./machines.scm" gives me "error: '-x' specified but no command given" <redacted>This is an intermittent issue I've been putting up with for a while. <redacted>Usually I'll just call guix deploy in a loop until it works. <czan>Is it possible there's an intermittent issue that's corrupting packets, or something? It seems like it's not related to connecting, so it might be happening when actually transferring data. <gabber>redacted: maybe you can watch the logs from the server side? <redacted>czan: haven't been able to figure out what that issue is, despite doing MTU tests and watching traffic with wireshark <redacted>gabber: yeah, I could try, though there's a lot of noise from intrusion attempts <czan>Wireshark on the client, or the server? <redacted>czan: I might have done both when last I checked. <czan>If it's a hardware issue then I'd expect wireshark to look fine, but the server to see something different to what the client (thought it) sent. <czan>s/wireshark/wireshark on the client/ <czan>If you've checked on the server as well then I properly don't understand what's going on. :) <redacted>I'll have to do a thorough investigation and document everything more thoroughly this time. <gabber>redacted: you could do something like `tail -f /path/to/log | grep $YOUR_IP` to filter live for interesting stuff <lh>the guix scheme indent rules dir locals don’t seem to work for me, but I can’t figure out why. the problem occurs in emacs -Q <czan>Do dir locals work for you elsewhere? Are you using scheme-mode? <lh>oh god it’s because I use my own emacs package to override the source field, and we apparently patch scheme-indent-function <lh>actually no I don’t think that’s it, that patch doesn’t affect the integerp case <czan>We don't patch it, do we? I thought we just add to the symbol's property list with PUT. <lh>the patch is at gnu/packages/patches/emacs-fix-scheme-indent-function.patch <lh>but I think it’s unrelated <lh>so to reproduce, run ‘emacs -Q doc/he-config-bare-bones.scm’ from the root of the guix repository, and press ‘y’ when prompted to apply dir-locals <czan>Mmm, I would argue that patch doesn't belong in Guix, but sure. <lh>I would tend to agree <lh>then go to the line with (guix-defaults and indent, and it indents by one extra space <lh>despite (get 'home-bash-configuration 'scheme-indent-function) returning 0 <lh>indicating the dir-locals are applied <czan>That's not what the indent function means. Quoting from help for lisp-indent-function: <czan>The property may be "an integer N, meaning indent the first N arguments specially (like ordinary function arguments), and then indent any further arguments like a body;" <czan>So it doesn't mean to indent by 0 spaces, it means "indent all arguments as if they were a body", which is a two space indent. <lh>right, I know the integer doesn’t indicate the indent width <czan>So then home-bash-configuration's arguments should be indented two spaces (despite the file in the repo only being indented one space). <lh>oh the guix files are wrong? <czan>Just tested: "guix style -f doc/he-config-bare-bones.scm" gives the result in the repo, which is different to Emacs' indentation. <lh>the guix style we control is different from the dir-locals we control!? T_T <lh>but I think you are right, emacs-xyz.scm is full of git-reference forms that are one space, but that has scheme-indent-function 0 as well <lh>but then I find it hard to imagine all the guix maintainers who are emacs users are constantly avoiding indenting with emacs, or just constantly having to guix style to fix emacs behavior <lh>whoever is adding all these scheme-indent-function props to dir-locals must have observed the discrepancy <czan>I expect you'll instead find inconsistency. <lh>yeah, you’re right again 🤦 <lh>I checked the blame and Hilton added a bunch of the dir-locals recently, including git-reference. Then I checked their commits mentioning git-reference following the dir-locals update, and they seem to be using guix style on everything <lh>I don’t get why we would bother to add a bunch of custom scheme-indent-function props that don’t even match guix style, but I guess the behavior without them must be even further away from guix style <lh>apparently emacs has lisp-body-indent which you could set to 1, but I don’t think we want that behavior for arbitrary scheme macros <lh>seems like it would be less invasive to tweak guix style to use 2 spaces for body forms rather than to try to pull emacs further away from defaults <lh>ok, guix style seems to use (guix read-print) which sets up 1 space indentation in %special-forms and %newline-forms <lh>I can imagine the rationale that the indentation can get to be a bit much with package forms, and they’re very common in guix, so it’s worth squeezing out extra columns where you can. but IMO 1 column isn’t worth having to monkey with custom logic to make emacs behave the same way <lh>actually only %newline-forms get 1 space. well, that’s enough of that rabbit hole, I can certainly live with the current behavior. thank you for the help czan! <czan>No worries. Sorry that the answer wasn't more satisfying. :) <lilyp>lh: I do wish this had been discussed beforehand, because from my POV it's `guix style` that's broken and making Emacs conform to it is not the right solution here <identity>the indent rules do not belong in .dir_locals, they should be part of ‘guix-devel-mode’ and mimic what guix style does… actually, no, last time i tried to guix style my home configuration it was horribly mangled <lilyp>`guix-devel-mode` has not and likely won't be a requirment for Guix development <lilyp>it can mimic what our .dir-locals does for other channels, sure, but .dir-locals *is* the right place for this; similar to editorconfig et al. <spencnet>Hi Trying to set up Guix SD for the first time and running into trouble following the update steps in the manual <identity>either way, it is really weird to have 3 separate places that control code formatting, and are in conflict with each other <identity>‘guix-devel-mode’, .dir-locals.el, guix style? or does ‘guix-devel-mode’ not do that? <lilyp>guix-devel-mode is not authoritative <identity>spencnet: by the way, it is not Guix SD anymore, just Guix System <lilyp>`hash guix` is a cautious measure anyway; you might be fine without it or you might need it depending on your shell's behaviour <identity>lilyp: but still, ‘guix-devel-mode’ provides indentation rules for a bunch of keywords, which makes the way indentation works extremely confusing: it works like this here, differently here, and guix style indents it differently once again (usually by not breaking lines at all and then crushing you code against column 80) <spencnet>Also, I installed guile-studio and it seemed to complete, but I don't see where it is or how to launch it <identity>i would guess you can just launch ‘emacs’ and that would be it <spencnet>Not sure how to launch emacs either though, `emacs` on the command line doesn't do anything and I don't see it in the Gnome search <spencnet>Oh, I just assumed it would have a different way of launching is all <spencnet>Is it a plugin for emacs that I start from inside it? <identity>ACTION pulls in seemingly every Elisp package in the repos to build guile-studio <spencnet>when does an app get added to the `Show Apps` GNOME Widget? Is there something special I need to do to place it there? <identity>spencnet: you may need to log out and log back in <identity>spencnet: the guile-studio executable is at $(guix build guile-studio)/bin/guile-studio <spencnet>Didn't seem to add it to the apps list BUT the guix system reconfigure is working now post-logout <identity>the usual way to fix environment problems is to log out and back in; you should not need to do that often anymore, though <spencnet>Huh, the giux build is doing a lot of work pulling substitutes even though it had guix install'd sucessfully earlier <spencnet>is that because the guix pull just went through so its getting a different version now? <Rutherther>spencnet: or you ran guix gc. Then you would redownload it all even without pull <csantosb>Hi ! Regarding #7453, I'm a bit surprised we provide all combinations of `guile-xxx`, ... but not `guile` itself. <civodul>csantosb: not sure what you mean by “xxx” :-) <kratacoa>civodul: it's been slow for me for like a month :( <kratacoa>for refrence, I am in a neighbouring country (to France) <civodul>oh a month? for me i’d say a few days but i don’t know <kratacoa>civodul: I didn't measure it, mostly a matter of feeling. More or less since 1.5 release <csantosb>civodul: we have guile-{3.0/pinned, 3.0, 3.0.11, 3.0-latest, next}; guile2 is even worst <Rutherther>There are times when it is fine and time it isn't, but I haven't noticed a chabfd <kratacoa>I'm speaking of guix operations involving substitutes <csantosb>kratacoa: It becomes unpractical, yes; specially, when you don't have a lot of disk space <csantosb>Then there is an issue with sw becoming more and more bloated with updates, which produces frequent timeouts in guix infra, see #7227 <civodul>Rutherther: the ci.guix front page is one thing (it’s always been slow), but here i was talking about substitutes <kestrelwx>Was trying to also complain about the Internet, but the message didn't go through. <futurile>kratacoa: if it's downloading substitutes have you tried any of the mirrors on the Wiki? <futurile>gabber: I'm doing a bit of research on Gentoo and Arch Linux governance today, to add some 'content/evidence/ideas' to the discussion <kratacoa>anyway, no. just using %default-subtstitutes <civodul>ACTION is curious what the “mirrors of the wiki” are <civodul>substitute delivery bandwidth would be a good topic for an internship <civodul>first setting up something to continuously measure things <civodul>and then experimenting with various options to improve on it <civodul>back in the day there was this experiment with the Amazon CDN <csantosb>Downloading 65 MB gcc-14.3.0 since 3 min, not even half of it <kratacoa>civodul: oh, I remember reading on guix-devel (I think?) discussions on how to set up new internships. Is there anything new on that front? <kratacoa>I would actually enjoy applying for such a thing but I'm not a professional nor do I have the skills of one, and I'm not even enrolled in a CS degree; not sure what would be the requirements <civodul>kratacoa: there are two things: Outreachy, to enroll people from underrepresented communities, and then i think Guix Foundation is considering setting up internships as well (i didn’t follow though) <kestrelwx>dthompson: `guile-3.0-latest` is the same as `guile-3.0.11` at the moment, so you mean `guile-next`? <gabber>kestrelwx: are you talking about # ... wait. did i not open a PR last night? <gabber>ah, forgot to click the button. #7458 <ekaitz>civodul: I told futurile about what GNOME is doing, did you take a look to that? it's very interesting <dthompson>kestrelwx: I guess? I'm unsure which guile variant is breaking the hoot build <kestrelwx>`guile-next` tracks latest main commit, `guile-latest` the latest tagged release, and Hoot doesn't build on the former. <dthompson>okay so it's a new build of guile-next that is the issue? <dthompson>I've used guile-next for hoot for a long time but I haven't updated my dev environment for a couple of weeks at this point so I didn't notice <kestrelwx>Yes, it broke after the `guile-next` update. <dthompson>okay so maybe it's a guile regression and not a hoot bug, then <kestrelwx>I think so too. Depending on `guile-latest` shouldYes, it broke after the `guile-next` update. <ieure>I have successfully integrated the Xorg modesetting driver tearfree patch in my system. Gonna PR that today. <acid-bong>evening. last week (March 18) codeberg-cli was discontinued by the maintainer, should it be kept in the Guix manifest? <futurile>acid-bong: if it builds and works it can stay I reckon - I use the alternative I think (fj??) <civodul>this ci.guix slowness issue is weird: if i wget one of the /nar URLs, it’s pretty fast <civodul>i cannot really exhibit the slowness problem by just running wget on one of these URLs <cbaines>civodul, what are you seeing when you think it's slow? <identity>acidbong: we should probably replace it with forgejo-cli at some point <civodul>cbaines: the “is somewhat slow” message and substitution being indeed slow <civodul>IIRC nginx doesn’t cache /nar requests at all now on ci.guix <civodul>we could try to reintroduce a bit of caching <cbaines>so my understanding is that these are very different things <cbaines>I think the "is somewhat slow" message appears where the time between trying to open the connection and finishing receiving the response headers exceeds 5 seconds <cbaines>then if the download speed is actually slow, that's all after that period when the response body is being read <cbaines>I often see the "is somewhat slow" message for both bordeaux and ci, although my assumption with bordeaux is that this is because bayfront is actually slow <cbaines>there's probably ways in which things could be improved in guix publish, e.g. force/flush when the response headers have been written so that they arrive sooner, or maybe enable using X-Accel-Redirect and have NGinx write the nar rather than starting a thread every time for this <lispmacs[work]>hi, how does one download the source for gcc through guix? my first thoughts were "guix build -S gcc-toolchain" or "guix build -S gcc" but this did not work <ieure>lispmacs[work], That's how you'd do it. In what way did those commands not work? <identity>«warning: package 'gcc-toolchain' has no source» <lispmacs[work]>if I try do it for package "gcc", I get error that there is no package gcc <kestrelwx> `guix build -S -e '(@@ (gnu packages commencement) gcc)'` I think. <ekaitz>idk if I like that gcc package is hidden <marty>Hey Guix folks! I'm trying guix out on my gnubooted T400. It's going well and I've noticed a few things so far. <marty>It uses ath9k. I've found it to be twice as fast as the ar92** options. <marty>it is available on second hand stores for $20 or so <lechner>Hi, how may I refer to a "lib" output inside file-append, please? This does not work: #$(file-append (list gcc "lib") "/lib") Thanks! <marty>I'm getting server is somewhat slow as well <marty>I picked XFCE but noticed that GNome is available. I think its Gnome 50 will have a systemd requirement. <lechner>Rutherther / I have to use search-input-file? <marty>That being said, is there a good DE that uses wayland and is available for guix? <marty>or xwayland might work as well <Rutherther>lechner: no, just use someting to ungexp the output, ie. (string-append #$gcc:lib "/lib") <lilyp>we also have plasma, if kde is "good" in your opinion <lechner>Rutherther / i get error: gcc:lib: unbound variable <lilyp>we're currently on gnome 48, and we were on 46 when 48 came around – hopefully we're faster on gnome 50, but I wouldn't want to throw money in a prediction market on that one <Rutherther>I am not sure if they work the same in this regard <lechner>Rutherther / that's only because that file is generated. (ice-9 pretty-print) produces the 'ungexp' <marty>does guix use only shepherd? <lechner>Rutherther / how can they be different? <marty>has anyone been able to put guix on a rockpro64? <Rutherther>lechner: see in the guix/gexp.scm file for the read-ungexp, it parses out the contents around ":", producing "(ungexp gcc "lib")", I think, I haven't verified it <lechner>marty / i believe so. grep the mailing list, please <Rutherther>ungexp != #$, the latter goes to read-ungexp and that does further parsing to pass proper arguments to ungexp macro <lechner>Rutherther / you are right! it works with #$. i'm not sure that's documented <marty>lechner / ty. i will join mailing list. <lechner>Rutherther / thanks so much for you help! <marty>lechner: thanks! that helps alot. I will try it out. I have u-boot on the spi already so maybe it won't be that difficult. <marty>oh boy... read to the bottom. no success. hmm <marty>i'll give it a go using the sd card. <marty>would love to have this working <marty>using debian, I can install to the sd card while booting the debian installer from the sd card. <yelninei>hmm the linux32 version of druntime has the same dwarf warning as my hurd32 version tripping up gdb with the difference that it does not segfault :( <lechner>yelninei / then maybe those two events are not related <marty>no joy using the default AArch64, although the boot splash screen shows up on HDMI but then disappears. <chris0ax>hey people, ive been trying out guix's offline way of exporting files and importing files from the store. For example i do a 'guix archive --export -r fish > /tmp/fish.nar', then scp this to the guix vm im trying this on. Then do a 'cat /tmp/fish.nar | guix archive --import'. The files come into the store but when i run 'guix shell fish', its still looking for substitutes and trying to build packages. I turned off internet to see if it <chris0ax>would work but it didnt. Any idea what id be doing wrong? <Rutherther>chris0ax: guix shell needs more than just fish, it needs the tools to build the profile <Rutherther>chris0ax: also there are now grafts everywhere, I am not sure if guix archive passes the grafted version or the ungrafted one, but for either it will just not be enough for guix shell to not try to download anything <Rutherther>chris0ax: lastly, but maybe most importantly, are you on the same guix revision on both machines? <chris0ax>Rutherther: on the guix revision part, i made sure that both are running the same channels <chris0ax>I checked the hashes between each like 'guix describe -f channels | sha256sum -' and they matched <chris0ax>Rutherther: aw, so a better option would be to build a profile on one machine, then export that so both of these cases would be covered? <Rutherther>chris0ax: fyi (profile (content (specifications->manifest (list "fish"))))* should be an expression that will build a profile when you supply that to guix build <Rutherther>or you could also use the path produced by guix shell, up to you <chris0ax>Rutherther: yeah i was thinking of using 'guix shell' or making a profile with 'guix package in a dir' <marty>after i installed on T400, i ran guix pull and its been doing stuff for like 30 mins <ieure>The first pull is always slow. <ieure>Did you install with a 1.5.0 image? <marty>good call. nope. 1.4.0. i tried guix a few months ago and reused the iso. <Rutherther>using 1.5.0 will not help with the first pull time, though <Rutherther>except if you used the default args and are downloading from savannah <Rutherther>no need to reinstall due to that... if it's still downloading from savannah, you see the progress bar, I think it's "receiving objects". Then just concel it and use codeberg instead <chris0ax>hmm..seems like even with the case where i build a guix shell profile on one machine, then transfer it to another, guix still really wants the network to be on. I tried for 'hello' and guix still really wants to pull and substitute a tar.gz file for hello, it doesnt notice its already got the store file it seems <Rutherther>chris0ax: do not use guix shell on the target, use the profile directly, source the etc/profile from it <Rutherther>if you use --no-grafts on both machines it could work <ieure>I understand the concept of grafts, I think, but am frequently surprised at the emergent behavior when they interact with other parts of the system. <chris0ax>so if i understand it correct, usually grafts are for security updates, and if a packages dependencies get a security fix, this patched dependency needs to be grafted onto the package <ArneBab>Are there any simple commandline audioplayers packaged that do not have atrocious dependency chains? I’m trying to build an appImage and just want to run a command to play an audio-file. <ieure>ArneBab, Decode the file to WAV and use aplay? <gabber>ArneBab: aplay should fit your description i guess? <gabber>or, if you feel funky, you could also simply write raw data to your devices (bad idea) <ArneBab>What I’d need is something like mplayer, but without video <ArneBab>I’m at the point where I’m pondering to use ffi to access libao … <gabber>part of ffmpeg, not really "small", though <gabber>also not very small, but capable <ArneBab>gabber: aplay fails with ALSA lib pcm_dmix.c:1000:(snd_pcm_dmix_open) unable to open slave … aplay: main:850: audio open error: file or folder not found <ArneBab>I’d guess because I have pulseaudio only <gabber>is pipewire out of the question? <gabber>i did a small PoC project at newyears (headless aarch64) and found that pipewire solved all my hardware config troubles and other headaches <gabber>IIRC aplay works fine with ALSA only <ArneBab>I need a player that just works on the hardware of others <ArneBab>For an appImage they can simply run. <gabber>yes, that's where pipewire shines <vagrantc>as of 2a2fe7705e9f881a8a76b4b74df96784f793af54 i cannot build successfully build guix as part of guix system build ... <gabber>ArneBab: ah. i lack any and all understanding of AppImages. i thought we talkes something like containers/vm images <ArneBab>as far as I know I can bring my own libraries, but I cannot run services. <ArneBab>there I thought ffmpeg might actually be viable, but it pulls in all of clang … <vagrantc>it ends with: guix system: error: open-file: No such file or directory: "/tmp/guix-tests/store/khhzdnrwpx3xrk2yz6pxxmarxlk2jr8z-substitute-me-not.drv" <vagrantc>although now substitutese for "guix" are available and it succeeds...