IRC channel logs
2025-02-25.log
back to list of logs
<meaty>I'm trying to get hires scrolling working on it, but it seems Solaar isn't working right and I suspec it's because it installs the udev rule it needs to work into lib/udev-rules.d instead of etc/udev-rules.d <meaty>if a package is installed just for a user (not for the system) will its udev rules be installed/work? I noticed there is no /lib directory in root <meaty>but I'm not sure how guix and udev interact <ieure>meaty, I think udev rules have to be in the operating-system to work. <ieure>*packages containing udev rules have to be in the operating-system for the udev rules to work <Aurora_v_kosmose>Oh, Ludovic (i.e. a core maintainer) is actually raising the topic? Nice. <mothbot>Is it normal for me to only be getting ~200kbps MAX downloading guix substitutes? I've been trying to install guix for two days now, and most of this time has just been spent downloading substitutes. What gives? <mothbot>I wish the commands at least told you how many substitutes are left to download so I knew where I was in the process <mange>I've experienced that at times. Where in the world are you, approximately? There are some substitute mirrors around the place that can be faster. <mothbot>us-east mirror was the first thing I tried, ran into a TLS error trying to run guix system init. I just ran the command again with no substitutes and --max-jobs=8. Hopefully this works. <mange>If you have TLS issues I think it will work with http:// instead of https://. This opens you up to a MITM observing the software you're downloading, but the signatures of the substitutes is still checked by Guix so a MITM can't modify the substitutes en route. <jacobk>Would it be correct to assume that because a program is in guix's repositories, any other programs it links to should also be free software (else it's a bug in guix)? <jacobk>So for example since the exercism package automatically opens exercism.org in some cases, the JavaScript on exercism.org should be free. <ieure>jacobk, Links to in the sense of program-compilation linking yes, links to in the sense of opening a URL, not necessarily. <jacobk>ieure: Do you know in which circumstances that would be considered a bug? For example if the site is not useful without JavaScript. <ieure>jacobk, I don't have a good example, but the specific thing you're talking about doesn't seem like an issue to me. There's tons of stuff in Guix that interacts with non-free services over the Internet. <ieure>You're welcome to start a discussion on guix-devel if you'd like more opinions. <jacobk>ieure: I think the JavaScript on exercism.org is free, but I'm not completely sure. I was hoping that the inclusion of exercism in guix would be an indication that it is free. <jacobk>If all of exercism.org were nonfree software (definitely not the case), then the exercism wouldn't be useful without nonfree software which seems like it would clearly be a bug, but I'm not sure how it works out if, for example, all but some of the scripts were free. <ieure>jacobk, Probably best to open a discussion on guix-devel if you have concerns, but I think it's unlikely this is a bug. <jacobk>I agree it's likely not actually a bug; I'm asking about a hypothetical. <jacobk>ieure: Yes, I agree it doesn't make sense to call services free or nonfree, but exercism.org is also software repository. The downloaded software is not a service. <jacobk>Maybe I shouldn't use the term "software repository" here, since I'm talking about JavaScript. <jacobk>to quote the page you linked: "the issue of the client software is logically separate from the service as such" <Aurora_v_kosmose>The logical conclusion of SaaSS is that the only reasonable networked options are distributed & peer to peer. <Aurora_v_kosmose>Anything else maintains the privilege necesary to deprive the user of agency, control & autonomy. <jacobk>Aurora_v_kosmose: Are there any examples of peer-to-peer SaaSS? <Aurora_v_kosmose>jacobk: I suppose, in that you can perfectly well host torrents with a seedbox. So you have peer & SaaSS which could decide to deny the user access at any point. <Aurora_v_kosmose>In terms of peer to peer or distributed program that nevertheless deprive the user of agency? There are a lot of "dapps" that are designed specifically to do taht. <jacobk>Aurora_v_kosmose: I don't think a seedbox is an example of SaaSS, since torrenting is about communication <jacobk>I would be surprised if any popular dapps are SaaSS though, since blockchain is so inefficient <Aurora_v_kosmose>"Service as a Software Substitute" is kind of build on the notion of a server or remote authority. <Aurora_v_kosmose>dapps frequently ensure authority ultimately lies with whoever pushed a given coin, token or whatever. <Aurora_v_kosmose>In general when one sees "blockchain" or "cryptocurrency" buzzwords appear, one can assume the program does not exist to serve the user. <jacobk>I thought most dapps were about negotiation though <jacobk>A public record of who "owns" a coin or token or something is communication, not SaaSS <Aurora_v_kosmose>That is why I phrased it as "In terms of peer to peer or distributed program that nevertheless deprive the user of agency". Because it is indeed distinct from SaaSS built around the server-centric paradigm. <Aurora_v_kosmose>The same desire to abuse the user is present, but the means are quite different. <Aurora_v_kosmose>To completely break the guarantees of any proof-of-stake model, one just needs to ensure the founders maintain a larger stake than anyone else (until they dump & exit, anyway). <jacobk>I am skeptical of the claim that dapps deprive users of agency but I've also never used a dapp so I don't really know <Aurora_v_kosmose>I investigated them because I like distributed stuff and saw them appear a few years back. I was generally unimpressed with what I found, with many examples that had functionality inferior to older programs. <ieure>Don't think much of web3/dapp stuff. I think they have correctly identified some problems, but the solutions are obviously unworkable. <Aurora_v_kosmose>Maidsafe is one of the more promising ones that disappointed me with token-centric design. <ieure>jacobk, You've never used a dapp because none worth using has been created. <Aurora_v_kosmose>Not sure what happened there either, it had been hyped for basically a decade, prior to the dapps & token craze taking off. <Aurora_v_kosmose>So they potentially changed course midway and I'm unsatisfied with the result. <ieure>Spritely Goblins is the distributed application thing I think is actually interesting and feasible. <ieure>I don't think it's built with P2P stuff in mind, but it also seems like such things could be implemented in it. And no blockchain junk. <jacobk>Oh, I think I've heard of Spritely Goblins. I did try running one of the examples, but I didn't see what was distributed about it. <Aurora_v_kosmose>Goblins is indeed one of the promising ones that hasn't been derailed. That's very much to do with the politics & user agency interests of the makers. <Aurora_v_kosmose>jacobk: Consider it a bit more like Erlang (but with a lot of differences). You can build distributed stuff with Goblins. <ieure>The kind of industrial-strength weirdware it's impossible to use in 98% of jobs. <ieure>Like Common Lisp in that way, a bit. <lfam>Spritely Goblins has a strong social connection to Guix <Aurora_v_kosmose>Massively-parallel immutable message-passing with potentially transparent cross-network messages. <ieure>Christine shows up here from time to time, and I think she was a more active contributor in the past. <jacobk>Cool, Exercism has a course on Erlang <Aurora_v_kosmose>Common Lisp is a lot more amenable to conventional software development than Erlang. <lfam>ieure: Yes, and some of the other contributors were very active in the earlier years of Guix too <Aurora_v_kosmose>It works but there are clear cases when a language built on another paradigm would do it either better or more succintly. <ieure>Yeah, the commonality between CL and Erlang is that they're bulletproof in ways other languages never thought to be (like hot reloading), and have features you never knew you needed. <ieure>Like CL's condition system, or CLOS. <ieure>I actually know CL, I don't really know Erlang. I'd like to, but haven't gottne around to it. <lfam>"We now have doc/guix.de.info-10, so additional [.git]ignore pattern is required." <Aurora_v_kosmose>Yeah, both Common Lisp and Erlang accommodate the use-case of "this must never be allowed to crash, nor to restart and must not require it to update either". <ieure>I like CL A lot, it's this wild combination of deep low-level, like you can eval (disassemble #'some-function) and get the asm listing for it; way high-level, like CLOS and conditions; and a huge gap in the center where the spec doesn't cover sockets or threading at all. <Aurora_v_kosmose>One could draw a lot of comparisions between the Condition System and Erlang's supervision trees. One could somewhat analogize it to procedural vs declaratives take on the same idea. <Aurora_v_kosmose>(Of course the introduction of internal and hardware-level parallelism promptly breaks the analogy) <Aurora_v_kosmose>Common Lisp was not really designed to handle parallel errors & conditions. <Aurora_v_kosmose>A meta level above the language itself is required, whereas with Erlang it's essentially built into it. <x8dcc>Hello, does anyone know why (set-xorg-configuration (xorg-configuration (keyboard-layout keyboard-layout))) isn't working? I think the (operating-system (keyboard-layout (keyboard-layout "..."))) is working because it changed on the TTYs <civodul>x8dcc: what do you mean by “not working”? <x8dcc>civodul: I mean that the keyboard layout changes in TTYs, but not in Xorg <civodul>x8dcc: ok, it should definitely work; what login manager are you using? <x8dcc>civodul: I am not sure how the login manager is called, just the normal TTY one. I have (service login-service-type) <x8dcc>With (a replacement of) 'startx', which calls 'xinit' <ngz>Hello. What is your value for useAutoBase in Guix’s .git/config? I keep getting errors about base commit when trying to use "git format-patch". <andreas-e>ngz: I work around it by doing things like "git format-patch HEAD^ --base=HEAD^", and am also interested in an answer to your question! <zkryh>Hi, I'm looking for help running Ardour with jack on my Guix system... Ardour is not giving me an option to start with jack as the backend. It is running into an undefined symbol error (libjack_audiobackend.so: undefined symbol: jack_client_stop_thread) <civodul>there seems to be dilution of responsibility here <ngz>andreas-e: Thanks. Maybe I need to set an upstream for the local branch containing the patches. I used a similar workarourd BTW. <ngz>andreas-e: While I’m at it, on another topic, did you manage to complete the Kanata package you started some months ago? <ngz>Gah, I was bitten again. These copyright lines are complicating patch processing. I think we should move attributions to an alphabetically-ordered THANKS file instead and leave modules headers alone. <andreas-e>ngz: I dropped it, since I bought a programmable keyboard and did not need the package to simulate a different keymap on my regular keyboard. <andreas-e>It could be revived from the email exchanges. <ngz>andreas-e: Oh, I see. I’ll see if I can do something about it. I’m still interested in using Kanata + Arsenik. <zkryh>Is anyone here using Ardour with jack backend? <ngz>andreas-e: It did ;) <ngz>andreas-e: Oh, IIUC, kanata is already complete. It just needs to be updated to its latest and turned into a patch. <andreas-e>ngz: There were problems with the update or one of its dependencies, this is why David sent a working patch with an older version. Something to be aware of, maybe this can be solved. <ngz>Ah, OK. Well, I’ll look into it but… Cargo… Thanks. <cbaines>any ideas why the gnutls output referenced by guile-gnutls doesn't match up with the gnutls package? <cbaines>the answer seems to involve package replacements/grafts, but I'm not sure why <cbaines>and I also don't know if this is actually connected to why I'm not getting gnutls debugging symbols in gdb <civodul>cbaines: most likely the grafted guile-gnutls refers to a gnutls without a “debug” output <cbaines>civodul, right, I see that now I'm looking at the grafting derivations <cbaines>so there's no guarantee of a one to one mapping of ungrafted output to grafted output when a package has multiple outputs <cbaines>civodul, given the debug outputs reference the out output by the full store filename, does this mean that gdb is never going to find the debugging symbols? <cbaines>civodul, hmm, maybe I've accidentally managed to work around this for the build coordinator, as I seem to be able to get debug information for it's segfaults <cbaines>but not for the qa-frontpage, maybe the build coordinator running from a gexp script makes the difference <cbaines>maybe I can hack the qa-frontpage package to reference the debug outputs and avoid this situation... <civodul>cbaines: you’re getting segfaults in GnuTLS? <cbaines>in the qa-frontpage, I think gnutls may be more involved, but as I say, I haven't managed to get a backtrace with debugging symbols yet <cbaines>I also didn't realise the qa-frontpage was segfaulting so often, it's like every 15 minutes at the moment <civodul>cbaines: is it thanks to shepherd that you noticed? <cbaines>civodul, I think the shepherd is more clearly displaying that the processing is segfaulting :) <civodul>cbaines: BTW, are you using ‘set-session-transport-fd!’, or are you passing it a port? <cbaines>I did also port over the code from the build-coordinator to enable writing the core files and to expose a Prometheus metric for when the file was written <cbaines>I'm using the Guile tls-wrap I think, which does (set-session-transport-fd! session (fileno port)) <civodul>cbaines: i attempted to restart nginx and it just hanged <civodul>cbaines: i guess i’ll have to reboot, WDYT? <civodul>shepherd is still doing things though, including receiving signals <civodul>it’ll have to be a hard reboot i guess <cbaines>I've restarted the bffe, it probably just started too early. The service doesn't require the build-coordinator since it doesn't have to be on the same machine <cbaines>it should probably just create it's pid file and then retry connecting to the coordinator until it succeeds <civodul>yes, it restarts because of ECONNREFUSED IIUC <Kabouik>Has anyone used `nbfc-linux` here? It's packaged for Guix but I see no instructions in the man page, and it does not seem to be something we can run without proper configuration in config.scm. <Kabouik>Florhizome is the one who packaged it, but I haven't seen them online in a lonk time. <ieure>I still see these push/pull errors at least once pretty much every time I `guix system reconfigure'. Happens with all substitute servers, even the one I run that's on the same LAN as the client. <civodul>ieure: can you tell if it’s usually while fetching from bordeaux.guix? <cbaines>civodul, I forget my previous attempts at tracking this down, but I think I was more trying to investigate what could be odd as bayfront as a client <ieure>civodul, It's not limited to bordeaux, I see it even with my local substitute server (which uses guix-publish-service). <cbaines>maybe being on the same machine as the NGinx that it's talking to, or the super slow caching involved in narinfo queries <cbaines>*maybe "odd with bayfront as a client" <civodul>ieure: you have a local substitute server serving over HTTPS? <civodul>cbaines: i think nginx or something must be closing connections prematurely, no? <ieure>civodul, nginx is on a different physical machine than `guix publish'. Both are connected with Ethernet. <civodul>i never experience it on my laptop, but bordeaux.guix comes 2nd in my substitute list <cbaines>civodul, if it is something to do with NGinx closing the connection, then that is being mishandled on the client side <garbados>hi! i'm having problems installing guix in order to install guile-hoot. i am having lots of different problems. is this a good place to ask for help? <civodul>so i thought this could be the reason <civodul>cbaines: oh yes, it’s definitely mishandled on the client side, hence the patch above <ieure>garbados, Yes, though help-guix is probably better if you have a lot of questions or they're detailed. <civodul>but still, it looks like things have gotten worse lately <garbados>#help-guix the channel? it's virtually empty <cbaines>civodul, I think http-multiple-get looks for the connection close header, but there's no handling of server initiated connection closing in download-nar in the substitute script <RavenJoad>Does Guix have a predicate for 32- vs. 64-bit machines, regardless of the ISA? Souffle (a datalog interpreter/compiler) can use 32- or 64-bits for int/float values, but I want to choose the word-length that is native for the architecture. <civodul>cbaines: the patches modify ‘process-substitution’ though, so that’s not the problem we’re seeing here <civodul>that is, the bug we’re talking about is in ‘fetch-narinfos’, not in ‘download-nar’ <civodul>(for ‘download-nar’ there’s commit fa70c141552c76cf4dc9666f577bf6b471fa0d50) <ieure>garbados, Yes, but please refrain from posting messages full of ANSI color escape sequences. <ieure>The URL leads to a mailing list post full of ANSI color escape sequences. <garbados>yeah i didn't post that. it contains the patch that fixes the installer <ieure>I don't know if that patch ever got sent to the right place. <ieure>garbados, If you want to check debbugs/guix-patches to see if it ever was sent in, and if not, contribute it, that would be a help. <garbados>otherwise i am googlng "debbugs/guix-patches" and praying to profane gods <ieure>garbados, See the contributing section of the Guix manual. <RavenJoad>Is gcc-toolchain a bad idea to use as an input for wrapping? I cannot seem to get Souffle to see <stdlib.h> despite it being in gcc-toolchain's include/. Souffle ends up using the gcc in the toolchain, which points to a GCC without <stdlib.h>. <RavenJoad>I can reproduce this by using the PATH and CPLUS_INCLUDE_PATH I wrap souffle with in a separate shell. <stdlib.h> _is_ in CPLUS_INCLUDE_PATH, but gcc cannot find it. <dstolfa>RavenJoad: gcc or g++? if former, you probably want C_INCLUDE_PATH or possibly CPATH <RavenJoad>dstolfa: Souffle produces c++, so g++ really. But the two wrap paths are the same right now. <RavenJoad>Turn around times on these builds are fun. I really hope Guix's CI builders have a ton of parallelism on them. The check phase takes a good 5-10 minutes on my 128 core machine. <dstolfa>RavenJoad: configure scripts are indeed where build parallelism goes to die <RavenJoad>The configure is fast. It's the check phase that is slow. 4k+ tests with some taking 2+ minutes to run just means that running the tests takes forever. <andreas-e>How about addings #:tests? #f while you are still working on the package? <RavenJoad>I can do my real work while I wait for the tests. <RavenJoad>Getting near the end now. ld cannot find crt1.o in the gcc-toolchain path. <meaty>Why is typst in Guix-science instead of upstream guix? What's the etiquette for getting it upstreamed <meaty>I submitted a patch series to rust-team for it before I knew it was in guix-science, but their packaging of it is much better <meaty>and I don't want to steal their credit by just copying from there <meaty>Should I just email the sole copyrighter on the guix-science typst module and ask if I can use some of their code + add their copyright? <ieure>meaty, Isn't guix-science GPL'd? <meaty>ieure: I mean, there's one guy with a copyright line on the file which contains their typst package <meaty>I presume they're the sole author <meaty>Would it be weird to email a stranger asking if I can use + credit their code? <meaty>or is it weirder to ask them to replicate my work (77 updates/new packages) by upstreaming it to guix <meaty>maybe both are more awkward than just using the piece of code and crediting the author without asking? <ieure>I think it would be polite to ask if they mind having it upstreamed, but isn't strictly necessary. <RavenJoad>meaty: I got a cold-email about this kind of thing recently and it wasn't weird at all. Just ask. <cancername>what's up! can I declare dependency on a channel in a directory with a "guix.scm", but without turning it into a channel? <ieure>I believe only channels can depend on other channels. <ieure>cancername, You can maybe use `guix-for-channels' in this case. <ieure>cancername, It's a procedure that gives you a Guix configuration pointing at some channels. I think a similar case came up here recently, but no luck looking in the public IRC logs. <meaty>if I have a very long patch series, what should I do if I have an update that only changes one patch? Send the whole thing again? Send a v2 of just the one patch? Or send a single-patch "amendment"? <zkryh>Hello to all, is anyone here using Ardour with jack as backend on a Guix system? <ngz>meaty: you should send a v2 with the whole thing <rekado>ci.guix.gnu.org seems to be slower than usual <rekado>I've been waiting for r-team to be evaluated for over 26 hours. <rekado>I meant "built", not "evaluated" <meaty>why does installing icecat also install all 100+ language packages for it <rekado>building the derivation manually on ci fails with: guix offload: error: open-file: Too many open files: "/var/guix/offload/10.0.0.8:22/0" <rekado>the aarch64 node at that address appears to be building rust things. <ieure>meaty, Huh, that is a little odd. Any Firefox-based browser pulls in all that stuff, but I don't think they all split them into a package per language. Could definitely see the benefit if you only have to install the subset you need, but if they all get installed anyway, I'm not sure what the rationale is. <meaty>ieure: correct, each language is a different output of a single package. Still, all of them are installed <meaty>I'd guess you're supposed to use --with-input? hopefully that doesn't trigger a local build <coyotes4ys>ok urmmm i have a soundcard not being detected it would seem <ieure>meaty, Well, I completely understand why you'd break a giant repo of every language into one package per, but if you're saying that installing icecat installs every lang package, that's the part that seems odd. <ieure>coyotes4ys, Do you get anything useful with `aplay -L'? <coyotes4ys> Discard all samples (playback) or generate zero samples (capture) <coyotes4ys> Default ALSA Output (currently PulseAudio Sound Server) <ieure>Please do not paste program output directly in the channel. <ieure>coyotes4ys, You can use a pastebin if you want to paste, but my question wasn't "what is the output," it was "Do you get anything useful." <ieure>`aplay -L' lists all the cards ALSA knows about, so, that'll tell you whether your card is detected or not. <coyotes4ys>in my hastebin paste, the first line says alsa "cannot find card '0'" <ieure>No, that's why I told you to run `aplay -L'. <ieure>coyotes4ys, What kind of sound hardware? <coyotes4ys>ok. i had never used aplay before, but now i understand you've said that it lists all the card ALSA knows about <ieure>So, yep, I agree with you, not finding the sound card. What kind of sound hardware do you have? <ieure>coyotes4ys, Are you on a laptop or desktop? Built-in sound or addon? <ieure>If on a laptop, many newer systems need proprietary blobs to make the sound work. <coyotes4ys>laptop built in. proprietary blobs needed for soundcards now?? <coyotes4ys>gpd is the make. model is p2max 2022 with n6000 processor <ieure>Yes, many manufacturers have been moving to sound hardware with DSPs that need firmware blobs. <ieure>GPD hardware is fairly niche, has all the stuff for it been upstreamed? I have an original GPD Pocket, I still use it sometimes, but lack of proper power management hurts. <ieure>Yeah, modern hardware is all made of WinModems now. Horrible. <ieure>Contributed to the Linux kernel. <ieure>Like many smaller manufacturers, GPD ships, at best, a hacked-up Linux kernel which it will never update or give source for. <coyotes4ys>whyyyyyy would hardware manuf's do this? winmodems i mean <ieure>Because it's inexpensive and they don't give a rip about much other than making money, which selling cheap junk allows them to do. <ieure>Anyhow, job #1 is figuring out how well supported your hardware is at all. Because if the vanilla Linux kernel doesn't support the hardware, no Linux distro ever will. <coyotes4ys>gpd isn't cheap junk. this thing has a processor that literally won't lad no matter what i throw at it as far as real life applications <ieure>Having owned two GPD computers, one of which completely failed within its first year and couldn't be repaired, I do feel that they make cheap junk. <ieure>They're less junky than a lot of cheaper no-name computers, but are not great. <ieure>But this is beside the point. <ieure>Oh, you don't have the Win Max, nevermind. <coyotes4ys>this computer has no lag at all with a four core non-hyperthreaded processor, when running several videos and a digital audio workstation simultaneously. it seems to be the highest quality computer i've ever owned, including zenbooks and custom built desktops. well equalling the desktops i'd say <ieure>Great, glad it works for you. <ieure>Anyhow, dig around ArchWiki, that should give you a reasonable idea what level of support exists for it. <coyotes4ys>ok yes i saw an arch result when i websearched my issue <ieure>ArchWiki is IMO the best Linux support resource on the Internet, whether you run Arch or not. <coyotes4ys>yeah iirc i found a couple very usefuls on there <coyotes4ys>i can't believe i finally have an external usb wifi so i can be the user of my own computer for the first time in my life, and now this