<raingloom>seems like tests/graph.scm fails on a recent commit. can't really investigate right now and my email client is broken, so can't send a bug either. ***pkill-9[m] is now known as itsme[m]
***jonsger1 is now known as jonsger
<apteryx>pkill9: has it been added already? I have it packaged locally, been meaning to send it. Also have a service (for which I must still write the doc) <pkill9>apteryx: i've sent a patch but it's not been added <pkill9>apteryx: could you send a link to the service? i want to add it as a service, but didn't get round to fixing the one i wrote <pkill9>i did it wrong because I think it killed the shepherd, lol <apteryx>pkill9: hehe. I'll write a quick doc for it, then append to your existing patch on guix-patches ***Drakonis__ is now known as drakonis
<apteryx>pkill9: hmm. Wondering in which service category (in the manual) a user space, OOM software should go. I was going to add a new "Linux Services", because it's explicitly tied to the Linux kernel. Any ideas? <plasma41>Are any of the Guix Days talks being recorded? <plasma41>I want to see them. I'm stuck on the other side of the pond *jackhill waves to the the other people on this side of the pond <drakonis>well they'll all be, fosdem records everything <jackhill>ah, cool. I know I've watch past videos, but didn't know it was universal. That's great. <apteryx>pandoc pulls a ridiculously huge amount of ghc dependencies <apteryx>pkill9: I finished the documentation, going to send in a couple seconds <apteryx>is (@@ ...) now prohibited in tests (due to Guile 3) ? <drainful>Is there a guix command to get the store path of a package? <raghav-gururajan>leoprikler Doubt! When use a particular source in package definition, if any of the source files contains non-free or unlicensed files, we should use that source, correct? <leoprikler>You patch the origin, so that no unfree sources exist even in `guix build -S` <nckx>raghav-gururajan: It's accpetable to remove (all, even ‘unused’) offending parts in a source snippett. <drainful>apteryx: I see. That makes sense, thank you. Is there a way to get a specific output from the build command? <raghav-gururajan>nckx I see. So if there is tarball that contains 90% files that are free and 10% files are unlicensed/non-free, and if we use that source with snippets to remove non-free parts, where will that 10% be stored in the users system? <leoprikler>if it is, you-ll have to do something like $package-$version$$, where $$ is a literal $ (end of line) <raghav-gururajan>leoprikler I am confused. Based on what you are telling, we should only use a source that contains 100% free parts correct? But nckx mentioned something about snippets to remove non-free parts. <leoprikler>If you use snippets as nckx says, you'll be left with a modified version, that contains 100% free parts. <apteryx>drainful: guix build will list them all. You could grep the result. <raghav-gururajan>leoprikler So in that case, the unmodified source is first downloaded, then patch applied to modify the source, correct? What happens to the unmodified source? <leoprikler>not sure if it's gc'd immediately or remains somewhere in the store, but either way it is not used and not returned by any guix-related command <vagrantc>pretty sure it stays in the store, like when building linux-libre, you end up with several linux-libre-VERSION.tar.* in the store ... i presume some are the cleaned sources and some are the originals <vagrantc>the linux-libre package in guix downloads upstream linux tarball, and then applies the linux-libre patches and scripts to do the cleaning <vagrantc>i think it was a while ago, but sometime in the last year or so switched <vagrantc>raghav-gururajan: "might as well can do that"? <vagrantc>it gives guix the freedom to update a minor point release before the linux-libre folks get around to it <vagrantc>raghav-gururajan: so poor i have no idea what you were getting at :P <vagrantc>since the patches and scripts for linux-libre rarely need many changes between point releases <raghav-gururajan>Wait a sec. So if linux-linux haven't got around a minor point release, that means their patches to deblob are not ready yet, correct? Then how come guix can use that patches when it is not ready? <raghav-gururajan>"since the patches and scripts for linux-libre rarely need many changes between point releases" <raghav-gururajan>vagrantc I think we should stop using vanilla linux as a source in package definitions. Because, we showing the end user where the unmodified/non-free source is. This also *gives* opportunity to use that source without patches/snippets. This is bad, right? <vagrantc>raghav-gururajan: you bring this up quite often <vagrantc>i know in the founding of guix it was discussed with the fsf and rms and it was deemed fine... <raghav-gururajan>vagrantc I don't know why. Whenever I work on pakages, this thing keeps bugging me like an OCD> <vagrantc>linux-libre does it as well ... just waiting for linux-libre would be hiring out the dirty work to someone else <raghav-gururajan>Ah I see. Was that discussion public? May the contents of that discussion could satisfy me. :) <vagrantc>libreboot also does the same sort of thing <vagrantc>it's a pretty standard practice for free-software respecting projects <vagrantc>*someone* has to do the sanitizing, if sanitized sources are to exist and be used as free software <raghav-gururajan>Is it? I didn't know. I thought projects that distributes free software should not convey where to get the non-free software/source. <vagrantc>it doesn't mean you have to put filters on your web browser to censor the internet <vagrantc>you have to not advertise to the end-user <vagrantc>to not allow automated systems to sanitize otherwise free software would mean you just can never use that software ***apteryx_ is now known as apteryx
<leoprikler>raghav-gururajan: the opportunity is already there even without doing so. See the channel that must not be named. <raghav-gururajan>leoprikler Yes, yes, I just got the clear idea. We can only control what we do (what we distribute), we cannot control what others do (what others do with what we distribute). <terpri>the linux-libre project doesn't have pre-librified tarballs available for all versions (iirc they sometimes delete them to save disk space) <terpri>which would be at least one reason to use the patches with upstream tarballs <bricewge>How to install libc manual, for example to get `man exec` working? <bricewge>leoprikler: It makes `info` works but not `man` unfortunately *bricewge still using an inferior documentation reader <terpri>bricewge, the man pages are (logically enough) in the man-pages package. i think glibc only has info docs <bricewge>terpri: Thanks, it works! I though that man-db and gnu-c-manual was enough. *kmicu wonders whether Guixers at FOSDEM have a good time. <oriansj>kmicu: people can have a good time anywhere at anytime; or a bad time anywhere at anytime. <oriansj>but generally people don't go to a thing that they hate or dread, unless they are forced by external circumstances. (Required for work, spouse demands, etc) <leoprikler>"My spouse forces me to use free software against my will" would be an interesting light novel. ***nly` is now known as nly
<mbakke>kmicu: I have it on good authority (unfortunately not my own) that the 30-something Guixers are enjoying themselves. :-) <oriansj>leoprikler: yeah and my wife could write it; probably end up calling it 50 shades of software freedom. <leoprikler>"Your safe word must contain at least one capital letter, one number, and one non-alphanumeric symbol." <dftxbs3e>I'll be at FOSDEM, couldnt make it to Guix Days <nckx>dftxbs3e: Greetings from the POWER9 hacking corner. <nckx>In which I try and fail to bootstrap Guix on POWER and jonsger tries to remember how he did it. <nckx>nckx: I'm at the Guix Days, unfortunately I won't be able to attend FOSDEM proper ☹ <dftxbs3e>nckx, run `docker run -it registry.gitlab.com/lle-bout/guix /bin/bash` <dftxbs3e>It's a container image that has a working ppc64le GNU Guix install inside it <dftxbs3e>that install can't build GNU Hello, but it brings you to the point where you can inspect to make it work <nckx>dftxbs3e: I already found the second link (thanks!), I'm trying to understand how to do this rather than ‘just’ run docker (which I don't have anyway). <nckx>dftxbs3e: Oh, I did that already. <nckx>guix show hello works, guix build hello doesn't. <nckx>But, I might be getting somewhere cross-building the binaries on my x86 box. Maybe. <nckx>Cool, it's building something. <kmicu>Do you have a POWER9 hardware there (e.g. on a Raptor conference stand) to deploy Guix on it? 😺 <nckx>dftxbs3e: That's exactly what I'm doing. <nckx>After the long-double-128 error and patching cross-base.scm to add it. <Draven>Does anyone else have an issue with a lot of rust drv's takin a lot of time? <Draven>My computer have been calculating rust drv's for the last 24H :/ <cbaines>I've built it successfully, but also had it fail to build <nckx>jonsger: /gnu/store/3cyrklgpakahjh342qpqg92r6n0dps9j-bootstrap-tarballs-0 <isBEKaml_>Hi, I believe this is a pretty common problem - how do you write shebangs in your scripts? <isBEKaml_>The path can't be some custom path, right? Or do we symlink to a well known path? <nckx>isBEKaml_: #!/usr/bin/env bash <nckx> /bin/sh will also work on most Guix Systems, but is less ‘portable’. <isBEKaml_>nckx: Yes, that's what I wound up with, eventually <nckx>isBEKaml_: Good, using /usr/bin/env is the current consensus amongst distroes. ***isBEKaml_ is now known as isBEKaml
<efraim>nckx: I can host them online if you need somewhere <efraim>The bootstrap binaries. I host the backup copies of the aarch64-linux ones and the WIP ppc32 and now wip Hurd ones <nckx>efraim: Any advantage to you hosting them over me throwing them up on tobias.gr? <efraim>nckx: not if you already have a spot for them <nckx>Ah, OK. Thanks for the offer 🙂 <nckx>They're pretty trivial to build so I'd rather not share them at all. It's a bit early to get back-doored. <mjw>hi guix room at guixdays <apapsch>How do filesystems in a operating-system relate to testing in a vm (or deploying via ssh)? Should I just assume there's a /dev/sda1? no reference by id? <bandali>leoprikler, hey, sorry, had already passed out. i have a question about ‘building emacs packages using emacs-next’ that we discussed yesterday. let me know when would be a good time for me to ask <bandali>but after adding “emacs-next-delight” to my manifest and installing it, it looks like it was built with emacs-minimal (which is 26.3) anyway <leoprikler>I think substitute-keyword-arguments does not work here, because no such argument is supplied. <leoprikler>Try adding the argument with (arguments `(,@(package-arguments p) #:emacs emacs-next)) <bandali>hmm, doesn’t it add it if it doesn’t exist? at least that’s what i think i saw happen in tests/utils.scm <bandali>also, if we just add it like that, what about the cases where #:emacs already exists? will the second instance of it (which we add) be used, or will the duplication be problematic? <scmguru>smithras: How are you this fine morning? <leoprikler>it should not be a problem, the latter will override the former <smithras>scmguru: I'm learning a lot at guix days :) <itsme[m]>is there a way to set proxy settings in guix? seems that the only way is on the command line <bandali>leoprikler, hmm, i see, will try that. there’s also an interesting use of it in the guile-emacs definition where they supply a new list on the fly. any reason to do it that way? (add #:parallel-build? and #:tests? there) <bandali>ha, so kinda like what i’m trying to do? ***coldpress_ is now known as coldpress
<bandali>thanks, so adding the emacs-minimal default helped. now i’m bumping into an issue with the name of the generated autoloads file for the package, where it has the “next-” prefix <bandali>guess i’ll have to look into the emacs build system again <mfg>Can someone explain where the android-ndk-build-system sets the src files for a given package? <bandali>leoprikler, so the culprit seems to be package-name-version->elpa-name-version from (guix build emacs-build-system). guess i could override that for myself in (bandali packages emacs-xyz) ? <mfg>Because i can't find any reference to the relevant variables LOCAL_SRC_FILES and LOCAL_SRC_FILES_linux <NieDzejkob>How can I check which package in my profile propagated a dependency? <guixy>I'm running the package manager on a bananapi. It's been slower than ever since switching to guile 3. What is the commit before they switched? <wingo>seems we don't have rr packaged ***ng0_ is now known as ng0
<leoprikler>bandali: you'd have to override the phase if I'm not mistaken <bandali>hmm. yeah a simple overriding of package-name-version->elpa-name-version in my module doesn’t seem to have helped <NieDzejkob>guixy: 8234fe653e61d0090138cbd4c48d877568355439? <joshuaBPMan>Hey debian! I've got a question for you. Which is "better": 1) Install Wordpress on debian via the Wordpress package ya'll provide....OR downloading the wordpress source code directly and using that? <bandali>leoprikler, can i “override”/“modify” a phase? or would my best bet be deleting the phase and adding a new one instead of it? <raingloom>hey, could someone help me debug evolution? i finally have some time to dig deeper into it but i could use some help. <raingloom>it doesn't show any of my emails but it does seem to successfully download them through POP. i tried using my existing config in a VM, but when i copied it over it just ignored it and showed the setup wizard. <mfg>Does anyone know a decent Tutorial which explains how to write a parser in guile? <raingloom>mfg: idk if there are tutorials but you should probably look into parser combinator or PEG libraries. <mwette>mfg: Can you be more specific? Guile comes with PEG and LALR parser generators; I have written a separate LALR parser generateor <mfg>i want to update the android tools. Since android8 they are using .bp files instead of makefiles. those files have a JSON-like structure. i need to parse them and (maybe) construct makefiles out of them. <mfg>i thought this was a better way than hardcoding everything <mfg>but maybe this should be solved in a completely different way idk <mfg>so someone already tried to update those packages ... <rupicapra>My system update is taking like 18hours now, building multple rust's <gnutec>rupicapra: You have to do this another day. <gnutec>rupicapra: You have to do this another day. <dongcarl>Hey all, did people decide on when and where for the dinner? <rupicapra>I was a but surprised coming home and seeing it <gnutec>rupicapra: I wait to upgrade another day when "sudo -E guix system..." doesn't work. <rupicapra>I get my laptop is old but it's not THAT old <gnutec>My notebook is a i3 with 2GB and is OK! <bandali>but it doesn’t seem to have helped, and the build still fails as before <leoprikler>apart from the naming issue, could this be a bug in delight itself? <leoprikler>the let inside the arguments seems really out of place <leoprikler>especially if you already have to construct the name earlier on <bandali>i don’t follow. i’m not using any of the three let-bound functions for the package name though ? <leoprikler>because you already need elpa-name to construct name, but you compute it in a way that you have to repeat your steps <bandali>(to be very clear, this is my first go at this, and by no means is this great code) <mehlon>here's a semi-convenient way to run Guix on nixos <leoprikler>How did you manage to make the same error TWICE? <bandali>i don’t know, maybe i’m stupid? or maybe because i’m trying to juggle like 5 other tasks at once <bandali>to be clear, you didn’t help with any of the let-bound functions, that’s my doing/stupidity *apteryx thinks guix deploy is amazing <bandali>leoprikler, can you give me some concrete feedback now? <leoprikler>I've managed to fix most of the build-system related stuff, but emacs-delight still won't build. <leoprikler>As I already assumed, that appears to be an emacs thing. <apteryx>leoprikler: could you please cool down a bit. You're coming off as aggressive. <dongcarl>I’m at great town square, where did everyone end up for dinner? <leoprikler>Okay, it appears as if the (file-exists-p generated-autoload-file) fails, even though generated-autoload-file is let-bound <leoprikler>Which is a weird way to fail, but it means we'll have to figure out something for that <leoprikler>(or perhaps its another file-exists-p that fails, checking) <leoprikler>more accurately, it's likely some internal reference being spilled out instead of the thing that it should refer to <leoprikler>bandali: as with emacs-next you ignored the case, where phases are unbound <leoprikler>the (commented) code I've pasted fixes that and adjusts the path correctly, but it still fails at generating the autoloads because that's an emacs bug <raingloom>i know i've asked this before, but is Evolution broken for anyone else? i deleted all its configs and reconfigured it manually and it still refuses to show my mail. <mwette> # is special read syntax; it's used for numbers quoting symbols etc That pastebin error does not make any sense. #~ is used for g-expressions; no clue what #g might mean <leoprikler>However, my gnome has some additional inputs, such as evolution and evolution-data-server <raingloom>hmm, looks like i have it propagated from jami <mfg>what does makefile syntax target: prereq := value mean? <thomassgn>I have wcalc installed and upgraded after guix refused my manifest... <bandali>leoprikler, i see. so the main two issues were: 1. me forgetting '%standard-phases for #:phases, and 2. emacs-next not generating autoloads correctly <bandali>do you know more about 2? and if it’s a reported bug? <leoprikler>Basically, the issue is, that autoloads is not loaded and the variable is hence set incorrectly. <mfg>leoprikler: so this sets a rule local variable? or what does it do? <leoprikler>thomassgn: could you rewrite this in terms of specifications->manifest? or perhaps simplify this by not resolving packages in such a way? <leoprikler>btw you can do guix package -m manifest1 -m manifest2 -m manifest3 and all will be merged into a single big manifest <leoprikler>I currently use the following to build my profile: `find ~/.config/guix -name 'manifest*' | sed -e 'i -m' | xargs guix package` <bandali>leoprikler, i see, thanks. please let me know what you find <leoprikler>bandali: I've sent a patch to the ML, waiting for the bug number <bandali>leoprikler, great, ty; will check it out shortly <thomassgn>leoprikler: the '(packages->manifest (map (compose list specification->package+output)' is what specification->manifest is defined as in gnu/packages.scm Already tried it, gives me the same result. <leoprikler>perhaps, but one of them is significantly nicer to read <thomassgn>I could also write them out plain, but I'd have to add a bunch of module imports, I find this easier. Though just now it's become harder... :) <thomassgn>haha. Yes. The spec->manifest requires an import though. :) I'll try again with it. <leoprikler>You can try splitting it up into multiple files as I already showed you. That keeps imports per file minimal in my experience. <thomassgn>yea, I might just have to do that. I just don't understand why this does not work anymore, been using it for a while now without problems. <leoprikler>or would this file still work with an old version of guix? <bgardner>Good afternoon guix; I'm trying to add guile-json to my system packages and getting weird conflict errors ("guix system: error: profile contains conflicting entries for guile-json") that are new to me. Any advice to resolve this? <bgardner>terpri: So says the error message, yes. Let me read that and come back <bgardner>terpri: This bug report seems to hint that "guix" is an installed package in the user's profile. I'm calling 'guix system reconfigure' with guile-json added to the package section, and I take that to be a different use case, no? <bgardner>terpri: Although possibly a distinction without a difference, I guess <mwette>Given a package is there a way to find it's guix package spec (in Scheme)? I get a hit from "guix search bytestructures" but I need the #:use-module part to add to my own package spec. <dctrud>Hi guix - had an interesting question from Compute Canada about an issue where they use Nix userspace and are running Singularity on HPC cluster <bavier>mwette: 'guix search' will output a "location" field that can be transformed straightforwardly into a scheme module name <dctrud>Compute Canada is using Lmod _and_ Easybuild on top of Nix, on top of CentOS <bavier>dctrud: reminds me I have an lmod package I've been meaning to submit <mehlon>I'm actually running Nixos right now but I'm using linux libre which surprisingly even has wifi on my device <pkill9>mwette: there is "specification->package" in gnu/packages.scm if that's what you're lookign for <mehlon>[6100:6100:0131/223850.981095:FATAL:platform_font_skia.cc(97)] Check failed: InitDefaultFont(). Could not find the default font <mehlon>I feel like I've had this exact conversation before <dctrud>bavier: what's the use case of having lmod vs using environments? build packages from source into arbitrary dirs and expose as modules? <bavier>there are several blog posts and papers regarding Guix in the sciences <mehlon>but no one mentioned the word deja vu before me so it's probably not a glitch in the matrix after all <dctrud>bavier: yeah - I've read through that periodically. My background is an HPC admin in academia (biomedical field). <mehlon>Drakonis: did fc-cache -fv but same error still.. I should mention I installed chromium with guix but am running NixOS <bavier>dctrud: some people have build processes that are run outside/on top of guix/nix that produces modules for end users <Drakonis>then you might want to follow a different guide <dctrud>bavier: fair enough, I guess if you have nix/guix base, easybuild making lmod module then you cover more ground in terms of applications available <mehlon>good point, maybe I need to get some fonts with guix first <bavier>yeah, kindof a crutch thing until everything has guix packages :) <dctrud>I'm one of the developers of Singularity these days. Just trying to puzzle out where containers come into it also for people who commit to this type of stack <mehlon>huh... I mentioned the matrix and now youre talking about singularity eh? hmm... <dctrud>I won't talk about the exact issue Compute Canada have with singularity here, as it's related to non-free things. <mehlon>Drakonis: thanks it worked. I just had to install font-dejavu and thats all <mehlon>maybe I'll install font-liberation too.. it sounds very... freeing <mehlon>it's not like a taboo or anything :3 <mehlon>ok I know I'm starting to sound crazy but I just realized that my issue that gave me deja vu was solved by installing... deja vu. <Drakonis>dctrud, i figure it has to do with HPC hardware such as nvidia? <mwette>bavier, pkill9 : I typed "guix repl" followed by ",use (guix packages)", but guile3.0-bytestructures is not defined. Should it be? <mwette>^ this after "guix search bytestructures" reports package w/ name guile3.0-bytestructures and location guix/pacakges.scm:880:11 <bavier>mwette: uff da, no, it's defined in (gnu packages guile). <bavier>The location must be thrown off the the use of 'package-for-guile-3.0' <pkill9>mwette: it's not in guix/packages, if you run `guix edit guile3.0-bytestructures` it returns you to a function used to generate the package <bavier>I thought we had fixed that behavior <bavier>mwette: if you want, a bug-report might be nice <bandali>bavier, do you have a minute? was wondering if you had a chance to have a look at my emacs-xyz patches? <bavier>bandali: not yet, but planning to tonight. sorry for the delay <bandali>bavier, gotcha, no worries, and thank you <bandali>they should be fairly straightforward and easily mergeable <nckx>Someone asked me today how #guix has grown over the years. I don't know if we can know that retrospectively, but boy can I run useless Perl scripts on the logs! <lfam>nckx: These "random" quotes... 😂 <Drakonis>probably a good time to reset the most referenced urls <Drakonis>sanitize it and get rid of the links from the bots <nckx>Let me censor those as is our want. <bandali>> nckx is either insane or just a fair op, kicking a total of 5 people! <nckx>bandali: Trick question, I'm both. <mehlon>oh, guixchan? I think there was a cute anime version of guix somewhere, let me find it <dctrud>Drakonis: sorry, got called away. It is due to NVIDIA stuff <dctrud>They have Singularity on a CentOS host where people use Nix environments, and Singularity uses ldconfig to find libs which may be in odd places *nckx → bed for 4.5 hours of sleep, thanks for believing in me. <nckx>I remain a real boy for another day. <dctrud>But they need the CentOS ldconfig for that... not the nix one <mehlon>so. immediate unanimous action to change the official Guix logo to that? <dctrud>Easy workaround make sure you don't have Nix PATH when you run Singularity installed on CentOS... but there was an ask for a config option <mehlon>I apologize and for this slip up I shall retire effective immediately <dctrud>I was just curious how people are, if at all, using containers if they've committed to Nix/Guix environments to work in *mehlon it's a 'The Good Place' reference <Drakonis>nix/guix have containers that pull from the store <dctrud>Yeah in this case they appear to be needing to routinely run Singularity containers that aren't anything to do with their Nix setup <bavier>dctrud: apparently some will produce singularity images with guix to be shipped off to a singularity runtime elsewhere <bavier>dctrud: 'guix pack --format=squashfs <packages>...' <dctrud>bavier: yeah, that's what I'm aware of. Here they are running Singularity images on the systems that have Nix envs on though <leoprikler>"The GuixSD distro uses the guix package manager (it is like APT in Debian) which is the best package manager ever, dealing with gazillions of file versions without squinting." <dctrud>not creating them to ship places without the reproducible envs <leoprikler>Well, at least they have accurate, if somewhat outdated, information 😉️ <dctrud>I guess it's covering the base of people who want to singularity run docker images, and not interact with Nix envs <Drakonis>i'm not sure if linuxreviews is supposed to b satire or serious <mehlon>it's just a bit of an interesting site, like 9front's main page <leoprikler>I also like the 10 most used words on IRC: package there build would think system should about could which <mehlon>I imagine vsauce reading this like an actual sentence <mehlon>I do like how linuxreviews always has kpop images for all their reviews <dctrud>"holy moly" were the exact words I used in a company chat :-) <Drakonis>doing hpc without this kind of setup would be so good. <rekado>I’m perpetually struggling with the Guile API for derivations and monadic values. <rekado>I find it terribly difficult to see the type of a value <dctrud>Drakonis: I found myself buying some Infiniband cards on ebay the other day, so I'll no doubt play at that with Guix. Heh. <dctrud>Not sure a Ryzen desktop and older Dell Optiplex count as HPC though :-) <rekado>often the comments or docstrings are confusing because they call something “derivation”, but actually it’s a monadic value <rekado>yeah, I’d love to have either type annotations or more magic <rekado>I also find mlet, mapm, mbegin and the like to be kinda annoying. <dctrud>bavier: It'll be nice to have to mess around with. There are many things I don't miss about being in an HPC group, but some of it was fun <rekado>the idiom (with-store store (run-with-store store …)) also isn’t very pretty <rekado>I wonder how this could be improved <Drakonis>i hope that's in the books with post 3.0 guile <rekado>I’m trying to make the GWL use an inferior Guix, and it’s been driving me mad