IRC channel logs
2022-11-20.log
back to list of logs
<ahriman>Hello, I am trying to package shortwave and during the build, meson calls cargo. Is there a way to pass #:cargo-inputs to meson-build-system? <singpolyma>ahriman: you shouldn't need #:cargo-inputs anymore anyway <ahriman>I am running into the problem that cargo.toml has as dependency rust-cast="0.16" but guix search rust-cast only shows 0.3 and 0.2 <ahriman>Is packaging something written with rust a very involved process? <ahriman>so guix has a package per version of every rust lib? <podiki[m]>wonder what the kernel changed, though mesa in master is pretty outdated at this point <apteryx>hmm, switching our mesa to latest (22.2.4) didn't improve anything with my opengl on top of linux 6.0.8 <ahriman>git-reference with (recursive? #t) is not cloning git submodules, is there another way? <lilyp>ahriman: that'd be cursed if true <lilyp>if you already have the checkout cached, you need to flip a bit to cause a rebuild <apteryx>nckx: uh, regarding linux kernel bug reports: "Locate the driver or kernel subsystem that seems to be causing the issue. Find out how and where its developers expect reports. Note: most of the time this won’t be bugzilla.kernel.org, as issues typically need to be sent by mail to a maintainer and a public mailing list." <FareTower>(answer: yes, says git, because fortune-mod contained "non-nice things"... whatever that means) <AwesomeAdam54321>It's a blocker to adding more GNUStep packages for Guix, unless the compiler should be changed to Clang for all of them <FareTower>my yubikey seems to work in a terminal, but in either chromium or firefox, it fails. Is there something special I need to do to configure fido? I do have pcscd running, but didn't configure it specially. I didn't do anything with udev either. <FareTower>pcscd seems to detect the yubikey when inserted: when started with -d, it outputs a Card ATR line. <podiki[m]>FareTower: not sure you saw my messages before, but I've had to include udev rules <FareTower>podiki[m], where do I find those udev rules? <FareTower>I saw that you mentioned pcscd and these udev rules, but it's not obvious what rules I need to put where <FareTower>Assuming I have a copy of this file---how do I convince guix to install it in /etc/udev/rules.d ? <podiki[m]>take a look at udev rules in services in the manual, but here is what you want: <podiki[m]>in your services: (udev-rules-service 'u2f libu2f-host #:groups '("plugdev")) <podiki[m]>libu2f-host is deprecated, I think you should just replace with libfido2 but I haven't yet done that myself <podiki[m]>will also need the (gnu packages security-token) module in your system config if you don't already <raghavgururajan>When a package name is changed in a new version upstream, do we usually change the name in existing package-definition or create a new package-definition with the new name and inherit from older definition? <raghavgururajan>FareTower: Under services in system config, you could create a `simple-service` and extend `udev-service-type` to include your new rules. <lilyp>FareTower: if you need fortunes, use daikichi with fortunes-jkirchartz – or package your own fortune database <jpoiret>been a long time :( trying to get back into it (and the 700 unread mails aren't helping) <unmatched-paren>i wonder how long it will be before i give up on reading Every Single Mail <jpoiret>tbh, I think managed to do that when I was reading them every day, but I haven't managed to do that for the past 2 months <unmatched-paren>i use aerc's V and mr commands to select and mark-as-read the ones i'm not interested in, and i occasionally do /search -u to see if any escaped <jpoiret>oh, that's what I do with notmuch in emacs as well <PotentialUser-30>But when i try to compile i get error messages, for ex. "no code for module (gnu package gst-plugins-good-qt)" <PotentialUser-30>Whats the right way? What did a miss. Is there a quick and dirty way without creating channels? Thanks again for your kind help ! <unmatched-paren>PotentialUser-30: if you want to do that without a channel (channel is highly recommended by the way :)) you need to create a directory for your packages <unmatched-paren>and put all your files in there, and make them all modules by using define-module <unmatched-paren>also, if you do that, you can remove the return value at the end of the files <PotentialUser-30>OK. I tried already with use guix install -L package-dir but didn't work. But I try again and report! <unmatched-paren>moke sure to replace the use-modules forms in each file with (define-module (FILENAME-WITHOUT-DOT-SCM) #:use-module (...) #:use-module (...) ...) <PotentialUser-30>BTW: I use the webclient for chatting. Is there a good chat client you guys would recommend? hexchat? weechat? Others? Emacs? ;-) <unmatched-paren>and to import one of the other modules, use #:use-module (FILENAME-WITHOUT-DOT-SCM) <PotentialUser-30>With guix build -L /home/joe/guix-pakete/ i get: no arguments specified, nothing to do <PotentialUser-30>I thougt that and tried: With guix build -L /home/joe/guix-pakete/ nheko-qt i got: failed to load '(nheko-qt) <PotentialUser-30>Is dpaste also OK? (besides the problem with the missing backslashes) Is there a registration needed for paste.debian.net ? <unmatched-paren>PotentialUser-30: you need to use (gst-plugins-good-qt) instead of (gnu packages gst-plugins-good-qt) <PotentialUser-30>Dont know if it's really a guix question: The new nheko-qt doesnt find the changed gstreamer plugin. Is there an automatism in guix for the plugin search paths or do i have to dig again in the build itself and patch? <unmatched-paren>PotentialUser-30: though i think you need to use (modify-phases (list #$@(@@ (gnu packages gstreamer) %common-gstreamer-phases)) ...) <omlet[m]>Whats is the oficial guix group in matrix.org? <PotentialUser-30>unmatched-paren: Thank you. Even more down the rabbit hole ... :-) But i think i should read the Scheme Primer first ! <user___>there is no advantage on xmpp or matrix over irc <omlet[m]>Are more advanced in my opinion than irc <omlet[m]>Nixos.org gnome.org, KDE.org. all with matrix servers <user___>the biggest difference as that you can use images/videos on matrix <unmatched-paren>the more advanced features can cause problems when interacting with irc <user___>omlet[m]: But what do you mean by advanced? <user___>I respect your opnion, just tryng to figure out <unmatched-paren>long messages are truncated and put on some matrix text paste site, which is quite annoying <user___>I already run my own matrix server and also irc, but never a jabber <unmatched-paren>the images are put on a media host thing, so it's exactly the same as if you just uploaded the image to the internet <omlet[m]>user___: Its free software too, for me its better use XMPP And matrix, irc its sam with revolt.chat, good for family, but not good for groups <user___>Matrix is good when you run your own server in a non federated way, or if some of your friends run this <unmatched-paren>as i said, the extra features of matrix are pretty much useless across the irc bridge <user___>the biggest problem with irc and even xmpp is that lots of people refuse to use <omlet[m]>Peertube GNU social pleroma akkoma Mastodon all woth same federation protoccol <user___>I think efnet doest federation for real <omlet[m]>user___: Its only package the CLI or tui cliente for guix <unmatched-paren>omlet[m]: ah, is the problem that there are no GUI irc clients in Guix? <omlet[m]>I think oar matrix u xmpp u agride the gnu project even <omlet[m]>I feel there are things that can improve in guix, learn from nixos and other distros <user___>What its the problem with old things man <user___>Your using TCP protocol to connect here <omlet[m]>user___: Why is very hard for me, i dont line the usablotty only this <unmatched-paren>there's quassel and konversation too, which are both Qt irc clients i think <omlet[m]>user___: The question os, have better alterbatives freedom software <user___>I still cant see how xmpp or matrix is better <user___>being created after is not being better <omlet[m]>user___: Why is better view the IRC group in matrix <user___>For guix help, I never needed SASL, I join here without even a registered nickname <yarl>Are files in /run/current-system/profile always read-only? I am trying to package a web application (web server + php + mariadb). What I have done so far is write a package definition (copy-build-system) that extract the zip into srv/http/app and write a system configuration (which uses the package and declares nginx, php-fpm and mysql services). It works fine using `guix system container -N ...` then running the co <yarl>ntainer but during the application configuration (execution from a browser of some install.php) it fails because /run/current-system/profile/srv/http/app is read-only. Am I doing it right? Do you have any advice? <omlet[m]>I think the problem os the different model <user___>yarl: Sry Idk.. I still noob at packaging <user___>omlet[m]: Whats exaclty the problem? <user___>For me, this channel works fine. I already have some problems in the past, I just joined here asked for help and it just worked <user___>Sometimes you need to wait, or fill a bug in mailing list etc <user___>But since you already can type here, doing this from any other software will be the same <omlet[m]>My english without translate is bad, but o speak for learning <omlet[m]>user___: The fórum group too its nonfree? <user___>the structure to solve problems will be the same regardless the software <efraim>this channel is for discussing Guix. If you'd like to discuss other chat protocols then I'd suggest heading over to #guix-offtopic <omlet[m]><efraim> "this channel is for discussing..." <- Link for Matrix please <efraim>I don't know how the matrix bridges work, but its #guix-offtopic libera.chat <nckx>apteryx: You asked if there was something more structured than an ML, and that's the only thing. <cbaines>when git am says: Patch format detection failed <nckx>Good thing I never send embarrassing private mails. <cbaines>thanks nckx, that does seem to help, but git still seems to not be able to apply the patch :/ <cbaines>I think I'll just ignore that series for now <cbaines>unfortunately the automation for applying patches has worked, but only for the single patch that was resent <nckx>Are you downloading them from issues.? <cbaines>yeah, I tried that for the first patch <nckx>cbaines: That's because the first patch ‘depends’ on libaec. <nckx>This is what happens when everyone adds their package to the end of files, instead of (ideally) alphabetically… <nckx>cbaines: Feel free to ask them to resend when that happens. <nckx>gabber: Which part doesn't live up to standards? <raghavgururajan>When a package name is changed in a new version upstream, do we usually change the name in existing package-definition or create a new package-definition with the new name and inherit from older definition? <nckx>unmatched-paren: I don't understand. <cbaines>nckx, right, thanks for the explaination <nckx>...you knew? :) Sorry then. <nckx>raghavgururajan: We use deprecated-package. <cbaines>nckx, gabber the patch was sent to the guix package in the bug tracker, rather than guix-patches <unmatched-paren>nckx: ignore that, i suppose it's fine to send patches to bug-guix issues <gabber>the part after ";; Trick the internal mechanism to load the parts" where i did some weird black magic to the code-base to get the thing to work <cbaines>unmatched-paren, that does mean that anyone just doing review on the guix-patches issues will miss those patches <gabber>unmatched-paren: i think my first post in the linked thread is the opened issue <raghavgururajan>unmatched-paren: Could you do the `deprecated-package` thing for go-github-com-go-ini-ini? <nckx>unmatched-paren: I see. I agree that guix-patches is the right place (my workflow is such that I don't notice) but I don't think it should be re-sent. <nckx>gabber: OK, if you you don't understand what your code is doing, thanks for being up-front about it 😛 <unmatched-paren>somewhere along the line they changed the import path to go-github-com-go-ini-ini <raghavgururajan>unmatched-paren: There is an existing older version of the package right? <gabber>nckx: i understand what it's doing -- it's circumventing some test which prohibited the application from /just werking/. Fritzing somewhat insists on being able to dynamically pull the parts from it's own repository -- which i think is somewhat agains Guix' philosophy of having bit-by-bit reproducible packages. the boards are included in the package but if the test in question fails the boards are not included or Fritzing fails or <gabber>some thing (not sure exactly what right now) <nckx>gabber: Just FYI, we don't include ‘update hash’ separately. It's implied by ‘update to…’ at the top level. This is just a convention. <nckx>You spotted the mailing list name. <raghavgururajan>You'd have create deprecation definition for the old package with old name; which mentions the new package with new name. <unmatched-paren>raghavgururajan: but the old package with the old name still exists :) <raghavgururajan>unmatched-paren: Same goes for `gnu: go-golang-org-colorful: Rename to go-github-com-lucasb-eyer-go-colorful.` <unmatched-paren>raghavgururajan: yeah, you did remind me that i should have added a deprecation definition for that <nckx>gabber: I admit to not really understanding why fritzing's pri/utils.pri can return a Guix/Nix-specific ‘base32’ hash. Who is consuming that? <raghavgururajan>> raghavgururajan: but the old package with the old name still exists :) <raghavgururajan>Yes, you'd have to replace that definition with `(deprecate-package` definition. <nckx>That hash encoding is not used outside of Guix/Nix. <unmatched-paren>we don't want to redirect the new package to the old package, because they're different versions and something might break <nckx>raghavgururajan: I haven't been following along, but deprecation is for when the actual *name* changes, not the version. <unmatched-paren>nckx: in this case, both the name and the version changed, but the old version is still useful <nckx>Go's definition of ‘name’ is insane. <raghavgururajan>> we don't want to redirect the new package to the old package, because they're different versions and something might break <nckx>I agree with unmatched-paren. <nckx>You could hide the old version if this concerns you. <unmatched-paren>raghavgururajan: i will send a v14 adding a deprecated definition for go-golang-org-colorful <nckx>gabber: That should have pasted src/version/partschecker.cpp, not .pri. <unmatched-paren>could i just send you a single amended patch for go-github-com-lucasb-eyer-go-colorful rather than resending the entire patch series? <raghavgururajan>unmatched-paren: I was gonna say, never mind. I'll attempt to add the deprecation myself. :-) <AwesomeAdam54321>ph03n1xaimverncc: I don't use the Guix System yet, so I don't know how to use xfce-power-manager with exwm <pinoaffe>ph03n1xaimverncc: i personally have not configured anything power-management-related, lid-close seems to work out of the box in my case <pinoaffe>I have a keybind set up to lock my screen using `slock` <lechner>unmatched-paren / Hi, has anyone reported that 'git send-mail' barks with --cc="( <paren@disroot.org>" Unmatched () '( <paren@disroot.org>' '' at /gnu/store/28lhnzpb2327mpxgm13fh2ixqicgq56f-git-2.38.1-send-email/libexec/git-core/.git-send-email-real line 642. <lechner>unmatched-paren / git send-email --smtp-server=localhost --smtp-server-port=25 --base=master --cc="paren@disroot.org" -4 <lechner>unmatched-paren / sorry, git send-email --smtp-server=localhost --smtp-server-port=25 --base=master --cc="( <paren@disroot.org>" -4 <nckx>If you're explicitly passing (, you have to explicitly quote it. <lechner>nckx / quote according to whose rules? <nckx>twopubsolar[m]: Not in the example above. <nckx>E.g., --cc='"(" <paren@disroot.org>'. <nckx>‘(’ is meaningful in mail headers, so if you don't want the special meaning, you have to quote it, is what I mean. <lechner>twopubsolar[m] / no problem. it was my error. sorry i was so curt. my annoyance with the RFC may have hit you there <nckx>You mean y'all don't read Internet Mail RFCs for fun? <nckx>The fun part is everything must remain backwards-compatible for basically ever. <nckx>unmatched-paren: Has your display name teased out any real bugs in mail software AFAYK? <a12l>Reading § 16.5.4 Formatting Code in the ref. manual. I'm a bit confused, where I'm supposed to find `./etc/indent-code.el`? <a12l>unmatched-paren: I thought that it would be in the repo, but I couldn't find any file with that name when I ran `fd indent-code`. <a12l>Thanks! Got it working :D <a12l>It there a simple way to pin channels? E.g. if HEAD@guix is broken, but I've my own channel that I want to use the latest version. <mirai>If a service has a 'account-service-type' extension that adds a group/user that's already defined in 'operating-system', does it overwrite the existing attributes of it? <unmatched-paren>and then replace %default-guix-channel with (channel (inherit %default-guix-channel) (commit ...)) <civodul>mbakke: thanks for the lld-wrapper fix in cd9161372aa3ec9b8902385d5937d04e1868495c <civodul>that fixes a tests/packages.scm failure <a12l>unmatched-paren: What's the pinning part of that expression? <a12l>Ah, missed your message after the link <a12l>unmatched-paren: Do I understand correctly that your suggested method is to enable setting a specific commit for only the guix defaulit repo, and not any other repo specified in `channels.scm` file? <a12l>I was thinking that I can't specify the other channels commits in `(channel (inherit %default-channels) (commit [sha256sum])))` <unmatched-paren>a12l: you can do e.g. (channel (name 'guixrus) ... (commit ...) ...) <a12l>And if I want to specify the commit for any other channel I need to do that in their own channel expression <a12l>But I think you confirmed my hyphothesis <lilyp>having an explicit channels.scm is best for reproducibility, but also required in this case i believe <lechner>unmatched-paren / Hi, thanks for your review. Why should any Golang "libraries", all of which are strictly sources needed to compile a static binary, be regular 'inputs' instead of 'native-inputs' please? https://issues.guix.gnu.org/59399#10 <unmatched-paren>lechner: it's possible for them to include references to other packages <unmatched-paren>(you can put references in a source, but it's definitely not a thing you should do) <lilyp>lechner: If you're compiling an application, it's all inputs, otherwise propagated-inputs <lilyp>native-inputs is for the compiler and toolchain <lechner>lilyp / unmatched-paren / i agree with the cross-arch argument and will make the change, but doesn't that result in a waste of runtime storage space? the source "libraries" are never used at runtime. <lilyp>guix is better at picking up what's used than you are. it tracks actual store references <unmatched-paren>and even though the source libraries aren't referenced, the source libraries' dependencies might be <lechner>unmatched-paren / can you give an example, please? <lechner>lilyp / so the source "libraries" won't be installed on a runtime system? <unmatched-paren>lechner: that substitute* means that any binary using go-notmuch will be linked to notmuch <unmatched-paren>if go-notmuch was a native-input, the notmuch used would be the host notmuch, not the target notmuch <lechner>it's just that the regular inputs in guix seem to conflate runtime prerequisites and target build prerequisites, and i am slow to adapt <unmatched-paren>it's just that things that aren't referenced won't be downloaded to the store either, i think... <jpoiret>unmatched-paren: propagated inputs are unconditionally installed along a package when building a profile with it inside, and for all other inputs the only thing that matters is whether the dependency is referred to by the store output <jpoiret>if some store output holds a reference to something else, then it will also be substituted, otherwise, you don't need it so it won't <jpoiret>that applies to native and non-native inputs i believe <jpoiret>although it's probably forbidden for something to have a reference to a native-input <lechner>jpoiret / so regular inputs may also be available as a runtime prerequisite but are not visible to the user because they were not installed in the profile, right? <jpoiret>propagated inputs are there for things that can't really be hard-coded into the packages <lechner>jpoiret / or something the user may also want to call directly? <jpoiret>yes, of course. Although if the user wants to call that directly, it would be better to explicitely install it <unmatched-paren>lechner: yup, they'll be added to the store, so any store paths pointing to them will be valid, but they won't be actually installed <lechner>jpoiret / that's something i have noticed in Guix that is unusual <jpoiret>that's also why many guix packages patch their sources to be able to directly reference a binary depedency instead of using one from the PATH: the latter would need to use propagated-inputs whereas for the former the reference scanner can kick in <lechner>jpoiret / some runtime prerequisites are not well tracked <jpoiret>it's harder to adapt to when coming from another distro, but it's *cleaner* on Guix because you have only what you asked for <jpoiret>i've never had any profile collisions ever, but i must be one of the lucky few :p <lechner>jpoiret / my emacs, for example, seems to require sqlite, but maybe that's because of some Elisp package I have installed. <unmatched-paren>so we avoid them by tracking down things like ``exec("bloop")'' and replacing them with ``exec("/gnu/store/...-bloop-tools-.../bin/bloop") <jpoiret>oh, sure. Did you install that package through Guix? <jpoiret>if so, it should be able to use sqlite itself without any issues <unmatched-paren>if the package requires the sqlite command it should be patched like i just said <lechner>jpoiret / maybe i misspoke. i am new to Guix and still adapting. <jpoiret>Guix packages are not impervious to errors and omissions though :) if one doesn't work properly, feel free to open a bug report <lechner>jpoiret / thank your for the explanations! <lechner>jpoiret / should fcitx-qt5 pull in fcitx for the user, for example? For some reason, I also had to install libcdio separately <lechner>jpoiret / unfortunately, i can't remember. maybe nothing <jpoiret>as for the first example, it doesn't look too odd to me. It's hard to judge what should be available from the get-go, but in general guix prefers to not pull in things by default <jpoiret>one other good example would be wireplumber not pulling in pipewire <lechner>jpoiret / or maybe xorriso, dvdbackup or dwd-rw-tools <jpoiret>then i'd say it's probably an omission in the offending package that should be fixed <lechner>jpoiret / would it be a propagated-input? <jpoiret>depends, if it's used as a python library then it would have to be <jpoiret>but if it's used through an executable, it would be fine as a simple input <ahriman>Hello Guix, I am trying to package something that builds with meson>=0.61.0 but meson-build-system is using 0.60.3, guix search shows that there is version 0.63.2, how do I get meson-build-system to use this version? <unmatched-paren>we can't change the default meson right now because it'd rebuild every single meson package <dirtcastle>(offtopic thanks msg) I have been trying to set up home server to selfhost like god RMS intended. but I didn't know how to expose the computer to the internet for SOOO many months now. I was told I need a public ip for which u have to pay. I came across pagekite recently and ever since I found that I was dying to rent a vps and selfhost it. Learning abt pagekite took me soo many months. And guess what u can't even find stuff like this in <dirtcastle>networking books. And now I learnt that ddns exist. None of my friends who were involved with networking never mentioned it. I don't know if they even knew abt it. And services for pagekite and ddns are already available in guix. Now I'm convinced I'm going to look through what every guix service does to see if there is any technology that I'm missing out. Thank you so much. I'm going to discover and learn abt so many new technologies from <dirtcastle>guix is already being what I expected and intend it to become - sysadmin/selfhoster's best friend. <nckx>What they probably meant by ‘public’ was ‘static’. Your current IP is public (although your IP might block ports and charge you money to unblock them), it just moves around a lot. That (and other things) makes it unsuitable for some services, like mail, but Good Enough for plenty of others. <singpolyma>For sending mail. Receiving mail requires some of the lowest reliability of any service due to long SMTP retries <singpolyma>I had a mailserver go out for three days once. Lost zero messages <ahriman>raider is building :D! I will send patch to mailing list once i re-learn git for the millionth time this month. <dirtcastle>o/ nckx. ive tried to learn abt email server in the past. I've seen ppl saying that convincing isp to unblock mail port isn't easy. <nckx>dirtcastle: That is, on average, true. It depends entirely on the ISP. Many will not unblock port 25 on consumer contracts for any reason. <nckx>It is the official port for hacking. <cbaines>guix archive --authorize doesn't seem to be writing to /etc/guix/acl (on a foreign distro) <cbaines>nckx, I think I remember you long ago looking at getting bordeaux.guix.gnu.org in to the install script in terms of putting the key in to the acl, do you remember? <apteryx>cbaines: perhaps 'strace -f -e file' can help <lechner>Hi, I do not login via Gdm but instead issue 'herd stop xorg' on a virtual terminal and then start X manually. Which packages should I install explicitly if I drop the gdm-service from my config, please? <vagrantc>or selectively pick which parts of %desktop-services you want directly <lechner>unmatched-paren / will i still have X? <vagrantc>been a while since i've gone that route, though <lechner>unmatched-paren / you do not use %desktop-services? <lechner>unmatched-paren / also, what do you get from greetd? <unmatched-paren>lechner: greetd gives you a login screen, but it's much simpler than gdm <unmatched-paren>greetd basically provides a daemon that manages the hard parts of login managers <lechner>i always wanted it, but then it seemed so minimal <unmatched-paren>and then you can write a program to display the login screen without doing any tricky PAM stuff or whatever <lechner>can it mount, or is that a seatd domain? <lechner>my home folder is not on my desk equipment (and it is encrypted) <unmatched-paren>i'm not sure whether greetd implements that. you'd have to ask #kennylevinsen <lechner>yeah, thanks. i know where to find him <lechner>the mingetty in your config is an escape hatch? <unmatched-paren>because the i18n_embed package it uses for localisation depends on it :) <lechner>unmatched-paren / do i have to disable any kind of login-service if i borrow that part of your config but otherwise stick with %desktop-services? <unmatched-paren>lechner: i think gdm is the only part you need to remove (along with elogind, if you plan to use seatd too) <lechner>unmatched-paren / thanks! i don't use networkmanager except for its default config to pull address assignments for eth0. maybe i can replace it with lower level network tooling in Guix <lechner>unmatched-paren / how come you use upower with seatd, please? <lechner>unmatched-paren / i'll have a look at %desktop-services later, but is there anything else that i should salvage from it when trying your setup? <antipode>I am currently on a kind of indefinite hiatus from Guix, please don't wait on responses by me on any currently open patches & bug reports. <mekeor[m]>antipode: have a nice hiatus - we'll be waiting for you. miss you already! :) <zamfofex>antipode: Whatever your reason is, I hope your break goes well! <antipode>(I'll still be watching, but not participating for now) <mekeor[m]>do i understand correctly that the "mu" package does not come with the guile-bindings? <apteryx>antipode: oh! I wish you well, enjoy the time off. I hope to see you again soon! <abhicherath[m]>is there a way to mark just one package as nonsubstituted? (I want to build php without the substitutes, but I also don't want to build gcc etc) <rekado>abhicherath[m]: guix shell php; Ctrl-D; guix build --no-substitutes php <apteryx>according to the Linux kernel's MAINTAINERS file <zamfofex>Conversely: guix gc -D $(guix build php) <zamfofex>(Though I suppose that doesn’t substitute build dependencies.) <abhicherath[m]>interesting, I don't see that error on x64, so unlikely to be that cert expiry thing I guess. <lechner>unmatched-paren / Hi, do you use guix home? I am worried it may not do so well with XDG folders after I drop %desktop-services. <civodul>mirai: weird; what locale are you using? <civodul>currently 'guix style -f' defaults to locale encoding <civodul>we should probably change that to UTF-8 <zamfofex>I don’t want to sound annoying, I know sometimes people are busy or just don’t have anything to say, but I’m currently feeling a bit anxious that no‐one responded to my email to guix-devel regarding npm. 😬 I suppose maybe it’s just not something people actually have a big interest on, given that npm packages aren’t useful outside of the npm environment itself. <mirai>civodul: 'operating-system' has it set to '(locale "en_US.utf8")' <civodul>zamfofex: i didn't see it but yet but npm is not among my skills :-) <mirai>oh, but from ./pre-inst-env within a guix pure environment "LANG" is unset <leelano64>Oh hello everyone, was wondering if I could get some advice, was going to use my old xmonad config however ghc couldnt find the installed modules, tried different ghc versions, no avail. Also attempted to use my old dwm binary, wasnt recognized, tried to recompile dwm, probably some confusion but can even find the xlib package that is needed. <zamfofex>civodul: I see. That’s fair enough! 😅 My hope is that it’s possible to avoid needing to traverse into all transitive ‘devDependencies’ by employing esbuild in an ad‐hoc way to build packages (instead of needing TypeScript, or whatever the current popular bundler is). That reduces the dependency tree of a package from several thousand packages to just a few hundred (at worst) and a few dozen (at best). <Guest54>Has anyone had trouble with home-shepherd only launching the first service specified in `services-to-start' in the `~/.config/shepherd/init.scm' file? I've been trying to debug this for a week or so, launching the services manually with `herd start' works fine. <lechner>zamfofex / Hi, does it really make sense to repackage all of npm's goodies in Guix? I used to run the lintian.debian.org service just straight off of npm's repo <apteryx>is it OK to use (guix diagnostics) in (gnu packages linux) ? <zamfofex>lechner: I think the big advantage is that npm does not have a mechanism to allow you to build packages from source, and Guix would help change that. <civodul>apteryx: technically yes, but in spirit it's not great <zamfofex>(And when I say “from source”, I mean including dependencies.) <apteryx>why so? is it OK then to use srfi-35 and simply raise a &message <apteryx>I prefer exception than (exxt 1) anyway, at the REPL for example <civodul>nicer as in: less tied to the implementation of diagnostic reporting <civodul>exit is strictly forbidden in (gnu packages ...) :-) <apteryx>this kind of insider's knowledge would definitely have a place in a coding style section :-) <civodul>there's probably too much "insider knowledge" like this <civodul>actually jpoiret did a great job demystifying a number of common issues at Guix10Years :-) <apteryx>it'd be nice to document the special cases, or make them disappear :-) <apteryx>a (guix diagnostics) usable the same on either host or build side would be fantastic <civodul>on the build side (building a derivation), it wouldn't be very useful <civodul>since normally nobody will see those messages <apteryx>we'd have a consistent way to abort the process with a useful message <apteryx>it's not so often perhaps on the build side, but it happens <civodul>downloaders print machine-readable status lines that are interpreted by (guix status) <civodul>which in turn uses (guix diagnostics) <apteryx>grepping for 'format "error:"' doesn't turn that many results, but there are a couple examples, of hand-formatted error messages <Kolev>Is there any way to add Guix support to GNOME Software? <Kolev>zamfofex: Oh, right. They should add per-user software installation. It would be good for Flatpak, too. <abhicherath[m]>zamfofex, where's this patch? It'd be really useful for my typescript-language-server package... I was having devdep fatigue <a12l>If I want to develop on a package in the repo I can create an environment with all the package's dependencies in `$PATH` with `guix shell -D <package>`. But how do I actually get the source code of the available package? <a12l>I checked the ref. manual for `guix shell`, but couldn't find any option.