<minima>ok, i might have to use svn-fetch then, i guess... <the_tubular>Could someone past the last one in chat please ? I think it has to do with a new python build system ? <rekado>don’t you see them with guix pull --news ? <the_tubular>I ran gc so I get : guix pull: error: profile /var/guix/profiles/per-user/root/current-guix does not have a previous generation <cnx>does anyone use ibus with an additional engine? <cnx>i put ibus et ibus-libhangul to home env and ibus-setup doesn't find hangul <lechner>cnx: i have used fcitx v4 with pinyin but it felt kind of old <cnx>i was asking because i wonder if some sort of composition needed to get external engine to work with ibus <cnx>in nix there's a specific config for input methods ***Dynom_ is now known as Guest5966
<tissevert>how is one supposed to run jack2 when the package doesn't contain the executable jackd ? <rekado_>it has jackdbus and it’s supposed to be started automatically <rekado_>all packages that still build with jack1 should be changed to build with jack2 <tissevert>and yet it doesn't show up in the output of ps <tissevert>(I tried jackdbus auto, of course, despite the warning, and can confirm it is not of much help) ***jeffbezos is now known as nckx
<rekado_>tissevert: what jack client program are you using? <rekado_>when jack2’s dbus service is installed it will be launched automatically as needed <minima>anyone sees anything obvious that i might be missing? ***jess-o-lantern is now known as jess
<Korven[m]>Does anyone know if just deleting my `~/.guix-home` folder would be enough to "stop" using Guix Home? <civodul>Korven[m]: hi! it wouldn't be enough <civodul>you would also need to remove /var/guix/profiles/per-user/$USER/guix-home* <civodul>and then you would need to restore your dot files <civodul>like ~/.bashrc, etc. so that they're no longer symlinks <civodul>it would be nice to have a tool to make it easy to leave Guix Home <Korven[m]>I wanna give multiple profiles (from the Cookbook) a try, because they seem more mature at this point. (+ the GC bug with Guix Home) <Korven[m]>If you do a `guix home reconfigure...`, and then a `guix gc` *civodul has been using Home for a while <Korven[m]>And then a `guix home reconfigure` again, it downloads *all* the grafts again <Korven[m]>civodul: I think so, lemme go find it really quick <minima>(i managed to make some progress with my clojure package, a '#:java-source-dirs' was missing) <civodul>Korven[m]: oh i see; not really a "bug", but an annoyance <civodul>(by "gc bug" i expected a data loss issue or something) <Korven[m]>Well it is annoying aye, esp since my download speeds aren't the best <civodul>you'll observe the same with regular profiles though <civodul>you can reproduce it with, say: "guix gc -D $(guix build libreoffice); guix build libreoffice" <nckx>apteryx, efraim: We've used ‘1400 CET’ to sync up in the past (now this strikes me as odd, though I'm not complaining :), but is that the canonical time? <nckx>s/partition/flash chip/, just so you don't go chasing wild geese. <nckx>Somehow, Guix's GRUB/efibootmgr is more prone to this, and I have no clue why. <nckx>Is it just because we reinstall the boot loader a lot, so it's more likely to be noticed? 🤷 <nckx>As a side effect, using ‘grub-efi-removable-bootloader’ might work around this. <qzdlns[m]>ah thanks @nckx, I'm not aware of this distinction -- if it matters, I'm in a previous system generation, but have a healthy mount on /boot/efi -- I had power-loss last night which prevented my from booting on the HEAD generation <nckx>Briefly (hah): UEFI stores preferences, including boot options, in an nvram chip, not on your drive(s). When you (re)install GRUB, it updates that preference to point to the ‘right’ GRUB. In our case that's Guix\grubx64.efi or something very similar. And for unclear reasons that nvram can get ‘full’ (some vendors are more succeptible to this than others) with no automated way to safely fix it. Switching GRUB to ‘removable’ mode insta <nckx>lls it to a different, default, file name, usually used for removable (USB) drivers, which doesn't require such a preference to be stored. <nckx>This issue has nothing to do with Guix, or Linux, or system generations. <nckx>We just call GRUB more often than most distros, causing this to be more noticable—I think. <qzdlns[m]>yo thanks so much for this info -- will pour over the manuals for grub and efibootmgr to get up on these concepts <nckx>Let me know if switching to ‘grub-efi-removable-bootloader’ has any effect. Will help me help the next person, and maybe even document this somewhere. Good luck. <qzdlns[m]>just one last uninformed question -- I can flash this from userland right? if I follow, given that it's some ROM, no unmounting needed of my boot is needed to write <qzdlns[m]>for sure will let you know; this is my work box, so need to get my fix on! <nckx>Yes, it's completely separat from your /boot, in no way connected. Linux mounts a *virtual* file system that *represents* these variables at /sys/firmware/efi/efivars, so you can remove and change these variables from user space. At your own risk—you can break things this way. But it's just a representation; in reality they are just some binary data on a chip. <qzdlns[m]>a success with applying `grub-efi-removable-bootloader` -- the linked install-bootloader.scm looks cogent; I will read up on the semantics of the pointer to the system volume vs nvram, but I don't expect any changes in my UX whilst booting <nckx>Thanks for reporting it! It would be nice to have a real (not manual) fix/work-around, but I can't justify buying UEFI hardware just for this… <nckx>…I wonder if one could trigger this in QEMU with sufficient battering. ***wielaard is now known as mjw
<vivien>Hello guix, While we’re all waiting for the openssl release, I’d like to present you a quick Minetest update, so you can relax a bit with beautiful dynamic shadows! https://issues.guix.gnu.org/58932 <unmatched-paren>2022-09-06 07:10:11 [E][module-portal.c:786 module_init()] Failed to connect to system bus <unmatched-paren>this is a home-pipewire-service-type, and it ``requirement''s on '(dbus)... <nckx>vivien: We should have a ⟳ hit refresh party… <unmatched-paren>vivien: some feedback: couldn't you use modify-inputs for irrlicht-for-minetest's new input? <unmatched-paren>(inputs (modify-inputs (package-inputs irrlicht) (prepend libxi))), i think <vivien>I didn’t know about that one to be honest <vivien>that seems like a very useful tool <vivien>Is there something similar for arguments, like modify-configure-flags and so on? <unmatched-paren>(substitute-keyword-arguments (package-arguments PKG) ((#:KEY SYMBOL) ...)) <vivien>I know about substitute-keyword-arguments, I wish it were by default in guile <vivien>What’s the point of keyword arguments if you can’t do that <unmatched-paren>vivien: btw, it's easier for reviewers if you use git send-email instead of using format-patch with attachments <pkill9>what's the proper way to change default.pa for pulseaudio on guix? <pkill9>by default it disables user configuration <pkill9>so i got to do it at the system level, tho I suppose I could test it by disabling those environment variables <vivien>Git send-email is unfortunately unusable as it can’t use an external mail user agent <lilyp>pkill9: if you want no system-level configuration, you can simply delete the service; otherwise you could try adapting it for guix home <pkill9>i kinda want it to be system-level, but only after testing it temporarily <pkill9>so 'commit' it to system-level after testing it on user-level <pkill9>lilyp: you mean delete it at system-level and then adapt it for guix home? <pkill9>the reason I ask about system level is if you can configure defualt.pa with service, or just give it a file <pkill9>but looking at /etc/pulse/client.conf, it hardcodes a path to default.pa <pkill9>which apparently can cause stuttering on bluetooth headphones, which I'm experiencing <unmatched-paren>lilyp: "Teams" and "Sending a Patch Series" currently look like subsections of "Configuring Git", which is really jarring, since they aren't really to do with "configuring" git, more using it <unmatched-paren>this also means the only subsection listed in "Submitting Patches"'s index is "Configuring Git", which is a little jarring and confusing <klm_>`CMake 3.22 or higher is required. You are running version 3.21.4` <-- how can I specify `cmake` instead of `cmake-minimal` in my package? thanks in advance <lechner>Hi, what is everyone's favorite size for a Guix root partition (including /usr and /var) for a 64-bit system nowadays, please? <klm_>unmatched-paren: that worked, thanks! for reference: … (arguments (list #:cmake cmake)) ***civodul` is now known as civodul
<raghavgururajan>Anyone have any suggestions on udev rules in guix system for connecting mobile devices? <lechner>Hi, I have a local checkout of mumi and am in a development shell. I also ran bootstrap, configure and make. How do I run mumi locally, please? <csepp>i'm working on some Guix related script in Emacs and running into a bug where even though I run it with pre-inst-env, GUILE_LOAD_PATH is somehow not set correctly by geiser. could this be caused by Guix's dir local variables? <csepp>i'm also not running emacs as a client, so it's not a case of the server running with different env vars. <csepp>also asked this in #guile but forwarding here for better visibility <csepp>oh, nevermind, i realized the path is correct. <csepp>but now i wonder why it worked previously... ***LispyLights is now known as Aurora_v_kosmose
<apteryx>mbakke: the new pyproject build system is a pleasure to use, thank you! <cnx>openssl 3.0.7 is out <nckx>(But, for some odd reason, mostly only CL users.) ***justache is now known as justPardoned
<nckx>cnx: Do you actually use it? Just curious. <cnx>i know i do on other machines but not sure on guix <cnx>does guix swap it out for gnutls? <nckx>Guix uses openssl for plenty of packages, just not 3.0. <nckx>Should have included the command: guix refresh -l openssl@3 ☝ <cnx>wait so not even like firefox uses openssl 3? <nckx>Has Firefox ever used openssl or gnutls? I don't think so. <nckx>The openssl actually used by $everything in Guix is 1.1.1q. <nckx>(Which I didn't update yet.) <nckx>I should be super clear about that not being a security update. Only OpenSSL 3.0 was the big noise of the week. <nckx>Or at least I don't think it is. That's what I was finding out :) ***sajith_ is now known as sajith
<nckx>Please read the git log first to get a general idea, then you can ask any specific questions/doubts you still have. <florhizome[m]>I am specifically confused since to me it appears as git adds a changelog itself <nckx>Maybe there's a script I don't use (all mine are 100% artisanal). <nckx>The (Guix) manual links to the (GNU) manual chapter above. <nckx>I've updated openssl@1.1 anyway, just in case, we'd get no end of flak after this widely-publicised debacle if it was a security fix 🤷 <nckx>florhizome[m]: I meant that about the ‘specific questions’ by the way, it's just not efficient to explain every single thing here. And most of it is most easily induced from the git log. <vivien>I would not be surprised if there were a gnulib module to do that <vivien>(disregard this, I read it wrong) <apteryx>florhizome[m]: you just invoke the script, with the changes uncommitted in your tree. if it's simple, it'll make a good basic commit message. if it's not, it'll crash <unfroq>Hey Guix! I am trying to build some software and it hast rust as a dependency. Rust quits the build with some errors regarding Udp "Address family not supported by protocol". I am pretty sure this is ipv6 related, since I disabled it via sysctl. Does anyone have an idea how to get around this? <unfroq>I like having ipv6 disabled, but I had issues with building another software as well. Is there a better way to disallow ipv6, but not breaking certain software builds? ***mark_ is now known as mjw
<apteryx>florhizome[m]: not even staged... it commits for everything it finds <vivien>unfroq, while disabling ipv6 will very likely cause a lot of problems for you in the future, the most likely problem is that rust tries to download stuff from the internet in the build container. The build container is not supposed to access the internet, so that the build is reproducible. You have to package rust things with the cargo build system to do that. If nobody can help you here, you should consider sending an email to the <jab>I wonder how many people here use the scheme shell <vivien>jab: “I'll only ask another 100 times.” <jab>vivien: as in you are encouraging more people to try to use scheme shell? Or that is a commonly asked question? <vivien>I just tried to run it, and then I wanted to quit it, and it said this to me. <unfroq>vivien: yeah, at some point I might run into more problems. Until now I just had issues with some Go based builds, because some dns resolver does not do anything, if ipv6 is not present. And the other thing is this rust dependency, which is quite annoying atm. Besides that I am quite happy to not have ipv6. But I did not think about the possibility, that the container tries to connect to the internet.. I will watch out for that. <unfroq>vivien: Do you think enabling ipv6 and blocking it via nftables would be smarter/cause less trouble than do disable it completely? <jab>I actually found a blog post about various sysctl hardening options for linux that I just applied. Kind of cool. <Obikawa>Hi. I am playing around with a (package) definition in order to gain experience with Scheme and Guix. While breaking up the code into separate (define) expressions, I have encountered a behavior which I do not understand. I am not able to break out any piece of code out of (arguments '(#:phases (...))). No matter what I do, the guix builder always <Obikawa>finds something to complain about. Are the (package (arguments ...)) always unbreakable, or is it a case of me not knowing enough of Scheme/Guile/Guix? ***devmsv_ is now known as devmsv
<jackhill>(may not be the best example, but I'm at least familiar with that code). Anyway, if you can ask some follow up questions it could help me understand, thanks! <Obikawa>jackhill: I mean pulling the code out of the body of (arguments ...). Like here - (home-environment (services (list ...))) -> (define list-of-services (list ...)) (home-environment (services list-of-services)) <jackhill>Obikawa: sure, it's all just a scheme data structure (i.e. a list) <Obikawa>jackhill: My issue is that I am not able to pull out anything from (package (arguments [here])). I don't know why and want to learn from it. <Obikawa>jackhill: Is it [sexp], [gexp], or something? I am not that advanced yet, but knowing the name would help in the future. <Obikawa>So far, I have understood how records and macros like to stay in a single body. <jackhill>I beleive that arguments is a list is a list where the odd positions are keywords (like #:make-flags, #:phases, etc) and the even positions are values for those arguments. There some helper funcitons, like substittue-keyword-arguments, wich is defined, uh, I have to look, to help manipulate that list. <jackhill>ah, in guix/utils.scm. The docstrings in there too should provide a little bit of help. <Obikawa>jackhill: Uh. Fine. I think I will have to satisfy myself with "It's magic - keep it single" answer. Thank you. <elevenkb>hello there, suppose one wants to build a guix image to test in a _vm_ whether a certain ``operating-system`` configuration works on one's hardware? how should one do this? <jackhill>Obikawa: perhaps for now. I'm sure one day you'll wake up and be able to see behind the curtain (or have the pleaser of talking with someone who is better at explaining/showing :)) <Obikawa>jackhill: Exactly. I am now looking at the nginx example and already can see a meaningful difference from my own code. I will play with it a bit more. <jacobk>It says "This project aims to automate software vulnerability scanning of packaged software to protect users against possibly dangerous code." but then there's not really much more information that isn't just about Guix in general. <lispmacs[work]>hi, I was trying to see if I could package this go program, even though I don't know anything about go: <lispmacs[work]>I see there is an importer for go, but I don't know how to see if kineto can be imported <Andronikos>If Linode does support Guix why is it necessary to do all those steps to get it running and not like others distro that can just be changed? <sneek>Welcome back Andronikos, you have 3 messages! <sneek>Andronikos, lechner says: There is also an off topic channel, I think, but people have tolerated much more banter here. Everything is cool & Cali is great <sneek>Andronikos, lechner says: The Guix devs read the back logs later, and our conversation makes it harder for them to catch up with real bugs <sneek>Andronikos, nckhexen says: That said: flaming: absolutely not. If someone flames you for talking about the weather, they are out of line, not you. <[>nckx: How did you make the commit upgrading OpenSSL 2 days before the patch was released (Oct 30), but after newer commits (the commit before is from Nov 1)? The date seems to be wrong. <jacobk>Are nckx and nckhexen the same person? <podiki[m]>as for the time question, you can have whatever time on the git log I believe, just really cares for parent commit for ordering <civodul>jacobk: first time i see this NlNet project; is it a proposal or a newly-funded project? <jacobk>This says calls ended October 2021. I'm curious what the funding actually led to though. <minima>the error seems to be "namespace 'compiler' not found after loading '/compiler'" <minima>which looks pretty obscure to me... <civodul>jacobk: if nobody here heard about it before, that may be because it didn't go anywhere <jacobk>civodul: Maybe, or maybe they just didn't interact with Guix officially for some reason. <jacobk>Or maybe the people working on this project were already part of Guix and didn't need to communicate in terms of the sub-project name. <jacobk>Does Guix do any sort of automated vulnerability scanning currently? <civodul>jacobk: there's "guix lint -c cve" but that's about it <jacobk>Hm, and all that would have been before the NLnet funding. <jacobk>Knowing about `guix lint -c cve` and `guix health` might turn out to be helpful though. <jacobk>Ah, I guess lint only looks at metadata and not the source code itself. <xtremeMonkey>my associate membership with the FSF expired, and i may commit to being one again if i manage to donate on a monthly basis, as a student <civodul>jacobk: right, there's no tooling looking at the source <pkill9>gnome decided to make 'Console' it's new terminal apparently <lilyp>gnome-terminal still exists and the two have non-overlapping feature sets <lilyp>or well, only partially overlapping, but also partially distinct ones