<slimjim>seeing some weird behavior with 'guix pack -R', can someone tell me if I'm doing something wrong? <slimjim>two pack commands give different behavior <slimjim>first one: guix pack -R python2 -S/bin=bin -S/etc=etc <slimjim>second one guix pack -R python2 bash -S/bin=bin -S/etc=etc <slimjim>in the one that only contains python2, if I untar the resulting back in a directory, I can run ./bin/python2 fine <slimjim>but in the one that's also packed with bash, if I untar and ./bin/python2, it crashes with an exception <slimjim>but if I cd into bin and ./python2, it works fine <slimjim>so, something about including 'bash' in the pack has an effect on relative file paths or something *nckx impatiently refreshes. ***ChanServ sets mode: -o nckx
<nckx>raghavgururajan: /foo:/bar <nckx>Sure, but no need for " ": " ", just : will do. <nobodynogroup>hey, looking through logs and searches, it is a little unclear to me if there is a way to install guix sd on a raspberry pi 4. the downloads section says that Guix System is availible on AArch64 but i dont see a download for an image. I saw several questions about proprietary blobs being required to book pi's and potential workarounds using uefi firmware, has anyone got this working? <roptat>you'll need another (assoc-ref %outputs "out") <roptat>or you'll hard-code /lib instead of $out/lib <raghavgururajan>nckx: (assoc-ref %outputs "out") "/lib/pipewire-0.3") ":" (assoc-ref %outputs "out") "/lib/pipewire-0.3") <nckx>raghavgururajan: Almost: (string-append (assoc-ref %outputs "out") "/lib/pipewire-0.3:" (assoc-ref %outputs "out") "/lib/pipewire-0.3") <roptat>nobodynogroup, I don't think we can boot guix system on the raspberry pi (any version) because of the propriatory drivers needed to run it <raghavgururajan>(string-append (assoc-ref %outputs "out") "/lib/pipewire-0.3:" (assoc-ref %outputs "out") "/lib/pipewire-0.3") <roptat>raghavgururajan, no need for the two string-append, only one is needed <nckx>(string-append "-Dc_link_args=-Wl,-rpath=" (assoc-ref %outputs "out") "/lib/pipewire-0.3:" (assoc-ref %outputs "out") "/lib/pipewire-0.3") <nobodynogroup>roptat: thanks, i will look into it. i guess I will have to check the next arm board I buy for compatibility <nckx>blendergeek: Thanks! Typo in gnu/local.mk (nextxcloud-fix-filenames.patch). I'm a bit confused by the rest of the patch: e.g. qsyncthingtray is already in Guix in (gnu packages sync). While we sometimes include package-specific variants, yours looks identical. <nobodynogroup>i didnt know that the pi needed proprietary blobs when i bought it, i should know better by now than to trust any hardware without looking it up <nckx>nobodynogroup: There's no bootable aarch64 installation image, but you can boot another distribution, install Guix on it, then ‘guix system reconfigure’ to turn it into Guix System on the next reboot. ***victor is now known as Guest69238
<Guest69238>Hi, I was here a few days ago and I said that I needed to run a VMWare image in GNOME Boxes. <nckx>blendergeek: I've sent some random comments to keep you busy. Don't forget to try your hand at a GNU-style commit message. <Guest69238>Someone sent me a link that showed how to convert VMWare images that are compatible with GNOME Boxes. <Guest69238>I was wondering if that person could repost that link again. <nckx>Guest69238: I was going to boast about the newly-working search box at http://logs.guix.gnu.org/ but it still seems to be borked. Still helpful if you remember roughly when it was. o/ <nckx>Just don't trust the search results, they are stale af. <raghavgururajan>I did `scp ~/webkitgtk.diff bayfront.guix.gnu.org/home/raghavgururajan/secondary` <raghavgururajan>I get "cp: cannot create regular file 'bayfront.guix.gnu.org/home/raghavgururajan/secondary': No such file or directory". <blendergeek>nckx: Do I just reply with a new version of the patch to fix the typos and weirdness? <bdju>anyone else having mpv + youtube-dl issues? <bdju>I get ffmepg input/output spam trying to open a youtube video and then it fails in the end with "error in the pull function" <stallion>Evening all - I've been poking about at Guix manifests and I can't seem to find an answer on this. Is it possible to add shell hooks to a manifest such that one could source a Python virtualenv or similar? I'm thinking of something similar to nix-shell if possible. <branjam>I've usually been able to find what I need to problem solve from the logs/manual/ML, but maybe I'm just off today with dbus-system-services. <branjam>Not sure if that exactly gets you there but maybe it can inspire a solution ***familia__ is now known as familia
<ryanprior>I hate how `guix build -f <some package I've been working on>` can end up building the world if I've run guix pull since I last worked on the package <ryanprior>like damn I didn't ask you to update my profile, why are all these things unrelated to my package (like Inkscape) building now <ryanprior>it really punishes you for running git pull regularly, maybe I should keep a separate profile for my packaging work that I just keep out of date <bavier1>ryanprior: you can `guix pull -p <hacking profile>` and use that to work on your project, while pulling to your default profile as usual <bavier1>i.e. `<hacking profile>/bin/guix build -f <some package>` <ryanprior>Yep that sounds like a smart thing to do, it might be a little less surprising :) <PotentialUser-74>hello! how do I get the full path of an installed package in the /gnu/store from within a package? I added it as input and then used (assoc-ref inputs "openssl") and it's telling me: unbound variable inputs - I don't get it! I've seen many packages do this and it works for them! <Guest69238>Hey guys, I get this error when I try to install the guix package manager on popos <Guest69238>running the command that the output suggests doesn't resolve the issue. <nckx>Guest69238: Run gpg as the same user that's running the script (you're probably running it with sudo; try sudo gpg). <nckx>PotentialUser-74: Are you using (lambda* (#:key inputs #:allow-other-keys))? If you're using just (lambda _ …), or if inputs is missing from the #:key list, it won't get passed in. <nckx>If you're referring to inputs outside of a phase (e.g. #:configure-flags) you have to use the slightly uglier %build-inputs. <nckx>Guest69238: And with sudo -i? <nckx>The previous message was for PotentialUser-74 🙂 <nckx>Guest69238: Sorry, I have to ask: you're using ‘| sudo -i gpg’, not ‘sudo wget’? <nckx>PotentialUser-74: Then you need to use (assoc-ref %build-inputs "foo"). <nckx>Your secret's safe with me! Good luck. <PotentialUser-74>nckx: `(#:make-flags ,@(string-append "WITH_OPENSSL=" (assoc-ref %build-inputs "openssl") " WITH_LUAJIT=" (assoc-ref %build-inputs "luajit")) <PotentialUser-74>also the software doesnt have an install script, and I am confused on how to install by hand, there's only an executable binary to copy <nckx>PotentialUser-74: Drop the ,@, or you're trying to evaluate %build-inputs on the ‘host’ (not ‘build’) side, where it is indeed not defined yet. `(#:make-flags (list (string-append "WITH_FOO=" (assoc-ref %build-inputs "foo")) (string-append "WITH_BAR=" …)) …) <luhux>I want to use Guix to cross compile the raspberrypi kernel, what should I do? <nckx>PotentialUser-74: You can add a phase that simply calls (install-file …). Look for examples in gnu/packages. Now I must go. ***apteryx is now known as Guest68538
***apteryx_ is now known as apteryx
<luhux>Or to put it another way, how should I ensure that the package must be compiled by the cross toolchain? <efraim>raghavgururajan: you missed the ':', should be 'scp ~/webkitgtk.diff bayfront.guix.gnu.org:/home/raghavgururajan/secondary' <luhux>I have used this, but I want to do this in xxx.scm to force it. <NieDzejkob>you could try limiting the supported systems of the package <Kimapr>I have segfaults in cargo build scripts <Kimapr[m]>here is a backtrace (10 lines long, should i paste it on a pastebin or here would be okay?) <Kimapr[m]>it is when i was building my crate with libloading as dependency <Kimapr[m]>i also had segfaults in a few dependencies (build scripts too) of hypermine ***ss2`` is now known as ss2`
<nckx>klys: It's certainly the most reliable source. The best person to ask would be vagrantc, who frequents #guix. <sneek>I think I remember vagrantc in #guix one day and 11 hours ago, saying: athune: sometimes takes a while for the response to come through. <nckx>blendergeek: Don't send an updated patch yet, I'll build it first. <nckx>‘error: corrupt patch at line 37’ — how did you generate/mail the patch? ***sputny1 is now known as sputny
<bdju>lotta icecat tab crashes lately! <bdju>anyone had issues with their screenshot bind not working on sway recently? I've got grim taking geometry from slurp bound to super-x, but it doesn't do anything when I press the keys recently. I copied the command out of my config and ran it in a shell and it works still. <rah>I'm in the process of installing Guix System on an old x86 box <rah>unfortunately the graphical installer doesn't display properly on a VGA console <rah>a lot of the text and critically buttons/controls are off the screen <bavier1>rah: the installer currently assumes a certain minimum screen resolution; I don't know if that is being looked into or not <nckx>bavier1: Interesting, I wasn't ever aware of that. I don't see it mentioned in any bug report subjects. <guixy>There's a log that dumps over tty1. What is that log? <guixy>It looks like the kernel log. <ss2>is it possible to use hibernation in Guix? <ss2>Haven't found much regarding that. <nckx>guixy: It's Linux's notion of the ‘console’, the default TTY (kind of). It's mostly the kernel log but also other things. And while none of it is *guaranteed* to be saved anywhere (fun!) most of it is, nowadays, in /var/log/messages. <ss2>oh, so it is still been worked on? <ss2>I remember reading a bug report somewhere. <guixy>nchx: is there a way to remove it? <bavier1>nckx: thanks for checking. I have a few details, and maybe even screenshots, so I'll try to file a bug report soon. <nckx>I basically have a completed patch but it's stalled due to unrelated kernel bugs making it impossible to test all the use cases I really want to test. <Kimapr[m]>hmm, by the way, every time i switch tty from X to a terminal i get a "nouveau: 0000:09:00.0: bus: MMIO write of XXXXXXXX FAULT at 6013d4 [ IBUS ]" <Kimapr[m]>and it shows up in whatever terminal i switch to <ss2>nckx: is there any discussion to be found online? I've got bug#25031. <guixy>Kimapr: I think it's the same issue I'm trying to resolve <Kimapr[m]>i don't think so. Do you want to remove tty1 logs? <guixy>I at least want tty1 usable. <guixy>without being interrupted by logs <nckx>That's a different question. dmesg -D disables printing of kernel messages to console. <nckx>I feel pretty bad about the age of that thread. Debugging kernels takes time (at least for me) and I haven't had any lately. <guixy>the shepherd graph says there's a service called term-auto. What is that? I cannot find that string anywhere. <nckx>(I'm typing this from a Windows XP machine at work, in case you're wondering why I seem to have some 🙂) <Kimapr[m]>hmmm. i see the issue i have now. *any* code i compile with cargo (or maybe just rustc too) segfaults <ss2>nckx: not sure how to help out with that issue though. I'm still at my basics here. <ss2>Anyone willing to tell me one gets the multimedia keys working? <nckx>ss2: The code is (IMHO) finished, I just want to test it in a few failure modes before we unleash it on users. I don't think it will work with the default Guix kernel. So I appreciate the offer to help but can't in good conscience recommend it to a newer user (welcome!). <nckx>Currently the only (mandatory) failure mode is ‘my laptop reliably hard-resets when I close the lid’ :-/ <nckx>ss2: That's often very desktop-specific. I can help you if you use i3. <nckx>s/desktop/desktop and hardware/ <ss2>but true, with i3 it should be already, but haven't tested that yet. <ss2>Anyways, how do you get crisp font rendering? In xfce it's okay, but isn't yet in i3. <nckx>ss2: AFAIK there isn't a big ‘make fonts gud’ switch on Linux. I have my own .config/fontconfig/fonts.conf that I've carried with me through the years. <nckx>XFCE does most of it for you but i3 does not. <ss2>yeah, I know. This is an ever recurring theme of me to hunt down. I'll work through that article. <ss2>But now I'm just wondering where the Xorg binary is. I installed xinit, and may disable gdm at some point. <Guest69238>I installed the guix package manager onto PoP! OS and install GNU IceCat but I don't see the icon anywhere. How do I run it? <NieDzejkob>Kimapr[m]: is that some weird architecture or setup? the rust and rust:cargo packages in Guix work fine for me. <NieDzejkob>Guest69238: I'm not sure how desktop integration works on foreign architectures. You could try running 'icecat' in the terminal for now. <Guest69238>just type "icecat" in the terminal. I did that but it threw an error. Should I use "guix icecat"? <vagrantc>Guest69238: you may need to add your guix profile to PATH <Kimapr[m]>no. there is no such thing as /guix/store (it is /gnu/store). /gnu/store is where everything is stored. you usually have a link to your guix profile (which is stored in /gnu/store) in your user dir <vagrantc>Guest69238: starting at the paragraph "For each user, a symlink to the user’s default profile is automatically created in $HOME/.guix-profile" ... <ss2>wow, rebuilding profiles can take a loong time.. <rah>does guix compile packages then? <rah>this installation is literally taking all day <rah>I don't know what substitutes are <Kimapr[m]>substitutes are pre-compiled store items so you don't have to build them <Kimapr[m]>just tried building a hello world package in rust with `guix environment --pure --ad-hoc rust:cargo rust:out coreutils` <rah>why would there be packages with no substitutes on the server? <Kimapr[m]>looks like there is something fundamentally broken in my setup <Kimapr[m]>rah: because the substitute server doesn't have infinite time nor infinite power <roptat>rah, when we update a package, it's immediately available to our users, and our build server, at the same time. If you update to that package before the build server had the time to build it, you'll have to build it yourself <roptat>it could be also that you didn't allow substitutes, so your machine is building everything <Kimapr[m]>now cargo doesn't accept certificates in container <Kimapr[m]>> guix environment --pure --ad-hoc --container --network -P rust:cargo rust:out nss-certs coreutils <apteryx>would someone know how to get rid of: "error: all build users are currently in use; consider creating additional users and adding them to the `guixbuild' group". Can't Guix queue the builds instead of failing? <rah>how do I find the xorg log? <apteryx>NieDzejkob: in my machines.scm file I specify (parallel-builds 2), but otherwise I don't give the --max-jobs argument explicitly <rah>rah@teasel ~$ ls /var/log <rah>debug gdm/ guix-daemon.log lastlog maillog mcron.log messages nscd.log secure wtmp <rah>rah@teasel ~$ sudo ls /var/log/gdm/ <rah>alas, not on mine :-/ <NieDzejkob>> Please also check the log file at "/var/lib/gdm/.local/share/xorg/Xorg.0.log" <NieDzejkob>Kimapr[m]: not sure what could be happening that makes rust not work in your --pure environment <nckx>apteryx: (guix-configuration (build-accounts 20) …) <NieDzejkob>do you have any special files in /lib or something maybe? <NieDzejkob>if you want to debug the container workaround: is $SSL_CERT_FILE or $SSL_CERT_DIR set? <apteryx>nckx: perhaps it's caused by grafts. I have 11 grafts to be made, and I think the default number of build accounts is 10. <rah>NieDzejkob: what are you quoting from? <rah>[ 194.423] (EE) Failed to load module "mga" (module does not exist, 0) <NieDzejkob>rah: I'm quoting from my log file in /var/lib/gdm <nckx>apteryx: FTR, I'm not sure which (edge?) cases trigger that error either. Just ‘locks or build slots’ don't always cut it, that's for sure. <nckx>Thanks for the grafts hint. <apteryx>yeah, even specifying guix build --max-jobs=1 doesn't help <apteryx>probably a bug w.r.t. grafts not being queued as different builds or something similar <rah>do I have to run "guix system reconfigure" as root? <apteryx>I've renamed my /etc/guix/machines.scm file to confirm it's not a bug with offloading but in general <nckx>rah: sudo guix system reconfigure …. But not guix pull! <ss2>how do I actually disable gdm from starting? I've disable xorg in use-service-modules and xfce-desktop-type in services. <ss2>but it still auto loads. <rah>nckx: why not guix pull? <rah>nckx: also, the manual said that to run guix as root using sudo, you have to use "sudo -i guix ...", is that not right? <NieDzejkob>rah: if you want to run root's guix, you need sudo -i, but you generally don't want to run root's guix <NieDzejkob>with just sudo, it will use the pull you did as your normal user <nckx>Because you need to update your user's Guix. You don't need to be root to do that (and being root would either screw up your own permissions or update root's guix, in any case bad things). <NieDzejkob>ss2: I think gdm is included in %desktop-services <nckx>rah: The manual is right, but you don't *want* to run root's Guix (unless you do, but then you'd know 😉). <rah>I don't understand at all :-/ <rah>if I'm updating my user's "Guix" then how does that effect running "guix system reconfigure" as root? <nckx>Then just do as we advise for now. ‘guix pull && sudo guix system reconfigure …’. It's the recommended way. <rah>I'd prefer to understand <nckx>Because ‘sudo [-E] guix …’ still runs your user's Guix. <nckx>At least on Guix System. <nckx>Other distributions' sudo may be configured differently, it depends. <rah>I don't understand what you mean by "Guix" <rah>then I don't understand why updating my user's guix effects running "guix system reconfigure" as root <nckx>To see the difference, compare ‘sudo which guix’ to ‘sudo -i which guix’. ‘guix pull’ will update the former, ‘sudo -i guix pull’ the latter. They are two separate guixes. <nckx>Because running ‘sudo guix system reconfigure’ runs your user's guix. <rah>that's doesn't help me understand <rah>you're just repeating yourself <apteryx>rah: the PATH is preserved on Guix System when running sudo. Nuff said. <nckx>Has anyone perchance packaged dracut for Guix? *vagrantc had just assumed guix was guile-all-the-way-down-as-it-should-be <Kimapr[m]>it has a daemon whih is not guile but afaik that's it ***sneek_ is now known as sneek
<nckx>vagrantc: klys asked: hi, I was interested to find a .deb for guix and came across https://salsa.debian.org/vagrant/guix would it be advisable to clone from there and do dpkg-buildpackage and install the result? <vagrantc>klys: in short that should work on debian's testing release (bullseye), but you'll probably need to disable tests as several fail <nckx>Hm, ‘guix gc: freed 140,713.22960 MiBs’ while ext4 /gnu/store has 77G free. <rah>apteryx: that doesn't help me understand either <apteryx>rah: guix pull refreshes the 'guix' program itself, which comes with the package definitions. When using that 'guix' program to run 'guix system reconfigure', it will rebuild your system using the latests package definitions it knows about. <apteryx>using either 'guix' or 'sudo guix' uses the same guix package (the one availabel as your user), but using sudo runs that same guix as the root user, which required to change the state of the system (otherwise any users would be able to wreak havoc) <Kimapr[m]>ooooh adding curl to my environment container ad-hoc fixes the problem and now cargo can fetch stuff yay! <nckx>Sounds like rust needs a native-search-path. <nckx>By installing curl, you're piggy-backing on curl's. Which works fine… but isn't great. <nckx>Kimapr[m]: What's a (ideally one-line) cargo command I could run in a pure environment to reproduce the failure? <Guest69238>I set the path variable to /gnu/store but I still can't open my installed version of icecat. ***terpri__ is now known as terpri
<nckx>Guest69238: That would only work for /gnu/store/icecat, which doesn't exist. $HOME/.guix-profile/bin should already be part of your $PATH. <Kimapr[m]>nckx: cargo build on any package with dependencies which need to be fetched <nckx>Guest69238: If not, run ‘. . ~/.guix-profile/etc/profile’ and try again. <rah>apteryx: that's more helpful, thanks <nckx>Translate that to non-cargo-person 😛 <nckx>Guest69238: Installing Guix should have created /etc/profile.d/guix.sh which should set up those variables for you automatically when you log in. <PurpleSym>I’m currently looking at how to integrate pypy3 properly into guix. What’s the rationale for having python-wrapper? I mean, it’s not possible to have python2 and python3 in the same env anyway due to how PYTHONPATH works. <vagrantc>you often need to set PYTHONPATH on a per-application basis ... i think <nckx>PurpleSym: IME, to alias ‘python’ to ‘python3’ without having to patch the hell out of every package that uses the former (which would be error-prone). What the description says. <nckx>Many packages use/check only for ‘python’. <PurpleSym>Sure, but the python package could just install the symlink itself, right? <vagrantc>it's discouraged for use with python3 ... <nckx>It seems like a step in the wrong direction to me. <PurpleSym>I don’t know, but it would make my life a little easier I think. <nckx>I mean, why can't you just use ‘python-wrapper’? <PurpleSym>I had to patch python-build-system’s function site-packages to choose the correct search path based on the python input (which can be python2, python3, pyyp3). Now it literally lists /lib/ and looks for a subdirectory site-packages. But that does not work if the input is python-wrapper. <Kimapr[m]>nckx: most cargo crates have external dependencies, which have to be fetched from somewhere (crates.io in most cases). i'm building hypermine, but now i have an error and looks like it needs a C compiler to build a dependency. <apteryx>we don't have a pdftk package, right? I'm looking for an option of encrypting a PDF with a passphrase. <PurpleSym>As far as I see I don’t have access to any of the package’s properties at this stage. <PurpleSym>Otherwise I would’ve gone for (string-append out "/lib/" (package-name x) (python-version x) "/site-packages/") <nckx>Damn, ‘guix environment --pure --ad-hoc rust:cargo -- cargo install hypermine’ works fine, probably because I can't container. <rah>it looks like this is unusable on a machine with limited ram <rah>where "limited" == 512M <Kimapr[m]>i use `guix environment --pure --container --network -P --ad-hoc rust:cargo rust:out --ad-hoc nss-certs coreutils guix curl gcc-toolchain` right now <rah>running "guix pull" brings the machine to its knees <Kimapr[m]>and then `alias cc=g++; cargo build --bin client --release` in there <nckx>Kimapr[m]: Yeah, I tried that but my kernel doesn't support it. <nckx>Kimapr[m]: One without user namespaces &c. <apteryx>rah: you can try --cores=1 but even with this it will be pretty limit <apteryx>swap may help too, but it'll take a long time. <apteryx>nckx: I was wrong about grafts; even using --no-grafts, it still fails. <nckx>apteryx: When offloading? <apteryx>I've removed my machines.scm file to test <apteryx>I'm using --quiet --keep-going and --check if that matters <NieDzejkob>nckx: shouldn't the search paths of libraries be 'propagated' to the packages that depend on them? <lfam>I'm looking forward to when all our bugs are >30000 <nckx>Anyway, ‘right thing’ discussions aside (I agree, they should be propagated, libva suffers from this too): Kimapr[m]: https://paste.debian.net/plain/1156522 should fix your issue, but I don't know how to test that hypothesis. <lfam>I was *just* doing some search-paths stuff <lfam>I was wondering, where all the file-types enumerated? <lfam>I mean, where are they enumerated <nckx>NieDzejkob: I didn't dive in far, but note that libcURL is its own special thing: AFAIK the library *doesn't* handle those variables, and it is indeed the consumers themselves that do. I might be misrembering, but that's a standard disclaimer to be slapped on anything I say. <nckx>Upstream says ‘libs handling env vars bad’ and that's the end of that. <apteryx>sneek: later tell rah you can configure swap if you like, in your operating-system declaration. <nckx>What I don't remember is why (native-search-paths (package-native-search-paths libfoo)) didn't work. <nckx>I think it's better to propagate them like that in a controlled manner than automatically propagating way too much. <nckx>But I didn't think that through. <lfam>I found that in certain cases, using package-native-search-paths makes Guix crash <lfam>Specifically, if you do (native-search-paths (package-native-search-paths curl)) and put it in the Transmission package definition, things will break <NieDzejkob>nckx: yes but we carry a custom patch that fixes this <lfam>I figure there's a dependency cycle but I didn't investigate <nckx>NieDzejkob: I didn't think upstream's reasoning was unreasonable but OK 😛 Fine by me. <nckx>I wasn't aware of that (newish?) patch, good to know. <nckx>Kimapr[m]: Perfect, thanks! That's the kind of idiot proof copy-paste I wanted. <NieDzejkob>nckx: would adding the search-patchs to rust trigger a rebuild? <lfam>IIUC search-paths only affect profile generation and so should not affect the build environment <nckx>NieDzejkob: No! That's the awesome part. <nckx>…I see my secr^Wadministrative assistant has already answered the question, again <NieDzejkob>hmm, what mechanisms set the env vars in the build container, then? <nckx>Kimapr[m]: Bah, it fails when building Ring even without the patch. Probably because I cannot container hard enough. Although it makes me wonder which filthy impurity is making it succeed. *nckx sudo mv /etc/ssl{,.bak} <nckx>Containers, I have made one. <nckx>Kimapr[m]: If you have time to throw away your ~/.cargo, it should be fixed on master (= no ‘curl’ needed). <Kimapr[m]>i don't need to throw away my ~/.cargo because i use guix environment container <Kimapr[m]>looks like it is thrown away for me every time i run it <nckx>Hand-crafted artisanal configure scripts are the bane of my existence. <Kimapr> --share=SPEC for containers, share writable host file system <Kimapr>so that stuff in the container doesn't get lost every time <nckx>lfam: I both saw your previous posting of that link and subscribed to the thread 😅 <nckx>If it was anyone else I'd think they were trying to stir drama. <apteryx>lfam: search-paths do affect build environment. That's how the useful environment variables get set there too. <lfam>I don't want anybody here to join that discussion! <lfam>But it's interesting to watch. And germane from a packaging perspective <lfam>It's clearly not a situation that requires feedback from outsiders <nckx>Oh, no, please don't, anyone. <lfam>Working on Guix, it's easy to assume that all projects are like this one <lfam>But, that's not the case <mbakke>modifying native-search-paths should change any dependent packages, I'm surprised it did not trigger a rebuild of rust 1.20 and later <nckx>I verified that it did not. <nckx>…although I'm keeping an eye on CI just in case. <mbakke>ah, it's because the rusts have OpenSSL as input, which adds those same variables to the build environment <nckx>NieDzejkob: No, doesn't seem to matter. <nckx>If it had triggered a rebuild here, it certainly wouldn't have occured to me to change the ordering. <mbakke>oh derp, I fixed a bug and forgot to mention it in the commit message, again :P <bdju> http://ix.io/2rIr anyone have an idea what's going wrong here? other youtube videos work as long as they're not from this same youtube channel, I can youtube-dl this video fine, just not play it in mpv, and someone else verified with the same mpv and youtube-dl version that it worked for them. <mbakke>the term "relocatable" is funny in Guix, you can copy binaries out of the store and they will work, except if they are using $ORIGIN in RUNPATH (touted as a "relocatable build") <nckx>bdju: Looks like yet another certificate problem? Works here. <bdju>ah jeez. what should I do about this? why would other videos work? <lfam>Did you ask in the #mpv channel bdju? <nckx>I wonder if it's a CDN/caching issue, so you keep getting directed to something broken for that one video. <nckx>Having actually read the rest of your message, Guix certificate weirdness seems unlikely. <lfam>What did they suggest bdju? <bdju>same thing, that it's vaguely a cert issue <bdju>and that the video worked for them <bdju>so thought maybe it was a guix problem <lfam>And it works for other videos that use https URLs? <bdju>but 3 of that same user's videos don't work, I believe <bdju>and I just tried a video uploaded 3 weeks ago so it's not even that videos after a certain date are broken <nckx>‘tls: Error in the pull function.’ can also just mean 0 bytes were received (hence, unable to set up TLS). <lfam>Well, playing Youtube videos with mpv and youtube-dl is not supposed to work, in general. So maybe you are seeing some technical countermeasures <nckx>It's just weird that mpv has to be involved. <lfam>I mean, I have no particular reason to believe that's the case. But using Youtube like this is a game of cat and mouse <lfam>Youtube-dl is updated constantly to work around "it's not an API" API changes <bdju>actually, I think the problem videos don't work on invidio.us either. they have to be played on the normal site <nckx>Especially since youtube-dl seems to be taking a short release holiday. But I haven't actually had problems. <bdju>I see "The media could not be loaded, either because the server or network failed or because the format is not supported." in invidious <leoprikler>youtube-dl + gstreamer works better than actually browsing the site tho, at least for me <nckx>lfam: It's not an A—oh, you want to pay us for an API key? Why didn't you say so! <lfam>For me, mpv works but invidious does not <lfam>I understand the motivation behind invidious but it will lead to the end of youtube-dl and the like <lfam>It's better to stay under the radar with things like this <lfam>A web page is too conspicuous <bdju>I still prefer mpv because it's actually responsive, pauses the moment I tell it to, I can see how much is cached, easy to make screenshots and webms, playlist support, etc. <lfam>Yes, it's very convenient <lfam>But not technically allowed :) <bdju>I believe it, though I can't say I'm aware of the exact rules in this case <lfam>The rules are just that you are supposed to watch ads <lfam>They aren't doing it for free <guixy>tty1 by default receives the kernel log, which is a distraction. <Guest69238>Hey guys I have the Guix package manager installed on PoP OS. I installed icecat using "guix install icecat". How do I open icecat, typing in "icecat" at the command line doesn't work. <guixy>I was able to silence most of it with dmesg -D, but when I log into another tty it logs that login/logout to tty1. Is there a way to stop this? <NieDzejkob>Guest69238: also, did you use the installer script, or did you install manually? <nckx>Guest69238: I answered you above. IceCat isn't in PATH, OK. Installing Guix should have created /etc/profile.d/guix.sh which should set up those variables for you automatically when you log in. Does it? <nckx>If it doesn't, you can run ‘. $HOME/.guix-profile/etc/profile’ manually, and you should add that to your /etc/profile (or put it in an /etc/profile.d drop-in) so it gets done whenever you log in. <bdju>I asked a friend in another city (but the same state) to test that video in mpv and he said he gets the same error. I wonder if it's a local-ish network issue. we're not using the same ISP, though. <mroh>mbakke: ty for the python-tinycss2 fix! <nckx>Maybe it is a CDN fluke after all. <mbakke>mroh: np! the gssapi one is more tricky though, I looked into it before without much luck, IIRC it has to do with a dependency that does not support Python 3.8 *mbakke wanted to go on a bug fixing spree, but real life calls, bbl <nckx>No such file? Interesting. <nckx>Guest69238: Interesting. If you manually ‘. /etc/profile’ in your terminal, then ‘echo $PATH’ again, does anything change? <nckx>Guest69238: Thanks. That's the standard file that guix-install.sh installs, looks good. <NieDzejkob>Guest69238: could you also post your /etc/profile ? <nckx>I see Pop OS uses GNOME. I'm not familiar with it, but it/GDM still sources ‘/etc/profile’, right? <mbakke>the guix.sh script looks wrong, it should source ~/.config/guix/current after ~/.guix-profile <nckx>Guest69238: OK, that should load the guix.sh drop-in. <nckx>Guest69238: > If you manually ‘. /etc/profile’ in your terminal, then ‘echo $PATH’ again, does anything change? <Guest69238>yes manually . /etc/profile made icecat come up, but it's throwing errors about not being able to load modules. <nckx>OK, but at least we know that your OS is failing to load /etc/profile at all :-/ <nckx>Guest69238: Hm. Did you install Guix, install icecat, and try to launch it all in the same session? <nckx>I'm looking at the script and it seems to assume you will. <nckx>At least log in & out again. <Guest69238>Okay will do, thank you, at least icecat is loading now. <nckx>I think it should be changed to print a message to that effect. *nckx also changes | gpg to | sudo -i gpg. <victor_>@nckx Okay I rebooted, but ice cat is still having the same issues loading modules. <nckx>victor_: Welcome back. No! Let's debug that now 🙂 What exactly does it say? <nckx>I don't even know which ‘modules’ IceCat is talking about. <nckx>I hope someone more familiar with foreign distros can help with that part. <nckx>I can't believe that nobody is using Guix's IceCat on a foreign distro. At least I hope not. <lfam>I can try reproducing the bug <nckx>victor_: Sorry about your first experience. It's not a good look. The ‘problem’ is that most Guix contributors run Guix System, where at least IceCat mostly works. <victor_>I'll stand by on what to do next, since it's not me just misconfiguring something like usual lol. <nckx>Browsers are complex & hard to fix & maintain, while at the same time being a popular ‘let's try guix!’ first package :-/ <lfam>I would try using the Epiphany browser, or just using a browser from your host distro <lfam>Epiphany is sort of simple compared to Icecat or Chromium but it does work and gets security updates <victor_>@nckx no need to apologize, Guix has been a revelation to me, so all this birth pain is worth it to me. If I could I'd just run Guix all the time but so much of the software echo system is proprietary <lfam>You could also use ungoogled-chromium <nckx>Bah, but I can't disagree. <victor_>Okay thank, the reason I use Icecat is because it comes prepackaged with librejs, which allows me to watch YouTube. Can I install libre js on epiphany? <lfam>I'm not sure but Youtube does work on epiphany <lfam>Without having to install anythiug <lfam>Well, you might have to install some gstreamer plugin packages. But those are Guix packages, not browser plugins <guixy>ungoogled-chromium might take forever to build if a substitute isn't available. <lfam>I'm not really sure why epiphany requires gstreamer plugins to be installed manually... <guixy>I noticed that more recently a substitute is not available :( <victor_>can I put an ad blocker on epiphany? <nckx>victor_: It's not the same thing, but you can watch youtube videos with ‘guix install mpv youtube-dl’, then ‘mpv <URL>’. <nckx>There's also youtube-viewer but it may or may not still work (proprietary YouTube API nonsense). <nckx>Lot's of recurrent themes in #guix today. <lfam>victor_: I checked and epiphany isn't working with Youtube anymore. Sorry, I'm not sure what's up <guixy>tty1 by default receives the kernel log, which is a distraction. I was able to silence most of it with dmesg -D, but when I log into another virtual terminal it logs that login/logout to tty1. Is there a way to stop this? <victor_>I guess I'll just have to install icecat the regular way . <lfam>guixy: I don't know the answer to your question. Did you try googling it? <lfam>victor_: Epiphany works with Youtube if you also install gst-plugins-base, gst-plugins-good, and gst-libav <mroh>same with nyxt and/or qutebrowser etc... <guixy>lfam, I have searched for the issue, but my searching skills have never been very good. Google is not my friend. <nckx>I didn't have more luck on DuckDuckGo though. <lfam>I would just Google the answer, but I don't want to involve your enemy ;) <lfam>Basically, there are kernel arguments you can use to make it quieter, and you can also configure TTYs to be silent <lfam>I'm sure there are lots of knobs and switches to turn here <nckx>lfam: This doesn't sound like a kernel message though. <guixy>I'm building my configuration. When it's done I'll grep for tty1 to see if there is something accidentally configured to send output to tty1. <nckx>But yeah, ‘just make tty1 not the console anymore’ is probably the answer. <guixy>I like having tty1 available... I use all 6 virtual terminals. <nckx>guixy: I doubt it has anything to do with you. *nckx .oO ‘All 6’? There are 12! You're missing out! *raghavgururajan dances after building wpewebkit and webkitgtk sucessfully on bayfront. <guixy>nckx, yeah, I guess I could hack my configuration more. If I get time I'll make it so there's no x server until I call startx. <nckx>raghavgururajan: Yes, and now I even read them. <nckx>guixy: That's not actually trivial but should be possible. <guixy>gnome-keyring is also abusing tty1 <guixy>Is there a tty0? When I go to tty1 and press alt+left it takes me to a console that conveniently receives kernel log. <guixy>If I can identify that, would I be able to use it for all system logs? <nckx>guixy: I use tty1 for X, using (service slim-service-type (slim-configuration (vt "vt1") …)). <klys>guixy, that's probably tty6 <nckx>Not for this reason (I just prefer it that way), but it would happen to solve your problem. <klys>guixy, it should be your last configured tty <guixy>ctrl+alt+f12 sends me there... <guixy>My last configured tty? Is there a way to see/change how many are configured? My shepherd graph only lists 1-6. <klys>it's the highest number that was used with openvt() <guixy>Is there a way to configure which ttys are mapped to which function key? Could I abandon tty1 and map a different tty to f1? <klys>that's done in the kernel <klys>if you want to look at the linux vt logic, that's under CONFIG_VT <guixy>When I did LFS, sysvinit had a script that handled switching ttys. I thought that was the responsibility of PID 1 <vagrantc>klys: but taking a long time is entirely unsurprising <nckx>klys: That looks normal. <mroh>nckx: "2 short patches for guix-install.sh" did you forget the patches? ;) <nckx>mroh: No, it takes a while for mails to arrive lately. <nckx>I didn't busy-wait for the bug number, but thanks for the ping! <vagrantc>klys: like nckx says, looks normal ... any cpu load or indication it's doing anything? <klys>vagrantc, I have been monitoring top and iotop and it doesn't look like anything relevant is using the cpu <nckx>I've noticed that it takes several hours for newcomecs to even make it to the guix-patches moderation queue. <vagrantc>klys: you can also DEB_BUILD_OPTIONS=nockech dpkg-buildpackage ... <nckx>The past few days at least. <klys>thanks, I'll break out and try that <NieDzejkob>heh, rust 1.45 has some new parser tests that involve shebangs. I wonder if/how Guix is trampling them <NieDzejkob>is there a way to offload builds but have the remote machine keep the failed directory? <victor_>does anyone know how to get a vmware image to work in GNOME-boxes. When I try to open them they keep failing. <victor_>I'm installing virt-manager now because I read online it can run VMWare images. If I get stuck can anyone here help me? <ss2>hey, thanks everybody! Your help is very much appreciated! Early on I managed to break my system and after reinstalling and configuring with my fresh config.scm just made my day. :) <nckx>victor_: I hate to say this but… I think virt-manager requires a ‘system service’ to work, and Guix on foreign distroes can't… hehm. <lfam>Can't you use qemu-img to convert between formats here? <lfam>I should say, rather, that I think you can do that <lfam>Then, you can use QEMU or anything based on it <nckx>I wonder if that was victor_. Maybe the ‘someone’ was you? <nckx>I couldn't find the message they were referring to but the search is still borked. <nckx>Anyway, they quit right before my reply. <nckx>klys: Yes, because of ‘substitute: guix substitute: warning: ACL for archive imports seems to be uninitialized, substitutes may be unavailable’. <klys>so should I break out and try to enable substitutes? <nckx>Did you mean to say no to substitutes? Also, what's the context here? A manual binary installation? <klys>I have a full-on debian system <klys>I don't think substitutes should be off <nckx>klys: Are you following the ‘Binary Installation’ section of the manual? <nckx>Then see point #7 (‘To use substitutes from …’). <nckx>Note that you're expected to be root at that point. <nckx>(While we have nothing against third-party guides, we can't support them.) <nckx>And you don't seem to be using the script since you created the build users manually. <apteryx>seems to occur only on Guix: error: ‘round’ is not a member of ‘std’; did you mean ‘round’? <NieDzejkob>apteryx: is that a C++ program? Could you try making a minimal example? Also, which version of G++? <apteryx>it builds fine in Ubuntu 19.04 (ran with Docker on my Guix System) <nckx>mbakke: I wonder if ba1e4ce256 should be reverted after all, unless Rust packages are relatively cheap: <https://ci.guix.gnu.org/eval/15095>. Despite rust itself not changing, it does seem to propagate changes ‘outside’ of the hash to their build environment. <nckx>klys: It does. Your first paste was downloading tonnes of source tarballs to build on your machine. This one is downloading binary packages. <nckx>(Well, I don't know what you thought substitutes meant, to be fair.) <NieDzejkob>klys: hmm, rusts can only be built on/for x64 for now <klys>and I guess I should be asking why it's trying to build rust <NieDzejkob>so if you're on 32-bits or some ARM then it's gonna try to build it and fail <NieDzejkob>oh, icecat has some components of it written in rust <klys>if it just needs some rust libs mebby I can install them with apt? <nckx>klys: Guix builds are completely isolated. <nckx>They don't see your host system. <klys>if I was building google's new os from scratch, perhaps I could see compiling that many copies of rustc, this seems ridiculous however. <nckx>I'm going to revert ba1e4ce256 because this is too suspicious. <nckx>klys: Rust can only be built with the previous version (you can skip some, but the argument is the same). <klys>is this what the "guix package manager" is good for? "builds...that don't see your host system" ? <nckx>It's one of the properties that make it good at what it does. <klys>I had thought I'd read that it "works with your existing package manager." <nckx>It does, you can use both at the same time. You can't tell apt to link against Guix's libraries, or Guix against pacman's. <nckx>That would be an unreproducible mess. <nckx>Everyone would have slightly (or wildly) different files depending on in what order they installed which version of what. <nckx>Madness! Confusion! Gentoo! <klys>where does guix install binaries to? /usr/bin ? I had thought someone was thinking long and hard about those kinds of problems when guix was developed. <nckx>/gnu/store (inspired by /nix/store). Indeed they were, as were the creators Nix, the direct inspiration for Guix. <nckx>If you ‘guix pull’ and reinstall icecat, the evil Rusts will be gone. <klys>you can "use both at the same time," then. and you will have two completely separate systems on your disk. <nckx>It sounds like you might want (or have expected) something else, but that was very much on purpose. <klys>"a package manager that you can use with your existing apt package manager" sounds like a different idea. <nckx>It both accurately describes Guix and wildly different things that aren't Guix because language is vague. <apteryx>NieDzejkob: I think it's a problem where some round from GNU overrides round from the std c++ headers (perhaps yet another ordering issue) <nckx>I personally think ‘alongside’ would be better. Where's that quote from? guix.gnu.org? <apteryx>or shadows. Just guessing so far, but cmath from gcc correctly defines std::round at least. <klys>sounds like someone was trying to be genearal accurate and simple at the same time. yeah that's probably where I read it. <nckx>All I can find on the home page is ‘Guix won't interfere with the package manager that comes with your distribution. They can live together.’ <nckx>klys: Thing is, I can't find any search results for (some basic variants of) your ‘works with your existing package manager’. Either on guix.gnu.org or elsewhere. <nckx>I don't see anything promising integration in the way you expected, but please let us know if you find anything on the Web site misleading. <nckx>It's hard to read things with fresh eyes. <branjam>Guile/guix on a few of my computers fails to recognize a certificate file with non-ascii characters while building standalone images. Would I be able to solve that with config changes or should I forward the issue to the mailing list? <nckx>branjam: Guix builds should be almost independent of configuration. Please send a bug report to bug-guix at gnu.org, with the failing command(s) and as much output as you think is reasonable (or more).