IRC channel logs

2024-07-17.log

back to list of logs

<ieure>podiki, Never seen it, anything matching that number in the output of `curl -I' on the download URL?
<ieure>Or maybe `curl -v', could get a different payload if you make a HEAD.
<podiki>i see this in curl -v "strict-transport-security: max-age=31536000; includeSubDomains"
<podiki>under HTTP/2 302
<podiki>or if i follow the redirect: "cache-control: public, max-age=31536000;"
<podiki>well i don't know the answer but in the mean time the webarchive has the file and eventually guix tries tehre
<eikcaz>I'm trying to authenticate my guix clone, but I get "make: *** No rule to make target 'authenticate'. Stop."
<eikcaz>Ah, looks like that changed to "guix git authenticate"
<spenc>yeah that got me too, you gotta look at the /devel/ version of the manual
<spenc>unrelated, but does anyone here have a take on Unison lang?
<spenc>seems like its doing a similar thing to Guix, but incredibly more granually, so each function of code is it's own hash addressed store
<jungy>Anyone know how might I see xorg logs and conf? Trying to enable SWCursor.
<RavenJoad>jungy: You can find the config file X is using with "ps aux | grep X". It will be after the -config portion of the command. On Guix, almost all logs are in /var/log. I have /var/log/Xorg.0.log.
<jungy>thanks!
<jungy>I couldn't find anything under /var/log so I was assuming it was deeper, but it's possible it got cleared out.
<isf>I found a video of richard stallman fighting and I add to it subtitles
<Oleand3r>civodul: I found out what was causing the bootloader error. The guix-defaults for the home-bash-configuration were adding some code to ~/.bashrc that was already defined in my local-file, so I had to disable the defaults as explained here: https://issues.guix.gnu.org/67652
<Oleand3r>This also fixed the shutdown issue.
<OwOday>ok earlier I was asking about packaging some nightmare weirdware python 3.8 thing into guix that had no toml
<OwOday>my solution is to make it build locally with pyenv and poetry then use pyinstaller to make a binary and package that. Thoughts?
<spenc>for upstreaming, or for your own use?
<OwOday>I guess both?
<OwOday>Is there a problem with upstreaming a package that uses a binary?
<OwOday>also, weird question, but can quix distribute stuff thats meant for integration with init systems?
<OwOday>I know guix-system has its own thing, but what if you're on debian or arch or something, does it rule out packages that make use of init systems or what?
<iyzsong>OwOday: build package from binaries have problems: it can be broken due to ABI mismatch between its dependencies (changes frequently) and the binary, it may have security issues which are hard to find and fix. so it's not accepetable for upstream. For support other init systems (than shepherd), ackages can ship unit files, but require manual setup (eg: symlink those unit files into /etc/systemd/system) to actually
<iyzsong> use them.
<bost>Hi. Can anybody explain me why `guix package --list-installed=emacs-with-editor` is twice the same?
<bost>emacs-with-editor 3.3.4 out /gnu/store/fh7d3yhp6ahgmxm0a91mkgagvzlchc89-emacs-with-editor-3.3.4
<bost>emacs-with-editor 3.3.4 out /gnu/store/fh7d3yhp6ahgmxm0a91mkgagvzlchc89-emacs-with-editor-3.3.4
<bost>returns twice the same
<atweedie>I just tried a physical guix install w/ an encrypted root & home partition
<atweedie>now I'm getting asked a password 4 times at boot
<atweedie>IIUC once by first stage grub (both partitions), then again both partitions by guix proper
<atweedie>is that expected behaviour? /boot is not encrypted
<atweedie>the resulting system doesn't boot successfully, but that's a separate issue, I'm sure
<jpoiret>atweedie: the kernel and initrd reside in /gnu, so it's unfortunately expected
<jpoiret>usually /boot only contains the grub configuration
<dariqq>currently testing the 6.10 kernel (a slightly adapted to get things working) and facing a problem that gdm crashes the whole vm i am running . If i switch to a tty quickly things are fine. On my greetd+sway laptop things are also working as expected.
<dariqq>any ideas how i can start troubleshooting this? in the logs i only have a lot of null characters
<dariqq>starting gnome-session via dbus-run-session also freezes and crashes so not really a gdm issue
<civodul>friends, i586-gnu is broken on core-updates
<sneek>Welcome back civodul, you have 1 message!
<sneek>civodul, RavenJoad says: The VM should have enough RAM, yes. I gave the machine 8GiB. The weird part is, I cloned the UEFI VM to test with BIOS too, and the BIOS one works fine. All I did was clone the UEFI VM, change the "Boot firmware" to BIOS, and reinstall.
<civodul>janneke: hi! do “unknown type name 'kernel_boot_info_t'” and “implicit declaration of function 'proc_reauthenticate_complete'” ring a bell? :-)
<civodul>this is while compiling hurd-minimal-boot0 on core-updates
<civodul>(natively)
<civodul>apparently might be related to the hurd/mach upgrades back in Feb: https://ci.guix.gnu.org/build/3408522/details
<atweedie>thanks! maybe I could get grub to stop decrypting /home twice?
<civodul>atweedie: hi! /home is decrypted after Linux has booted, GRUB is not involved
<civodul>(and thus only once)
<civodul>but perhaps your root file system is encrypted too?
<dariqq>werid when gdm/gnome-shell is on x11 everything works. Something goes horribly wrong on wayland
<dariqq>ok nvm ti just hangs at a later point
<dariqq>but at least the system does not crash
<OwOday>iyzsong: of course I agree, even worse the program contains a web server that would belong to the init system and multiple entry points that use the web server/clipboard. But still, I'm curious what it would take release it. I wish their was the equivalent of a community repo I could put it in. Is it feasible to use guix to run pyenv pyinstaller on
<OwOday>the clients machine as a build process? To install the weird code as a binary? Whats the vibe check at work here to upstream packages, no hacks?
<attila_lendvai>OwOday, no such hacks are accepted upstream. guix packages are built reproducibly in a sandbox, i.e. no internet connection.
<attila_lendvai>OwOday, what i do for such cases is i create a unix user for the service, install the dependencies for that user, then install the unpackaged app into a directory, and then start up the service from shepherd from that user and dir
<lfam>Howdy guix
<attila_lendvai>OwOday, there are some channels that allow binary releases as packages, but they still expect the package to be reproducible, i.e. the binary source (i.e. the tarball, or zip file) must have a hash code that identifies it.
<OwOday>thank you
<OwOday>do you have an scm that demos the user creation process
<civodul>hey lfam!
<lfam>I've been testing core-updates on x86_64 and, at least for me... not bad
<lfam>I reported some failures of leaf packages, but nothing too deep in the package graph
<lfam>Epiphany fails its test suite on core-updates, although it does work as a web browser when the tests are skipped: <https://issues.guix.gnu.org/72048>
<lfam>I read that it's not as pretty on other systems besides x86_64, though
<lfam>Regarding Epiphany, it seems that dbus is killed by SIGSEGV in the build environment. Is this expected? If not, any hunches as to why it gets killed?
<attila_lendvai>OwOday, i do that by hand. just create a new user, guix install [deps], then git checkout the repo, do the app specific install dance, and then start the app from there
<OwOday>oh, so its not automated by guix?
<attila_lendvai>OwOday, this is the shepherd service that i use to start it: https://paste.debian.net/1323547/
<attila_lendvai>OwOday, this is a guix channel that has examples for packaging binary releases: https://github.com/attila-lendvai/guix-crypto
<OwOday>oh thanks
<atweedie>civodul: yes, root is encrypted as well. does that explain why /home needs to be decrypted twice as well?
<atweedie>I've also noticed that decrypting the first time takes over 30 seconds for each partition
<ieure>Yes, it's dirt slow.
<atweedie>ah, got it
<jpoiret>atweedie: the crypto implementation of GRUB is not very good, also doesn't run in actual 64-bit mode
<jpoiret>only real mode iirc, not long mode
<jpoiret>ah, maybe i'm wrong, UEFI bootloaders apparently start in long mode
<RavenJoad>jpoiret: atweedie: I think GRUB is single-core, which is part of the problem. I think GRUB does NOT use hardware cryptography extensions, which is the bigger issue.
<RavenJoad>civodul: Sorry to ping you about this so much later, but did you see my response about my VM setup?
<dariqq>when you try to fix something for hours only to realise that you cant fix it because it is an upstream problem
<RavenJoad>Could someone look at #71825? It is needed to fix virt-manager for non-GTK desktop environments.
<peanuts>"[PATCH] gnu: Fix virt-manager for systems with no GTK cursor." https://issues.guix.gnu.org/71825
<lfam>RavenJoad: Do you have a hyperlink to the source of the patch file?
<lfam>It's helpful to include this in the patch file as a reference
<RavenJoad>I have a patch generated out of the git commit. Virt-manager has not had a tagged release in quite some time. https://github.com/virt-manager/virt-manager/commit/cc4a39ea94f42bc92765eb3bb56e2b7f9198be67
<lfam>Thanks
<dariqq>my problem was solved by switching from qxl for video in my vm
<lfam>I'm building with your patch now RavenJoad
<lfam>Can you confirm it fixes the issue for you?
<RavenJoad>Give me a moment and double-check. I remember it worked 2ish weeks ago.
<lfam>Thanks!
<lfam>It builds for me and I added the hyperlink and some other annotation
<lfam>Ready to push
<RavenJoad>Just built locally, and I can actually do things in the UI. Seems that the cursor patch fixes everything.
<lfam>Great
<lfam>ACTION g2g
<b0ttymac`>How is the rainbow highlighting for scheme code parenthesis in the html manual done? I'd like to recommend doing so in the mcron manual as well.
<lfam>Hm, something new in linux-libre-6.10. A set of errata patches that must be fetch and applied. It's easy to apply them to the upstream linux-sources, since our code already has the capability to accept optional patches for that. But we will grow the ability to apply patches to the deblob-scripts
<ieure>Would anyone like to review this patch series which updates Librewolf? https://issues.guix.gnu.org/71832
<dariqq>lfam: so that you dont waste ressources too i also needed to use the latest dwarfes package to build 6.10 earlier today (otherwise libbpf fails eventually)
<dariqq>* dwarves
<lfam>Thanks for the hint dariqq
<lfam>Is there a patch for that yet?
<dariqq>no not yet, i have been struggling with a 6.10 regression that kills my testing vm upon reaching gdm
<lfam>I see. It's fairly typical for the mainline release to be buggy like that
<dariqq>workaround was to change video output of the vm to not qxl. But took me a while figure that out as things were fine on my laptop
<lfam>ACTION writes the most naive non-schemer Scheme pseudo-code and crosses fingers
<mirai>is it ok to simply encrypt /home with LUKS? The performance for a FDE setup seems awfully slow
<b0ttymac`>What ist the prefered way to search for package or service modules?
<b0ttymac`>There has to be a better way than searching the git source or guessing names until reconfigure works.
<lfam>`guix search` for packages, and `guix system search` for services
<b0ttymac`>guix search shows me where a package is located, but is there an interface for finding out which module provides lets say fail2ban-service-type?
<b0ttymac`>ahh thx
<lfam>Cheers
<lfam>I'm working on applying the deblob script patches. For now I have this diff: <https://paste.debian.net/1323579/>. However, when I try to build Guix (`make`) it crashes with "gnu/packages/linux.scm:381:10: error: (if errata-patches (patches (search-patches errata-patches)) (patches (quote ()))): invalid field specifier"
<lfam>I'm not a strong Schemer. Does anyone see what's wrong?
<lfam>Hm, it works if I "reverse" the scope of (if ...) and (patches ...)
<lfam>I suppose that's enough to keep going for now
<RavenJoad>b0ttymac`: Because of the way Guile modules work (of which Guix is one), if you get a result from a search, the module is the "location" field with / replaced with spaces and .scm removed.
<RavenJoad>For example, "guix system search fail2ban" produces location: gnu/services/security.scm:378:2 on my system, which means fail2ban-service-type is in module (gnu services security).
<b0ttymac`>Ive got it, same as the package modules.
<RavenJoad>b0ttymac`: re manual's syntax highlighting: it is a combination of generated HTML <span>s and CSS :hover pseudo-classes.
<civodul>lfam: maybe (map search-patch errata-patches) and have #:errata-patches default to '()
<civodul>that is, you must always supply the ‘patches’ field
<civodul>and its value must always be a list
<civodul>but you can make it the empty list
<lfam>civodul: That seems like a handy way to accept 0, 1, or more than 1 patches. Rather than my solution which assumes a single patch
<lfam>I appreciate the advice
<civodul>or you can have #:errata-patch (singular) that would be either #f or a file name
<civodul>and then: (patches (if errata-patch (list errata-patch) '()))
<dariqq>lfam: how would thinks work if both scripts require patching? Do we know if these would be seperate patches foreach script?
<lfam>civodul: Yes, I had got it working with that 2nd option. But I like the use of map
<lfam>dariqq: Then we'd have two option parameters, like "deblob-errata-patches" and "deblob-check-errata-patches"
<lfam>s/option/optional
<lfam>I don't know if it's better to prepare for that now, or to take it as it comes, so to speak
<dariqq>i guess we could always split the scripts patch manually
<lfam>dariqq: Like, if a single patch file was supposed to patch both the deblobbing scripts and the deblob-check scripts?
<dariqq>yes
<lfam>Yeah, I think it would be okay to handle it manually. I already had to adjust the name of the deblob script file to be patched, since we rename the deblob script file names. And we would also have had to add (patch-flags '("-p2")) in linux-libre-deblob-scripts
<lfam>There are pros and cons to fetching patch files at build-time rather than including them in our repo, but I think it's okay to include them for things like this that will probably be short-lived
<lfam>A kernel series usually lasts a few months
<lfam>ACTION rebots
<lfam>My mail provider is having trouble, so my patches for the deblob errata scripts will be delayed
<DinGoAussie>Good day all. Wow, a lot of members in here.
<lfam>Howdy
<DinGoAussie>Hey Ifam.
<DinGoAussie>I wish I could say I was chatting from a Guix host, but unfortunately I am not, at least not yet.
<lfam>All good
<DinGoAussie>Trying to install it on a system that it appears does not have free software support.
<DinGoAussie>Is it a common thing for new users to face this? For me it appears to be network and video card driver related. Intel NIC, AMD Athlon APU.
<lfam>Yes, it's unfortunately common
<lfam>I think that in the past, consumer-level ethernet PHYs did not require blobs, but that's changing recently. Wifi and GPU have required non-free firmware for a long time
<civodul>oh NICs as well?
<lfam>I don't have details but I seem to recall hearing about it. Worth looking in to for DinGoAussie
<lfam>Hopefully I am wrong!
<DinGoAussie>well, when I tried on one of my systems yesterday, I got: "Device not supported by free software were foumd on your computer: - Intel Corporation Wi-Fi 6 AX200 (networking device)
<DinGoAussie>Would you say there are workarounds for it using the same hardware, or it is just not possible.
<lfam>Yes, it's expected that the wifi does not work with free software
<lfam>As far as I know, the most recent wifi hardware that works with free software are the Atheros chips that use the ath9k driver, which is 802.11n wifi
<lfam>They are okay (I was astounded at how fast 802.11n was back in the day) but considered relatively obsolete
<lfam>The only "work around" is to use non-free software / firmware
<DinGoAussie>So I can still install GUIX on the system but I must provide the proprietary drivers at the time it asks?
<DinGoAussie>Ok, it looks like I have some background work to do in order to install it. https://wiki.systemcrafters.net/guix/nonguix-installation-guide/
<lfam>Right DinGoAussie, that's what you'd have to do
<ieure>DinGoAussie, Yes, all normal stuff, very irksome, Guix System is very difficult to install because it only does netinst and doesn't support most wifi. I understand why it's like this, but find it very annoying.
<DinGoAussie>I am beginning to see that now. Thank you ieure, so you feel my pain, but it is all good. I thought about NixOS instead, but I want to not have to rely on systemd.
<DinGoAussie>bbiab
<theesm>lfam, thanks for looking into fixing the build failures of the 6.10 kernel; noticed CI failing shortly after it started to pick up kernel-updates with the recent patch... it seem's like the way linux-libre adds its deblob comments in some files became invalid? (haven't tried to build the kernel locally before sending the patch, so I didn't catch the erroneous behaviour early on)