IRC channel logs

2022-05-05.log

back to list of logs

<mekeor[m]><djeis> "Channels should be able to refer..." <- you said that channels should be able to refer to modules from other pulled channels. i first pulled babariviere's emacs-guix channel, then i changed my personal channel to refer to it, but it fails with "no code for module ... (emacs packages melpa-generated)"
<mekeor[m]>lets see if (emacs packages melpa) works
<mekeor[m]>no, it doesnt :/ same error
<djeis>Is this after pulling your personal channel?
<mekeor[m]>this is while pulling my personal channel
<djeis>Hmm.
<djeis>Oh, do you have a dependency on the other channel in your channel definition?
<djeis>i.e. https://guix.gnu.org/en/manual/devel/en/html_node/Declaring-Channel-Dependencies.html#Declaring-Channel-Dependencies
<luke-jr>kinda annoying that the so-called "minimal" config pulls in X, texlive, etc in the first place :/
<mekeor[m]>djeis: thanks for the hint. lets try :)))
<mekeor[m]><luke-jr> "kinda annoying that the so-..." <- hm where did you find that minimal config?
<luke-jr>mekeor[m]: /etc/configuration/bare-bones.scm
<mekeor[m]>im not sure where that comes from, luke-jr:
<mekeor[m]>here's a minimal bare-bones configuration in the guix repository: https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/system/examples/bare-bones.tmpl
<luke-jr>looks like the same one
<djeis>X is a dep of the %base-services? Huh.
<djeis>Doesn't seem to be an intentional one, given the comments at the top.
<djeis>Maybe it's a build dep of something that supports X?
<horizoninnovatio>Who'd like to help?
<horizoninnovatio>Installing on a thinkpad r61, installer 1.3 errors before even starts! 1.2 & 1.1 both install successfully, then on reboot: failed to start service file-systems, file-system-/boot/efi, etc. Gtp file system, multi-boot with other linuxes.
*horizoninnovatio uploaded an image: (154KiB) < https://libera.ems.host/_matrix/media/r0/download/matrix.org/OzMLvxflMueStcGOhiuesRGS/20220505_092725_4066574866070673040.jpg >
<djeis>What's your file-system form for /boot/efi?
<horizoninnovatio>Fat32
<djeis>Ah, no, that's not quite what I meant... Are you just trying to boot the installer iso right now?
<djeis>Oh, no, I see.
***rekado_ is now known as rekado
<djeis>Hmm... I'd need to see the contents of the guix config file to help, essentially, and I forget where the interactive installer puts that at the end.
<horizoninnovatio>Grub installin to sda, /boot/efi on separate partition "grub-bios"
<djeis>Right, yea, and in your guix config file there's an entry for your /boot/efi mount. Likely, something about that is wrong.
<djeis>Oh, wait, grub installing to /dev/sda?
<djeis>That'd be for a bios boot grub, not an efi one... I think, anyway.
<djeis>Or do you mean that grub is installed on the efi partition of sda?
<luke-jr>djeis: any easy way to view the dep graph?
<horizoninnovatio>This is the "bootloader" section of config.scm:
<horizoninnovatio>(bootloader... (full message at https://libera.ems.host/_matrix/media/r0/download/libera.chat/483d9a5821bd75038738fd04b4cf8a8e3dd8510e)
<djeis>luke-jr: Of a whole system config? Not that I'm aware. You can get that for a given package pretty easily tho.
<djeis>horizoninnovations: That looks like you've installed a bios grub instead of an efi grub.
<djeis>But also, do you have any way to verify that the `(uuid "0092-DC63" 'fat32)` in the `(file-system ...)` form for boot/efi is correct?
<djeis>With an efi grub you'd have `(bootloader grub-efi-bootloader)` instead of `(bootloader grub-bootloader)` and the `targets` list would have `"/boot/efi"` instead of `"/dev/sda"`.
<horizoninnovatio>So, for a gtp file table I'd need to edit to install grub to /boot/efi rather than sda (mbr?) and then it should boot? I can add the relevant section of grub.cfg to another linux grub and boot and it will have the same result.
<djeis>Well, clearly you're booting it one way or another. So, fixing the bootloader won't fix your boot problem.
<djeis>Not being able to start file-systems is a separate issue.
<horizoninnovatio>It's just the services failing
<djeis>I'm just also pointing out that, once you've fixed that, your bootloader definition doesn't match up with you having an efi partition.
<horizoninnovatio>OK, thanks
<djeis>So yea, if the problem is actually file-systems, the first thing I'd check is if all of those uuids are right.
<djeis>With the ext4 one, I'm not sure you need to specify `'ext4` in the call to uuid: ext4 uses a standard uuid, most file systems do, it's just fat uses a short one.
<horizoninnovatio>ok, I'll check. I've been trying this for a couple of days and pretty sure all the uuid are good.
<horizoninnovatio>The config.scm was generated by the installer
<djeis>Hmm.
<horizoninnovatio>Puzzeling indeed. I had guix running on this machine a few years ago, found an old config.scm using lxde as the desktop (not available now!)
<djeis>I admit, I've never used the graphical installer.
<djeis>One way or another I've always written the config myself and used guix system init
<djeis>That config looks reasonable though.
<djeis>Well, apart from the bootloader discrepancy.
<djeis>luke-jr: Oh yea, I think it's ssh is built to link against a bunch of X11 stuff. Probably needs to build said X11 stuff in order to provide that to the linker or such.
<horizoninnovatio>I've also tried with guix system init. Used graphical to generate the config.scm then abort to cmd and contiinued from there. Either same result or grub not installing (not liking file system and blocklinking etc) and having not grub files at all. Alway though with a failing boot.
<luke-jr>djeis: but (openssh-configuration (openssh openssh-sans-x)
<djeis>luke-jr: Oh, indeed.
<djeis>Allow me to look at the package graph for that...
<horizoninnovatio>I'm going to try without /boot/efi on another partition if it isn't needed. Just read that if a partition isn't mounted (and therefore not there, file-system will fail!
<luke-jr>kinda surprised the system config isn't considered a package itself XD
<djeis>System config's are a different sort of derivation than packages.
<djeis>In theory I could look at the derivation graph of the system config, but mapping that back to nice friendly things like packages would be really tricky. And, that graph would have a lot of noise in it.
<djeis>Also, openssh-sans-x still build-depends on x indirectly because it depends on eudev, which depends on python, which depends on Tk, which depends on X stuff.
<djeis>Well... the path actually goes https://paste.debian.net/1239970/
<djeis>(you can get a rendering of the whole dep graph of a package by not passing `--path foo`, comes out in dot format)
<djeis>I suspect a lot of this is lighter if you accept substitutes b/c they don't require build-only dependencies to be downloaded.
<djeis>So, you might want to host your own substitutes for other machines on your networks.
<luke-jr>Python upstream doesn't require Tk :/
<djeis>Depends on how it's built, I think, and the plain python package in guix does.
<luke-jr>sigh
<djeis>I bet python-minimal doesn't depend on it, but it seems that python-minimal has a build dep on the full python package.
<luke-jr>lol
<luke-jr>Gentoo has USE flags for this kind of thing :x
<djeis>Yea, guix's packaging infrastructure as it stands doesn't have a lot of support for doing cross-cutting variants of the entire package tree.
<djeis>It could be done, I suppose, but it'd need a good bit of work to implement.
<djeis>And it'd presumably introduce a bit of combinatorial explosion for the substitute servers...
<djeis>I wonder if grafts could lessen the load there.
<horizoninnovatio>Out of curiousity, can another linux install (like debian) be converted to guix inc init system to shepherd?
<djeis>horizoninnovations : sortof.
<djeis>If you have a guix (the package manager) install in the other distro then you can `guix system init config.scm /` to install guix in-place on top of that other linux distro.
<djeis>However, that's only kinda a conversion.
<djeis>It's destructive, and it really is more installing the config in that `config.scm` to `/` than "converting" that other linux install.
<horizoninnovatio>I presume shepherd init is managed by the guix package manager. Can a systemd init system be "converted" to shepherd? I haven't seen any how-to's!
<djeis>Depends what you mean by convert.
<djeis>Like, is there a tool to generate a guix config that sets up a shepherd which does everything the systemd was configured to do? No.
<horizoninnovatio>I've "swapped out systemd for openrc and runit successfully. Still learning about guix so was wondering if systemd could be swapped out of shepherd as an init system
<sneek>wb atka
<atka>thanks sneek
<horizoninnovatio>* "swapped out" systemd
<AwesomeAdam54321>horizoninnovatio: I did it, but by hand
<horizoninnovatio>AwesomeAdam54321: That sounds like fun 😂
<djeis>I mean, I suppose you could set up shepherd as the init system of a debian distro with enough work...
<djeis>It'd be a lot of work though.
<djeis>Or, you could use guix to install guix system on to the partition where you had debian installed.
<djeis>In principle you could do that without needing to leave debian, but next time you booted you'd boot in to guix system instead.
<horizoninnovatio>djeis: That's exactly what I'm looking for!
<djeis>If the guix config was written carefully it'd probably keep around any users and user-data. Probably most of the file system would be left unchanged, actually.
<djeis>Ah, that's what I meant by doing `guix system init config.scm /`.
<djeis>You'd first install guix to debian, have systemd run the guix daemon and such.
<horizoninnovatio>Could config.scm install/initialise shepherd?
<djeis>Then you'd write a guix system config.scm that matches your existing partitions.
<djeis>Ah, yea, shepherd is the init system that guix system uses.
<djeis>Any deployed guix system config would run shepherd.
<djeis>When I say `config.scm` I'm talking about the sort of file that the graphical installer generates. That file, with its `operating-system` definition at the end, gives `guix system` everything it needs to configure your machine. `guix system`, in turn, sets up shepherd as the init system when it boots.
<horizoninnovatio>So, in therory: install debian, install Guix package manager, write config.scm, guix system ini reconfigure config.scm, reboot into guix system?
<horizoninnovatio>s/ini/init/
<djeis>No `reconfigure`, but otherwise essentially yea.
<AwesomeAdam54321>yep, the partition will have its space taken up by 2 systems though
<djeis>`guix system init config.scm /`
<djeis>`guix system reconfigure config.scm` is for updating the config of an already installed guix system.
<djeis>`guix system init` otoh is the "real" guix installer, the wizard is mostly just a partitioning tool a generator for the config.scm.
<horizoninnovatio>ok. I might give it a go, migrate a debian system to guix. I'll probably start with devuan as it has openrc/sysvinit rather than systemd!
<djeis>This is actually how `guix deploy`'s digitalocean support works, btw: it uses digitalocean's API to provision an ubuntu droplet, ssh's in to have it install the guix package manager, and then it uploads a system config and does a `guix system init ... /`
<horizoninnovatio>djeis: That's interesting. Wonder if their script is avaiable?
<djeis>I've also done something similar to install guix using another linux distro's installer I had lying around: boot in to that installer's live environment, install the package manager, partition and mount the target system's drive, `guix system init config.scm /mnt/foo/`
<djeis>*install the guix package manager
<djeis>When I couldn't be troubled to write guix's own installer iso on to a drive.
<djeis>And yea, one sec..
<djeis> https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/machine/digital-ocean.scm#n192
<djeis>Oh huh, they actually do it by a guix system reconfigure.
<horizoninnovatio>djeis: thanks
<djeis>In a comment they reference https://wiki.pantherx.org/Installation-digital-ocean/ which looks like a more step-by-step guide.
<djeis>Keep in mind the bootloader and file-system stuff in that config.scm are specific for a digitalocean droplet ofc.
<horizoninnovatio>Yeah, thanks. Maybe a good stating point.
<horizoninnovatio>Thanks for your help. Off out now with Mr7 to the park 😂
<djeis>\o
<vagrantc>anyone else able to "guix build guix" on aarch64 with b80ca672de936a76368de6e6ea0b28505e74d420 ?
<vagrantc>i keep getting test suite failures
<vagrantc>this is a tricky one to bisect...
<the_tubular>Talking about podman, can someone find the commit in which podman was updated to 3.4.4 ?
<the_tubular>I usually check new commits everyday and I must have missed it
<horizoninnovatio><horizoninnovatio> "I'm going to try without /boot/..." <- Removing /boot/efi and partition from config.scm worked. I can now boot in guix desktop - yay!
<horizoninnovatio><djeis> "Keep in mind the bootloader..." <- This is interesting as the old /etc/ needs to be replaced and some info moved across. The rest of the file system stays the same.
<littlebobeep>horizoninnovatio: Any reason you are choosing to start on a foreign distro if you want full Guix System?
<luke-jr>is it normal for `guix pull` to build a bunch of stuff?
<luke-jr>or is there a way I can defer that for system reconfigure?
<apteryx>luke-jr: it depends on the availability of substitutes, but yes, it's normal
<horizoninnovatio><littlebobeep> "horizoninnovations: Any reason..." <- Just seeing what can be done. Also, as another option for an existing system to move away from systemd without having to "re-install & configure" everything like new. I like to explore what the possibilities are. Also looking to put on the #pinephone to take it mobile.
<horizoninnovatio>So now guix is installed with xfce. However, usb mem sticks are not mounted, gvfs is installed but thunar is not "seing" it. Any solutions please
<horizoninnovatio>* solutions please? Update, can mount manually but not automatically.
<AwesomeAdam54321>horizoninnovatio: Do you use polkit for authentication?
<horizoninnovatio>AwesomeAdam54321: good question, not at far as I know!
<horizoninnovatio>* I know! Umm, would polkit stop appimages from working?
<AwesomeAdam54321>You would need an automounter if you don't already have one, like udiskie for example
<AwesomeAdam54321>horizoninnovatio: Maybe, polkit is meant to provide the system privileges to programs when needed, and not running the whole program as root
<luke-jr>apteryx: can I defer it to reconfigure?
<AwesomeAdam54321>horizoninnovatio, I don't think polkit would be a problem though
<horizoninnovatio>AwesomeAdam54321: There's nothing in the file specifically for auto mounting
<horizoninnovatio>AwesomeAdam54321: thunar -volman is meant to take care of mounting.
<allana>Hi #guix! Curious, are there any openconnect users here? I'm thinking of creating a guix home service that will connect me to a VPN with split tunnel. This could be a little complicated. I'd would be thrilled to know if there is anyone doing something similar.
<allana>I have hit my first roadblock. I tried adding openconnect as a setuid-program, which works as expected, but openconnect also calls a bash script (vpnc-script) that fails to mkdir-p "/var/run/vpnc". I'm sure that I can work around that a number of ways, but I'm first curious to see if anyone else is doing something similar.
<allana>Also, I am using a secondary script "vpn-slice" which helps to configure the split tunnel, but unfortunately it doesn't quite work the way that I want it to. The split DNS part doesn't quite work for linux in its current state. There is a pull request on the project that *might* work that relies on resolvconf. I'm thinking of trying to make it work with dnsmasq. But this is unknown territory for me.
<ennoausberlin>Hi. I wonder, what is the best way to package a docker container, which is built outside from guix (newer tensorflow etc.). I can copy and run it manually, but would be nice to have it managed in a more guix way
<ennoausberlin>To put it into a store using a channel maybe
<the_tubular>How can I tell which package pulled which binary ?
<allana>ennoausberlin: That's curious, I'm usually using guix pack to create docker images, not the other way around. Is the purpose to get newer versions of packages that are available in a docker image from somewhere?
<the_tubular>I mean, to go the other way around would me that every packages in the docker container would be available in guix, after that it's just a manifest away
<the_tubular>Guix will never pull random binaries
<ennoausberlin>allana: It is mostly because of the outdated tensorflow. I do unterstand that packaging/bootstrapping tensorflow for guix is a pain. I finally decided to deploy all possible software with guix and put the problematic parts into an external docker container
<the_tubular>Well the main channel that is
<ennoausberlin>I am not an AI guy by any means. When my colleagues tell me, they need tensorflow 2.7 I have to accept it.
<ennoausberlin>I guess, I can somehow use some copy-build-system (don't know the correct name). I am just curious how other people manage this use case
<the_tubular>I use a container
<the_tubular>I also wish I could get rid of those
<ennoausberlin>the_tubular: Can you provide a specific example?
<the_tubular>Well I use podman
<ennoausberlin>the_tubular: I'll have a look. Never heard about podman before
<the_tubular>It's like docker, but made by redhat
<ennoausberlin>the_tubular: Thanks
<dc>allana i'd be interested to see what you do for split tunnel VPN. i've never done that but it it looks like the dnsmasq-configuration in guix would be sufficient.
<dc>i'm not sure how you would dynamically switch between configurations for services like dnsmasq without updating packages.
<jpoiret>allana: you could just add mkdir-p in an activation-service-type extension, that's usually where those things are done iirc
<sneek>wb apteryx
<allana>jpoiret: thanks!
<allana>dc: vpn-slice (which I have packaged but not added to guix yet) does the trick for split tunnel. But in its current form I need to specify all hostnames that I need to reach inside the VPN, which is not practical for me. On other platforms it supports setting "domains to query with vpn dns" and there exists a pull request on that project for linux support (via resolvconf) but there has not been any action on that for months now.
<civodul>Hello Guix!
<jpoiret>hullo
<dirtcastle>WesterWest[m]: I have installed guix on bootable ubuntu usb. I have created guix partition and mounted it to /mnt/guix. and then ran mkdir -p /mnt/guix/boot/efi inside the partition. there was /mnt/guix/lost+found on it already.
<dirtcastle>what configuration should I do to correctly install bootloader on boot.?
<WesterWest[m]>dirtcastle: iirc lost+found is created automatically for every block device with filesystem
<jpoiret>WesterWest[m]: not for my BTRFS at least
<dirtcastle>yes. I created a ext4 fs
<dirtcastle>do I need btrfs?
<jpoiret>no no :p
<dirtcastle>okok
<WesterWest[m]>jpoiret: yeah it is more of a ext thing
<jpoiret>i was just mentioning that it depends on the fs
<jpoiret>dirtcastle: well, you'd need to mount the efi partition to /mnt/guix/boot/efi, possibly by bind-mounting
<jpoiret>other than that, the default efi examples of the manual should suffice
<WesterWest[m]>dirtcastle: the default configuration is almost exactly what you need
<dirtcastle>through fstab? jpoiret
<jpoiret>no need to modify your current fstab
<WesterWest[m]>guix system init creates fstab from your configuration automatically
<WesterWest[m]>for temporary mounta you use sudo mount
<dirtcastle>ok? but efi should be installed on fat32 right? so I need to make that?
<jpoiret>you'd just need to know which device corresponds to the efi partition and mount it there
<jpoiret>oh, my bad, you're on a bootable ubuntu usb...
<WesterWest[m]>dirtcastle: if you already had the efi partition then you don’t
<jpoiret>then yes, you'd have to create a new partition for it, fat formatted
<WesterWest[m]>aka if it is the ubuntu disk
<jpoiret>if you had another OS installed on that computer it's likely you already have an efi partition though, you can re-use that one
<WesterWest[m]>if you format an existing efi partition it will destroy the system, but if it is a new disk, then you can go ahead
<allana>Hi everyone. I am considering joining for the Guix meetup in September in Paris, but would love more information. It would be great to see some more details, especially something like a count of attendees/who is planning to go.
<allana>Is there a place where information is being collected? I have only seen what is on the mailing list.
<WesterWest[m]>ennoausberlin: i had some issues with podman, so your mileage may vary, docker is working for me, only thing is that you mustn’t forget to start the docker service
<dirtcastle>jpoiret: but my plan is to uninstall ubuntu. so I made a new partition. WesterWest[m] . should I mount the new fat32 to /boot or /boot/efi?
<dirtcastle>jpoiret: but my plan is to uninstall ubuntu. so I made a new partition. WesterWest[m] . should I mount the new fat32 to /boot or /boot/efi?
<dirtcastle>sry idk if the msg went through.
<dirtcastle>sry if I sent twice
<civodul>allana: hi! we'll publish more information and a web page soonish, hopefully--stay tuned!
<dirtcastle>what is bind Mounting? i just ran mount on it? jpoiret
<jpoiret>no need to bind mount, i was mistaken
<jpoiret>allana, civodul: i haven't answered yet because i don't know if i'll be in france then, but if so i'll definitely go as well :)
<jpoiret>dirtcastle: i'd say mount it to /boot/efi, that's the usual setup
<dirtcastle>jpoiret: did that exactly
<civodul>jpoiret: would be great to meet in person!
***dustfing` is now known as dustfinger`
***dustfinger` is now known as dustfinger
<f1refly>dirtcastle: bind mounting mounts a part of the file system somewhere else, in that process mount options can still be applied. that way you can substitute uids/gids, make the contents ro for a specific purpose etc etc
<f1refly>you can check the "bind mount operation" part of the mount manpage if you want, it's really nifty
<WesterWest[m]>dirtcastle: /boot/efi
<zacque>Is Emacs 28.1 available on Guix already?
<zacque>If not, does anyone know where to get it?
<zamfofex>zacque: I think you should be able to run ‘guix time-machine --branch=emacs-28 -- guix install emacs’ to install it.
<zamfofex>Or, err… ‘guix time-machine --branch=emacs-28 -- install emacs’, I think.
<dirtcastle>f1refly: noted.
<dirtcastle>i can't install guix of persistent bootable iso. is this normal
<dirtcastle>usb*
<dirtcastle>ubuntu usb
<dirtcastle>mv: cannot stat '/var/guix' : input/output error
<dirtcastle>ok. the problem was with my usb.
<attila_lendvai>could it be that the unpack phase for zip files changes the working directory?! i add a phase before 'unpack and after, and pwd prints a different dir (one of the subdirs).
<attila_lendvai>but how, when extraction is done using (invoke "unzip"), i.e. in a subshell?
<attila_lendvai>a (chdir "../") in an add-afte 'unpack fixes the symptom
<attila_lendvai>err, wrong. invoke uses system*, i.e. no shell is involved. could it be that unzip's behavior changes the cwd?
<zimoun>civodul: about merge train, one difficulty is that we do not have a single green/red but one commit implies many derivations which can green or red. For instance, how to consider https://ci.guix.gnu.org/eval/283615 where 134 Succeeded abd 112 Failed. Is it green or red in the context of merge train?
<zimoun>for reference https://toot.aquilenet.fr/@civodul/108245799863530769 :-)
<attila_lendvai>yep, replacing the unpack phase with (system (string-append unzip ...)) fixes the issue
*attila_lendvai facepalms
<attila_lendvai>the unpack phase has a horrendously wrong DWIM: (and=> (first-subdirectory ".") chdir)
<zacque>zamfofex: Can I know why do I need `guix time-machine --branch=emacs-28` instead of `guix install emacs-28`? Can the package installed this way be upgraded normally to emacs-28.2, 28.3, 29.0, ...?
<zamfofex>zacque: It is not currently merged into the ‘master’ branch yet, so you have to use the ‘emacs-28’ branch.
<civodul>zimoun: hi! yeah, you're right, it's more nuanced than just green/red
<civodul>so, tricky
<civodul>but i find it inspiring
<civodul>clearly, if the core-updates/staging/master rule could be automated, that'd be wonderful
<zacque>zamfofex: I see, is there a way to keep track of the progress? I guess I'll wait until it's merged into the 'master' branch
<zacque>Ah, this one should be it: https://issues.guix.gnu.org/54829
<zimoun>civodul: yeah, somehow we need more policies as this set of package must be green for that arch, as still many grafted packages, etc.
<cbaines>the Guix Data Service should be able to provide enough insight in to changes on a branch and comparison between branches to apply policies
<zimoun>cbaines: ah cool!
<cbaines>as part of trying to get package builds for patches happening again, I did setup https://data.qa.guix.gnu.org/ (just a DNS change from data.guix-patches.cbaines.net) which seems to have been doing OK keeping up with branches
<zimoun>cool!
<zimoun>instead of series-xxxx, would be possible to get the debbugs number?
<cbaines>it's possible to relate the branches created for patch series back to a debbugs bug, but it's a many to one relationship, as you can send many patch sets to a single bug
<zimoun>yeah, I agree. That’s just from https://data.qa.guix.gnu.org/ I need to click to https://data.qa.guix.gnu.org/repository/1/branch/series-11529 then click to https://data.qa.guix.gnu.org/revision/2ed4828915e0a9afc87b22d71453f67fcfc57505 then click to https://git.guix-patches.cbaines.net/guix-patches/commit/?id=2ed4828915e0a9afc87b22d71453f67fcfc57505 then check with my emails
<zimoun>it could be neat to have the debbug number on https://data.qa.guix.gnu.org/
<zimoun>maybe another field, as «name debbug date commit status»
<civodul>cbaines: i had completely overlooked that data.qa.guix.gnu.org was up and running, sweet!
<civodul>given that Arun has been looking at mumi lately, i think we should check how to "bridge" the two
<cbaines>civodul, the change was just a DNS one, but one I wanted to make before reconfiguing the build nodes to also fetch derivations from data.qa.guix.gnu.org
<civodul>ah alright
<cbaines>civodul, indeed, I want to get back to working on the patches stuff, just need to make some time
<civodul>"bridging" could be just adding a link or a badge on mumi pages
<civodul>that would already help a lot
<civodul>cbaines: BTW, did you already have code to extract patches from emails/mbox data?
<civodul>or is that in Patchwork?...
<cbaines>that's in Patchwork
<civodul>ok
<civodul>the other day i started working on a "guix review" command and then wondered whether some of that code was already available
<civodul>maybe mumi has relevant helpers too
<cbaines>I think there's something in Mumi, but it's a hard problem, as at some level, you're trying to parse free text
<civodul>yeah
<civodul>Patchwork does a great job at it
<zimoun>civodul: b4 from Linux kernel folks is doing a good job for extracting the patches from the thread. And public-inbox (again from Linux kernel folks) is a great tool (see an instance at https://yhetil.org/guix) and lei+b4 is “guix review” IMHO.
<f1refly>I have an issue with thunar (or any other compenent of xfce4) not saving settings on my de-less setup. this thread (https://forums.freebsd.org/threads/thunar-not-saving-preferences.82428/) suggests that the '/var/lib/dbus/machine-id' file is required by xfce4 components to save their respective settings. Any idea how I can get that? As far as I know dbus is part of %desktop-services, so it should
<f1refly>have done its thing by now, shouldn't it?
<zimoun>civodul: Well, instead of reinventing the wheel, since extract patches from a thread is a difficult task for the general case, it appears to me better to rely on external tools already doing a good job.
<jpoiret>cbaines: what would be the most promising interface, the patchwork instance at https://patches.guix-patches.cbaines.net/project/guix-patches/list/ or data.qa.guix.gnu.org?
<jpoiret>it seems to me that the patchwork instance already does a lot of nice things
<cbaines>next, I'm trying to focus on getting builds for packages happening, and then getting that information in to useful places
<cbaines>probably Patchwork and Mumi, plus anywhere else
<cbaines>there's already linting information that is available just from data.qa.guix.gnu.org
<mekeor[m]>what's the container technology that "guix shell -- container" uses?
<jpoiret>mekeor[m]: plain linux namespaces
<jpoiret>the same as other "container technologies"
<mekeor[m]>cool, interesting, wow
<jpoiret>note that it's also how the build daemon isolates itself
<mekeor[m]>would be nice if there was a function "isolate" so that you can write (operating-system (services (isolate (service foo-service-type)))) or doesn't that make sense?
<mekeor[m]>i'm not entirely sure what those linux namespaces do
<AwesomeAdam54321>mekeor[m]: You would need to specify what would be provided in the namespace, unless there's a default
<mekeor[m]>sounds good :)
<apteryx>if others are annoyed that siging in gitlab fails from GNU IceCat, please consider liking or commenting that you can reproduce here: https://gitlab.com/gitlab-org/gitlab/-/issues/345328
<apteryx>or investigating the issue further to provide useful insights :-)
<f1refly>so the issue was that I didn't install the xfce-desktop-environment itself, now that the system is configured to run xfce4 thunar works as expected.
<f1refly>however, because I now have xfce installed, slim will offer me to log in with xfce instead of my preferred session. the slim configuration has an option to set a list of sessions slim should make available, but the documentation says it isn't modifyable from guix's configuration system.
<civodul>zimoun: b4 + public-inbox make a lot of sense, thanks!
<civodul>we even have a package for that
<luke-jr>failed to compute the derivation for Guix (version: "b80ca672de936a76368de6e6ea0b28505e74d420"; system: "x86_64-linux";
<zimoun>civodul: public-inbox at 1.6 and the upgrade to 1.7 (containing lei) seems problematic; see message https://yhetil.org/guix/877d70wtd7.fsf@kolabnow.com
<jpoiret>luke-jr: does the error message mention &nar-error or something about substitutes? sometimes it's just due to transient network issues, and retrying may simply work
<zimoun>civodul: and a version of b4 is returned by “guix show b4” :-)
<f1refly>when I inspect the slim-configuration object, it says there is supposed to be a 'extra-config' field that is empty by default, but when I try to use it guix tells me it's an extraneous field initializer, what's up with that?
<zimoun>civodul: you could be interested by these recent threads https://yhetil.org/guix/87h76l8yp5.fsf@systemreboot.net/#r https://yhetil.org/guix/87bkwf956h.fsf@gmail.com/#r
<luke-jr>jpoiret: I use --no-substitutes
<luke-jr>don't see &nar-error anywhere
<civodul>zimoun: perfectly à propos, thanks :-)
<civodul>i had started writing code that all the messages corresponding to a bug number, with one file per message
<civodul>that's the easy part
<zimoun>civodul: the tandem public-inbox+lei+b4 is great because you can subscribe (and apply) only the patches you are interested in. For instance, only the R pacjages in (gnu packages bioinformatics) of the last 3 months https://yhetil.org/guix/?q=dfn%3Agnu%2Fpackages%2Fbioinformatics+AND+b%3Ar-build-system+AND+rt%3A3.months.ago..
<zimoun>having a kind of “guix review” extension allowing to glue the various things, it would help for «incentives for review» and for «formalizing team»
<civodul>zimoun: yeah, i agree
<civodul>i was thinking that anything that lowers the barrier to review, even just a little bit, would be an improvement
<civodul>there's been a drop in review activity lately and i thought that alone was a good incentive to do something about it :-)
<apteryx>gitlab sign in issue resolved by changing the user-agent... is this 2022?
<f1refly>in the future, everything runs chrome
<apteryx>this works: 'Mozilla/5.0 (X11; Linux x86_64; rv:83.0) Gecko/20100101 Firefox/83.0' (I used the User-Agent Switcher extension - https://gitlab.com/ntninja/user-agent-switcher) to switch the user agent
<apteryx>but this (the default in our current GNU IceCat): "Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101 Firefox/91.0" doesn't
<apteryx>so the cloudfare crap used by gitlab is filtering based on browser version... ahem.
<muradm>hello guix! are you down for me or there is a common problem?
<luke-jr> 1) Failure: refs::revparse::date [/tmp/guix-build-libgit2-1.3.0.drv-0/source/tests/refs/revparse.c:31]
<attila_lendvai>is there something like wrap-program but instead of env vars, it adds command line args?
<civodul>apteryx: i thought we had a user-agent that read "Windows" etc., to be in a large anonymity set
<civodul>maybe that was with the older versions of IceCat
<apteryx>that's no longer the case
<apteryx>it was removed at some point in IceCat due to not actually helping (much) with the fingerprinting score
<AwesomeAdam54321>Would having GNU/Linux in the user agent also not help?
<apteryx>it would not help in hiding in the mass, no :-)
<apteryx>but if you'd like to propose a change to GNU/Linux you could do it to the IceCat project, they may consider it, given they've already moved away form Windows user agent
<jpoiret>resisting fingerprinting goes way beyond modifying UA, generally making any manual modification singles you out anyways
<alanz>Stupid question: does "guix pull" have to clone the entire repo? It is pretty slow, especially the first time.
<roptat>alanz, yes it has to, at least from the last verified commit it knows about, which I think is pretty old the first time
<alanz>ok. Just checking, as I set up guix on a new debian machine.
<f1refly>I tried defining a package because I wanted to try developing with guix environmnet -l, but ruby won't find the defined dependency. Am I missing something? http://ix.io/3X31
<jpoiret>f1refly: aren't you missing ruby in the inputs?
<f1refly>it's included by (guix packages ruby)
<f1refly>s/guix/gnu
<f1refly>oh i think i set the gem path wrong
<f1refly>every directory in GEM_PATH is empty. hm. isn't this supposed to be set up right by guix environment?
<jpoiret>f1refly: no but your package should have ruby itself in its input
<jpoiret>inputs *
<f1refly>now I get a warning that ruby is imported from both licenses and ruby, but otherwise it works. neat!
<f1refly>..is that warning bad?
<roptat>the warning is fine, but you can get rid of it by specifying what you want to import from licenses, or specifying a prefix like most modules do
<roptat>like #:use-module ((guix licenses) #:prefix license:)
<roptat>now all licenses need to be prefixed, so it's no longer ruby but license:ruby (for the licenses), the package is still ruby
<f1refly>makes sense
<lechner>vivien: Hi, thanks again for your files. why did you use dkimproxy instead of opensmtpd-filter-dkimsign, please?
<akonai>Since Shepherd is a Guile program, is it possible to directly invoke a herd action from a gexp without externally calling the herd program?
<drakonis>theoretically it should be possible
<akonai>I think I figured it out yeah
<akonai>seems like you can do it via the (shepherd comm) interface
<the_tubular>Someone yesterday shared a link with a few emacs packages that aren't available in guix proper.
<the_tubular>I lost the link :/
<the_tubular>Found it : https://github.com/babariviere/guix-emacs/
<the_tubular>I'll try and test a few
<vivien>lechner, I wasn’t aware it existed
<lilyp>the biggest sadness regarding emacs is that we're still lagging behind on emacs 28.1
<podiki[m]>the_tubular: oh that looks cool
<pashencija[m]>Any ideas about this error? I'm stuck https://gist.githubusercontent.com/shlyakpavel/3ac9a2dbcd84d747c486590466588b36/raw/7642295ceaadcc1efff61c6fa7c74dc0ec305756/qmm2sy120igyh485c0d3bhclym5k4i-guix-1.3.0-25.c1719a0.drv
<pashencija[m]>guix/cpu.scm:94:2: In procedure cpu->gcc-architecture:
<pashencija[m]>aarch64 host, aarch64 target
<pashencija[m]> * Any ideas about this error? I'm stuck
<pashencija[m]>In procedure struct-vtable: Wrong type argument in position 1 (expecting struct): #f
<pashencija[m]> https://gist.githubusercontent.com/shlyakpavel/3ac9a2dbcd84d747c486590466588b36/raw/7642295ceaadcc1efff61c6fa7c74dc0ec305756/qmm2sy120igyh485c0d3bhclym5k4i-guix-1.3.0-25.c1719a0.drv
<jgibbons[m]>Is there a reason mono 4.4.1.0 is the latest version in guix? Did something unacceptable happen to the license or something as of version 5?
<civodul>pashencija[m]: hi! what command did you type?
<civodul>jgibbons[m]: i don't think so, it's probably just that the work needs to be done
<civodul>you might want to check https://issues.guix.gnu.org in case there are patches floating around
<the_tubular>podiki[m] I'll have to test it. But yeah, I think it's a good place to start.
<rekado>do we have tools to show the impact of upgrading a rust package?
<rekado>“guix refresh -l” doesn’t help
<civodul>rekado: hey! no, unfortunately Rust packages are out of control
<f1refly>i finally found out why pulse was eating so much cpu (and battery)
<f1refly>there's an issue with firefox where it launches its tts engine when a website announces that it may us it, then firefox continues to feed silence into pulseaudio for as long as they both run
<f1refly>can be turned off with a key in about:config, that's a whopping 10W of power consumption gone
<civodul>jgibbons[m]: for mono, i found this: https://issues.guix.gnu.org/52592
<apteryx>wicd looks pretty much abandoned
<apteryx>and debian has dropped it
<apteryx>I guess I'll stop trying to build the latest 3 years old commit on python 3
*jonsger[m] used wicd back when he was younger :)
<z4kz>I remember wicd
<sneek>z4kz, you have 1 message!
<sneek>z4kz, jpoiret says: the build farm (berlin) uses guix iirc
<z4kz>thanks jpoiret
<lechner>Hi, is guix gc too aggressive?