IRC channel logs


back to list of logs

***wehlutyk is now known as wehlutyk_
***wehlutyk_ is now known as wehlutyk
<mbakke>cbaines: glad patchwork is still working for you :-) I only ran the system test.
<PotentialUser-12>Hello, can someone please try and build mailutils on i686 and see if it works
<sneek>Welcome back PotentialUser-12, you have 1 message!
<sneek>PotentialUser-12, nckx says: Typo: guix → gnu.
<nckx>PotentialUser-12: Sorry about that, sneekbot doesn't understand the numbers good. I can build mailutils using --system=i686-linux, that is using qemu binfmt emulation. Test suite passes. I don't have any old i686 hardware. What happens there?
<nckx>This is mailutils 3.10 from Guix c9a17d8f (9 October).
<PotentialUser-12>Thanks. Wanted to switch back to this distro, mailutils frustrated me when installing Emacs and made me quit
<nckx>Welcome back. I'm sure we have plenty of broken packages to keep things interesting.
<PotentialUser-12>But shouldn't the ci report broken packages thus avoiding putting those packages up
<PotentialUser-12>Or how does it work?
<nckx>We don't have CI like that. Master is master, and it's what we push to and what others pull. No robots inbetween.
<nckx>People are expected to test their commits, of course... Still, accidents happen.
<PotentialUser-12>I see, is it possible to just get security updates only from the stable snapshot
<PotentialUser-12>and avoid new broken packages
<zimoun>PotentialUser-12: You can try ‘guix weather package’ before ‘guix install package’ to see if the substitute is there. If not, the package is probably broken, well you can check at for example. (Package python-umap-learn)
<PotentialUser-12>Thanks for the tip
<zimoun>and no, you cannot only pull security updates. AFAIU.
<nckx>You could maintain your own security branch, keep an eye on guix-commits, and cherry-pick only those patches from Guix master that you consider vital.
<zimoun>when you pull, you pull guix/ (new features) *and* gnu/ (new packages or services or security updates). The both are in the same tree.
<nckx>(If you think ‘well that's a jolly lot of work’, you're right, and that's why nobody else has volunteered either.)
<zimoun>nckx: :-)
<PotentialUser-12>If guix weather can somehow predict if a package is broken, can it be configured to run before guix pull and filter things out or am I overreaching
<zimoun>guix weather informs you if the substitute is available on the build farm, that’s all
<nckx>PotentialUser-12: You're not overreaching in principle: it's possible, but the code/API to know whether a package failed to build, or simply hasn't finished, does not yet exist.
<zimoun>PotentialUser-12: somehow it is up to the user. You could do: ‘guix weather --display-missing -m manifest.scm && guix pull’ where manifest.scm tracks the packages you care.
<nckx>We should eventually get to a point where no broken commits are pushed, so ideally no user-side hacks would be necessary.
<nckx>C'est TODO, as the French say.
*nckx → 😴 soon, good night all.
<zimoun>nckx: good night
<PotentialUser-12>Given guix current size is it honestly hackable to someone just starting out or are users at the mercy of the core developers like all the other distros?
<zimoun>PotentialUser-12: we know which packages are failing. The point is: you cannot exclude them when you ‘guix pull’ because all live in the same Git tree.
<PotentialUser-12>nckx: nyt & thanks
***catonano_ is now known as catonano
*apteryx continues mozjs packaging adventure
<apteryx>rather, adventures
<apteryx>has someone ever seen this cargo error when packaging rust things? error: checksum for `pkg-config v0.3.9` changed between lock files ?
<apteryx>these are while building crates bundled along mozjs; the patch phase of the gnu-build-system caused many hashes to change so I had to use the (guix build cargo-utils) `generate-all-checksums' to fix that, and now the above error
<apteryx>oh, the cargo-build-system suggests I can simply delete those apparently problematic Cargo.lock files.
<pkill9>it would be neat to have a desktop environment designed around guix, so that all the settings work with guix's config file
<apteryx>perhaps more like a graphical Guix System configurator?
<pkill9>yea pretty much, but as part of the desktop environment, so part of the control panel or whatever where you have all settings, from the guix config stuff like users, down to things like desktop background
<apteryx>these things are more *user* settings rather than *system* settings, so in its current form, wouldn't be suitable for the Guix config.scm file (which is system-centric).
<Gooberpatrol66>Does (gnu services nix) allow you to install nix packages on guix?
<apteryx>it'd be nice to extend configuration to allow user configs as well
<apteryx>Gooberpatrol66: I think so, although I've never tested it!
<kozo[m]>apteryx: thank you for the suggestion of the etc service. I was able to create a deploy that copies the .scm to the remote machine after deploying
***iyzsong-- is now known as iyzsong-w
<pkill9>yea i wouldn't put those in the ocnfig.scm
<apteryx>kozo[m]: eh, I don't know if it was me who suggested that, but I'm happy it worked out :-)
<kozo[m]>Gooberpatrol66: you need nix service and nix installed on profile
<pkill9>i just mean have a control panel that also handles system settings, which on gui would neccessitate reading and writing the guix config
<pkill9>i just like the idea of a distro like ubuntu, but based on guix
<pkill9>as guix has things that make it so much more reliable than those distros
<pkill9>e.g. functional package management
<pkill9>but these are just shower thoughts, ah well, can always dream
<apteryx>I think if we hook the guix package manager in the Gnome Apps (or whatever it's called) application and add some graphical tool to automate/make updating Guix easy for end users, we wouldn't be very far.
***iyzsong-- is now known as iyzsong-w
***sneek_ is now known as sneek
<Black-Mug>hi there
<Black-Mug>does guix use gnutls or openssl? and wonder why it doesnt use hardened openssl called libressl?
<apteryx>it's now called Guix System, FYI
<Black-Mug>ah cool, thx for telling me
<apteryx>there was a name change around version 1.0
<apteryx>there are many packages in Guix, and they use the SSL implementation of their choosing.
<apteryx> guix search . | recsel --count -e 'dependencies ~ "openssl"' -> 403
<apteryx>s/openssl/gnutls/ in the above command returns 115
<apteryx>libressl has only 4 dependents
<Black-Mug>yes, better to switch to libressl if we are talking about security in case of versus openssl
<Black-Mug>gnutls as well not security focus until now they are like proud to support tls 1.0 , 1.1 i dont know whats the good thing with that
<wleslie>guix also packages libsodium, which is even nicer than openssl
<apteryx>supporting it is one thing, using it by default another one
<wleslie>ah maybe libsodium doesn't expose an api for TLS
<apteryx>libsodium is not an implementation as TLS, AFAIK
<apteryx>of TLS*
*apteryx zzz
<wleslie>good GMT evening, apteryx
<wleslie>how do I run a build -j1 while developing?
***amfl_ is now known as amfl
<lafrenierejm>Running `herd load root ~/.config/shepherd/init.d/transmission.scm` as my normal, non-root user exits 1 after printing "error: connect: /run/user/1000/shepherd/socket: No such file or directory".
<lafrenierejm>What is the correct way to run per-user shepherd services?
<lafrenierejm>For reference, I'm attemting to follow along with
***amfl_ is now known as amfl
<sss2>hi all, is it possible to use multiple profiles with manifests and not loose everything installed during guix gc ?
***iyzsong-- is now known as iyzsong-w
<divoplade>Hi guix, I still cannot seem to run cuirass :( I have this error ( for this config ( I had received good help from Mathieu Othacehe by mail up to this point, but I still can't overcome this error.
<divoplade>Does someone know what's wrong?
<divoplade>There was someone private messaging me a config that worked in the wild but my client does not remember who that was and what was the discussion, sorry.
<xelxebar>hrm. bombadillo is failing with "permission denied" in install-license-files phase...
<civodul>Hello Guix!
<Black-Mug>where can someone ask for request features?
<janneke>hello civodul!
***nikotine is now known as ping
<maav>hi, everybody, is this working?
<maav>this is my first time with erc :)
<wleslie>greetings, maav!
<xelxebar>maav: Looks like you're good!
<maav>thanks, wleslie and xelxebar
<maav>was pretty easy then, let's see if the first issue of the day it's the same
<maav>i'm gonna check the installer language selection
<civodul>hi maav!
<maav>hi civodul! :)
<civodul>you must be the one i think you are!
<maav>hehehe, surely, you too
<civodul>great you're looking into the installer
<maav>i haven't answered yet because I was too busy this week, but now it's weekend for me
<maav>so i'll take a look
<Brendan[m]2>i never got a response about the fact the installer fails for me at installing grub
<civodul>Brendan[m]2: oh, did you open a bug or was it on guix-devel?
<maav>that step is later, but i'll take a look to... but i cannot grant a response ;)
<Brendan[m]2>civodul its at the bottom of the same email you read where i commented on the translation
<Brendan[m]2>i get that error on two different old laptops. i think they use BIOS and maybe it assumes its got some more modern stuff?
<civodul>oh sorry
<civodul>hmm dunno, it may be worth a bug report
<Brendan[m]2>no problem
<civodul>it's less likely to be lost
<Brendan[m]2>ok ill do that
<maav>btw, civodul, i already tought about a patch that i think will work, but i'm reproducing it first to be sure
<maav>Brendan[m]2 error seems to be related with partition table selection/detection
<Brendan[m]2>also, i think the manual should be scanned for 404 links before release. I just found one
<Brendan[m]2>there must already be such a tool for doing that?
<civodul>Brendan[m]2: ah, so for that we need to fix htmlxref.cnf, which is what tells where external manuals live
<civodul>and yes, there are probably tools
***iyzsong-- is now known as iyzsong-w
<cbaines>There's a package called linkchecker in Guix
<Brendan[m]2>linkchecker fails to build
<maav>civodul, good news, i found the issue :)
<maav>(about languages)
<maav>my patch wasn't complete, but was a good starting point
<maav>the thing is: setlocale is not conserved through exec calls
<civodul>maav: ah!
<maav>but gnu/installer.scm does set en_US.utf8 before exec-ing installer real
<civodul>ah so it was running in the C locale and thus LANGUAGE was ignored?
<maav>i also made a simple test case
<civodul>well done
<maav>i'm cleaning it up and i'll take a look into the partition table issue, just to get more confortable with the installer :)
<civodul>heh, awesome
<civodul>we haven't installed your Git access yet, have we?
<maav>one thing, shouldn't we remove then installer/installer-real?
<civodul>i think you have 3 people vouching for you so we're ready to go
<civodul>hmm lemme see
<maav>no, but thank you :)
<maav>i didn't get it at first, too early... (facepalm)
<civodul>there's a comment explaining why there's "installer-real" but it's unclear
<civodul>it's worth trying to remove that installer-real thing
<maav>btw, how could i generate a vm image for the installer?
<maav>i built the gexp directly for this, but i'd like to try the whole thing...
<maav>(to check it properly)
<civodul>to test the UI, i just do "./pre-inst-env guix system vm gnu/system/install.scm"
<maav>would be with installer-os from gnu/system/install.scm?
<civodul>and if you want to actually install from there, you need a full-blown image, so "disk-image" instead of "vm"
<maav>neat, thanks :)
<civodul>maybe "disk-image -t iso" actually or similar (not yet familiar with the new -t)
<maav>i've translated it, but i don't remember exactly, i belive it's iso9660
<maav>yup, confirmed, thanks again :)
<civodul>the nice thing when translating is that you get to know every option and every API :-)
<maav>i learned a lot about dns with that, yes :)
<civodul>ah ah
<rekado>I upgraded R (locally) to 4.0.3 and the weird irreproducibility problem still exists
<rekado>here’s the gist of it:
<rekado>0x20 becomes 0x0a in subsequent rebuilds
<rekado>that’s all
<rekado>space vs newline
<rekado>this has been puzzling me for years now
<rekado>any ideas?
<Brendan[m]2>so strange
*kmicu indents with tabs, aligns with spaces 😺
<civodul>rekado: did you find the code that generates those rds files?
<Brendan[m]2>guix build: error: some outputs of `/gnu/store/1imn2lsph7zvryybziqqrvciayr6fxy6-r-minimal-4.0.2.drv' are not valid, so checking is not possible
<Brendan[m]2>i notice that happens when the output is empty, for example, see guix build --check --no-grafts rust-clippy
<Brendan[m]2>actually maybe without --no-grafts
<civodul>Brendan[m]2: you can't use --check if you don't already have it in store
<Brendan[m]2>is that what the invalid outputs mean?
<Brendan[m]2>btw the output says "output ... differs" but it doesnt provide a link to the substitute version to compare
<divoplade>Scheme is such a beautiful language that it is impossible to indent it with tabs.
<civodul>Brendan[m]2: yes, "invalid outputs" means there's no r-minimal in the store, so nothing to check
<civodul>that message is not as clear as it could be...
<rekado>civodul: yes, is complicated…
<rekado>rds are the results of R object serialization
<rekado>and that looks pretty sane to me
<rekado>I also couldn’t find any special case handling that would result in *one* of the spaces to be turned into a newline
<rekado>actually, the file src/main/serialize.c does contain something special…
<rekado>case '\n': … /* GROSS HACK: … */
<rekado>but that’s for unserialize, so for reading the files, not writing them
<maav>i'm no r-hacker, but ub can poison everything with a gross enough hack
<maav>about languages... removing installer-real, didn't work, i'm trying to figure out why...
<rekado>there is, of course, a chance that serialize and unserialize are called repeatedly as the file is generated
<nefix>hello! How does dbus work and how am I supposed to make it work in my system? Isn't %desktop-services enough? Thanks! :)
<PurpleSym>I’m trying to package JASP, but it fails to run inside a `guix environment -C`. However, if I run it with strace, it works. How do I debug that?
<Brendan[m]2>lots of stuff crashed and all this in dmesg, wth:
<civodul>rekado: without looking any closer, the "GROSS HACK" seems like a likely culprit to me :-)
<civodul>maav: i think you can translate the "guix search text editor" example because "guix search" search in localized descriptions
<rekado>I’m wary of changing the GROSS HACK, though, because that would affect *all* serialized R files. I guess for this reproducibility problem I can get away with another ‘gross hack’ (lowercase, because it’s not so bad).
<civodul>ah, good then :-)
<rekado>the problem appears to be that line continuation in DESCRIPTION files is wacky in some cases; so I’ll patch the DESCRIPTION file to simply not have a line break. Maybe that fixes it.
<civodul>oh i see
<nefix>also, is there any reason why minikube isn't packaged?
<maav>civodul: that's true... but nobody is translating the packages (i just updated it because some translations were plainly wrong)... and that would mean another commitment
<civodul>yeah right
<maav>rekado: I've read GROSS HACK... yup, it seems quite bad indeed, but not on the ub side but on the 'whaaaat' side
<civodul>perhaps it would be useful to guide translators here, for example so they translate only important leaves
<maav>hmmm, yes, even though the content is already quite well ordered, as contributing is the first (so you know the steps to contribute) and the introduction later (you start knowing better the system) and the bootstrap process at the end :)
<civodul>you're referring to the manual, right? (i was referring to the translation of package descriptions)
<maav>oh, sorry, you're right
<Brendan[m]2>seems like something that should be collaborated amongst all other package managers with such info
<rekado>looks like I fixed it
<civodul>you can share this story with rb-general :-)
<jayspeer>question to users running guix system - what mpcie wifi + bluetooth card can you recommend?
<jayspeer>also can you point me to a channel better suited for such a question?
<jayspeer>I'm using the WB222 now (based on AR9462) - wifi works, but bt requires blobs (yikes!)
<wleslie>so I think I got my libgcc build confused. for some reason I have host=i386-unknown-capros and build=x86_64-pc-linux-gnu in the libgcc configure.
<wleslie>the gcc makefile passes build, host, and target, so maybe this is "correct" and I need to support 'host' being the target in libgcc/
<rndd>hi everyone! how i can build iso image from guix source?
<wleslie>alternatively if you happen to know a better place to ask wild cross-compiler questions, please let me know
<xelxebar>rndd: guix system disk-image --image-type=iso9660 foo.scm
<xelxebar>rndd: What are you wanting to use the image for?
<rndd>xelxebar: i just wanna know, can i build guix disk-image from source on gnu/linux machine that don't have guix installed. or i have to firstly install guix and then run guix system disk-image?
<xelxebar>rndd: Oh. That's a very different question. I'm sure it's possible in principle, but my guess is that it's way easier to simply install guix locally and build. It's ridiculously easy to nuke guix later if you don't want it.
<wleslie>you can run the guix command from checkout
<wleslie>it's a bit more complicated, though
<rndd>xelxebar, bytheway, how disk-images are assembled? always wanted to know
<rndd>i know that i can download disk-image, mount it locally, copy everything from it to another directory and then use disk utils to produce new image from directory
<rndd>but how first one created 0_o?
<wleslie>you don't have to run guix system to use guix
<xelxebar>wleslie: Nice! That idea had escaped me.
<xelxebar>rndd: Depends on the format. If you *really* want to know, the repo is there in all it's glory for you to look at :)
<rndd>xelxebar: \o/
<zimoun>roptat: website patch updated with civodul comments.
<wleslie>looks like the build is correct, but the host should be the same as build
<wleslie> has --host=${target_alias}
<wleslie>phew, that's an easy fix
<wleslie>wow, this is specified nearly 70 times
<rekado>jayspeer: I’m using a USB dongle for bluetooth (on one machine).
<mfg>I have a strange error currently. Yesterday I plugged in a new monitor and sway sent it into suspend mode while I was eating, back at the machine the monitor didn't want to resume operation. I tried to reboot but the machine didn't power off, so I cut the power after 30 minutes. Now the whole system freezes when I try to launch sway and a monitor is plugged in via Display port.
<mfg>dmesg suggests something nouveau/ACPI related, I have never seen those errors before, so i guess it's possible I corrupted something.
<mfg>How do I verify that for example the bootloader is not corrupted?
<rekado>mfg: Guix reinstalls the bootloader on every reconfigure
<rekado>mfg: so if you’re worried about the bootloader I’d just reconfigure once more to install it to disk again.
<mfg>rekado: okay I'm trying that out :)
<mfg>rekado: that doesn't make a difference so I guess I grub is fine. Will have to dig deeper
<mfg>directly after the bootloader, where I need to type in the password for my drive, I'm in guix's initrd environment, right?
<mfg>Can't read it atm to many artifacts ...
<divoplade>I suppose you already know it, but the install-license-files from the go build system introduced yesterday (I think) breaks a few packages, for instance docker-cli
<divoplade>I get the exception here:
<civodul>uh i started the Julia build ~2h ago and it's still going
<civodul>efraim: did you see the issue divoplade reports? ↑
<divoplade>(or runc)
<civodul>divoplade: if you have ideas on how to fix it, we could go ahead and push the fix
<civodul>otherwise we could revert until efraim is available
<civodul>it seems the source directory is incorrectly determined
<divoplade>civodul: Unfortunately I don't; I'm not familiar with anything go-related
***jonsger1 is now known as jonsger
<divoplade>That src/... directory is present after unpack
<divoplade>civodul: maybe the current working directory was "build/docker-linux-amd64", so it should be "../../src/"
<divoplade>Do you know a go package that is not related to docker?
<divoplade>go-ipfs works though
<civodul>divoplade: the install phase of docker-cli chdir's somewhere else
<civodul>so presumably we need to chdir back to the source
*civodul tries the simple fix
<divoplade>You mean the "chdir ("build")" line in the install phase replacement?
<divoplade>It's a different problem for runc then
<divoplade>Wait no
<divoplade>It has a "chdir" too
<civodul>the fix for docker-cli works
*civodul tries runc
<divoplade>umoci too
<divoplade>I'm just searching for "chdir" in virtualization.scm and trying to build all the packages ^^
<divoplade>skopeo too
<divoplade>I'm a little worried that it might be a problem for a huge number of go packages
<divoplade>Not only in virtualization.scm
<civodul>divoplade: using with-directory-excursion instead of chdir (what i'm doing) can't hurt
<civodul>but if we find that this is not enough, we can revert the offending patch
<divoplade>More precisely, it can only hurt really broken packages^^
<civodul>hmm what happened with build users on berlin?
<efraim>civodul, divoplade: Just signed on, I can take a look at it
<efraim>civodul: yeah, the julia test suite takes hours
<efraim>ah, I see. At the end of the install phase for docker-cli it changes to build and then never leaves
<civodul>efraim: awesome; i'll push the with-directory-excursion fixes i have and then you can check if there's more
<zimoun>efraim: you are build #44004 on Berlin to update Julia to 1.5.2?
<efraim>zimoun: I don't have access to berlin
<ani_>zimoun: I fixed the issues in git message, and run git format-patch
<ani_>it generated same file name
<ani_>so I can rename it to 0001-doc-Fix-typo.patch from this to v2-0001-doc-Fix-typo.patch and send it to
<ani_>not to
<ani_>isn't it?
<civodul>zimoun: i'm building Julia for that one, yes
<civodul>almost done!
<civodul>well, i hope :-)
<zimoun>ani_: yes, cool! Thanks
<zimoun>civodul: nice! v1.2 will be so much featurefull :-)
<zimoun>I have tried to reviewed the patch but then I remembered that building julia is like cross the ocean by plane, energyly speaking :-)
<civodul>i had forgotten it was *this* long
<divoplade>gnunet has a pretty long test suite too
<divoplade>But it does not use any CPU...
<divoplade>I'm starting to think the test suite is just a long sleep()
<divoplade>efraim: you also need to fix runc, umoci and skopeo
<divoplade>And maybe others
<zimoun>efraim: ah I thought you had :-)
<aitzkora>Hi! someone knows how could I deactivate parallem compilation when using gnu-build-system ?
*zimoun thinks it will be faster to first properly fix haunt/website and then build and rebuild the website :-)
<maav>aitzkora: doesn't -c/-M work properly?
<maav>one question about the language selection... because I've been driven mad until I noticed
<maav>should we apply the language selected from a clean environment? instead of deriving from the last choice, I mean
<aitzkora>maav : the author of the makefile did bad dependencies so i must compile in sequential
<maav>aitzkora: ahh, to enforce it at package level
<aitzkora>maav: yes! maybe a good #:make-flags will do the job ,
<ani_>So I wanted to change couple of commit messages in repository. Repository I cloned from git does have some modification which are not commited Now when I run "git rebase -i HEAD~3" I get a error saying that,
<ani_>error: cannot rebase: You have unstaged changes.
<ani_>error: Please commit or stash them.
<ani_>How shall I proceed?
<divoplade>ani: run git status to check whether the unstaged changes are important or not
<divoplade>ani: are you trying to modify HEAD~3 with the unstaged changes?
<ani_>okay but how would I know, whether the changes made are important message? What does .go file mean? Does it belongs to golang?
<divoplade>ani: .go is either Go source code, or Guile binary objects
<ani_>I guess, I will troubleshoot first.
<divoplade>ani: in any case, if you have staged your changes, you can do "git commit --fixup THE_COMMIT_ID_YOU_WANT_TO_FIX"
<mfg>also you could just stash the changes debase and stash poo them again, you might have to resolve conflicts in this case
<mfg>amazing I meant rebase and pop :D
<divoplade>(stagint changes means "git add ..."
<ani_>staging I know
<civodul>efraim, divoplade: i pushed the fixes (and that overlapped with what efraim pushed minutes earlier :-))
<ani_>but shall I add those files which are not modified by me.
<divoplade>If you did not modify them, don't add them
<efraim>civodul: thanks :)
<civodul>aitzkora: there's #:parallel-build? and #:parallel-tests? that you can set to #false
<divoplade>If they are not source code, and git detects them as changed, it is suspicious
<divoplade>ani: are you working on a guile project?
<aitzkora>civodul: dankon!
<civodul>nedankinde! :-)
<ani_>divoplade: no I aded cran package to cran.scm I am on guix project
<divoplade>So you have .go files in a R project?
<divoplade>That's quite surprising ^^
<divoplade>OK so you are writing guix package recipes
<divoplade>Then it's suspicious that you have .go files
<divoplade>If "git status" gives you a list of .scm files that you have modified as "Changes not staged", then you can add them with "git add ..."
***zimoun` is now known as zimoun
<divoplade>If the list of files is in "Changes to be committed", you can do git commit --fixup COMMIT, where COMMIT is the commit ID you want to fix
<ani_>divoplade: I did use git keyring to
<divoplade>(use git log)
<ani_>now everything is okay
<ani_>once I use git stash
<ani_>Now I have run git rebase -i without error
<ani_>edited three last commits ready to send the patch.
<divoplade>ani: wonderful
<ani_>lets see, my first actual patch for guix. technical I mean. :)
<ani_>thank you.
<civodul>maav: ¡ahora puedes acceder al repo! (etapa n°3 :-))
<civodul>¡bienvenido compañero! :-)
<civodul>i still need to send you and co-maintainers a signed message
<civodul>but for that i need to fix my Gnus + GPG setup, which broke with Emacs 27
<divoplade>In any way, if your git repository badly damaged, you can always use the XKCD trick:
<ani_>what I learned at guix IRC or contributing to it is importance of Manual Reading. :)
<maav>merci beaucoup, civodul :)
<ani_>divoplade: Sent it. :)
<divoplade>ani: which is a shame, because with guix we could expect Automatic reading!
<divoplade>(sorry, I'm not good at making jokes in a foreign language)
<maav>I have the patch for the installer ready, and I have an idea to test for the partition issue (as it seems that the case gpt + bios is not really taken into account when the partions are deleted)
<maav>this evening/tomorrow morning I'll continue with that, now it's drums time B-)
<maav>have a nice time, everybody!
<vits-test>sneek: seen botsnack?
*vits-test that nick is a best hack of the year.
<ani_>divoplade: :D
<civodul>cbaines (or anyone web-savvy): i tweaked the nginx config to have an expiry on /static stuff, but apparently we receive an "Expires" header that contradicts "Cache-Control"
<civodul>"Cache-Control" does what i want, but "Expires" suggests immediate expiration
<civodul>IceCat thus ignores Cache-Control, AIUI
<mfg>where do I find exact differences between linix-linre and vanilla Linux? I haven't found a list or smth comparable... :(
<civodul>i'd check their web page for an overview
<mfg>hm I see, i guess I have to read deblob-check and the likes ...
<civodul>their release announcements often list newly deblobbed parts
<civodul>but i don't know if there's a complete list somewhere
<OriansJ>grand summary of guix source tarball availablity: 33GB of 85GB; 7,554 files out of the 74,881 attempted could be downloaded or only 10.088% of source tarballs listed in sources.json could be downloaded.
<sneek>OriansJ, you have 1 message!
<sneek>OriansJ, nckx says: <I am starting to wonder what percentage of guix developers don't use substitutes> I don't, for one. I doubt I'm alone (mhw, while no longer very active, never did). Why?
<zimoun>OriansJ: thanks for these numbers. Is it possible to have the list of the failing ones? Then I could investigate in sources.json, maybe issues are there…
<OriansJ>anyone have a suggestion on where to put a 88,581,812byte logfile
<divoplade>Make a torrent at this point!
<zimoun>88G right?
<zimoun>no 88M
<OriansJ>zimoun: correct megabytes of log; for downloading only 33GB of tarballs
<OriansJ>I could use nc to host the file on the server for a day
<zimoun>could be nice. Otherwise, you can use “split -b 10000000 | mail” (or something like that) and send mail privately?
<civodul>cbaines: actually Cache-Control and Expires are fine, but IceCat seems to ignore them (Chromium is fine)
<civodul>actually no, IceCat is fine too
<civodul>go figure
<civodul>browsing does feel faster now, but there's a psychological component to it
<roptat>10% sounds very low...
<mfg>civodul: thank you :)
<zimoun>what are the packages instead of the big texlive to produce pdf from TeX files?
<OriansJ>you can download the report from on port 7557 (it is a tarball)
<OriansJ>roptat: well that is the result of doing wget -ci ~/foo from the very server the report is available on
<zimoun>tarball named?
<OriansJ>nope, just hosting it with: while true; do netcat -l 7557 < report.tar; echo "there is a download" ; done
<roptat>so anyone downloading grabs a lock on the download, nice :)
<OriansJ>roptat: well if you want me to setup apache to serve it instead; I can
<OriansJ>it is just a 95,170,560byte file
<OriansJ>I could compress it with xz if someone wants it smaller for faster downloads
<zimoun>OriansJ: yes please, compress
<OriansJ>doing that now
<OriansJ>with basic xz compression: 23,450bytes
<roptat>OriansJ, looking at it, you have a bunch of failed: Network is unreachable.
<OriansJ>switching out the nc command to serve the xz file instead
<roptat>so I think you had network issues on your side, rather than the source disappeared
<roptat>are you using the guix system on that server?
<OriansJ>roptat: possibly, I am just reporting what happended from the server in question issuing the specificed commands
<OriansJ>standard ubuntu server
<roptat>mh... ok, weird
<roptat>I'm also seeing a lot of these and I was starting to wonder if that was related to Guix somehow
<OriansJ>the only changes I did was update the iptables rules to as follows:
<roptat>actually, there aren't that many errors like that
<roptat>what is sources.json exactly? is it the current set of sources, or all the history?
<civodul>roptat: the current set
<zimoun>roptat: generates every X hours when the website is rebuild.
<OriansJ>I can provide that file too if there is a desire to see it as well
<roptat>no it's fine :)
<roptat>I see a lot of File ‘YAML-1.30.tar.gz’ not modified on server. Omitting download. and similar, does that mean you already had the file, or does it skip download entirely?
<zimoun>this file is used by SWH to ingest the tarballs
<OriansJ>roptat: it means wget already obtained a file by that name
<OriansJ>there was a restart on day 2 to set a timeout of 30seconds
<OriansJ>and the shuf the order of the contents of the file (as previously it was sorted)
<OriansJ>as 48 hours only obtained 28% of the files (because servers had long timeouts)
<OriansJ>or more precisely only attempted 28% of the links thus far
<roptat>interesting, thank you!
<Zambonifofex>Once the file was compressed to 23 KB, you should have been able to use some kind of pre‐existent hosting.
<nly>OriansJ: I have a patch that allows me to use external my own sources for packages, maybe this is interesting to you
<nly>for example if i am building guix, will check out source from file:///archive/git/guix.git
<OriansJ>Zambonifofex: feel free to upload it any hosting site you think appropriate and share the link here. So I can turn off that server as it has generated us the useful information we needed.
<pineapples>Can anyone explain as to why Guix interprets this line of code as the root of the GNU store, not the package's store path? - (package (string-append (assoc-ref %build-inputs "package") "/"))
<pkill9>is it possible to start gnome from terminal with no display manager running?
<pkill9>i try but it fails
<pkill9>well, no display manager service added
<raingloom>hey, i'm lazy and don't know TeX, what package do i need to make pandoc produce PDF through LaTeX? it keeps complaining about lmodern.sty
<raingloom>...there should probably be a "texlive for pandoc" package...
<raingloom>hmm. texlive-lm might be it...
<raingloom>it wasn't texlive-lm :(
<jsoo>raingloom: if all else fails, can you use the giant catch-all texlive package?
<raingloom>i can. but i don't want to.
<OriansJ>pkill9: startx?
<pkill9>now i need to get startx to execute gnome-shell or whatever
<OriansJ>well traditionally it would be edit ~/.xinitrc to have exec gnome-shell
<OriansJ>or exec $1 if you wanted to use F1 to select the session with slim
<jsoo>raingloom: me too. I gave up having a minimal texlive set so long ago though
<pkill9>i can't get it to work
<pkill9>i have /bin/startx added by a service
<pkill9>i run sudo /bin/startx vt6 :1 with 'startxfce4' in my user's ~/.xinitrc
<pkill9>maybe needs to be added to root
<OriansJ>pkill9: root isn't require to run startx
<jsoo>pkill9: does startx have setgid?
<jsoo>it took me a long time to get startx working
<pkill9>it's not even running root's .xinitrc
<pkill9>jsoo: idk, I just added the startx service
<pkill9>dunno why it requires so much setup
<pkill9>oh hmm i dont have that service added
<OriansJ>pkill9: does xinit work?
<pkill9>so idk why i have /bin/startx
<pkill9>yea xinit from a package works
<pkill9>i did it a few days ago
<OriansJ>startx is just a front-end for xinit
<pkill9>the problem with xinit is it switches to the X, then ctrl+alt+f<number> doesn't work
<OriansJ>does something like: startx /usr/bin/i3 work for you?
<pkill9>i'll test
<OriansJ>If the binary takes arguments, they need to be quoted to be recognized as part of the first parameter of startx:
<pkill9>i ran /bin/startx startxfce4 and it said unrecognised argument
<pkill9>unrecognised option*
<OriansJ>If the binary takes arguments, they need to be quoted to be recognized as part of the first parameter of startx: eg startx "/usr/bin/application --key value"
<clone11>I'm trying to package a patched program, so I added a list of patches to it. The problem is one of the patches requires an additional .c & .h file. Is there a way to get guix to download them too?
<pkill9>wayyy i got gnome-shell running with `dbus-run-session -- gnome-shell --display-server --wayland`
<pkill9>argh it fails to exit, and it fails to run the settings control panel thing
<mfg>leoprikler: C-x C-c n ?!
<raghavgururajan>Hello Guix!
<pkill9>i like gnome, but it's quite monolithic
<pkill9>just seems like if things go wrong it's hard to fix
<pkill9>maybe Im generalising about something i don't understand
<divoplade>I guess the GNOME folks prefer to say their apps have a good desktop integration ^^
<pkill9>it's just things like, there not being a single entry point to start gnome
<pkill9>i can't just run it from terminal
<pkill9>it wants me to run it with dbus
<pkill9>then the control panel doesn't work
<pkill9>now it does run when i run from login manager
<bdju>trying to fix an adb udev issue and now I get a whole mess of errors on reconfigure. here's the output. and here's my config.scm:
<bdju>last changes were the modification of %desktop-services + udev stuff (copied from a config on another machine where this was working in the past)
<pkill9>oh well
<ss2>Is there a site that explains how package dependencies are resolved by calling service-types? Or where can I see that this and this service-type would install the necessary dependencies? I'm not so sure that I've understood this correctly yet.
<brettgilio>is there a way to force guix pull to operate over ipv4?
<divoplade>ss2, I've had difficulties trying to understand this, but I gave up so I don't have any problems anymore ^^
<roptat>disable ipv6?
<divoplade>ss2, basically there's the store monad... and my brain stops working at the word "monad"
<divoplade>After that there's GEXPs
<divoplade>All of this is in the manual though
<brettgilio>roptat: well duh lol. i was wondering if there was a flag for it like there is in git. but disabling at the system level is fine too
<divoplade>brettgilio, no this is not fine!!! I have a lot of trouble setting up my server just so that people with only IPv4 can access it, and I hope I can disable all this nonsense as soon as possible ^^
<brettgilio>i only found out today my isp doesnt support ipv6 dns
<brettgilio>which is just... dumb
<divoplade>My ISP does not even let me choose my own DNS... I think dumbness is pervasive in the field!
<morgansmith>Anyone having trouble building guix from source rn?
<rekado>morgansmith: what error do you see?
<morgansmith>some documentation stuff. I noticed I do have a good number of local changes so I'm stashing everything and doing a good git clean -fd
<divoplade>Are there .go files?
<morgansmith>wait, this likely has to do with /bin/sh being dash, not bash
<morgansmith>nope, still doesn't like me... hmmm...
<morgansmith>the texi files have nonexistent nodes... maybe I have to redo everything with /bin/sh == bash
<morgansmith>yep. it works now. maybe we should put a check in the build system to tell people to use bash
<morgansmith>or we could use POSIX shell. that would also be a good solution
<roptat>morgansmith, if you can fix it, go ahead :)
<morgansmith>no... I'm literally dying from school...
<morgansmith>on a related note, someone wanna take a half finished libguestfs package from me?
<morgansmith>like it builds, but the binaries don't validate and I dunno if it works. It took a lot of work to get it to that state. So much so that my use case actually disappeared
<morgansmith>Can I just post it to the debbugs? What do I label it? Not WIP because I'm not working on it...
<mfg>hm having things more POSIX so that it's not necessary to use bash would be awesome, but I don't know how much guix depends on bash... it's deep down in the dep graph I think...
<morgansmith>I mean that's fine. but if your shebang is sh, then use POSIX not bash.
<morgansmith>/bin/sh implies POSIX
<morgansmith>would we have to install a /bin/bash on guix system to change the shebangs though?
<leoprikler>pkill9: I think I get what you're saying in that debugging gnome-session is improbably complicated, but I'd like to compare that to debugging guix.scm ;)
<leoprikler>Like, you can write "perfectly sane" scheme, that results in a perfectly insane system description
<mfg>morgansmith: I see.
<mfg>I have git:send-email in a fuix environment but git send-email says send-email isn't a known script. Is there any configuration missing?
<bdju>Can someone look over my config.scm and tell me what error I made between line 67 and 72?
<divoplade>I had trouble with this because git send-email and git were not in the same profile
<mfg>hm okay
<leoprikler>bdju: uhm, wtf?
<rekado>bdju: modify-services returns a list of services
<rekado>on 48 you have append, which expects one or more lists
<rekado>but what follows is a list containing some services but also another list of services
<rekado>make sure that your list is a flat list of services and the error should disappear
<leoprikler>plus %desktop-services already contains services, so adding %base-services as tail is redundant
<apteryx>morgansmith: I`d just send it over to guix-devel
<leoprikler>it also already has elogind, udisks and dbus iirc
<leoprikler>so any configuration thereof should go into the modify-services part
<OriansJ>if you wanted to manually list all of the services; here is the minimal list:
<morgansmith>apteryx: rip, just sent it guix-patches
<bdju>oh heh I'm not using desktop-services
<bdju>I'm cherrypicking lines from the config of a past machine where I was using desktop services but removing gdm from it
<leoprikler>That's not what modify-services reads as tho
<leoprikler>apropos commenting out rottlog: Who here should clean up their /var/log the most? I have guix-daemon.log.29
<bdju> here's the old config I'm looking at which may clear up confusion, I had this %stripped-desktop-services thing which is maybe why I had to use modify
<apteryx>leoprikler, yeah, I saw this too. I think this means 29 weeks worth of concatenated logs IIRC.
<bdju>anyway I just need android udev stuff to work, not sure what I'm doing really but it worked on the other machine so I was hoping I could get it going pretty quick, but the configs may be too different
<leoprikler>bdju: that config makes sense tho, because %stripped-desktop-services is at the tail, as it should be
<bdju>yeah. I had gotten a lot of those lines from people here. I'm afraid I don't really grok guile at all
<rekado>bdju: this can be fixed
<bdju>I removed the modify stuff but now I get guix system: error: =>: bad use of '=>' syntactic keyword
<rekado>the “services” field wants to have a single flat list of service values
<rekado>show us the edited config
<roptat>my server is back online! apparently cable issue... :)
<bdju>edited config:
<bdju>I'm guessing the => thing was specific to the modify stuff that I'm no longer doing
<leoprikler>bdju: you still want to modify-services %desktop-services, you just want none of the "(append (list" nonsense (use cons*) and you want to only add services, that don't already exist in %desktop-services
<bdju>oh jeez
<bdju>well, I'm not using desktop-services, I wrote that by mistake
<bdju>I am just using base-services on this machine
<bdju>but do I want to modify that then, I guess?
<roptat>only if you modify stuff from %base-services
<bdju>and alright I can try to swap out the append list thing, but I swear that was the default generated by the installer
<leoprikler>oh, well, if you insist on basing this on %base-services, then no
<bdju>yeah, I don't use desktop-services on this machine because it gave me trouble in the past and it was simpler to not have it so I don't have to disable gdm and such
<bdju>I was blindly copying a bit too much
<leoprikler>yeah, the default config is sadly a bad example 9/10 times
<leoprikler>for udev-service removing the line "config =>" should do the trick
<leoprikler>anyway, /me → afk
<bdju>I feel like I'm getting somewhere but no it says there's an unbound variable and hints about a use-modules form without suggesting what it might be I'm missing
<bdju>ah wait, it gives a line number, and it's the line where I start the udev-service-type
<bdju>tried (gnu services udev) and that wasn't right
<bdju>I don't see a use-modules line for udev in my other config, but it's maybe just because it had desktop-services
<rekado>what’s the error?
<bdju>`no code for module (gnu services udev)`
<rekado>that’s correct
<rekado>there’s no such module
<rekado>why did you try (gnu services udev)?
<rekado>what did the error message say?
<bdju>/home/brad/dotfiles/guix/config.scm:67:6: error: config: unbound variable
<bdju>hint: Did you forget a `use-modules' form?
<bdju>tried it because the line matched up with the udev service and all other use-modules looked similar so I tried to work off of them
<rekado>bdju: the error says that “config” is undefined
<rekado>it’s nowhere defined in Guix; why do you use it?
<rekado>kinda hard to assist when we can’t see the current version of the file
<bdju>I'm not aware of using config
<bdju>I'll upload the file again in a sec
<bdju>ah, I see it now
<bdju>(inherit config)
<bdju> I just removed a bit but there's another thing further down that says "config"
<bdju>as for not being defined in guix, I don't know why my other use-modules are what they are either, to be fair
<bdju>ah I see now the manual had the version of this that modified %desktop-services, so I probably got it from there. I don't see an example for %base-services
<raghavgururajan>roptat: Any luck with #43575? :-)
<mfg>how does the pypi importer build the pypi-uri? I can't find it in the source... it seems to generate invalid URLs --- at least for the pyinstrument-cext pypi package