IRC channel logs

<stephen>per the info page, I can put build machines directly in the operating system definition -- and it built fine when I did, but then trying to call guix offload test it complains about /etc/machines.scm not existing -- is there a way to call test without passing a machines.scm file?

<stephen>Oh, I see, the rebuild created that file but the regular user has no access

<old>so I recently update my home configuration and now I get the following when trying to compile a buffer with geiser:

<old>Unknown meta command: geiser-load-file

<old>does that ring a bell to someone ?

<old>hmm I seem to only get that when working on system/repl/command.scm so perhaps that's an old issue

<stephen>Is there some easy trick to use a remote builder as a substituter? I'm going to have nearly identical packages across three systems that are all going to do the same slow pull.

<bavier>stephen: can you run `guix publish` on your builder?

<lechner>old: Is that in the ?

<lechner>old: Is that file in the Guix repo?

<stephen>bavier: what's that do? (apparently yes)

<stephen>ah! (reading the info)

<SurfingOnBytes>do I have to package `gmp-mpfr-sys` for my basic kalker packaging attempt, https://bpa.st/DUULU ? Or am I missing something from the manual

<SurfingOnBytes>I may be confusing nix's cargo install method with guix... :S

<old>lechner: nah in Guile

<apteryx>first time I see Unicode text printed when building stuff with ,build, that looks strange (building kmail with ,build: https://paste.guixotic.coop/repl-cyrillic-characters.html)

<apteryx>would anyone know how to fix plasma-workspace on the qt-team branch? I think if we get this one fixed KDE may be unblocked.

<apteryx>updating it is not an option unfortunately (larger piece of work)

<Heterochromium>apteryx: Did something break in a recent version of Plasma?

<apteryx>Heterochromium: yes, on the qt-team branch, with the Qt update to 6.12

<apteryx>I mean, with the current version of Plasma in Guix, which is probably not that "recent"

<apteryx>was it possible to use time-machine with an AGit-created PR?

<apteryx>does not look like, at least in a single command pointing to the remote repo

<cnx>apteryx, does the specifying the commit hash not work?

<olivial>is there an way to trigger build phases manually in a devshell? I often find myself wanting to do this when debugging packages, for example if unpack/patch-shebangs takes a long time and I tried a failing build with '-K'

<olivial>I figure I can write a helper for it without a ton of work, but that I'm probably not the only person who has wanted to do this :)

<csantosb>olivial: Are you aware of 9.1.4 Debugging Build Failures ?

<apteryx>good, fixed plasma on qt-team

<apteryx>magit is kind of broken for me, at least with ivy/counsel; e.g. push to a specific refspec never let me input what I used to be able to, it rejects it. Now merging a branch only proposes origin/master as the single candidate, it looks like I can't input a local branch.

<apteryx>anyone else?

<apteryx>seems to be ivy related; ivy-mode to disable it and then I could select a local branch in 'm' (merge).

<apteryx>wait what, I just rebased qt-team on master, and now qtbase has to be rebuilt

<sham1>You chould be able to accept typed in input instead of a suggestion even in ivy

<folaht>Hey there, I created a guix channel and when I try to `guix pull` my channel I get this error: guix repl: error: failed to connect to `/var/guix/daemon-socket/socket': No such file or directory

<untrusem>folaht: how are you pulling the channel

<untrusem>can you share your commands and error log in a pastebin

<folaht> https://paste.i2pd.xyz/?3a59d5a8048d9d4d#GYin3sSfVYpnXWcKQ6J4pdV19ZJEYDkFgMaehbWhdjAh

<untrusem>lol i2p mentioned

<folaht>i2p?

<dariqq>is your guix-daemon running?

<untrusem>i2p is a network protocol and i2pd is an daemon program

<untrusem>also use a plain text pastebin, I can't curl it and it requires js, and it is good that its e2e but no advantage here

<folaht>Okay, https://paste.debian.net/hidden/337ebdf6

<folaht>dariqq, it's running on root

<folaht>untrusem, is it visible now?

<untrusem>yes but I don't see how your channel would be related to this

<untrusem>can you run

<untrusem>file /var/guix/daemon-socket/socket

<folaht>result: /var/guix/daemon-socket/socket: socket

<gnufairy>Has anyone been able to get boot-up penguins on a custom kernel on GNU Guix System?

<untrusem>folaht: is your channel in a repo

<untrusem>can you try to restart your guix-daemon while your are at it

<folaht>untrusem, yeah, it's on codeberg.

<folaht>untrusem, tell me how to best restart it. I've tried it already.

<untrusem>what commit are you on?

<folaht>sudo herd restart guix-daemon, right?

<folaht>Oh, let's see... I'm trying to pull guix again, it says 7a58889

<untrusem>hmm, you can get the current commit via

<untrusem>guix describe

<folaht>Hey, it's missing

<folaht>Missing in guix describe

<untrusem>wait

<untrusem>your channel or guix?

<untrusem>because the later shouldn't be possible

<folaht>My channel is missing in guix describe

<folaht>guix a720900

<untrusem>also I don't see any 7a58889 commit in guix repo

<folaht>Oh sorry, that's my channel's commit

<untrusem>i still don't think its related to your channel

<folaht>Well I'm gonna do a guix pull again then, without my channel and with my channel.

<folaht>Two guix pulls

<untrusem>you can use time-machine for this

<folaht>Yeah, a roll-back?

<folaht>Okay, if it still goes wrong I'll try a roll-back

<untrusem>(info "(guix)Invoking guix time-machine")

<untrusem>no, I didn't meant rollback

<untrusem>also see `guix pull --help`

<untrusem>you can exclude and include a certain channel with `-x` and `-e` respectively

<folaht>untrusem, oh cool, I'll try that then once my guix pull without my channel is finished.

<folaht>Then I no longer have to comment out my channel all the time.

<dariqq>folaht Is something in your channel calling the daemon directly upon evaluation? In the build chroot there wont be an accessible daemon socket

<folaht>dariqq, no, it's just very strange.

<dariqq>folaht: with a stopped daemon: failed to load '(yrop packages etchit-io fetchit-desktop)'

<dariqq>folaht: There is a git crate source in fetchit-desktop-Cargo.lock. Youd need to delay evalution of the cargo inputs (either inlining it in the package, wrapping it in a thunk, delay/force, ...)

<untrusem>dariqq: ohh nice trick with the stopped daemon

<folaht>dariqq, okay thanks. I'll look into that.

<folaht>I thought I eliminated that one.

<untrusem>next time you can do `guix repl -i <culprit file>` to find some mistakes

<untrusem>though it can't catch everything

<Heterochromium>I cannot resize Emacs windows on Wayland Plasma. Has anyone else ever had this problem? I'm wondering if it's a known bug, or just a peculiar issue on my rather Linux-unfriendly hardware.

<kaelyn>hello #guix! I hope everyone is doing well!

<untrusem>tusharhero-xmpp, ^

<untrusem>you use emacs on plasma wayland

<untrusem>right?

<untrusem>hi kaelyn :)

<hako>how can we update the nss-certs package? is adding a new pacakge and make it the "nss-certs" available to CLI + %base-packages sufficient?

<untrusem>umm is it about emudhra

<dariqq>hako: No because gnutls trusts the store path of nss-certs (I am not a fan)

<untrusem>ACTION is currently updating, also nss-rapid

<untrusem>updating librewolf

<dariqq>the same is true for java

<Noisytoot>Does IceCat not use OneCRL? IceCat is still trusting certs signed by emSign Root TLS CA - G3 but LibreWolf isn't

<Noisytoot>After updating my system a few days ago I noticed that font rendering changed in several ways. Some text (specifically the kde plasma window decorations and filenames in dolphin) now has a heavier font weight, and in librewolf and icecat the horizontal lines on "f" and "t" are now two pixels wide instead of one pixel. Has anyone else noticed something like this and how do I change it back? (I'm using Noto Sans as my font)

<Noisytoot>left is post-update, right is pre-update: https://x0.at/9h4w.png

<Noisytoot>changing the font style from regular to light fixes that, but makes other the font weight stuff where font rendering remained unchanged too light (which I also don't like)

<Noisytoot>(the wider/more blurry horizontal lines are making text hard to read)

<hako>font-google-noto was using variable fonts as the default output, now (since 2026-10-01) they're moved to the "variable" output. perhaps revelant?

<hako>relevant*

<untrusem>hako, dariqq, Noisytoot: I opened 11954 regarding eMudhra

<untrusem>please label relevant teams

<Noisytoot>I don't think I have the ability to add labels

<Heterochromium>untrusem: I noticed from the logs that you said "you use emacs on plasma wayland↵right?". Was that question directed towards me?

<untrusem>nope to tushar

<AidenIsik>Does Guix have updated CA certs from after eMudhra had their (absolutely stupid) private keys exposed?

<AidenIsik>I would check codeberg but they're blocking my vpn at the moment

<olivial>csantosb: yeah, I've read 9.1.4 Debugging Build Failures, but didn't see the thing I'm interested in there. like, if I wanted to reproduce the buildPhase in the shell that way, I would need to manually reconstruct the build commands and make-flags and such

<olivial> https://jade.fyi/blog/building-nix-derivations-manually/ <- here's an example of the kind of thing I'm thinking of in nix

<bjorkbjork>AidenIsik: https://codeberg.org/guix/guix/issues/11954

<wingo>hi all!!

<wingo>is there a short-term workaround for the CA whose privkey is 4?

<olivial> https://codeberg.org/guix/guix/issues/11954 exists but no full workaround afaik

<olivial>I suppose you could combine the patch with https://codeberg.org/guix/guix/pulls/7968 if you wanna rebuild a lot of stuff

<pdpnw>wingo: please migrate to "5"

<pdpnw>do people consider it a compromise or is that obviously some sort of intentional backdoor

<wingo>hehe

<wingo>the comedy value is amazing

<pdpnw>as a douglas adams fan I would have appreciated a 42, but anything less that 256 is still worth a chuckle

<olivial>I mean if it was a backdoor surely they would have picked something higher-entropy and just handed the value to the cops (or equivalent)?

<lasersgalaxy381>which value?

<lasersgalaxy381>(sorry for my missing all context)

<untrusem> https://logs.guix.gnu.org

<untrusem>you can see irc history here

<olivial> https://emudhra.we-have-a-problem.com/

<pdpnw>olivial: I remember back in the day people were worried uncle sam had some magic values for eliptic curves. using "4" seems like a huge time saver when the alternative is looking for breaking points in ECDSA

<olivial>lmao

<pdpnw>I refreshed that page and the number of packages grew yikes

<olivial>oh I guess that page is missing the crucial comedic detail that the private key is 4

<lasersgalaxy381>i think that even without those magic values, what uncle sam does have, is likely at a level equivalent to magic (in terms of fascination to the world outside). i mean by this that there is from the 1950s (give or take 20 years) a report by the NSA, that even then new mathematicians struggled to come up with anything that had not been tried already. the world has changed and the u.s. might be not be

<lasersgalaxy381>as much of a "hegemony" as it used to be, but it seems very unlikely, that they had not discovered something in all that time funding mathematics.

<pdpnw>I was only able to find a few mastadon posts referencing the actual findings, perhaps not entirely true

<pdpnw>lasersgalaxy381: it's very much a "one-way street" between higher education and that apparatus with a lot of valuable things disappearing over the fence

<lasersgalaxy381>yes

<lasersgalaxy381>i think that higher education might be used not only because it signals competence, but also, possibly, because from a security perspective you kind of get a better leash on the person. i do not mean this in a cruel sense but know no other way to put it. those people also are more stable.

<pdpnw>I think the advantage guix has here is the scope of whats impacted can be (I would think) immediately quantifiable

IRC channel logs

2026-10-20.log