IRC channel logs

<aswjrisp>Franciman: have you tried booting a previous generation from the grub boot menu?

<Franciman>aswjrisp: i was suggested to do a rollback

<Franciman>i was learning to do it

<aswjrisp>Rutherther: podiki: It finished building and when I run sway I get a different error. So this looks like progress to me. The new error says "[sway/server.c:312] Unable to open wayland socket".

<aswjrisp>I think booting a previous generation from the grub boot menu is a rollback, but maybe someone here will correct me if I am wrong.

<aswjrisp>Rebooting did not solve the sway error I am still getting the error unable to open wayland socket.

<aswjrisp>How is Guix on Hurd?

<xelxebar>Ugh. Apparently libsixel just got updated and now it's bash completion script is broken, causing errors at source time.

<xelxebar>In particular there's no `have' command, so line 1 of share/bash-completion/completions/img2sixel errors.

<xelxebar>Appears that libsixel is a propagated-input of mpv, so it's likely to cause issues.

<xelxebar>efraim: Oh, it looks like this is from your latest commit just a bit ago.

<ieure>xelxebar, That's from #72180 ?

<peanuts>"[PATCH 0/2] gnu: mpv: Update to 0.38.0 and include sixel support." https://issues.guix.gnu.org/72180

<xelxebar>Yeah, looks like that's it.

<ieure>It was suggested that I send a refreshed (rebased) patch to an issue I opened in March, and let QA process it. I've done that, but the QA status is still "unknown," and clicking through says "Issue not found."

<ieure>Is QA overloaded, do I just need to wait? Or do I need to close and reopen? Or... what

<podiki>aswjrisp: that wayland error sounds like you don't have wayland running. if you are using something like gdm that should automatically work (wayland) but other login managers or manually i don't know

<podiki>ieure: QA can take a while; how many dependents?

<ieure>podiki, It's a service, so I'm not sure that applies. Looks like it's there now. https://qa.guix.gnu.org/issue/69692

<apteryx>would someone know where xdg autostarted .desktop files (~/.config/autostart) will log errors upon failure?

<apteryx>I'm using GNOME with GDM

<podiki>sometimes there's an ~/.xsession-errors maybe there?

<apteryx>that's on wayland, too

<podiki>then no, may be up to gdm logs and/or compositor?

<apteryx>I looked under /var/log/gdm/, but couldn't see anything related to my .desktop file failing ot run

<podiki>ieure: not sure if QA will tell you anything useful then, just that patch applies cleanly? i don't know for services

<apteryx>doesn't the gnome-keyring daemon knows to read pubring.kbx to find my public GPG keys collection?

<apteryx>seems I'd have to re-import them all, which is redundant and will turn my config GNOME-specific.

<aswjrisp>podiki: I think I will try sddm which I found running `guix system search wayland`.

<podiki>i think i've only tried wayland from gdm; i don't know for sddm

<podiki>ACTION away for a while

<aswjrisp>sddm did not work so I am going to try gdm.

<aswjrisp>I have added (service gdm-service-type) and reconfigured.

<aswjrisp>It is installing hplip. Why would gdm need hp printer drivers? The computer that is reconfiguring is not connected to a printer.

<aswjrisp>gdm needs mariadb?

<podiki>you'd be surprised what depends on what these days

<podiki>you can use guix graph to see, e.g. guix graph --path gdm mariadb

<aswjrisp>podiki: thanks for sharing that command.

<aswjrisp>podiki: I am just trying gdm to see if the code snippet from Rutherther to see if it gets sway working. I think the code snippet has already resolved the missing i915_dri.so.

<aswjrisp>I think you were talking about the guix's mesa getting i915_dri.so in the future.

<apteryx>aswjrisp: 'guix graph' to the rescue

<apteryx>aswjrisp: guix graph --path gnome mariadb

<apteryx>GNOME requires Qt 5, which is hilarious. I've reported that before.

<podiki>and does qt depend on gtk too now?

<podiki>is the circle complete??

<robin>aswjrisp, must be the special login mode where you print out the login form, enter your username and password, and then scan it back in :P

<apteryx>podiki: I fail at gpg-agent using GNOME/ŵayland, apologies

<podiki>apteryx: no worries! i've really bungled some gpg emails in emacs (complicated by the protonmail bridge)

<podiki>which reminds me to sort that out still...

<apteryx>podiki: it does on GTK 3 for theme support; I believe the day it wants GTK 4 for that we'll have a cycle

<podiki>oh boy

<apteryx>because GTK 4 has support for mediastreamer plugins that pull in Qt in their current configuration

<apteryx>we should filter these out, perhaps make Qt-less variants if feasible

<apteryx>(qt-less variants of the mediastreamer plugins to be used by GTK 4)

<apteryx>I haven't had time to delve into this but it sounds like a worthwhile adventure

<apteryx>the closure of GTK 4 could be made much smaller.

<bent_fingers>this is an active channel

<bent_fingers>i should try this distro

<Rutherther>aswjrisp and there is no error prior to getting that sway cannot start the wayland socket? Also preferably try from tty and as non-root user

<robin>[...hours pass...] and now i have my channels in the correct order and can proceed to install guile-next:debug to make gdb useful for my program(s)!

<apteryx>podiki: I finally found how to have my own gpg-agent start, instead of some kind of gnome-keyring man-in-the-middle

<Rutherther>apteryx how?

<apteryx>in a nutshell: https://paste.debian.net/1328851/

<apteryx>these were XDG autostarted, they exist in the $system-profile/etc/xdg/autostart

<apteryx>now my own xdg autostarted shepherd works correctly, as it used to when it was started from ~/.xsession

<Rutherther>Ah, so you just made sure Gnome keyring doesnt autostart if I understand correctly

<apteryx>yes

<apteryx>the Hidden=true lines disable them

<apteryx>ACTION tries to send some gpg-encrypted mail as a test

<Lumine>Here's how I achieved gpg-agent working through home.scm configuration: https://paste.foobarcode.com/QjJAQK

<apteryx>ugh. I'm still getting 'No encrypt key for...'

<apteryx>Lumine: thanks for sharing, looks easy!

<Lumine>No problem :)

<Lumine>Just need the pinentry, pinentry-tty and pinentry-gtk2 in main config, but I'm not sure whether less packages is sufficient

<Lumine>But it's fully working for me

<civodul>Hello Guix!

<Lumine>Hello!

<janneke>Lumine: nice thanks; this should allow me to remove gpg-agent.conf from dotfiles

<Lumine>janneke: nice, and no problem

<Lumine>Although, for me it has generated gpg-agent.conf in .gnupg directory automatically

<Lumine>So I'm guessing the home.scm occupies it

<Lumine>It has the same parameters

<janneke>ACTION adds (ssh-support? #t)

<janneke>err #f

<efraim>hello guix!

<janneke>so actually removes: (ssh-support? #t)

<Lumine>:)

<Lumine>Hi efraim

<efraim>o/

<mfg>Hey, since the core-updates merge the glibc-locales on a foreign sdistro don't seem to be of the correct version even though i upgraded. Did i miss something?

<mfg>~/.guix-profile/lib/locale just contains 2.35 and not 2.39

<haugh>Am I tripping balls or did the web routes at guix.gnu.org change? Do I actually have to go fix all the links I've ever made to manual pages?

<mfg>Ah nvm, i actually had the package installed in the default profile (which i ususally don't use at all) so that's the problem.

<civodul>haugh: i think they’ve been changes lately, hopefully in a backward-compatible way

<civodul>i mean, we still carry redirects from 10 years back

<Rutherther>haugh the urls were not suppossed to be changed, only more restrictive now. It was possible to have anything before /manual, that was fixed. But many urls also from search engines have something like /en/manual, it should be /manual/en

<Franciman>hi, i tried rolling back to a previous generation, but i still get a segfault

<Franciman>this happens right after it tries to pull from a channel

<Franciman>so maybe it's that channel giving problems

<jpoiret>Franciman: it's guix that's segfaulting, did you rollback the guix profile itself?

<Franciman>uhm jpoiret how do i do it? I tried to boot a different generation

<Franciman>in grub

<jpoiret>you can manually do that by changing the `/var/guix/profiles/per-user/<user>/current-guix` symlink to an older current-guix-n-link

<jpoiret>you should be able to do `guix pull --rollback` to do that but it may segfault instead

<jpoiret>for some reason I can't find the thread about it on yhetil.org/guix/

<Franciman>jpoiret: it worked!

<Franciman>it went from gen 66 to gen 65 and now it is updating

<Franciman>what should i do after the guix pull?

<Franciman>will it create a new gen 67?

<jpoiret>yes

<Franciman>great, thanks a lot

<jpoiret>you can delete the generation with `guix pull --delete-generation`

<jpoiret>just like with profile and system generations

<Franciman>i should learn more about generations. Guix was so stable for me

<Franciman>that i never bothered

<jpoiret>heh, that's supposed to be a killer feature of Guix though :P

<Franciman>eheh yes

<jpoiret>i gotta say it's been very useful to test possibly breaking updates without worrying

<Franciman>a question, when i do guix pull --roll-back which generation is rolled back?

<Franciman>profile/system/home?

<Franciman>everything?

<Rutherther>~/config/guix/current is changed only

<Franciman>uhm

<Franciman>that's the profile then?

<Rutherther>Yes, it where profile with only guix is, nothing else

<Franciman>ah ok nice

<Franciman>thanks

<apteryx>podiki: your key was simply outdated in my keyring (and in Savannah)

<apteryx>please update it there :-)

<apteryx>it had expired, and Gnus/epa unhelpfully would only say the key was 'missing'

<apteryx>yay, success

<apteryx>is there a way to have Emacs colored like the rest of a GNOME running in dark mode? It's keeping its 1990's gray menu bar swag for me.

<apteryx>uh, this does it: export GTK_THEME=Adwaita-dark:dark

<apteryx>not sure why it's needed

<henrix>Hi! I've created two packages updates bugs (my first guix patches) a few days ago

<henrix>I was wondering how long should I wait for getting feedback

<henrix>(and what should I do if I don't get any :-) )

<attila_lendvai>does this mean i need to reboot? In procedure dlopen: file "/gnu/store/pgjyl3fn4sflk6xy63qd5anrhqwylpgw-guile-gnutls-4.0.0/lib/guile/3.0/extensions/guile-gnutls-v-2.so", message "/gnu/store/ln6hxqjvz6m9gdd9s97pivlqck7hzs99-glibc-2.35/lib/libc.so.6: version `GLIBC_2.38' not found (required by /gnu/store/9mkcil1rl450r84hn1hcbny5pi5js8ig-gnutls-3.8.3/lib/libgnutls.so.30)"

<apteryx>attila_lendvai: I guess so!

<attila_lendvai>ACTION tries it then

<attila_lendvai>apteryx, unfortunately it didn't go away. it happens when `make -j` in my guix checkout.

<attila_lendvai>after a `make clean`, it's getting worse, earlier: ld: /gnu/store/ln6hxqjvz6m9gdd9s97pivlqck7hzs99-glibc-2.35/lib/libcrypt.so.1: undefined reference to `__snprintf@GLIBC_PRIVATE'

<efraim>what if you rerun the bootstrap script and re-configure?

<attila_lendvai>i am pulled and reconfigured after the core updates merge

<apteryx>attila_lendvai: ah, didn't know this was in the context of building from source

<apteryx>I think efraim's suggestion is a good one

<attila_lendvai>efraim, i think that was it, thank you! (the build seems to be running)

<aswjrisp>apteryx: For emacs you can turn off the toolbar, menubar and scrollbar. Making the gnome theme not relavant for emacs. Then you can apply whatever emacs theme you want.

<mehrad>Hi all, I have a project that is using guix time-machine and the guix shell to create development environment. I copied the manifest and channels files to a new computer, installed guix and ran the same guix time-machine comment and this time around I'm getting this cryptic error: 'guix time-machine: error: execl: No such file or directory'. Any ideas on why?

<mehrad>just to be clear, on both machines, guix was installed on foreign distro

<mehrad>I am not calling execl anywhere myself, so whatever happening is hapenning inside Guix

<apteryx>mehrad: I turn all of these off except the menu bar, which I like to quickly browse what a mode can do

<aswjrisp>Rutherther: As root I get the error about the wayland socket. As a non root user I get the error "XDG_RUNTIME_DIR is not set in the environment. Aborting." Maybe the not having the xdg runtime dir set in the environment is why sddm just hung after putting in a non root users credentials.

<aswjrisp>Rutherther: There are no other errors only the wayland socket error when I run sway from the linux console as root.

<Rutherther>aswjrisp no log before the socket? Thats strange. Yes the error Xdg run time dir is expected since you dont use elogind, you need to set it yourself

<Rutherther>Last week I said its pam that sets it, which is only partially correct. Elogind sets it via pam, I went through the code since.

<aswjrisp>Rutherther: From playing with different attempts to get sway to start I think seatd and logind conflicted. I will try adding elogind and removing seatd and greetd.

<jpoiret>seatd does conflict with elogind

<jpoiret>but greetd works with elogind

<sepeth>Hi Guix, a tool idea: I am trying to add a Rust package to Guix, and finding myself endless cycle of adding/checking/upgrading dependencies of dependencies from crates.io. Since it is Lisp, it should be easy to go from cargo.toml to Scheme. And Guix generally uses consistent naming for Rust packages, such as rust-foo-0.1, so it should be easy to check if the package&version exists before adding. One

<sepeth>complicated bit could be doing the package inherit bit if an older version exists, but this can be ignored initially. Please let me know something as such or better or worse exists, or if this is a good or bad idea.

<Rutherther>aswjrisp right, seatd is a replacement for elogind, so they conflict. But it doesnt have all features of elogind, like setting that env var

<sepeth>It could output to stdout, which then can be directed to crates-graphics.scm, crates-io.scm or whatever.

<Rutherther>Did you try guix import?

<sepeth>I have not :O

<sepeth>Hehe this is embarrassing, guix import looks great.

<sepeth>Rutherther: fantastic, thank you.

<efraim>I prefer to think of it as proof that your idea was worth implementing

<sepeth>True ^-^

<unfroq>Hi Guix! I have some trouble for a longer period now: If I reconfigure and boot the new generation, it won't start my greetd-wlgreet-sway-session (default-session-command) on tty2. However I can login to another tty and run commands as my user, but when I try to try to escalate using sudo, it doesn't let me switch to root (e.g. sudo -s). In that

<unfroq>case I receive: "sudo: /run/current-system/profile/bin/sudo must be owned by uid 0 and have the setuid bit set". Similarly, when I try to start sway from the terminal, I receive the following error: "00:00:00.022 [ERROR] [wlr] [libseat] [libseat/backend/logind.c:317] Could not activate session: Failed to execute program org.freedesktop.PolicyKit1:

<unfroq>Permission denied". Any ideas whats causing this or how to dig further?

<Rutherther>You cannot execute setuid stuff from current system, has to be through security wrappers. So the wrappers are probably not setup correctly. Are you sure you have not removed those from your config? Should be in base services iirc

<civodul>unfroq: the ‘sudo’ binary is supposed to be in /run/privileged/bin, where it is setuid-root

<civodul>do you have that one?

<civodul>(or /run/setuid-programs/)

<unfroq>Rutherther: what wrappers do you mean exacltly? I've set (privileged-programs %default-privileged-programs) in the os-config. I also have set %base-services with modifications suggested for greetd: https://guix.gnu.org/manual/devel/en/guix.html#index-greetd_002dservice_002dtype , e.g. (modify-services (delete login-service-type)).

<xelxebar>Poked around a tad bit with the issue in libsixel's bash completions and submitted a bug: https://issues.guix.gnu.org/73142.

<unfroq>civodul: Since the system is unusuable, I needed to boot another config, so I can't look it up. I will check after rebooting, any other points to look for?

<attila_lendvai>some packages are installed into my user profile. then i started using guix home, and some packages are installed in my home profile. many of these packages are installed in both, and the ones in my user's profile are not updated. should i uninstall the packages from my user profile?

<xelxebar>Not sure if I'll have more time until quite a bit later, so just threw it out there.

<podiki>apteryx: received thank you! will update on savannah, must have forgotten when i renewed subkeys earlier

<podiki>apteryx: might also use your email later to check i can send encrypted response :)

<Rutherther>attila_lendvai yes, remove the ones from.guix-profile if you want to manage those with home instead

<attila_lendvai>Rutherther, thanks! since then i was experimenting, and it looks like the packages in the user's profile are shadowed by the home profile, but they remain as obsolete GC roots if they are not removed.

<apteryx>podiki: great! I have some old emacs mode that doesn't seem to work, jlencrypt, that's supposed to automated using GPG encryption when the gpg key of the sendee is available

<apteryx>I should look into it

<podiki>yeah i should see what is available, i just call the various functions by hand but it is always confusing what is the right one

<aswjrisp>Rutherther: podiki: I was finally able to start sway (not i3, comfirmed by pgrep). I went back to basics using (services %desktop-services).

<aswjrisp>Rutherther: I also used your first code snippet to get i915.

<attila_lendvai>so, i have removed emacs from my user profile, and it's only in the home profile. but now it fails to start with: guix-emacs-find-autoloads: Opening directory: No such file or directory, /home/[user]/.guix-profile/share/emacs/site-lisp

<attila_lendvai>i guess reinstalling emacs into my user profile will solve this, but is there anything else to be fixed?

<aswjrisp>Rutherther: podiki: Is there anything else you would like me to test with this hardware for mesa. Would be nice for i915 to get into guix's mesa.

<rynn>Has anyone tried to port the Cosmic Desktop alpha release yet?

<unfroq>civodul: I just found out, that the problem exists already, directly after the reconfigure. Here is a few commands I tried to gain sudo access: https://paste.debian.net/hidden/57b22ef0/

<unfroq>I just can't use 'sudo' anymore.. /run/setuid/programs/ exists, but I am not sure if they are used correctly, since the sudo command refers to a different item in the store. /run/privileged/ exists, but I cannot see its contents.

<civodul>unfroq: so can you run /run/setuid-programs/sudo ? does that one work?

<unfroq>civodul: $ /run/setuid-programs/sudo

<unfroq>bash: /run/setuid-programs/sudo: Permission denied

<unfroq>civodul: also: $ bash

<unfroq>bash: /etc/bashrc: Permission denied

<aswjrisp>unfroq: Can you run `su - root`?

<unfroq>aswjrisp: $ su - root

<unfroq>Password:

<unfroq>su: Authentication service cannot retrieve authentication info

<civodul>hmm!

<civodul>weird

<civodul>and /run/privileged/bin/sudo?

<unfroq>civodul: permission denied

<Rutherther>I think that is related to you removing the login service

<Rutherther>But I would have to inveatigate more to confirm and I am only on phone

<reedm>Hi All, is there an example of a guix home service that would add some directories to the XDG_DATA_DIRS environment variable? In particular, I don't want to declare the entire variable; I just want to append a couple of directories to it

<unfroq>Thanks, I'll put the login-service back in again and see what happens. Need to reboot though, since I can't switch back bc of permission errors.

<reedm>I tried creating a home-environment-variables-service-type service, but it did not seem to work

<reedm>Sorry, I was disconnected. here is the home service definition I used. https://paste.debian.net/1328885/

<reedm>It seems to have no impact on the XDG_DATA_DIRS environment variable.

<cwebber>I feel like this is something I used to know...

<cwebber>how do I check what the command that a shepherd service would run is?

<cwebber>specifically, it seems like openssh isn't running even when I try to run it, even though shepherd says it is

<cwebber>but I want it to spit out what the command it would have run is so I can try myself

<cwebber>oh whatever, nevermind

<cwebber>I was confused

<unfroq65>civodul Rutherther still no luck. it wasn't login-service apperently.. I am not sure, but are those permissions for /run/privileged/bin correct?

<unfroq65>drwxr-x--- 1 root root 308 Sep 9 15:51 bin/

<Rutherther>unfroq65 the privileges are important on the sudo executable. But the issue doesnt seem to be executing sudo, it seems to be related to what sudo does, so some pam rules etc.

<Rutherther>Did you relog after adding login service? Maybe even reboot. Though yeah, maybe I am wrong and its something else

<unfroq65>Rutherther: I did a full reboot. I just tried a very similar config in a vm, and it does not seem to have the problem. Also the permissions on /run/privileged/bin are 'drwxr-xr-x' in the vm, but not on my main host system.

<unfroq65>I am basically stuck with a config from 1,5 months ago.

<Rutherther>Yeah the perms seem wrong, you cannot read its co tě t

<Rutherther>contents*

<unfroq65>Weird, how do I fix those permissions?

<Rutherther>Is your /run tmpfs?

<unfroq65>'/run/user' is, but not /run

<Rutherther>Hm, okay, I am not sure if it should be or not. But then probably just setting the permissions manually will work as Guix doesnt seem to touch the folder permissions when it exists

<unfroq65>I'll try that.. I also saw that I have set umask manually to 0027 which correlates with drwxr-x---

<unfroq65>So I'll try to unset this as well

<Rutherther>Like as your user? What do you mean you have to set it?

<Rutherther>Hm, maybeif you run reconfigure with that it will put wrong permissions. But if it does, its a bug

<unfroq65>Rutherther I've set is in a few files which get executed by my shells.. like fish and bash. But since I am doing 'sudo guix system reconfigure ...' it might pass it along. And yeah, if so, I would consider it a bug.

<aswjrisp>I just started to install guix on a laptop. When I selected the shell installation I got a message saying that the [AMD/ATI] Picasso is not supported by free software.

<unfroq65>Rutherther thanks again for the support. I'll reboot now and try a few things.

<aswjrisp>When I look up that device it looks like it is a soc cpu and vega gpu.

<aswjrisp>Does that mean that there would be a problem with the vega gpu. If so would I be able to start a graphical environment like sway or i3?

<aswjrisp>I am assuming that the cpu would work.

<aswjrisp>If I can not use the laptops gpu are there other ways to get a graphical environment working?

<aswjrisp>Maybe use a desktop gpu as an external gpu?

<unfroq>Rutherther removing the user umask settings from my shell configs fixed the whole issue! Thanks for helping out. Not sure, but the issue could be related to https://issues.guix.gnu.org/51442

<Rutherther>So you removed umask, then reconfigured and it was fine?

<unfroq>Rutherther yes. Just unset the umask settings, did a reboot to the working config from july, then reconfigure with a new revision. After reboot everything worked perfectly. The permissions were fine everywhere.

<Rutherther>Okay, great!

<futurile>afternoon Guixers

<Rutherther>Afternoon

<aswjrisp>Rutherther: With your code snippet for the mesa i915 not only is sway working but so is qutebrowser. I am going to try out some other graphical programs.

<Rutherther>Did you try the graft as well?

<Rutherther>Happy to hear that. So its feedback for podiki that it indeed still works, just is disabled

<aswjrisp>Is there a way that your the important part of your code snippet will make it's way into guix's mesa or do you think I should just keep the parts of your code snippet in my config indefinately.

<aswjrisp>Rutherther: the grafts was the second code snippet? I had not tried that one yet as you said it might break things.

<Rutherther>aswjrisp I think podiki is going to add it to mesa updates, if they didnt add it yesterday already, yes. Its probably just a mistake that its missing

<aswjrisp>okay

<aswjrisp>Rutherther: would it be helpful for me to also test the grafts?

<Rutherther>aswjrisp if it will work it will probably be very helpful as you dont have to recompile everything that uses mesa

<podiki>aswjrisp and Rutherther: fixes to mesa already on mesa-updates, builds i915 for x86 again (did check locally that the i915 driver built)

<podiki>if you want, you can guix pull from that branch, sub coverage is already decent for x86_64

<Rutherther>Nice. Also sorry for the change logs in the patches, I will have to read more about the rules for future contributions

<efraim>it's too bad there's not a faster way to run 'guix gc'. I basically have to run it in a loop so can do the occasional build in between each round

<podiki>it takes a while to figure it out, but not bad when you get th ehang of it. i always look at examples. and the last patch was trickier anyway

<podiki>but no worries, thanks for the contribution!

<podiki>ACTION away again

<Rutherther>efraim why do you need to gc that much?

<efraim>I have my main machine offload to a number of other machines to do builds for other architectures and I figured after the core-updates merge was a good time to actually empty out my store

<efraim>I estimate my store at ~1.4 TB before btrfs compression

<ieure>That's pretty large.

<Rutherther>Why isnt running it once enough?

<efraim>it takes too long to release the lock on the sqlite database so I break it up into smaller chunks of about 10GB

<Rutherther>Oh I see

<simendsjo>Trying to package foo which requires coreutils at runtime. Adding it to inputs doesn't help. `guix shell foo coreutils` works though. How can I make foo also include coreutils?

<efraim>you might need to wrap the binary with coreutils or patch it to use absolute paths

<simendsjo>I cannot "inject" the "out" of other packages into this package?

<ieure>simendsjo, You can, but I don't think that's the correct approach here. I agree with efraim, put coreutils into your package inputs and patch your package to refer to the absolute paths within that input.

<simendsjo>I see, thanks. I can patch the usage I see, I'm more afraid that there might be some usage I don't see ;) And that I'm matching too much/too little. But I'll do this to avoid requiring coreutils passed in too.

<PuercoPop>I see some emacs packages use "This package ..." in the description while others use "@code{foo}" where foo is the name of the package. Is there a preferred style?

<reedm>Just wanted to report back with the fix for my issue, in case someone finds themselves reading the logs for this channel later with the same issue. I needed to enable the home-bash-service. I think this was required for it to source ~/.profile using bash as the login shell

<reedm>but I'm not completely sure

<ieure>Okay, who can help me figure out why `git format-patch' refuses to detect the base of my changes?

<ieure>I have a branch called `staging-nss-updates'. This is cut off `origin/master', which is the main Guix repo, at commit b82c8ceae1.

<ieure>Sorry, upstream/master is the ref, but same commit.

<ieure>The branch has its upstream set, and `git pull' will fetch commits from upstream/master.

<ieure>If I run `git format-patch --cover-letter --base=auto -o outbox/', it produces zero patch files.

<ieure>If I run `git format-patch --cover-letter --base=upstream/master -o outbox/', it produces zero patch files.

<ieure>This has *never* worked for me, I don't understand why, it seems very simple.

<fnat>ieure: Does adding something like '-1 HEAD' help in any way?

<ieure>fnat, it outputs the most recent commit. What I want is to output one patch for every commit added to this branch vs. the one I cut it from.

<ieure>I do not want to count how many commits that is and pass that as an argument. I just want all new commits on the branch and to never have to think about it.

<ieure>Like how I can push a branch and open a PR and the diff in there is only the new work I've added vs. the branch I target.

<Rutherther>why do you think specifying the base should do that? reading the docs it seems to me --base is just for correctly storing what the base commit is, not for saying what commits to do the patches from

<Rutherther>if you want every on this branch vs the ones you cut from then it should be just something like this specification of commit range: "my-branch...origin/master"

<Rutherther>okay, sorry, the other way round, so "origin/master..my-branch", alternatively use HEAD instead of my-branch if you are just on the commit you want

<ieure>Rutherther, Well, I guess the man page indicates this does something totally different and not helpful to me, but when you're creating a diff and there's an option to specify a base commit, that intuitively seems like the bottom end of the range you want the diff for.

<ieure>This works: `git format-patch --cover-letter upstream/master -o outbox/'

<Rutherther>ah, right, it automatically just goes from current commit

<ieure>Though it's pretty bunk that it doesn't look to see what the configured upstream of the branch is and use that if no other specification is present.

<ieure>And also that failing to provide a specification does nothing and returns a 0 status, which makes it hard to understand what happened.

<Rutherther>I think the base is there just to make sure the patches apply cleanly on the base - it's the target base commit of the produced patches, not the one they are coming from. Though they usually do match

<PuercoPop>ieure: --base=auto should use the upstream if configured. You mentioned it doesn't work for you?

<ieure>PuercoPop, It does not work.

<Rutherther>what is the command you use and what is the output?

<ieure>Rutherther, I gave two commands I tried earlier, they produce no output on stdout, and no patches on disk.

<ieure>`git format-patch --cover-letter --base=auto -o outbox/'

<Rutherther>okay, I think I've explained why they don't. But you probably don't believe me, so okay

<ieure>Rutherther, I believe you, sorry, I meant to mention PuercoPop.

<PuercoPop>ieure: What is the output of git rev-lead HEAD ^upstream/master? Does it list any commits?

<ieure>PuercoPop, git: 'rev-lead' is not a git command.

<PuercoPop>ieure: sorry typo, git rev-list

<ieure>PuercoPop, It lists all the commits added on my branch.

<PuercoPop>ieure: Yeah, that makes sense. But then I'm not sure why git format-patch is not doing anything. If should consider those commits as the ones to generate patches for

<ieure>PuercoPop, It does if I manually specify upstream/master, if I don't specify anything, it outputs nothing.

<ieure>> git version 2.46.0

<Rutherther>See this for example: I made a simple repo, I made a list with one, two, three, and made that into initial commit. Then I added four in one commit and five in another. Now if I do git format-patch -1, I can choose the base be either the initial commit or the commit adding four. When I choose the base be initial commit, git format-patch adds prerequisite-patch-id, to signify it cannot apply cleanly on base commit, and first another patch would be...

<Rutherther>... needed. Now when I make patch for the second commit, and obtain patch id via git patch-id, it has this id

<Rutherther>So I think the base is really just to let the maintainer know how to apply it properly. Not to say what commits to make patches for. For that there are other arguments like -<number> or a specific commit, range etc.

<ieure>Yes -- I simply don't want to have to specify -1 (or any other argument), because Git can figure out the base for me based on the branch's upstream configuration.

<ieure>Just surprising to me that there's no way to do that.

<Rutherther>I understand what you want, I am just explaining why --base doesn't give it to you

<ieure>Yes, I understand why that is now.

<ieure>Okay, another question. Pretty much every time I send a patch series, I get prompted: To whom should the emails be sent (if anyone)?

<ieure>Is this how it is, or is there something broken in my setup? The manual section on sending a patch series doesn't mention this or say what email address it should be directed to.

<ieure>Only the issue number after you've sent the cover letter patch to open the bug -- but it seems to imply that send-email will know what address to send that to, which doesn't seem to be happening for me.

<ieure>So many problems that just wouldn't exist, but for the email patch workflow.

<ieure>Okay, submit@debbugs.gnu.org is where I need to send it. I can never remember this / should not have to care about it, ever.

<Rutherther>for a patch? The documentation mentions guix-patches@gnu.org

<Rutherther>I would be surprised if submit@debbugs.gnu.org works for guix patches as debbugs is for multiple projects and I can't imagine it deciding which project to put it to

<ieure>Ah, so it does, okay, so I have to send the first to a totally different domain. wild

<cwebber> https://puckipedia.com/7hkj-98sq/qixt does this affect guix?

<meaty>is there a way in guix to make it such that a certain file heirarchy can only be modified using a certain program? I want ot make it such that only my music library management software (beets) can edit the music directory

<cbaines>cwebber, potentially, the nar restoring code looks simple enough that it might work, I'd suggest emailing guix-security@gnu.org if you haven't done already

<meaty>also, is there any way to make the process names shorter? the store system makes e.g. 'top' a lot less helpful

<Rutherther>meaty: have you tried -c option or pressin c?

<meaty>Rutherther: holy shit thanks

<emenel>Hi everyone. I'm trying to setup guix as an os for the first time and am having trouble with the guix home conf. I'm trying to set my account shell to fish in my home conf, but can't find any examples and haven't been able to figure it out. It feels like it should be simple, so I'm wondering if I'm also missing something fundamental here...

<aswjrisp>Rutherther: I just tried opening a couple of graphical programs that I use, with your code snipet rebuilt mesa with i915. They all opened.

<aswjrisp>I did run into one piece of unexpected behavior.

<Rutherther>you can switch to the mesa-updates branch, and you don't have to use the snippet

<aswjrisp>Rutherther: Okay I will read the manual to figure out how to pull from a branch.

<cwebber>cbaines: emailed, thx

<aswjrisp>The unexpected behavior is that in sway the cursor icon is missing. The mouse is functional, there just is no indication of where it is without a cursor icon.

<aswjrisp>Does anyone know what package I need to add to get a cursor icon in sway?

<Rutherther>it should just be visible even without further configuration or programs installed. My guess would be it's a package without the mesa override that is responsible for adding the icon

<aswjrisp>Rutherther: Next I will test your second code snippet with grafting.

<aswjrisp>Maybe the rust-wayland-cursor package will give me a cursor image.

<aswjrisp>Nope, I could not even reconfigure with rust-wayland-cursor in my config. Even though I double checked spelling and used the module it came from.

<divya>Hello, how does one mount a file system in profile/system configuration with specific read/write/execution permissions?

<divya>So, for example when I add this into system.scm:

<divya>(file-system (mount-point "/home") (device "/dev/sda3") (type "ext4"))

<divya>I cannot specifically decide what permissions this mount should have.

<divya>One could do this when adding a mount in /etc/fstab

<attila_lendvai>Rutherther, thanks for your arm gcc toolchain work! my original need is gone, but i may play a little with it nevertheless to see if it works for compiling that 2040 flasher i tried (libreboot stuff)

<divya>The manual says that the filesystem-independent mount options in `man mount` should go into the `flags` symbol. But doing `flags '("rw" "exec")` doesn't work.

<futurile>evening

<fnat>\o/ futurile

<futurile>fnat: I'm in London - my hotel is so close to St Pauls I can hear the bells heh

IRC channel logs

2024-09-09.log