IRC channel logs


back to list of logs

<jgart[m]>lfam: Can you call up Katherine Cox-Buday and ask her to fix it for us? lolz
<negligiblebeans>Hmm I'm still stuck on packaging github-cli. Outside of guix, the package can just be built with "go build cmd/gh", but I can't figure out how to pass the path "cmd/gh" using the go-build-system (by default it seems to just run "go build" with no path. It seems like a trivial problem but I'm having no luck
<singpolyma>Can you clone the subfolder or cd before running build?
<jgart[m]>negligiblebeans: Is this the one that you are trying to package?
<jgart[m]>or this one?
<negligiblebeans>the second one
<jgart[m]>singpolyma: Is your question for negligiblebeans or a general one?
<jgart[m]>I'm guessing the former
<singpolyma>Maybe if you can't pass the path you can just be at that path so that go build works
<negligiblebeans>singpolyma I tried added a new build phase beforehand which called "chdir", but I couldn't get it to work - it was throwing an error saying the directory didn't exist
<negligiblebeans>is it possible that the repo had not been cloned at that stage maybe?
<negligiblebeans>Since manually "cd cmd/gh" then "go build" does work outside of guix
<jgart[m]>negligiblebeans: Are you interested in packaging gomuks? I'd greatly appreciate it if so;()
<jgart[m]>I packaged everything except gomuks itself here:
<jgart[m]>I made an empty "wish list" module for gomuks here:
<negligiblebeans>jgart to be honest I'm completely new to packaging things with guix, so not sure how much i could actually help
<jgart[m]>I think gomuks will be easier than github-cli
<jgart[m]>Also given that it is mostly packaged
<jgart[m]>actually I partially take that back:
<zacchae[m]>Alright, why did I need to add --substitute-urls to the guix-daemon invocation in guix-daemon.service in order to get substitutes from I'm on a debian-based foreign distro, so it would make sense that the daemon might be out
<zacchae[m]>Ah, that service has /usr/bin/guix-daemon hardcoded, and so doesn't respect any invocation of guix pull
<zacchae[m]>unmatched-paren: I solved my problem: out-of-date guix-daemon was running
<nallastro>does anyone know which package am i supposed to add to inputs if i'm getting "‘FILE’ is defined in header ‘<cstdio>’; did you forget to ‘#include <cstdio>’?"
<apteryx>are you using gcc-toolchain or just gcc? use the former
<apteryx>oh, is down
<nallastro>i'm using gnu-build-system and compiling a pure C++ package with readline as only dependency
<nallastro>am i supposed to explicitly add gcc?
<nallastro>i'm using gnu-build-system and compiling a pure C++ package with readline as only dependency
<apteryx>sneek: later tell nallastro in this case no, there's no need to add gcc manually, it's provided by the gnu-build-system
<sneek>Got it.
<lilyp>sneek later tell nallastro to add to what apteryx said, you might have to add a newer gcc to native-inputs if your package doesn't compile with GCC 10
<sneek>Will do.
<bumble[m]>how can one view the guix analog of /usr/share/kbd/consolefonts/?
<bumble[m]>the system config file used on this machine includes this expression and it would be great to view the contents of consolefonts/... (full message at <>)
<jpoiret>bumble[m]: you can look at it by looking into the directory `guix build font-terminus` returns
<jpoiret>sneek, later tell negligiblebeans: you can always replace the build phase by your own, with the proper command line invocation!
<unmatched-paren>hello guix!
<Guest70>good morning, everyone
<unmatched-paren>does anyone have a working emacs smtpmail configuration i can use?
<unmatched-paren>i keep getting either error 500 or error 553 when i try to send from mu4e
<unmatched-paren>oh. i just got it working, apparently. never mind! :D
<unmatched-paren> i didn't. :(
<iyzsong[m]>i use sendmail-send-it as send-mail-function, and the send-program is from opensmtpd (was msmtp)
<gabber>hi! i'd love to have `man socket` and `man strcpy` and the like handy in Guix. unfortunately the package linux-libre-documentation only serves the info manual. are the man-pages part of the linux-libre-headers package? or where can i find/how can i install them?
<Guest70>I have installed nix via guix install nix
<Guest70>warning: the group 'nixbld' specified in 'build-users-group' does not exist
<Guest70>error: file 'nixpkgs' was not found in the Nix search path (add it using $NIX_PATH or -I)
<Guest70> at «string»:1:25:
<Guest70> 1| {...}@args: with import <nixpkgs> args; (pkgs.runCommandCC or pkgs.runCommand) "shell" { buildInputs = [ (cowsay) ]; } ""
<Guest70>dave@host ~$ sudo nix-shell -p cowsaywarning: the group 'nixbld' specified in 'build-users-group' does not existerror: file 'nixpkgs' was not found in the Nix search path (add it using $NIX_PATH or -I) at «string»:1:25: 1| {...}@args: with import <nixpkgs> args; (pkgs.runCommandCC or pkgs.runCommand) "shell" { buildInputs = [
<Guest70>(cowsay) ]; } "" | dave@host ~$ sudo nix-shell -p cowsaywarning: the group 'nixbld' specified in 'build-users-group' does not existerror: file 'nixpkgs' was not found in the Nix search path (add it using $NIX_PATH or -I) at «string»:1:25: 1| {...}@args: with import <nixpkgs> args; (pkgs.runCommandCC or
<Guest70>pkgs.runCommand) "shell" { buildInputs = [ (cowsay) ]; } "" |
<Guest70>does anyone know what I am missing in nix?
<gabber>Guest70: please use a web paste service for multiple lines of pastery
<gabber>why do you want to have nix-installed? why don't you use `guix shell foo`?
<Guest70>my error for nix is here:
<Guest70>gabber I dont know actually. Not installing it is better?
<gabber>Guix is pretty much a rewrite of Nix in GNU Guile, so you have most stuff ready in Guix
<gabber>if you tell us what you are intending to do we might tell you how you can achieve your goals in Guix (or refer you to the relevant sections in the manual)
<gabber>ACTION is away for a couple of minutes
<zimoun>about core-updates, the update of mpfr breaks julia and I do not what I can do… For instance, “Evaluated: "0.1" == "0.10000002"” or “Evaluated: "0.1" == "0.0999999999999999999999999999999999999953". efraim, any idea?
<rekado>grpc is failing on core-updates
<rekado>python-grpcio also fails, and it looks like it’s not trivial to fix
<rekado>it contains a cython-generated cpp file, and it looks like to generate it from source we need a matching version of gprc.
<rekado>(ultimately this all breaks tensorflow, which I want for guile-aiscm)
<unmatched-paren>Guest70: you'd need to use the nix-service-type, but as gabber says if you don't need it specifically don't use it
<mroh>Love, Peace & Guix! Good morning!
<apteryx>good morning!
<unmatched-paren>(this may seem like an overreaction but i have been wrestling with emacs smtpmail since yesterday morning so...)
<unmatched-paren>apteryx: hello!
<bjc>is there a way to tell guix to farm out build jobs to another machine? my desktop is pretty anemic, but i do have another machine that's pretty powerful
<bjc>i just leave it off most of the time because it sucks down 250 watts at all times
<unmatched-paren>bjc: <- looks like it's possible! :)
<bjc>brilliant. thanks for the pointer
<bjc>that'll make crunching through core-updates easier =)
<apteryx>ACTION makes heavy use of the guix offload facility
<apteryx>main rig being a vintage 2006-assembled desktop ^^'
<bjc>that's impressively old
<bjc>i wonder if i can get guix going on some old powermac g4 or g5 machines i have about
<bjc>sneek: seen nckx
<sneek>nckx?, pretty sure was seen in #guix one month and 2 days ago, saying: Like the Chinese mechanical typewriter I had as a child..
<apteryx>mirai: any clue how I can use the "users" group for the mpd-service-type?
<apteryx>I tried: "(group (system-group (name "users")))"; should that work?
<mirai>hmm... It should but I think not in the way you expect
<apteryx>ah, no, system-group is a (gnu system shadow) local syntax
<mirai>remember the past issue, where a duplicate user would compete with it?
<apteryx>I'll try with: (user-group (system? #t) (name "users"))
<mirai>I think if you create a "users" group that way, it will compete with the “built-in” 'users' group
<apteryx>but if it's exactly the same? it should deduplicate at some point perhaps?
<mirai>You could define the users group in a variable and use it in both (groups …) operating-system field and in mpd
<mirai>it's still not 'eq' afaik
<bjc>it looks like the mpd field takes a string?
<apteryx>yes, but that's deprecated
<mirai>do you get a 'duplicate' group warning?
<apteryx>I'll try it now
<apteryx>seems like I should start using guix home soon though... as that'd be a more natural fit for it there
<apteryx>or perhaps I could leave mpd-service-type running with user mpd, and relax user permissions on the music directory it should use?
<mirai>you could change the group for the directory?
<apteryx>I guess it looks into ~/Music by default; how can I configure that to something else?
<mirai>or change the mpd user-account to have membership in other groups
<mirai> (music-directory "/media/Music")
<mirai>so I never had to tinker too much with the user-accounts/groups
<mirai>note that the group and user fields of mpd are unrelated
<mirai>they only set the process user/group
<mirai>you can have (user-account (name …) (group X) … ) and (user-group (name Y) …) and it will still work
<apteryx>thanks, I'm reconfiguring with this on top of the staging branch
<gabber>since there's some more activity here i just ask again: does anybody know how to get `man socket` and the like (c-standard calls man-pages) in guix? they don't seem to be part of linux-libre-documentation.. are they part of glibc:debug? any ideas?
<apteryx>for linux kernel interfaces, try 'guix install manpages'
<apteryx>err, man-pages
<apteryx>for glibc documentation, use 'info libc'
<gabber>thanks apteryx !
<unmatched-paren>there's also man-pages-posix for the Np pages
<gabber>actually, package `man-pages` was what i was looking for (it also has `man strcmp` and `man memcpy` although they are part of glibc -- or am i mistakes there?!)
<unmatched-paren>linux provides the man pages for glibc
<unmatched-paren>glibc only comes with an info manual
<gabber>unmatched-paren: yes, but unfortunately not in the Guix packages case ... :)
<gabber>ACTION sometimes wishes there was some kind of collection of guix packages for specific tasks (like C/kernel development)
<unmatched-paren>gcc-toolchain is essentially the collection for C
<gabber>well, linux is installed but the man-pages in question were not provided -- or was i just missing the `man` binary? not i'm not sure anymore
<unmatched-paren>linux does not provide its man pages
<unmatched-paren>oh, right, i see what you mean
<gabber>unmatched-paren: yes, but there's still some stuff missing. and for development i need documentation, maybe a linter and a text editor -- i build my apps with specific manifests
<unmatched-paren>the problem is that there are lots of different linters and text editors :)
<gabber>i was more thinking of a loose collection of all-the-packages-one-could-possibly-need for a task X -- so that i could search the web and eventually stumble upon a package called man-pages (which then might give me a hint that i was looking for that all along)
<gabber>lots of linters?
<gabber>for C development? in Guix?
<unmatched-paren>you could use a traditional linter, ctags, lsp...
<gabber>ctags is a linter?
<unmatched-paren>doesn't it help vim do linting?
<unmatched-paren>or is it just go-to-definition etc?
<gabber>WDYM with "traditional linter"? i was looking for one version of `lint` or the other... but `guix search lint` gives a bunch of stuff (and it's close to impossible to filter for the C language)
<gabber>i use it mostly for jumping code.. haven't heard of it doing actual linting (but i'm not really an expert in the field)
<unmatched-paren>anyway what i mean is that there will be plenty of different things for different editors, preferences, levels of complexity
<gabber>hrmpf.. not sure we talk about the same thing. in the C world there used to be `lint` which made programmers aware of anti-patterns in their source code. this has little to nothing to do with coding style and white-space arrangement. maybe static-code analysis is the right keyword? anyhow, i was having trouble finding such a tool (in Guix)
<unmatched-paren>probably the best tool for that would be LSP
<unmatched-paren>it's available for vim, and built in to neovim and emacs
<unmatched-paren>then you can install clangd or ccls
<gabber>do you mean this: ?
<bjc>there's cpplint, but i don't see anything that looks like the old-school unix ‘lint’ program
<gabber>bjc: i did some research and IIRC `lint` has been abandoned(?) and was superseded by splint?
<bjc>ah, that makes sense. i couldn't even find it on my bsd boxen, which i'm pretty sure used to include it automatically
<bjc>it's not something i've used myself in probably a couple decades
<gabber>yeah.. i figure gcc (nowadays) does a bunch of what lint used to provide exclusively, but gcc's linting capabilities are somewhat limited
<bjc>exactly. gcc covers a lot of ground now
<unmatched-paren>ccls and clangd should provide that for LSP
<unmatched-paren>given that the latter is literally using clang... :)
<bjc>gcc has static analysis built in now, too, though i'm not sure how good it is
<bjc>the problem is setting up lsp for c and c++
<gabber>yeah, in the safety/security world gcc's static-analysis capabilities is somewhat frowned upon (my boss actually refers to some proprietary linter -- which i hadn't seen so far, so....)
<gabber>ACTION just found an earlier attempt of packaging splint for Guix
<gabber>aaaaand it doesn't compile (or else it would have been upstreamed already)
<Guest19>Hi, I sent a patch and got a review. When I now want to reply, should I send the mail to AND to the reviewer, or will the reviewer get informed automatically when I mail ?
<gabber>Guest19: i'd CC the reviewer, they won't get a notification otherwise
<Guest19>okay, thank you
<gabber>you're welcome
<unmatched-paren>Guest19: use --cc for that fyi :)
<Guest19>unmatched-paren: alright :P
<gabber>is there an option for gnu-build-system to build single-threadedly (aka `make -j=1`)?
<gabber>gabber: parallel-build?
<unmatched-paren>gabber: #:parallel-build? #f
<unmatched-paren>gah, too late :)
<gabber>thanks, you two! ;)
<gabber>how can i match a line starting with # and ending with \ in a (substitute*) expression?
<gabber>"#(.*)\\" and "#(.*)\\\\" don't seem to work
<bjc>i was going to suggest the latter, but with anchors =/
<bjc>ie: "^#(.*)\\\\$"
<gabber>no dice :(
<bjc>it works in fwiw
<bjc>maybe some guile string escape weirdness?
<gabber>the sad part is i don't know which part of the expression is not workign (the wildcard or the special characters)... and i'm not sure how i could test it
<bjc>it looks proper from ‘display’
<bjc>throw an error when its matched. that's what i tend to do
<bjc>(substitute* "some/file.c" (("regexp" all) (throw 'matched all)))
<nom[m]1>unmatched-paren: >the files you need write permissions for should not be managed by guix home anyway
<nom[m]1>how should I manage them? Part of the thing i'm trying to do, is adapting all my dotfiles configuration to work within guix, and certain programs like my web browser need the writable access
<nom[m]1>I guess I'll have to make a script to run stow after the guix home configuration is done
<bjc>if they need to be written to, they can't be managed by guix
<unmatched-paren>nom[m]1: web browsers are tricky ones, that's true...
<bjc>the store is immutable, and (basically) all files managed by guix live in the store
<unmatched-paren>nom[m]1: however, there is a stow home service available somewhere on the mailing list
<nom[m]1>I've been looking through the mailing lists, figure this is the one
<gabber>bjc: not sure how/where to throw that
<unmatched-paren>nom[m]1: that's it, i think
<unmatched-paren>there was also one titled "Add home dotfiles service type" or something along those lines
<nom[m]1>seems like it's just a way to make stow work tree compatible with guix, but it still goes through the store
<nom[m]1>are any of these actually implemented upstream?
<unmatched-paren>no, they haven't been merged yet
<bjc>gabber: wherever you're doing the matching. you can use whatever scheme you want, afaik, when a match is found; so you can just throw an error to see if its matching at all and what data it matched
<gabber>inside the (substitute* ) s-exp?
<gabber>ACTION is even unable to match the starting #
<gabber>ACTION or any # for that matter???
<gabber>i think i'm doing something completely wrong
<gabber>i guess i need a break
<gabber>ACTION thanks everybody for their helpful input and wishes a good night
<jlicht>Hey folks
<unmatched-paren>jlicht: hello!
<apteryx>eh, 'sudo herd restart syslogd' terminates the graphical session
<bjc>a surprising number of things do that
<apteryx>a prompt would have been nice :-)
<oriansj>a reconfiguration can wipe out a session too
<oriansj>even withno changes
<apteryx>I've never experienced the later
<apteryx>I think reconfigure only starts *new* or stopped shepherd services, not restart already running one, so this shouldn't happen as far as I understand.
<apteryx>ungoogled-chromium spews a lot of junk to dmesg; e.g.: [423612.755268] nouveau 0000:07:00.0: gr: 00200000 [] ch 2 [0027a20000 chromium[23315]] subc 4 class 502d mthd 0860 data 42700000
<apteryx>is it just me?
<apteryx>even segfaults: [421212.760503] nouveau 0000:07:00.0: fb: trapped write at 00006ea000 on channel 2 [27a20000 chromium[23315]] engine 00 [PGRAPH] client 0b [PROP] subclient 0c [DST2D] reason 00000007 [WRONG_MEMTYPE]\n[421466.671943] type1[8373]: segfault at fffffffffffffffe ip 00000000004015e5 sp 00007fffffff7a90 error 5 in type1[401000+2000]
<mvnx>Anyone know what the Guix equivalent of NixOS's `hardware.opengl.enable = true;` and `hardware.opengl.extraPackages = [ pkgs.mesa ];` would be?
<unmatched-paren>mvnx: opengl should be enabled by default
<unmatched-paren>iirc that setting is supposed to ensure that everything is completely reproducible even on different graphics cards
<unmatched-paren>guix just concedes that this is pointless
<unmatched-paren>(if you don't want opengl, just don't install things that use it...)
<mvnx>Should the mesa drivers "just work" then?
<mvnx>I needed `hardware.opengl.extraPackages = [ pkgs.mesa ];` on NixOS for my particular Wayland compositor to run.
<unmatched-paren>yup, they'll Just Work
<Guest60>hey, is there a convenient way to remove propagated inputs from packages? e.g. install gnome but without epiphany
<unmatched-paren>Guest60: you'd need to create an inherited package definition.
<Guest60>ah that makes sense
<ngz>Hmmm "guix time-machine --branch=core-updates -- weather -m manifest.scm" raises an error "substitute: In procedure write_wait_fd: unimplemented". It looks odd. Is there a workaround?
<mirai>apteryx_: inspect the shepherd graph?
<mirai>perhaps syslogd is an ancestor of the graphical thing?
<apteryx_>it must be indeed
<ircuser>who manages packages?
<apteryx>guix contributors and committers
<ircuser>some packages are very way out of date. one package for example toxic what would it take to get that package updated?
<apteryx>a patch :-)
<ircuser>how do i submit a patch?
<ircuser>i do not know how to patch aj guix biuld though
<apteryx>see the 'Contributing' section in the Guix manual
<apteryx>e.g. info '(guix) Contributing'
<random-stranger->i manually check every package i use and see if out of date, it might be good to just have stale packages automatically removed
<random-stranger->to prevent old out of date software from being installed
<random-stranger->of course if something just was not updated because no need to update, can just be tagged as fine. i mean removing out of date packages
<random-stranger->i found so much software i was using was out of date
<apteryx>why remove stale package that works? it's best to update them
<random-stranger->true, if someone updates it though
<apteryx>it can be you
<random-stranger->im not a programmer
<random-stranger->otherwise i would
<apteryx>you don't have to
<random-stranger->don't have to?
<apteryx>'./pre-inst-env guix refresh -u something' goes a long way, then ./pre-inst-env guix build something
<apteryx>if it still builds and pass tests, done :-) often a 2 lines change
<random-stranger->im going to look into those commands
<random-stranger->but that makes no sense to me. you mean run that in the directory of the env?
<random-stranger->what would the "something" be?
<apteryx>it'd be the name of a package you want to update
<random-stranger->i can try to figure out
<random-stranger->but would be helpful if i saw an example
<random-stranger->for the info
<apteryx>and you'd run this in the git checkout of Guix, after having built it following the instructions in the manual (mostly just: 'guix shell -D guix && ./configure --localstatedir=/var && make -j'
<apteryx>there may be videos on the guix website, I forget
<apteryx>otherwise feel free to ask questions here, the channel exists for this purpose
<random-stranger->thanks one more question for now
<random-stranger->using luks encrypted system on guix. i want to move the boot to an external drive to protect from evil maid attack. but cannot identify  where the boot partition is, grub, on this install
<random-stranger->as in boot off of usb drive, with encrypted partition still on the internal drive
<random-stranger->where is the boot partition on guix install? or how to find?
<apteryx>I think the graphical installer must not create a separate boot partition by default
<apteryx>so everything is under the root partition by default
<apteryx>and the /boot would be encrypted too
<bjc>it has to create at least an esp for efi systems
<random-stranger->even if i cannot move to external drive, i'd like to be able to get recursive hashsum of the unencrypted part so i can audit it frequently
<random-stranger->so trying to locate where grub, the boot part is
<bjc>lsblk may help you. i'm not sure, since i don't use luks
<apteryx>the executable GRUB boot code? if it's an efi system it'll be save to that partition I think
<apteryx>if it's a legacy BIOS machine it'll be spread in the MRB + grub_bios partition or something like that
<apteryx>err, MBR, master boot record
<apteryx>does someone know where messages such as: [12420577.725729] SSH_brute_forceIN=eno2 OUT= MAC=b0:26:28:b7:9d:0b:00:09:0f:09:00:12:08:00 SRC= DST= LEN=64 TOS=0x00 PREC=0x00 TTL=35 ID=47750 DF PROTO=TCP SPT=36680 DPT=22 WINDOW=229 RES=0x00 ACK PSH URGP=0 come from?
<apteryx>it doesn't seem to be from sshd itself
<random-stranger->lsblk shows
<random-stranger->sda   256 GB
<random-stranger->|--sda1 2MB
<random-stranger->|--sda2  256 GB
<random-stranger->   |-cryptroot 256 GB
<apteryx>it appears in 'dmesg', so I'm guessing some component of the kernel?
<apteryx>bjc: fail2ban is not running on that machine (yet)
<bjc>that explains the timestamp
<ieure>apteryx, It's logging from iptables in the kernel.
<random-stranger->so maybe sda1
<apteryx>ieure: thanks
<random-stranger->with cryptrot under /gnu/store/
<bjc>random-stranger-: may be sda1? what's ‘fdisk -l /dev/sda’
<apteryx>ah yes, berlin has some custom firewall with that iptables --log-prefix
<random-stranger->bjc /dev/sda1 type says "BIOS boot" and /dev/sda2 "Linux filesystem"
<random-stranger->maybe some of sda2 might be unencrypted, i'd have to check
<apteryx>fail2ban should look into /var/log/messages fo a sshd jail on Guix System, right?
<apteryx>it's set to /var/log/secure by default, which doesn't seem to work
<bjc>no idea. i've only used fail2ban on freebsd
<apteryx>yeah, there's almost nothing in /var/log/secure; just PAM stuff
<bjc>random-stranger-: that's your boot partition, then. it'll have the grub install in it
<bjc>iirc, for legacy boot grub just copies an executable image to that partition, so you won't be able to mount it
<random-stranger->i was wondering why i was nto able to mount it
<random-stranger->so grub might actually be on file system?
<apteryx>it tracks [sshd] messages such as: Nov 28 09:16:03 srv sshd[32307]: Accepted publickey for git from port 57904 ssh2: DSA 36:48:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx, per its test suite
<random-stranger->and if so, wonder if i can take hashsums of that grub image, wherever it may be.
<apteryx>which show up in /var/log/messages and not in /var/log/secure
<bjc>there'll be some stuff in that partition (effectively, the same thing that'd be in the esp on in efi system), and some stuff will be read from /boot when grub starts up
<apteryx>OK, it comes from a Guix-specific fail2ban-paths-guix-conf.patch, which sets syslog_authpriv = /var/log/secure
<apteryx>which seems accurate... perhaps another log should be defined as well
<mirai>those messages are from nftables I think?
<mirai>is there any value to fail2ban for ssh anyways?
<mirai>especially if you use things like pubkey auth, it's not like they're going to bruteforce the pubkey
<glenda>Where is the source code of IceCat that is used on guix?
<Guest19>I have  a performance question about generations.  My guix home is currently in the 70 generation ( i have the old ones still on the disk) and it takes 5 minutes to create a new generation with my old HDD.  Will more generation impact the performance or is it just my HDD that may be defragt or something?  Since in the past it was quick and took
<Guest19>max 1 minute, now it takes couple of minutes
<bjc>previous generations have no impact on the current one
<jpoiret>do you know which step exactly is taking a long time?
<Guest19>okay, good.  Other question.  If I use pacucontrol my HDD is making cracking noises.  Only if I use pavucontrol.  Is this normal?
<bjc>i should say: previous generations can make the new build faster, since existing derivations can be reused, but it shouldn't make anything slower
<Guest19>jpoiret it takes some time with the last output being building profile with 34 packages...  That message sits there for like 5 minutes
<Guest19>rest is normal speed like I used to have
<jpoiret>do you have any big packages in there?
<jpoiret>but that's definitely a bit weird
<Guest19>do you mean big by size? that would be texlive with 3gb
<jpoiret>that might be the issue then, were your profiles fast to build when using texlive before?
<jpoiret>it also has a lot of files
<Guest19>my HDD also makes cracking noises everytime i am in gdm.  Really weird.  The HDD itself can't be the problem since it is since I switched to Guix and it is a relative new one
<glenda>jpoitret: thanks
<jpoiret>I don't think Guix would influence that
<Guest19>jpoiret hmm, I can test that.  If I remove texlive and add intel-vaapi-driver, will that be enough information for you if I see a time difference?
<glenda>jpoiret: But in guix it's version 102, while on that site - 60.2.0
<Guest19>jpoiret takes some time, i will remove texlive from my home config and add intel vaapi (already in store), also it is faster to type out guix home recongiure home.scm than using guix home reco and autocomplete
<jpoiret>glenda: has newer versions
<Guest19>also I have a weird shepherd behavoir.  I have 1 HDD connected with Guix without UEFI, sometimes I disconnect that HDD (no power, no sata) and connect a different HDD with MS W11.  Now if I disconnect that drive and connect back to my HDD, Guix will boot normally and behave normally but it can't shutdown normally.  sudo halt doesn't really do
<Guest19>anything and sometimes it says "service root is not running".  In the logs I see that shepherd is having a timeout and getting "grace periof of  seconds is over sending -417 sigill"
<jpoiret>Guest19: HDDs in general are not very gux friendly
<Guest19>jpoiret could you explain the technically way? I thought guix is just changing the env vars basically
<Guest19>hmm, seems like now that mpv and intel vaapi is in guix home, vaapi isn't working anymore. interesting
<jpoiret>it's also moving lots of things around since building profiles for example is making a lot of hard links
<jpoiret>brb building qtbase@6.5.0 and don't want it to fail
<Guest19>jpoiret okay, it is because of packages that have tons of files (papirus, texlive).  thanks for helping out.  Do you may also know why mpv does not detect intel vaapi in guix home? If I have intel vaapi installed with guix install intel-vaapi-driver it works although mpv is installed with guix home reconfigure.  But if I add intel vaapi it does
<Guest19>not detect it.
<unmatched-paren>Guest19: perhaps you need to resource ~/.guix-home/setup-environment?
<Guest19>unmatched-paren ah thanks, now it works.  I thought it does that automatically like Guix if you install something with guix install.
<unmatched-paren>Guest19: it's not possible for a program to source something in your shell
<unmatched-paren>source/. isn't a program, it's a shell builtin that modifies the environment of the shell
<unmatched-paren>so there's no way for guix to "auto-source" the script other than to add it to the login file so that it gets sourced next time you log in
<jpoiret>And since vaapi drivers created a new environment variable, you probably need to relogin for the change to take effect everywhere
<Guest19>Ah okay, so it is a little bit different than just packages?  If I for example install htop I can directly use it
<unmatched-paren>Guest19: no, it's the same
<unmatched-paren>if you install htop, there's already an env var PATH that points to ~/.guix-profile/bin
<unmatched-paren>when you do the guix install, it doesn't change the var
<unmatched-paren>it changes the path
<jpoiret>Same if you already had another package that relied on whatever env var vaapi uses
<unmatched-paren>so the shell will recognise the change because it doesn't require a new environment variable
<unmatched-paren>whereas if you installed vaapi, assuming it's got some variable like VAAPI_PATH, if you don't already have a package that defines that var, it'll need to be added to the etc/profile
<Guest19>ah I see, I now have LIBVA_DRIVERS_PATH, and because it is a new env var I need to resource it.  Got it
<unmatched-paren>but there's no way for guix to get all programs to automatically setenv
<unmatched-paren>(of course)
<unmatched-paren>Guest19: right
<unmatched-paren>did you install vaapi with guix install before you last logged in?
<Guest19>Awesome, thanks for the explanation.  Makes stuff so much easier if you know how it works
<unmatched-paren>if so, then that would have been added to ~/.guix-profile/etc/profile then, to be sourced by .profile
<unmatched-paren>same with home
<unmatched-paren>except it adds that to ~/.guix-home/profile/etc/profile
<Guest19>I installed intel vaapi several days ago, so yes.
<unmatched-paren>and uses the directory inside ~/.guix-home/profile, not ~/.guix-profile
<unmatched-paren>Guest19: right. when you first installed it, it wouldn't have worked
<Guest19>I wonder, is that the reason sometimes that "hint source ..." comes up?
<unmatched-paren>i think guix produces that hint whenever it detects that it had to add a new env var
<Guest19>ah smart, well shouldn't that be in guix home as well?
<unmatched-paren>possibly, yes
<Guest19>thanks for taking the time to explain it to me
<unmatched-paren>no problem :)
<glenda>unmatched-paren is a hero
<Guest19>do you may know why vt 7 is the default in guix and not 1?
<glenda>unmatched-paren: By the way, it's not smtp trouble on my mail service provider end.
<glenda>They tested the same configuration as I did, and it worked.
<glenda>Why doesn't works for me? A mystery.
<Guest19>maybe to at a little bit extra information to my question, I wonder if hit has to do because of historically reasons, since why should one start with 7
<glenda>Guest19: vt7 is default for X everywhere (almost)
<Guest19>ah okay.  Well my last one was Gentoo and they had had it on 1.  Okay my bad.  Didn't knew that distribution changed it
<glenda>not sure about gentoo
<glenda>Trisquel use vt7 too, I think gdm just use vt7 always.
<glenda>dietpi, freebsd, openbsd, use vt1
<Guest19>on vt12 there is log output, is that just syslog?
<glenda>I would put dmesg there if I used it
<glenda>brb, the hero unmatched-paren will answer you for sure
<Guest19>okay, thanks
<jpoiret>damn, I'm getting "fatal error: Terminated signal terminated program cc1plus" while building qt, no idea how to debug this
<jpoiret>a single file that requires 4GB to build??
<jpoiret>like .cpp to .o
<unmatched-paren>Guest19: i don't know, sorry :(
<Guest19>unmatched-paren no problem
<tex_milan>jpoiret: you might run out of memory. some link optimizations can be huge. whole application is linked together in one step and then optimized.... takes lot of memory... (link time optimization)
<Guest19>jpoiret maybe not enough RAM or disk space?
<mirai>are there any examples of transient services?
<jpoiret>apteryx: nah, it's just a single object file that's being produced i think
<jpoiret>oops, wrong @, i guess they left and i didn't notice, sorry
<jpoiret>yeah it's definitely an out of RAM error, although my OOM killer didn't report anything out of the ordinary
<apteryx>I'd like to have our syslog service put its config at a static place, e.g. /etc/syslog.conf, so that we can hot reload it
<apteryx>would a one-shot service for creating /etc/syslog.conf and added as a requirement to syslog-service-type be a good way to accomplish this?
<apteryx>ah, for other services (e.g. pulseaudio), that's done via extension of etc-service-type
<apteryx>that's probably better
<glenda>unmatched-paren: What email client do you use?