IRC channel logs

<oriansj>sneek: sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek

<oriansj>sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek sneek

<leoprikler>I think you might have a problem.

<nckx>Easy now. Flooding's still flooding.

<mehlon>well we only need two sneeks in order to make them loop and instantly get banned from freenode

<mehlon>wait no that wouldn't work

<apteryx>nckx: indeed for QUERTY, Wikipedia has it: https://en.wikipedia.org/wiki/Magic_SysRq_key

<oriansj>sneek: tell sneek I sneek you

<sneek>You just did.

<leoprikler>In regards of sneek's source, I found http://savannah.nongnu.org/projects/bobotpp/

<mehlon>a nonfree bot? blasphemy!

<mehlon>oh

<nckx>apteryx: Really not to be pedantic, promise, but because ‘U’ is quite… significant in SysRq land to say the least: it's QWERTY.

*nckx doesn't actually know what SysRq-W does, if anything.

<oriansj>mehlon: on an nonfree IRC server too

<oriansj>how will we survive/

<nckx>oriansj: Is sneek part of the bootstrap chain yet? If not, what's your excuse?

<nckx>Why do you hate sneek?

<mehlon>sneek: compile guix

<apteryx>nckx: oops. I've even forgotten how to spell it.

<nckx>sneek: Package npm.

<mehlon>sneek: tell s‎n‎e‎e‎‎k to release source code

<sneek>You just did.

<nckx>sneek sure has attitute.

<nckx> https://en.wikipedia.org/wiki/Magic_SysRq_key has a Dvorak mapping, nice.

<leoprikler>sneek later tell sneek to release source code :)

<mehlon>ah, it even ignored the secret unicode I hid in that sentence

<sneek>Weirdo.

<apteryx>nckx: works! and the SysRQ action is logged to /var/log/messages

<apteryx>I'll be able to do what should be part of Linux's job, and calling the OOM killer when it oughts to be called.

<nckx>apteryx: Yay! It's in dmesg (--follow) too, which doesn't rely on a working syslog + file system, in case you ever find yourself in that unfortunate situ.

<oriansj>nckx: it never will be. I enjoy reading books too much

<nckx>☹ But I read that bots would soon write all our software.

<nckx>Then I looked around at the quality of all our software

<nckx>I just naturally assumed it would be sneek.

<oriansj>nckx: and realized they already do...

<mehlon>I doubt bots will write software though

<leoprikler>sneek write the source code for sneek

<mehlon>once you have machine learning why even bother with source code?

<mehlon>just add more data

<oriansj>mehlon: unless you know what a virtualized private static inherited templated smart pointer is used for

<drakonis>mehlon: i'm having some issues installing guix that i assume you've dealt with

<mehlon>if computers do write software they won't need programming languages, they'd just output machine code directly

<oriansj>and why it is used 577 times in a single source file

<drakonis>i cant add the public keey

<mehlon>for gnupg for verifying the iso?

<drakonis>yes

<drakonis>i should probably comment out the function that invokes iso verification on the script

<drakonis>that typo tho

<drakonis>shameful

<nckx>drakonis: What did you try, and which error messages did you get, if any?

<nckx>oriansj: ‘Smart pointers‘. You'd better have made that shit up, young man.

<nckx>Or we will have words.

<nckx>*swords

<drakonis>'gpg --keyserver pool.sks-keyservers.net --recv-keys 3CE464558A84FDC69DB40CFB090B11993D9AEBB5'

<mehlon>I think that's something like garbage collection but different

<drakonis>'gpg: keyserver receive failed: Server indicated a failure'

<drakonis>there

<mehlon>or something like Rust's ownershiop but different

<drakonis>that prevents me from including ludo's signing key

<nckx>drakonis: Interesting, I pasted that exact command and got: https://paste.debian.net/1124684/

<nckx>on 2 machines.

<drakonis>mind you that i'm on nixos right now

<nckx>Are you behind some spoopy firewall?

<nckx>Hm.

<drakonis>that happened home too

<drakonis>no spoopy firewall over here

<nckx>Wish I could say more than ‘works for me’.

<drakonis>a shame

<drakonis>i need to cook up a functioning iso

<apteryx>nckx: well, that didn't work well. SysRQ - F wouldn't help getting out of a hang caused by building a Guix package. I had to resort to the RSEIUB trick

<apteryx>guix search earlyoom

*nckx always types REISUB is this the new emacs:vi.

<nckx>apteryx: Did you disable swap?

<leoprikler>RSEIUB makes more sense, because then programs can still corrupt your system like you'd expect them to ;)

<nckx>That's much more UNIX, thank you leoprikler!

<apteryx>nckx: I don't use swap (and don't want to). This is a tiny 55 GB drive.

<nckx>Sure.

<apteryx>I'd just like the kernel to do its job at killing fat processes that use all my RAM when I run out of it, but it seems that this isn't one of the strength of the Linux kernel.

<nckx>It likes to go through an utterly pointless and moronic ‘what if I wantonly drop all caches and wildly flush every last byte to disc’ phase first.

<apteryx>I'll look at packaging earlyoom, I've read reports of it being useful as a userland out-of-memory killer.

<nckx>Cool.

<nckx>apteryx: You can also try setting overcommit_* to 0 if you haven't yet.

<leoprikler>where would one do that?

<apteryx>it reads good: https://github.com/rfjakob/earlyoom

<nckx>I dunno, lemme search the net.

<nckx>Actually I just searched /proc, it's slightly smaller: find /proc/sys -name \*overcommit\*

<apteryx>there's also this from Facebook, that has already have the kernel component merged since 4.20: https://github.com/facebookincubator/oomd

<nckx>Hm, it's already at 0 here. Maybe not then.

<nckx>This is just ridiculous.

<apteryx>it requires "PSI" to be enabled in the kernel, I don't yet know what that is or if we have it in the default kernel config of Guix

<nckx>BSD: not unfairly mocks Linux for even having an ‘OOM killer’. Linux: adds a user-space daemon for it.

<leoprikler>how is OOM handled in BSD?

<nckx>apteryx: It's enabled by default in Guix.

<apteryx>yay

<nckx>leoprikler: Oh, I dunno. I got tired of their ranting. For all I know they reboot the box [they don't] and are proud of the brutalist elegance of that code block.

<nckx>I'm pretty sure there's no common BSD solution anymore.

<leoprikler>sounds like something *BSD would do

<nckx>Skimming https://svnweb.freebsd.org/base/head/sys/vm/vm_pageout.c seems pretty oom-killy to me. Who knows, perhaps I was told a half-truth *crowd gasps* by BSD elitists *‘surely not!’* *someone faints*.

***daviid is now known as Guest82422

<nckx>‘The pagedaemon threads randlomly select’.

<ng0>it's almost as if documentation on this exists by the systems and some share details etc.. https://www.netbsd.org/docs/kernel/uvm.html

<nckx>Yes, excuse me for getting sidetracked by the biggest BSD, by far, first.

<nckx>To be fair my toaster never OOMs.

<ng0>i can only guess that freebsd did this because they implement lots of linuxish things

<nckx>My sentiments exactly.

<nckx>ng0: Now see, that's actually interesting! Thank you.

<PotentialUser-52>how can I rebuild a package? (ignoring substitutes?)

<ng0>i mean, i hope my reply didn't read too sarcastic? this wasn't my intention

<ng0>lots of work happened since 4.4BSD, and details and implementations are shared, but not so much in all places

<ng0>but i guess you know that

<nckx>There

<nckx>is

<nckx>room

<nckx>for

<nckx>impro

<nckx>vement

<nckx>in

<nckx>UVM’

<nckx>s handling

<nckx>of

<nckx>this

<nckx>condition.

<nckx>The

<nckx>BSD

<nckx>VM

<nckx>system

<nckx>“handles”

<nckx>this

<nckx>condition

<nckx>by

<nckx>ignoring

<nckx>it.

<nckx>If you

<nckx>run

<nckx>a BSD

<nckx>VM

<nckx>system

<nckx>out

<nckx>of

<nckx>virtual

<nckx>memory

<nckx>the

<mehlon>what's all this then

<nckx>system

<nckx>will

<nckx>deadlock.

<nckx>Oh no, I'm so sorry.

<nckx>I just copied that from the PDF.

***ChanServ sets mode: +o nckx

***nckx was kicked by nckx (nckx)

<nckx>Only fair.

***ChanServ sets mode: -o nckx

<nckx>PotentialUser-52: Does --check (with --no-grafts for good measure) do what you want?

<ng0>this is the '99 dissertation right

<bandali>the fuck

<bandali>:p

<ng0>because uvm(9) suggests there has been development since then, in case you want to read more.

<nckx>Yes. So I was going to paste that as a ‘obviously things have changed’ illustration.

<bandali>nckx, ERC?

<ng0>:D

<nckx>bandali: No, I blame IceCat's PDF viewer, not my client (HexChat).

<PotentialUser-52>nckx: guix --check?

<nckx>PotentialUser-52: guix build --check --no-grafts the-package.

<PotentialUser-52>heh I'll let you know (it's gcc-toolchain, so I expect it to take awhile)

<nckx>This will rebuild it and complain if the result doesn't match what you already have (substituted or not).

<bandali>nckx, ha :D

<PotentialUser-52>odd it was quick...

<nckx>PotentialUser-52: gcc-toolchain is a meta-package, so it probably just created the tiny merge union again.

<PotentialUser-52>ohh, yeah. I'm getting segfaults while building... so I'm trying to rebuild stuff w/o substitutes

<PotentialUser-52>woefully understudied in guix though

<nckx>If you actually want to rebuild gcc, use gcc on the command line instead of gcc-toolchain.

<PotentialUser-52>"guix build: error: gcc: unknown package"

<nckx>Ah fuck of course.

<nckx>I always forget.

<pkill9>heh

<nckx>PotentialUser-52: Use ‘-e "(@ (gnu packages gcc) gcc)" ’ instead of ‘gcc’.

<nckx>gcc is hidden to protect users against themselves *mumbles*

<mehlon>well see ya guixers

<nckx>o/

<PotentialUser-52>ahh. thanks. that appears to be building something :)

<PotentialUser-52>hmm.. that fails @ "error: C compiler cannot create executables". I'm not sure if this is my issue, will try some other things first. thanks for the commands

<nckx>PotentialUser-52: This really feels like a hardware or corruption issue.

<nckx>‘sudo guix gc --verify=contents,repair’ will scan your entire store & check it against stored checksums.

<PotentialUser-52>yeah I'm thinking it's the substitutes I recently enabled

<nckx>I'd be very sceptical of Guix happily unpacking corrupted substitutes. Plus the build far would be segfaulting into obvlivion if the ‘source’ binaries were corrupt.

<nckx>But who knows.

<nckx>s/far/farm/

<PotentialUser-52>ok running that

<nckx>It'll read your entire store so will take a while. 'spesh if you have a rotating HDD.

<PotentialUser-52>I don't, but have a quick (10m) task to complete so it's perfect :)

<PotentialUser-52>that exited 0 (not much output except it downloaded a substitute). there isn't much in my store + I gc'd recently

<nckx>I'm still betting hardware. We have a ‘memtester’ package in Guix that can test your currently unused RAM while your OS keeps running, or a more comprehensive memtest86+ package that requires you to reboot and probably assumes you know your way around the GRUB or kexec command line (I haven't checked).

<pkill9>i'm gonna take the titanic approach and assume guix substitutes are uncorruptable

<leoprikler>I think you could verify them based on their hash

<raghav-gururajan>Folks! While building gnome-photos, at build phase, I get error "failed to load "../gnome-photos-3.34.0/src/./../data/dnd-counter.svg": Couldn?t recognize the image file format for file ?../gnome-photos-3.34.0/src/./../data/dnd-counter.svg?"

<raghav-gururajan>How do I make the meson build system to recognize 'svg' file types?

<PotentialUser-52>nckx: memtester install fails w/a segfault. I'll see if I have memtest86+ available on any of the grubs I currently have

<leoprikler>raghav-gururajan: check inputs, perhaps something with svg is missing

<leoprikler>if gnome-photos doesn't deal with svg by itself, try gdk-pixbuf

<raghav-gururajan>leoprikler I tried libsvgtiny and librsvg as input. Did not work. Gonna try gdk-pixbuf

<raghav-gururajan>Oh wait. leoprikler I already have gdk-pixbuf as input.

<leoprikler>and is that built with svg support?

<raghav-gururajan>I have to check.

<raghav-gururajan>I think it need cairo. let me try that

<raghav-gururajan>nope. didn't work.

<raghav-gururajan>leoprikler Yes, it is built with svg support.

<leoprikler>okay, let's play the old game where you share your package again

<raghav-gururajan>just a sec

<raghav-gururajan>leoprikler https://bin.disroot.org/en?ef4a19fe64ca56fd#B7WVitsbXnpXG4vgzWqCR4XT8f1YFu9qwPziU6cyXUmd

<leoprikler>your pastebin server hates me

<raghav-gururajan>lol ?

<nckx>‘Could not get paste data: server error or not responding’.

<nckx>(Yes, I'm secretly playing along at home.)

<leoprikler>sneeky spy listening on a public and logged IRC channel

<raghav-gururajan>leoprikler nckx https://bin.privacytools.io/?fada7f7265e39b8d#nlcFOQNKCO8q9mRBgE4qFbR5+ALaId41G3ys6Rt2HPg=

*nckx goes back to beating the installer with fork instead.

<leoprikler>raghav-gururajan: that's not going to work

<leoprikler>you've just introduced a circular dependency

<raghav-gururajan>leoprikler Which one?

<leoprikler>gtk <-> gimp

<leoprikler>ah i mean gnome <-> gimp

<raghav-gururajan>you mean in use-modules?

<leoprikler>yep

<str1ngs>raghav-gururajan: did you need tell the plugin to generate pc files?

<pkill9>does anyone have an example configuration for running a web server hosting a static website?

<raghav-gururajan>str1ngs Nah, I added some missing dependencies and updated the version. It worked. ;-)

<str1ngs>raghav-gururajan: ah that's good

<leoprikler>raghav-gururajan: the bug appears to be missing info in glib:bin concerning the to-pixdata preprocessor

<nckx>pkill9: Does the Guix Web site itself count? It runs Haunt to update the site hourly, although configuration is split up between the maintenance (hydra-berlin.scm) and artwork (website/.guix.scm) repositories.

<nckx>s,hydra-berlin,hydra/berlin,

<leoprikler>and I got it past that step

<nckx>pkill9: That uses nginx for everything. It might still be overkill if by ‘static website’ you mean ‘my hand-crafted handful of HTML files’.

<pkill9>where is this config nckx ?

<pkill9>oh nvm found it

<nckx> http://git.savannah.gnu.org/cgit/guix/maintenance.git/tree/hydra/nginx/berlin.scm#n497 is a simple static Web site.

<nckx>In the ‘bunch of HTML’ sense.

<nckx>(Ignore the wild complexity that surrounds it; I just don't know of a similar, public, super-simple single-site example.)

<raghav-gururajan>leoprikler I have no idea what that is.

<leoprikler> https://bin.privacytools.io/?8fd7f259a5f6961e#2jHgVssxqoOiPyTm6aEgX3vrl+sQmfhxuOYWcQK5IyA=

<leoprikler>gets until test, there it lacks python-dogtail

<leoprikler>your call whether to package that or disable the test

<raghav-gururajan>leoprikler Thank you

<leoprikler>np

*leoprikler → 😴️

<goldenshimmer[m]>Hi! I'm new to Guix, and a few questions have arisen I don't find obvious answers to in the manual/mailing-list... My goal is to configure Guix to ensure that all installed packages, and their direct and indirect dependencies, starting from the binary bootstrap seed, are always available in source code form in my computer.

<pkill9>can anyone provide an example configuration for `guix deploy`? I can't login with ssh, do I need to permit root login or something?

<pkill9>as in, guix deploy can't login with ssh, i can manually

<goldenshimmer[m]>First, is this possible? Is it fairly simple/maintainable? I'd basically just like "guix build --sources=transitive [everything-installed]" to be automatically ensured. Second, does "guix build --sources=transitive" for a given package make sure everything's present down to the binary bootstrap seed, or does it skip anything?

<nckx>goldenshimmer[m]: Hullo, and welcome to #guix! If nobody has any ideas here (we tend to be busier in the Euroday, and not everyone uses IRC), feel free to copy <help-guix at gnu.org>.

*nckx → 😴 themselves.

<pkill9>yes i think this is why

<goldenshimmer[m]>nckx: Ok, thanks!

<goldenshimmer[m]>Are the sources used to build things (after cleaning; the ones from "guix build --sources") always real source code, or might for instance a package maintainer for a Web app include a minified JQuery file rather than the source code? Is there a policy on this? Thank you!

<wigust>goldenshimmer[m]: generally we try to unbundle such things, but if it requires a lot of work then it could be delayed and accepted with bundles

<goldenshimmer[m]>Ah, good to know I should be on the look out for them; thanks wigust!

*raghav-gururajan is looking for any folks with commit rigths around.

<pkill9>when running `guix deploy`and you set it to build remotely, does guix send the .drv's to the remote guix and then the remote guix builds those?

<drakonis1>duh i know why it didnt work

<drakonis1>the problem is that i'm elevating the wrong command

<drakonis1>i didnt elevate gpg, i did elevate wget like a chump

<jackhill>pkill9: My reading of the manual is that the build-locally? field of machine-ssh-configuration controls where the build happen

<jackhill>speaking of the manual, isn't the version of it from master online somewhere now? I can't seem to remember where it went/find in in guix-artwork.git

<apteryx>earlyoom packaged. I guess now we need a service.

<oriansj>#bootstrappable needs more scheme programmers talk to our recruitment officer to learn more.

<alloy>Hello Guix! Is there a better way for modprobing acpi_call installed as a system package opposed to useing modprobe's --dirname option specifiying the system profile root (/run/booted-system/profile)?

<leoprikler>alloy – I assume you meant /run/booted-system/kernel?

<leoprikler>you could symlink /run/booted-system/kernel/lib/modules into /lib/modules

<alloy>so, when i invoke modprobe it finds modules in /run/booted-system/kernel, at least it says so. But when I install a package which contains modules they get packed to /run/booted-system/profile/lib/modules..

<alloy>But a symlink would be an option, thanks. Just thought there might be some kind of cleaner way to do it

<leoprikler>alloy: kernel only gets updated on reboot

<leoprikler>so if you reboot after installing your kernel module, you should be able to load it

<leoprikler>(i think)

<leoprikler>either that, or your package is actually not part of the kernel as guix understands it

<leoprikler>in which case you'd have to add it to that

<brown121407>Is there any way to hide /gnu/store/<long-string-thingy>- from Guix's output?

<brown121407>So instead of "building /gnu/store/blabla-profile.drv" it would show "building profile.drv"

<cbaines>brown121407, if you're using Emacs, there's a mode to abbreviate the store paths

*kmicu is happy to see some increased activity on #guix tag on the fediverse.

<numerobis>Hi #guix! Is there an equivalent to the archlinux 'pkgfile' command, that gives the package names associated with a certain package? (At the moment I'm looking for a package to give me access to 'htpasswd', and a quick 'rg' on the guix tree did not return anything.)

<str1ngs>numerobis: as far as i understand no. this would require some kinda cache on the publish server. you can though use tools like find to search /gnu/store eg find /gnu/store/ -type f -name bash. but this will not work for packages that have not been built/installed locally.

<pkill9>i have a problem with Guix System: the purpose of an operating system is to get out of the way of the user as much as possible, so that they can get on with doing whatever they want to do, and Guix System does this so well that I have to face the fact that I am wasting my time ☹️☹️☹️☹️☹️☹️

<leoprikler>pkill9: easy solution, install a gentoo vm on top

<mehlon>install bedrock linux

<numerobis>str1ngs: many thanks!

<numerobis>good enough!

<mehlon>you'll have plenty to do

<str1ngs>numerobis: to clarify I meant searching /gnu/store *will* only work for install/built packages

<nckx>numerobis: Someone is being paid actual (euro)bucks to work on that feature, amongst other things.

<nckx>(Not by us, by the Dutch internets.)

<apteryx>most of the services run as root in Guix, correct?

<apteryx>I see; the make-forkexec-constructor takes a user argument. Example allowing configuration of user in the configuration: rsync-service-type

<apteryx>I seem to recall there was a way to start the service in a chroot (namespace)...

<nckx>apteryx: https://guix.gnu.org/blog/2017/running-system-services-in-containers/

<nckx>?

<nckx>(Never done so.)

*apteryx reads

<pie_[bnc]>this is why i want guix to be good https://twitter.com/mietek/status/1213855758158352385 otoh would prefer a lazy language :(

<apteryx>hmm, wondering where the to-be-written earlyoom service should reside... in a new gnu/services/linux.scm file, perhaps?

<apteryx>it's tied to the Linux kernel

<nckx>apteryx: Sounds good!

*apteryx produces keyboard noises

<leoprikler>pie_[bnc]: just wrap anything in (lazy ...) 😉️

<nckx>pie_[bnc]: Hey now, we have undebuggable error messages too!!

<leoprikler>heh

<nckx>pie_[bnc]: That looks very much like a VM I have set up to explore the world that could have been. GNU could have been a Lisp system, & all that… Now I want to play with it again.

<civodul>nckx: GNU *can* be a Scheme system :-)

<kirisime`>Does guix have a plymouth service available?

<nckx>civodul: Running Guix is as close to ‘running GNU’ as I've ever felt. 🙂

<pie_[bnc]>nckx: hooray

<nckx>kirisime`: No.

<kirisime`>My laptop lid is about to fall off again and I don't think the screw threads are going to last much longer, so I thought that I'll just glue the lid shut and pretend it's a linux tablet when it can no longer be fixed in place, but I'd then need a way to punch in my disk encryption passwords without a keyboard...

<kirisime`>Well, since grub is on the crypted drive as well that wouldn't work anyways.

***kirisime` is now known as kirisime

<terpri>as it happens, last week i started working on packaging the vlm for guix (just needed an nfs service and a few other additions)

<terpri>1. make genera usable on modern gnu/linux systems 2. liberate the source code 3. ???

<mehlon>profit.

<nckx>terpri: Yeah, I was talking about the same thing but, y'know, non-freeness & all. But awesome!

<dftxbs3e>nckx, hey, got news from OSUOSL?

<mehlon>are we making lisp machines?

<terpri>fun fact: as of february 2018, the federal government apparently has full rights to the genera source code, so step 2 is theoretically possible if one finds the right person to petition...

<nckx>dftxbs3e: Yes! As it happens, I have some time reserved tomorrow to log in and lern me some OpenStack.

<nckx>terpri: No. Way.

<dftxbs3e>nckx, cool! you could've asked them to deploy a machine for you too

<nckx>I know, I know, what can I say.

<dftxbs3e>I don't know if they host an Horizon Frontend, but that's quite easy to use

<nckx>I want to on my hands.

<dftxbs3e>Great

<dftxbs3e>Well I hope that's going to finally unblock the GCC bootstrapping issue!

<nckx>I'll send a mail to g-d as soon as there's a machine up & running.

<nckx>terpri: I assume you saw the NFS service that Ricardo posted to an ML? Did that do the job?

*nckx disproportionally excited now.

<terpri>nckx, haven't seen it, do you have a link?

<terpri>there's a (gnu services nfs) module but it doesn't actually provide an nfs service :)

<nckx>terpri: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=38902

<janneke>dftxbs3e: there is a gcc bootstrapping issue?

<nckx>janneke: On POWER.

<janneke>nckx: ah, right; that's not on my radar yet; could have interesting impact on all bootstraps, though

*efraim finally got around to modularizing their os config files

<terpri>nckx, thanks, that looks perfect! hopefully with that i can get genera to boot under guix today :)

<terpri>incidentally, https://paste.debian.net/1124750 from the symbolics license boilerplate is the "february 2018" thing mentioned above, and https://www.law.cornell.edu/cfr/text/48/252.227-7014 is the relevant law afaict

<dftxbs3e>janneke, the issue is related to Glibc changing the long double float format for POWER8+ LE, so it's specific

<janneke>dftxbs3e: ah, that's unfortunate. i should probably learn something more about the power9 gcc bootstrap

<dftxbs3e>janneke, the reason why it's the most annoying is that none of the GNU Guix troubleshooting work yet. So I'm left in the dark.

<dftxbs3e>And fixing GNU Guix's troubleshooting tools needs fixing the GCC bootstrap issue. Chicken and egg!

<apteryx>hm, where is 'ungexp' defined?

<brettgilio>Hey Guix!

*alextee[m] has a patch for 3 new python packages coming \o/

<brettgilio>alextee[m]: do you have commit access? If not, when I get time I'll look at your patches :)

<janneke>apteryx: guix/gexp.scm

<janneke>hello brettgilio

<brettgilio>janneke: you are one of the founders of GNU Lilypond right?

<janneke>brettgilio: yes

<alextee[m]>make that 4

<alextee[m]>brettgilio: nope, will send the patches later and ping you :-)

<apteryx>janneke: I already have (guix gexp) in my use-modules, there must be something else at play; thanks :-)

<brettgilio>janneke: I used it throughout college when I was getting my first degree in Music. Such a great thing :)

<brettgilio>alextee[m]: sounds good

<janneke>apteryx: ungexp is only available inside gexp (like unquote)

<janneke>brettgilio: thanks! it was great to work on, sadly no way to make a living off of that, at least not at the time

<apteryx>janneke: ah, that makes sense. thanks!

<alextee[m]>should these 3 python libraries go under python-crypto? https://github.com/jborean93/ntlm-auth https://github.com/requests/requests-ntlm https://github.com/apple/ccs-pykerberos

<alextee[m]>and this under python-xyz? https://github.com/diyan/pywinrm/

<leoprikler>sounds good to me, but I don't have any authority

<alextee[m]>well, ill do that then, can change later

<apteryx>alextee[m]: perhaps python-requests-ntlm should go next to python-requests in python-web.scm? Otherwise crypto makes sense also.

<nly>hi guix

<sneek>Welcome back nly, you have 1 message.

<sneek>nly, janneke says: thanks for putting the manuals up, that looks great.

<brettgilio>raghav-gururajan: does your gnome-boxes patch work?

<brettgilio>Because it requires libvirtd

<nly>happy new year folks

<nly>sneek tell later janneke welcome! :)

<sneek>later, nly says: janneke welcome! :)

<nly>Can the file package be upgraded to version 3.8 from 3.3?

<alextee[m]>apteryx: oh good idea

<alextee[m]>how do you force guix build to rebuild a package if you've built it before?

<alextee[m]>--source?

<leoprikler>--check?

<nckx>alextee[m]: guix build --check --no-grafts

<brettgilio>Just ran gc for the first time in awhile. Let's see the damage

<brettgilio>47,546.89342 MiBd

<nckx>alextee[m]: Why do you ask?

<alextee[m]>thanks. nckx i built a package in another directory and im adding it to the guix sources to prepare a patch, just want to make sure it can find the dependencies and whatnot

<PotentialUser-26>hi everyone!!!!!

<PotentialUser-26>what about kde in guix ?

<pkill9>just needs contributers to work on it

<alextee[m]>brettgilio: patch sent

<NieDzejkob>alextee[m]: if it skips the build and gives you the /gnu/store path immediately, that means that the environment in which the package would've been built was exactly the same, so in this case there wasn't really any point

<alextee[m]>NieDzejkob: well, just to be sure, since there's 4 new packages in different directories. im still not 100% sure how the environment works

<brettgilio>Forgot ZNC also stops when I shut off my computer. Lmfao

<pkill9>i want to run guix system on a pinephone

<pkill9>or maybe just guix on a foreign distro, not sure

<castix>hi there, i'm trying to make a package, when i run `guix package -f file.scm` I get `guix/base16.scm:36:4: In procedure bytevector->base16-string:

<castix>In procedure bytevector-length: Wrong type argument in position 1 (expecting bytevector): "j77tkews4ukyabkdjhl3bciek2uf3bss6znai5u5d7vg5fj2iqwa"` what can this mean?

<castix>i got that hash from `guix hash -rx -f base32 .` in the git cloned directory, i inserted this hash into (sha256 call in file.scm

<terpri>castix, what does your sha256 call look like? iirc you can get an error like that if you put a plain string in the field

<terpri>it should be something like (sha256 (base32 "j77tkews4ukyabkdjhl3bciek2uf3bss6znai5u5d7vg5fj2iqwa"))

<castix>ok, then my issue is that if I keep base32 call I get `guix package: error: exception thrown: #<condition &invalid-base32-character [character: #\u string: "j77tkews4ukyabkdjhl3bciek2uf3bss6znai5u5d7vg5fj2iqwa"] 2f48570>`

<mbakke>castix: the hash used in packages is actually "nix-base32", which is the default if you do not pass the "-f" argument to 'guix hash'

<castix>ok, just got it, thanks both, issue solved :)

<NieDzejkob>How would one usually setup per-user shepherd?

<lxsameer>hey folks, is there any smaller installation image

<lxsameer>?

<lxsameer>for the guix SD

<brettgilio>lxsameer: not that I know of. The guix installation image is mostly big because of our bootstrapping situation with gnu coreutils and corelibs. The reduced seed binary project should make future images smaller.

<brettgilio>But in the mean time, that is about as small as it can get

<lxsameer>damn it. thanks any way

<kmicu>Hi lxsameer: what size limits are interesting to you?

<lxsameer>kmicu: under 1G

<brettgilio>lxsameer: what for,m

<brettgilio>?

<lxsameer>brettgilio: sure

<brettgilio>lxsameer: what are you wanting a smaller image for?

<lxsameer>brettgilio: I just have 1G flash driver available to me right now

<brettgilio>lxsameer: I see

<lxsameer>A netinstall image like debian can be really cool in the future

<bandali>+1 for that

<brettgilio>lxsameer: thing is, our images are already equivalent to the netinst Debian images in functiin. Just not in size.

<kmicu>You could assimilate a non‑guix distro iirc. Though that’s not as simple as a Guix System Stick.

<brettgilio>kmicu: somebody wrote a tutorial on this for Debian. I don't remember who

<brettgilio>nckx: bandali ^

<lxsameer>thanks folks, It seams that I have to wait

<kmicu>Procedure is similar to installing Guix System on a VPS https://lists.gnu.org/archive/html/guix-devel/2017-04/msg00139.html

<oriansj>brettgilio: well I have guides for multiple distros to be installed on arbitrary media with minimal space

<oriansj>including to get debian to have an encrypted /boot

<bandali>brettgilio, right

<bandali>feel free to link these here oriansj

<oriansj>well here is an example for debian with encrypted /boot https://paste.debian.net/1124791/

<bandali>nice!

<oriansj>here is one for arch: https://paste.debian.net/1124792/

<kmicu>[jokin’) No nice at all; no guix in the end! Terrible xD

<oriansj>you will not be able to download the guix manifest.scm in that one kmicu

<oriansj>nor this one you jerk ;-P https://paste.debian.net/1124793/

<kmicu>We could help lxsameer by figuring out how to (easily) install Guix System when only 1GiB of installation media is available.

<lxsameer>cool

<lxsameer>how

<oriansj>kmicu: well if you downloaed and read the guix.scm you would know exactly how to do it

<kmicu>./guix.scm or ./build-aux/hydra/guix.scm?

<oriansj>One only needs 847MB for a minimal guix system; you just need to trim the fat as is explicitly specified in that last link

<oriansj>kmicu: read https://paste.debian.net/1124793/ carefully

<kmicu>wget 'newsagg.pdp10.guru/guix.scm' ?

<lxsameer>thanks

<oriansj>kmicu: yes that is included in the instructions. good you are reading

<kmicu>‘guix.scm’ was ambiguous plus wget 'newsagg.pdp10.guru/guix.scm' hangs here. I cannot access it.

<oriansj>kmicu: anyone can access it if the know the correct way to access it

<bandali>oriansj, actually no, that address times out for me also

<drakonis1>downloading NARs is sure taking a while here

<drakonis1>there's no way to make it parallel right now, right?

<oriansj>bandali: did you follow the connect procedure?

<bandali>oriansj, i’m just trying to wget that file on my local machine

<bandali>Connecting to newsagg.pdp10.guru (newsagg.pdp10.guru)|159.203.137.80|:80...

<bandali>it just timed out after a few minutes

<oriansj>bandali: It is a hardened server

<bandali>?

<kmicu>Sorry lxsameer I have no idea how to generate a minimal image for but please keep asking for few days maybe we figure something out and you won’t have to wait to enjoy Guix System 😴

<oriansj>There is an authorization procedure you need to follow before it can even start talking to you or recieve anything you sent

*kmicu went 😴

<lxsameer>kmicu: don't worry mate, I'll install it tomorrow

<bandali>oriansj, what kind of authorization procedure is it? care to elaborate? should i not be able to wget that file from outside a qemu vm? i don’t see anything mentioned in there

<oriansj>authorization for network access are never included in system setup instructions for public distribution.

<bandali>okay you’re clearly not interested in actually giving any explanation as to what that guix.scm file actually is or how to obtain it. fine

<oriansj>bandali: you would need an issued certificate to send a cryptographic port knock at the tunnel system and a user issued certificate to send a different cryptographic port knock (which is checkeed again your user access) to gain access to the system; at which point your authorized access ports are open to your IP address for the duration of the session+ connect time

<bandali>oriansj, what’s the point of that if this newsagg.pdp10.guru/guix.scm file is just a ‘template’ or an standard configuration of some sort? if you’re sharing that paste.debian.net/1124793 snippet here, why not make that guix.scm file accessible too for folks who want to peek at it?

<NieDzejkob>brettgilio: Wait, why does the bootstrapping path affect the installation image size? Wouldn't the image only contain the run-time, and not the build-time, dependencies?

<oriansj>bandali: sure here you go: https://paste.debian.net/1124798/

<oriansj>you can skip the luks bit if you like

<bandali>i see, cool, thanks for the link

<oriansj>I dare anyone else to find a way to make a smaller guixSD install

<leoprikler>oriansj: drop wireless-tools and use cable only

<leoprikler>(iw too)

<leoprikler>also drop nvi emacs et al. and use ed

IRC channel logs

2020-01-05.log