IRC channel logs

2020-02-14.log

back to list of logs

<shtwzrd>looks like it http://linux-libre.fsfla.org/pub/linux-libre/releases/5.5-gnu/ not that it couldn't have other potentially hairy parts ofc :)
<shtwzrd>For my purposes though I made a package that just uses the tsys fork directly, and inherits linux-libre package. But still I haven't booted yet so I can't vouch for it.
<ashjkaell>Hello friendly guix people (guixopolites?) :-).
<ashjkaell>A small question; What would you say is the way to incrementally develop a service?
<ashjkaell>I have a trivial shepherd-service that I'm writing (just running a script at boot, mainly to get used to writing services), and I wanted to load it in shepherd at runtime to test it out.
<nckx>bandali: Just because I'm curious, did you try to search for gandi.cli or just assume it wouldn't be in Guix (which I'd understand, in a way)?
<ashjkaell>I know about `herd load shepherd xxx.scm` but this does not seem to include the modules for gexps. Conversely, I just need the register-services function from guile, but in which module it it?
<leoprikler>ashjkaell: see the shepherd documentation and source code, specifically (shepherd service) for register-services
<leoprikler>It may already be defined in the environment, that executes load, however.
<bandali>nckx, i searched :-) i think guix is pretty much at a point where i generally don't assume something isn't packaged
<leoprikler>(meaning it might not be necessary to use-modules it)
<bandali>i've even been surprised to find packages for some of the more 'obscure' things in guix
<bandali>s/surprised/pleasantly surprised/
<nckx>Yay.
<leoprikler>I know something that isn't packaged. But it's because that thing has a build system so broken, I hardly understand how people use it at all.
<bandali>^_^
<nckx>Everything I care about is packaged in Guix. All done.
<leoprikler>Oh, and one obvious thing: unfree software, hurr, durr.
<bandali>ye
<bandali>btw, i wonder, does guix handle fonts with multiple variants?
<leoprikler>ashjkaell: If you're thinking about running shepherd from guix, though, that sounds like the wrong way to do things (outside your config.scm)
<bandali>um, actually, that's not a sensible question; i retract it
<nckx>bandali: What's there for Guix to handle?
<nckx>Oh.
<bandali>nckx, :p a font i'll be looking to package soon has multiple variants in subfolders of its release zip archive
<leoprikler>just build it with other configure flags?
<bandali>i guess for one thing i was wondering if the fonts build system looks recursively in a zip archive
<leoprikler>or package all variants and name them accordingly
<bandali>right
<leoprikler>(all together)
<nckx>For ‘super customisable’ fonts like Input I just keep my own copy in ~/.local/share/fonts; little point in creating x^y variant packages.
<nckx>bandali: Just install them all?
<leoprikler>or if you still want to maintain it through guix have a channel for that
<leoprikler>ashjkaell: you should be able to run guix code from your init.scm or other shepherd-related-stuff.scm
<bandali>right
<bandali>this particular one i'm looking at has four variants
<bandali>at the very least, they have different file names. i'll check and if they have different "font names" as well, i'll probably just have the package install them all
<bandali>guess i should check fonts.scm to see what is more common, packaging all at once or having separate packages for different variants/flavours
<ashjkaell>leoprikler: My thanks! Trying from init.scm seems like a good idea, I'll try to hunt in the source as well :-).
<nckx>bandali: *Which* variants? E.g. light vs. regular, or ‘has a slashed 0’ vs. not, or…
<bandali>nckx, both, but my question is about the latter: https://github.com/rastikerdar/vazir-font/tree/master/dist
<bandali>'cause of course it almost always makes sense for light, regular, etc variants of a font to be packaged together :-)
<nckx>That's what I thought but hey, you asked 😛 ‘Variant’/‘flavour’ is just so overloaded as to be meaningless.
<nckx>bandali: I don't know the first thing about Farsi, but aren't these just subsets of the ‘main’ font?
<nckx>If so I'd just install the biggest one.
<bandali>nckx, right :-) re subset: sort of; at least some are. i think some like me might actually prefer the "smaller" versions that don't carry e.g. latin glyps, since hinting with persian fonts isn't always great, and i often much rather use different fonts for persian vs. latin glyphs
<bandali>and i could see each of the variants of the above being favoured by different demographics
<bandali>the above *font
<nckx>If Arch is right (…) they can just be thrown together: https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=vazir-fonts
<nckx>¯\_(ツ)_/¯
<bandali>true :-) yeah i'll probably do that
<bandali>also, YAY looks like libreoffice is finally rebuilt on ci.guix :p
<nckx>Happy to provide valuable & meaningful assistance.
<bandali>.oO such value much meaning very wow
<bandali>nckx, more seriously though, one last question: does the font build system collect font files recursively from subdirs or just the root of the archive/repo?
<nckx>bandali: It *should* recurse and install any file with a font extension but I haven't tested that.
<nckx>It won't preserve subdirectories: foo/bar.ttf & fuz/fiz.ttf → out/share/fonts/truetype/{bar,fiz}.ttf.
<nckx>So make sure your file names are all unique, but in this case they are.
<bandali>nckx, right, cool. yeah, was gonna say, in the case of these fonts we should be fine; but that's good to know
<nckx>OMG.
<nckx>bandali: LibreOffice updated on master 😃
<bandali>nckx, you gotta be #$%*@ kidding me
<bandali>anyway, not really a problem for me :p i just finished pulling the second-last build and reconfiguring my system etc
<bandali>not gonna be pulling for a good while :p
<vits95>what is common to all guix? we are pulling. we just can't not to pull our guixes. ("8-bit joke")
<nckx>lfam quit in shame. /s
<vits95>Ok, no more.
<bandali>lol
<bandali>yeah it was a bit hard conditioning "pacman -Syu" of guix out of me
<nckx>vits95: lfam is the evil prankster who updated libreoffice, I didn't mean your joke 🙂
<bandali>shame shame :p
<nckx>Yes, it's so memorable.
<bandali>nckx, *_* i JUST saw your reply to my patch
<bandali>how did i not see gandi.cli before *_*
<vits95>nckx: what is the symbol at the end of your post? seems like i'm need to add some fonts (i'm see a "long square")
<nckx>How did my ‘did u even surch bruv’ question even make sense.
<nckx>vits95: Just a regular smiley face.
<nckx>If you want, installing font-google-noto will provide emojo support (except in emacs).
<bandali>nckx, lol honestly. i *think* i just looked in python-xyz, this one time :p i will fall back onto my general `grep' practice
<vits95>nckx: thanks, i'd forget about google-noto.
<bandali>re (not) propagating, thanks for the feedback. my main reason was addressing test suite failures, and also run-time deps as seen in their docs
<nckx>bandali: grep's fine but don't forget guix search.
<bandali>nckx, thanks
<bandali>nckx, like, i'm pretty sure executable from at least one of the propagated inputs in my patch was used somewhere in the cli
<bandali>not sure if it was git, ssh, or openssl
<nckx>See the existing package for openssl.
<bandali>yup, saw that
<bandali>i think openssl was needed for tests, and the other two were mentioned/used in the source in form of calls to the executable
<nckx>Maybe ssh is used by a function I never used. git… I'm extremely sceptical about.
<bandali>let me check
<nckx>Thanks, I'm curious.
<bandali>cheers
*bandali too
<bandali>nckx, ssh: https://github.com/Gandi/gandi.cli/search?utf8=%E2%9C%93&q=ssh&type=
<bandali>imma take a closer look for git
<bandali>git seems to be referenced a bunch in their "paas" modules
<bandali> https://github.com/Gandi/gandi.cli/search?utf8=%E2%9C%93&q=git&type=
<nckx>bandali: I don't see any calls to ssh in there, if you do could you point them out?
<bandali>nckx, will do, one sec
<bandali>also, the arch package had those as deps, so i didn't think twice, but i should have. https://github.com/Gandi/gandi.cli/blob/master/packages/archlinux/PKGBUILD
<nckx>So git is in fact called (only) by gandi/cli/tests/commands/test_paas.py (I guess you could look at my package to see why the test doesn't fail — is it skipped?), but there's still no reason to propagate it.
<nckx>Very broadly, propagation means: when installing me, also install this package into the profile as if the user typed ‘guix install <me> <this>’.
<bandali>right, i understand
<bandali>re ssh: i saw https://github.com/Gandi/gandi.cli/blob/dbae8ad2b62d49fe0fa333b255353c8d4286dacf/gandi/cli/core/utils/unixpipe.py#L152 and https://github.com/Gandi/gandi.cli/blob/dbae8ad2b62d49fe0fa333b255353c8d4286dacf/gandi/cli/core/utils/unixpipe.py#L178 so far
<nckx>Whenever you can avoid it, do so.
<bandali> https://github.com/Gandi/gandi.cli/blob/dbae8ad2b62d49fe0fa333b255353c8d4286dacf/gandi/cli/modules/paas.py has a bunch too
<nckx>bandali: Interesting. Am I trippin' or was that missing from <https://github.com/Gandi/gandi.cli/search?utf8=%E2%9C%93&q=ssh&type=>?
<bandali>no i'm pretty sure i found them there by searching for ssh :p did you see next pages perhaps?
<bandali>or ah it may be that github's search preview doesn't show all occurrences in the file
<bandali>... if that's what you were looking for on the search page..
<nckx>Yes.
<bandali>right. it sucks, but yeah that's why
<nckx> https://github.com/Gandi/gandi.cli/search?utf8=%E2%9C%93&q=ssh&type= is 2 pages and neither contained ssh_call.
<nckx>Good to know.
<bandali>ha, yeah it seems ssh_call is only referred to in gandi/cli/core/utils/unixpipe.py
<bandali>executed in _ssh_master_cmd in there nonetheless
<bandali>which is used in is_alive: """Check wether a tunnel is alive"""
<nckx>So: this will all work as long as the user [also] installs ssh, and should throw a clear ‘ssh: command not found’ error if it's not, so the user can install it. If that's the case I tend to defer to the user. That's just me though.
<bandali>i see
<nckx>It wouldn't be *wrong* to (substitute …) that ssh call with /gnu/store/…/bin/ssh, either. There's no hard line when such optional dependencies are concerned.
<nckx>What do you think?
*nckx thinks: 😴 o/
<bandali>right. yeah i'm not quite sure honestly
<bandali>i've thought a bit about optional dependencies and guix in general
<bandali>but i don't really have any good ideas myself
<bandali>and in this specific case, so long as the command would indeed fail with a clear command not found error, i don't think it'd be a big deal, at least not in a supervised (read human-facing) setting
<bandali>that it's used for an is_alive check, i'm not sure how critical that is or isn't, and how easily the error would surface up to the user of the cli if they don't have openssh
<drakonis>optional dependencies is something that would be good to have, there's a variety of packages that cannot pick up extra functionality when present in guix
<drakonis>say, libreoffice and dictionaries, software with plugins
<bandali>right
<drakonis>it'd have to be built into the package definition
<drakonis>list out all the optional dependencies that could be made available if installed
<drakonis>hmm, i'll go back to figuring out udev now
<nckx>drakonis: That's already covered by search-paths.
<nckx>bandali: Yep, for me the line is at ‘prints a clear 'foo is missing' error’ vs. ‘fails mysteriously’. In the latter case, a hard dep it is.
<drakonis>hmm, i'm still running into some trouble with figuring out how to make a service include fixed udev rules
<drakonis>heck, a package even
<nckx>Optional dependencies don't make sense in Guix: dependency is a binary state. Parameterisation will cover what other PMs try to do (badly) with optional deps.
<nckx>drakonis: What do you mean by ‘include’ and ‘fixed’?
<nckx>I'll assume you've tried (extensions (list (service-extension udev-service-type foo-package))) and/or it's not what you want.
<drakonis>i have a service that needs to introduce non modifiable rules
<drakonis>the package doesnt come with udev rules by itself.
<drakonis>its a messy messy piece.
<nckx>What are modifiable rules?
<drakonis>declarable rules?
<drakonis>i would like to install a specific set
<nckx>Let me formulate it this way: what do you require beyond the above udev-service-type extension, and why?
<drakonis>hmm, i dont think i've understood how to use udev-service-type correctly
<nckx>I don't know if it's the best example, but it's a minimal one: take a look at sane-service-type, it does nothing but provide some udev rules from the sane-backends package and create a user.
<nckx>If that's not what you want, maybe you can explain it in comparison.
<nckx>s/user/group/
*nckx really 😴 now.
<drakonis>huh i think i figured it out
<bavier`>the patch-dot-desktop-files phase issues a misleading "warning of reference to '' which cannot be found" if the Exec program listed is an absolute file name
<bavier`>"/doc/guix.es.texi:20971: @samp missing closing brace"
<vits95>bavier`: do absolute file name should be like /run/current-system/profile/bin/SOMETHING ?
<commanderkeen>i get no pinentry with gpg. i tried installing several different ones in the repo and it still gives error for no pinentry. any ideas?
<lfam>commanderkeen: You have to set a value for 'pinentry-program' in the gpg-agent.conf file
<commanderkeen>lfam: thanks. that worked
<str1ngs>janneke: in emacsy the <internal-window> has a slot named orientation which is nice. but then I can not superclass it with <gtk-box> which has the same slot name. I've tried to avoid modifying emacsy upto now. but I wonder if it might be more useful to rename this slot.
<str1ngs>janneke: by superclass like so (define-class <my-internal-window> (<internal-window> <gtk-box>)) it avoids extra methods and slots for my custom class. hopefully this makes sense.
<str1ngs>janneke: <window> suffers from the same issue. WDYT about renaming this slots?
<civodul>Hello Guix!
<adamantium>Hi, where is the best place to get some help with shepherd? I'm trying to get it up and running as pid1 on a non-guix distro
<adamantium>I can boot it as pid one and get a hello world, but I can't find any minimal /etc/shepherd.scm init scripts to work from, the guix scripts I can find online, but are very complex looking and do a lot of things I don't think I need.
<adamantium>Does anybody know if there is anyone not using guix who might be able to share a simpler shepherd.scm file?
<civodul>hi adamantium!
<civodul>adamantium: this is a good place to ask for help about the shepherd
<civodul>i think the manual and the source tree contain simple examples that you could start from
<civodul>but then you need to define the right services for your system
<civodul>on Guix, we have services to mount file systems, start udev, start mingetty, etc.
<civodul>you'd have to find a minimal set of services that works for you
<valignatev>I'm kinda interested in running Shepherd on non-guix distro as well. I'm planning to switch to Artix from Arch and try it out since I'm sure I won't be able to tear systemd from the mainstream Arch
<zzappie>Hello Guix!
<valignatev>o/
<zzappie>adamantium: valignatev: I've been searching for examples too. I found govuk-guix repo to be a nice reference point
<civodul>roptat: there is/was(?) a bug with guix.es.texi, making "guix pull" fail: https://ci.guix.gnu.org/log/7pc1rbsai2p6b99spx11d56m3pj2p35p-guix-manual
<zzappie>they have many services defind and also some scripts.
<valignatev>zzappie: Whoa, cool, thanks! Who would've known that UK gov so much into free software ^_^
<zzappie>valignatev: Yeah that is preety cool. Another example is, i've heard French government uses matrx as their main messaging platform. They wrote their own client
<valignatev>I'm sure civodul spreads free software blessings all over Europe :D Btw, happy Free Software day!
<valignatev>I love you people <3
<civodul>heheh, love free software and the people who make it! <3
<zzappie>I did't know about #ilovefs thing :) fsfe has a new look today
<smithras>wow >20,000 commits from the UK gov!
<nixo_>Hi guix! Is it expected that guix build -K does not offload?
<raghavgururajan>Hello Guix!
<zzappie>Hi raghavgururajan!
<smithras>hi raghavgururajan!
<civodul>nixo_: it is! it's a recent feature
<civodul>apteryx asked the same question recently
<nixo_>civodul: thanks! I was expecting some magic like building remotely and copying back the /tmp/build dir
<civodul>nixo_: there's no such magic, and that was the motivation for having -K imply --no-offload
<nckx>nixo_: Copying the /tmp/guix-build alone is easy, but you'd need copy all its dependencies back to the client as well. Probably doable if you're interested.
<civodul>nckx: the dependencies are already in the client
<civodul>so you're right that it's "just" about copying /tmp/guix-build
<nckx>Oh? Then I misread something badly.
<vits95>"hint: Run `guix search ... | less' to view all the results." -- I'm got the sense of this sentence only after the search returned one result for a query "emacs". This is looks like a lot less work for developers to do: no need to check which output will be "too long for a particular terminal, so call the $PAGER".
<civodul>calling out to $PAGER is controversial, though :-)
<nckx>DTRT always is.
<nckx>civodul: If that is true after all, I don't understand your ‘Right’ here: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39387#21
<roptat>civodul: oh! I was able to make it though
<civodul>nckx: i'm not sure either, it's ambiguous :-) but the dependencies are sent from the client to the build machine, so they should be there
<vits95>civodul: i'm remember only systemd-tools do this automatically. Why?
<civodul>and Git, too
<roptat>Please revert if I broke something :)
<vits95>i'd searched the Web with different "why bad to use $PAGER" queries, but didn't find any answer. My guess is that it isn't portable or such. Also in kitty terminal emulator Ctrl + Shift + h calls less to allow the user to scroll the output up and down.
<vits95>Instead of Shift + pgUp
<nckx>vits95: It's not bad.
*janneke has export PAGER=cat
<vits95>nckx: "controversial" yields nothing too.
<nckx>Heh.
<nckx>People who defend the current behaviour (if they even exist) should be forced to use an ‘ls’ that returns only the first entry anywhere, with a cutesy ‘hint: you didn't use the magic word, add | less and try again, tee hee!’. See how they like it.
<vits95>nckx: but if the first entry will be too long!? first 255 letters. One tool -- one job.
<kmicu>One tool = all jobs. Emacs 🤭
<vits95>nckx: In another hand, if this is an ARM single-bouard with RAM and CPU limited resources... the behavior of guix search is simple and it is good. In theory... i don't know.
<vits95>kmicu: BTW, i'm yet not ended with scrolling of "emacs-" packages yet...
<kmicu>I can only display 8 lines on my RISC‑V SBC with ISP RGB screen.
<vits95>kmicu: you mean that default behavior steel requires a pager, correct?
*kmicu is blissfully unaware of current HOT PAGER CONTROVERSY. But it looks hot and important.
<vits95>"But it LOOKS hot...". catch the Cyborg, people!
<civodul>we got to this point because before that "guix search" would spit out entire pages of results
<civodul>and you'd be left with the least relevant results on the screen (because that's sorted)
<civodul>what we have now is a bit like "i'm feeling lucky"
<civodul>which is obviously pretty neat
<vits95>civodul: yes, but kmicu can SEE the temperature, its more important now... ;)
<vits95>breaking news, i dare to say!
*zzappie vits95 kmicu haha
<vits95>"adamantium has left" -- adamantium not impressed...
<vits95>but... this was a good joke! XD
<civodul>g_bor[m], nckx: did you see the GSoC call at GNU?
<civodul>well nckx mentioned it before
<civodul>are we doing ok?
<g_bor[m]>civodul: about what?
<civodul>g_bor[m]: a call to provide project ideas
<civodul> https://lists.gnu.org/archive/html/summer-of-code/2020-02/msg00008.html
<g_bor[m]>civodul: yes, it was already taken care about on Wednesday.
<civodul>oh awesome, thank you g_bor[m] :-)
<g_bor[m]>yw. I also forwarded my mail to the ml, it must have been lost in the storm...
<g_bor[m]>Bye, going back to work
<civodul>i guess it's just me lagging behind
<vits95>"emacs: woman-mode: browse man-pages ``W.O. man'' (in emacs, directly)". I'll sleep with the lights on...
<NieDzejkob>hmm, what do you think of this clarification? I've seen some confusion in this channel before... https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39501
<vits95>NieDzejkob: i'm think current wording is sufficient. And it's shorter, easier to read.
<nixo_>How do I solve this? error: depends on 'libstdc++.so.6', which cannot be found in RUNPATH ()
<nixo_>by moving gcc from native inputs to inputs?
<vits95>Hi nixo_. I'm can't help either, but what is the package name?
<nixo_>vits95: I'm packaging dart
<nixo_>that's version 1.24.3 but I hope to get the latest working
<civodul>nixo_: i'd check how that .so/executable is linked against libstdc++; it's weird that its RUNPATH is empty
<civodul>normally ld-wrapper adds things to the RUNPATH
<vits95>nixo_: are you sure that you need it ;)? "Dart tools may send usage metrics ... By downloading ... Terms of Service ... Privacy Policy ..."
<nixo_>vits95: we can disable it. Dart+flutter seems to be the only way to write android applications without using java :D
<vits95>nixo_: https://duckduckgo.com/html?q=%22error%3A%20depends%20on%20%27libstdc%2B%2B.so.6%27%2C%20which%20cannot%20be%20found%20in%20RUNPATH%20()%22
<vits95>there was something about FHS and disable checking... (/gnu/store was mentioned)
<nixo_>civodul: might it be this ldflags = [ "-rdynamic" ]?
<nixo_>no idea what rdynamic is
<nixo_>"The -rdynamic option instructs the linker to add symbols to the symbol tables that are not normally needed at run time. It means there are more, possibly many more, symbols that the dynamic linker needs to weed through at run time for symbol resolution."
<civodul>nixo_: i don't think that's the problem
<civodul>can you make sure the "ld" executable in the build environment is that from ld-wrapper?
<civodul>and then, can you make sure the link command-line either uses "g++" or passes an explicit -lstdc++?
<civodul>alternately, you can change that makefile to pass -Wl,-runpath=.../gcc-7.4.0-lib/lib
<kahiru>hey, is there an iso which would allow me to do headless installation? That would mean preconfigured password and sshd
<nixo_>civodul: the makefile is written in gn, I just found this: ldflags = [ "-Wl,-rpath=\$ORIGIN/", "-Wl,-rpath-link=", "-Wl,-z,origin","-Wl,--disable-new-dtags",]
<nckx>kahiru: No, but it's very easy (for a Guixer) to create one. The fact that I didn't keep mine is proof that it wasn't hard 😉
<kahiru>nckx: I expected it would be easy, given how easy it was on nix. but still, no work is better than a little work :)
<nckx>I used (and would recommend) SSH keys over passwords though.
<nckx>I'll see if I have the .scm file in any of my back-ups but I doubt it.
*nckx tries not to back things up that have (wi-fi) passwords in 'em.
<nckx>Nah.
<vits95>every time i'm reconfigure my Guix, there is a "building fonts directory". But to make newly installed fonts work i should use `fc-cache -r`. Why (and: it's correct way)?
<leoprikler>fontconfig has caches outside guix' view (including ~/.cache/fontconfig for instance)
<nckx>Because (by design) Guix only modifies your Guix profile. Conversely, fontconfig knows nothing about Guix profiles and creates its own per-user cache — under the assumption that fonts are global things in fixed locations — in ~/.cache/fontconfig, which is what fc-cache rebuilds.
<nckx>The font directory that Guix builds is something else. It's been said that's it's basically obsolete now that X no longer follows symlinks, but I haven't fact-checked that assertion.
<vits95>leoprikler, nckx: Just the Guix manual tells about fc-cache only in "Installation", but not in the "System Installation". And i was need to add the fontconfig to the sys.config (with lightweight-desctop.scm as a base). Thanks for the answers!
<nckx>My /etc/guix/system.scm has no mention of fontconfig, so I'm not sure what you mean.
<janneke>does guix-publish serve it's public key?
<janneke>*its
<nckx>janneke: I don't think so.
<vits95>nckx: somehow i didn't had `fc-cache` in my PATH. I'll see the first generation...
<nckx>vits95: fc-cache is in my user profile. It doesn't matter, although keeping your system profile slim will reduce the chances of some broken X package keeping you from updating your kernel.
<nixo_>civodul: Thanks! Fixed with: (substitute* "build/config/gcc/BUILD.gn" (("disable-new-dtags") "enable-new-dtags"))
<Guest54845>I really hope it's not because of you guys make fucked us over with sorting wildcard matches BTW
<Guest54845>(I found this channel looking for perf, just following links at top "reproducible" and not knowing about $(sort $(wildcard ...)) was gnu make's motivation)
<nckx>wat
<nckx> https://savannah.gnu.org/bugs/index.php?52076
<Guest54845>Yeah that's the one nckx I have no idea why that has a bug report
<Guest54845>Changing the order of things I glob for is arguably a bug, not like we ever hae to put object files in a specific order!
<Guest54845>And it wont make builds magically deterministic now
<bavier>better to let the filesystem dictate it?
<nckx>But literally 0 of those people are Guix people?
<bavier>right, it was a patch from openSUSE
<Guest54845>Glad to hear.
<Guest54845>bavier: better to not sort what could be a really big list
<Guest54845>Is it already sorted? Maybe, should it be sorted? $(sort) does that.
<jlicht>hey guix!
<Guest54845>Also it's not as if make was supposed to sort it and that got overlooked in the last 1.5 decades (3.81 is 2006)
<Guest54845>On top of the 4.x release which fucked up - prefixes (when you don't care about the return status of a command) and fucked up EVERY tool ever that scans and annotates ouput (you know the error/warn/info lines)
<Guest54845>ANYWAY that wasn't you guys, so hi there!
<vits95>nckx: i was lied, fc-cache was available by default.
<nckx>Guest54845: Maybe try ‘hi’ first next time, but hi, welcome 😛
<vits95>But after i'd booted to the current gen from generation#1, i was unable to login. Just in the gen#1 i'd user vits-guix with UID=1000 and later created guix-vits user with UID=1001. After booting in the current generation i'd guix-vits assigned UID=1000.
<NieDzejkob>vits95: I don't think the current wording is sufficient. That part of the documentation is clearly directed at new users, who don't understand why "root's guix command" is any different from your user's. In fact, there have been users who read the current wording and didn't get it.
<vits95>NieDzejkob: "Note: Note that sudo guix runs your user’s guix command and not root’s, because sudo leaves PATH unchanged. To explicitly run root’s guix, type sudo -i guix"
<vits95>Translated to Runglish, at least: "Press sudo -i guix to win".
<NieDzejkob>but you don't want to use sudo -i
<vits95>NieDzejkob: if you want to change something, i'm propose to point that: "root user is a system-wide things", "and not a word more". Almost nobody love to read more than less. Correct? (Why not use sudo -i?)
<NieDzejkob>"root user is a system-wide things" sorry, but I can't parse that
<NieDzejkob>"why say many word when few word do trick"
<jayspeer>hello guys
<NieDzejkob>if you used sudo -i, you'd have to guix pull twice, one as your user, one as root
<NieDzejkob>sudo without -i uses the results of the non-sudo pull
<jayspeer>I'm planning to install guix on my newly acquired thinkpad w530. how well does guix system support dual graphics devices? can anyone point to me to a config which will work out of the box?
<NieDzejkob>which is good, because then you know that you are using the same commit both for your user and system
<NieDzejkob>jayspeer: What is there for guix to support about it?
<vits95>NieDzejkob: I don't know, as i'm just add things to the sys.config, do `sudo -i guix system reconfigure /etc/lightweight-desktop.scm`, and this is all, for all the users (me and root).
<NieDzejkob>you should have most of your packages just in your user profile
<vits95>jayspeer: Linux-Libre kernel. My laptop works, but it is from 2013 year.
<vits95>NieDzejkob: probably; but what i'm installed for a root user is available for everyone on the system, so i'm just update this way.
<NieDzejkob>okay, then sudo -i works for your approach, but sudo without -i works for all of them
<apteryx>hello Guix!
<sneek>Welcome back apteryx, you have 1 message.
<sneek>apteryx, nckx says: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39592
<apteryx>nckx: I saw this, thank you!
<apteryx>So, I have an NFS mount that fails to mount at boot time... I can't get the logs of shepherd for that time, probably because the service responsible for logging (syslog?) wasn't started yet. Any clue as to how to debug this?
<vits95>NieDzejkob: i was thought that sudo without -i makes me need to issue `guix pull` as many times as i've users. If it's wrong, then probably the manual should be updated.
<apteryx>I'm wondering, would the network actually be up at the time the file systems are attempted to be mounted? Because obviously NFS requires this.
<vits95>apteryx: so, not only me missing systemd here? journalctl ... systemctl list --failed or such
<jayspeer>this machine is from 2012 so I'm not worried about hardware support (libre wifi card is already on it's way). but after installing it yesterday with "graphical" installer, it was unable to boot :(
<jayspeer>I figured I should ask you guys is there something specific I need to know about
<nckx>apteryx: Let me know if it actually works for you. Here, it changes the ‘logical’ resolution (limits drawing to, say, an 800x600 area) but not the resolution of the monitor.
<apteryx>vits95: I'm hoping I'm just missing something, but yeah, systemd does a good job at capturing everything.
<apteryx>nckx: ah. That doesn't sound too useful. :-)
<apteryx>I'm going to give it a quick try now.
<nckx>Maybe it's hardware-specific!
<nckx>Again, this thing is from 1999, I'm halfway impressed it does anything at all.
*sirgazil just finished packaging python-fuzzywuzzy
<nckx>uwu
<vits95>jayspeer: try manual install. i was told there, that the installation .iso is nearly 1 year old. If you'd used btrfs and zstd compression it is should fail to boot (too old GRUB)
<jayspeer>vits95: thanks, I'll try manual today. It's definately simpler than gentoo install. What I'm worried about is dual graphics. Does anyone use a device with such hardware?
<vits95>jayspeer: i do Intel integrated and the N*$Q!a GEFORCE 710m
<vits95>But i'd not tested this Gnome3 "Launch using dedicated graphics" thing.
<jayspeer>vits95: are you using optimus/bumblebee or something like that? or maybe prime offloading?
<apteryx>yikes, I need to be more careful with 'origin' in my git checkouts. Sorry for the new branch. Removing it.
<vits95>jayspeer: no, i'm just do surf the Web; have no idea which device is currently works with my graphics.
<jayspeer>vits95: thanks, you were of much help to me :)
<vits95>:trollface: You always welcome XD
<nckx>apteryx: Not to gitsplain, but I use ‘upstream’ for this reason. Helps with those zombie moments.
<apteryx>nckx: that's a good idea. 'origin' is hardwired in my brain and I don't think about it enough.
<nckx>Same. It's stopped me from pushing stupid things at least 5 times.
<nckx>For the other stuff, I have no excuse.
<NieDzejkob>civodul: it seems that the key you're using has expired at the end of January
<nckx>NieDzejkob: rsa4096/0x090B11993D9AEBB5 2014-08-11 [SC] [expires: 2020-05-05]
<nckx>This one?
<apteryx>nckx: there is -xres and -vxres, where -vxres is for the 'virtual vertical resolution'. Are you sure you were using -xres -yres ?
<nckx>Yes, I would've noticed the extra v.
<nckx>Let me double-check my bash history.
<nckx>sudo /gnu/store/v8qfw3s6w4nz15f16m1pl1i4p2cbny6g-fbset-2.1/sbin/fbset -xres 800 -yres 600
<jlicht>I recently (yesterday...) extended the validity of my gpg keys, and pushed them to *some* keyserver. Is there anything else I should still do to make sure verification of any commits signed by me will keep on working?
<nckx>jlicht: Did you update (paste) your key on Savannah?
<apteryx>also, perhaps you have to use --geometry and give both -[v]{x,y}res
<jlicht>nckx: stupid question, but can I just paste the freshly exported key there?
<nckx>Yeah, just paste the ASCII-armoured (public 😛) key.
<jlicht>and nothing else is required for guix pull's logic?
<NieDzejkob>nckx: huh. The key I've got imported matches the hash, but expires on 2020-01-30
<kahiru>how does zfs on guix work? I added zfs to the list of packages, but that only gave me to userspace utilities, not the kernel modules. Is there something I'm missing?
<nckx>Did you --refresh-keys lately?
<nckx>NieDzejkob: ☝
<NieDzejkob>I don't remember when I last did that :D That sounds like it
<NieDzejkob>yup, fixed it :D
<jlicht>nckx: thanks for pointing me to the right direction
<zzappie>can you break machine startup with bad service config? I mean if some exception ocurs in gexp will shepherd just die?
<nckx>Oh hell yes.
<apteryx>nckx: fbset doesn't do anything for me as well
<vits95>nckx: but there will be some emergency shell started ... yes?
<apteryx>after doing: 'fbset -g 3840 2160 3840 2160 32', as root, and then querying with 'fbset', I get teh same 'geometry 1024 768 1024 768 32'
<apteryx>too bad :-)
<zzappie>nckx: "Oh hell yes". Was it for me?
<nckx>apteryx: Pity. I didn't try more than the above command, maybe there is some magical incantation.
<nckx>zzappie: Yes.
<apteryx>seems others also have issues with it: https://www.linuxquestions.org/questions/linux-from-scratch-13/so-i-got-fbset-working-can-change-resolution-at-run-time-4175502019/
<zzappie>nckx: ouch
<apteryx>perhaps worth to add it anyway in case it come useful to someone 'in the know'.
<apteryx>comes*
<nckx>vits95: You'll get a very sparse initrd REPL that will make you want to commit crimes. And I say that as someone intimately familiar with it.
<nckx>It's not comparable to other distributions' busybox/dash/whatever shell. At all.
*apteryx proceeds to install a Gefore GTX 680 card to work around the useless AMD R7 240 card
*nckx wondered what Guest54845 was up to. Probably insulting someone about make.
<apteryx>Will Nouveau (the free driver for nVIDIA) work without custom config in Guix System?
<apteryx>I guess it's built-in Xorg or something.
*zzappie Guest54845 baad guy
<jayspeer>kahiru: I'm wondering about zfs myself. Can anyone provide any useful info? Should I define custom kernel package definition?
<nckx>jayspeer, kahiru: The zfs package provides the kernel module — *not* just the user-space.
<nckx>kahiru: The zfs package includes the ZFS kernel module. If you've added it to your system.scm and reconfigured, try ‘insmod /run/current-system/profile/lib/modules/`uname -r`/extra/zfs.ko’ or something very similar. I'm not sure about the directory, this package is taking forever to build.
<leoprikler>apteryx: it works for me, but it will probably not work well for all nvidia cards
<nckx>Copy-paste for the win. Sorry.
<apteryx>yes, nouveau is part of the %default-xorg-modules, so should work
<zzappie>nckx: whoa this is dangerous :) I did't think that you can kill shepherd so easily
<apteryx>leoprikler: I've checked h-node.org for that one I'm about to install: https://h-node.org/videocards/view/en/1792/GK104--GeForce-GTX-680---rev-a1-
<nckx>zzappie: It will improve over time, it's not a design flaw 🙂 Shepherd is already less suicidal than it was a few months ago.
<joshuaBPMan>Hey guix, thanks for implementing guix import! I needed a perl file that was not in guix...so I've set up a channel, imported the perl file, uploaded it to the channel, and now it's installed! Ya'll are awesome!
<joshuaBPMan>nckx: why was the shepherd suicidal a few months ago?
<user_oreloznog>joshuaBPMan: congrats!
<jayspeer>info about nvidia gpus on h-node seems a little outdated for me. I'd look here https://nouveau.freedesktop.org/wiki/FeatureMatrix/ for more up to date data
<nckx>joshuaBPMan: It was a bug that's now fixed. Can't find the link now.
<joshuaBPMan>nckx: may I ask what the bug was?
<vits95>joshuaBPMan, nckx: codename of that version was "Kenny"?
<vits95>"ERC"? So now i'm have a irc client too. Cool.
<zzappie>nckx: good news. I feel like service defenitions are bit complicated it took me much less time to get into packaging than ito service composition. It requieres you to read guile manual to understand things like records and pattern matchig and there are also gexps. It's fine with me cause Im into scheme but maby it's a turnarround point for many users.
<nckx>joshuaBPMan: Are you having actual problems with current Shepherd?
<nckx>zzappie: Oh, same, I think services still have an air (partially deserved) of wizardry for most contributors even.
<zzappie>I've mentioned govuk-guix today as a referene. I did't know at this time that they are not actually using it...
<nckx>zzappie: Is that cbaines by any chance?
<zzappie>nckx: right!
<zzappie>nice :D
<nckx>Chris definitely uses & works hard on Guix, even if (or not) the UK government hasn't yet seen the light.
<zzappie>Yeah great work it helped me a lot
<apteryx>hmm, dead video signal from GTX 680. Arf.
<nckx>I took only a brief look a the repository but it was just a bot committing bumps.
<apteryx>perhaps the card is dead, or not enough power, though I doubt the later.
<civodul>NieDzejkob: perhaps you just need to refresh it?
<civodul>(the key)
<nckx>Well, that took long enough :-/ kahiru: The zfs module is in the zfs:module output, after installing it ‘sudo insmod /run/booted-system/profile/lib/modules/`uname -r`/extra/zfs/zfs.ko’. You can also use current-system before rebooting although it's slightly less robust.
<nckx>civodul: That was it.
<civodul>cool
<apteryx>ah, the GPU blackout seems related to using UEFI mode: UEFI says: There is no GOP (Graphics Output Protocol) support detected in this card. 'BIOS UEFI/CSM Mode' settings in BIOS will be changed to 'Disabled'.
*apteryx and now the card boots \o/
<apteryx>still needed that 'insmod all_video' trick in GRUB though.
<apteryx>perhaps because of the huge 3840 x 2160 pixels display connected
<apteryx>I'll see if I can include this module in our default GRUB configuration
<apteryx>now I get to the GDM greeter, but can't login, perhaps because of my ~/.xsession file?
<apteryx>I'm trying to use GDM + ratpoison (starting the session using my ~/.xsession file, which I read in the commits being honored by GDM).
<vits95>apteryx: try to login in tty, it should print the error.
<vits95>also: maybe you willing to try sway? it need cleanup of i3wm configs from ~/ and a elogind.
<vits95> https://wiki.archlinux.org/index.php/Sway#Autostart_on_login
<apteryx>could see the /var/log/gdm/greeter.log in a tty. I just see errors from xkbcomp. That makes sense, I've not installed my manifest yet.
<NieDzejkob>when I get a review like this, is it reasonable to fix the issue and commit to master, adding the S-O-B line? https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39501#8
<NieDzejkob>(mbakke: ^)
<mbakke>NieDzejkob: we usually only add the s-o-b line when committing on behalf of someone else
<NieDzejkob>okay, good to know
<NieDzejkob>another question: should "trivial" patches like this one add a Copyright line? https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39594
<mbakke>NieDzejkob: ludovic posted a good summary of when to add copyright here: https://issues.guix.gnu.org/issue/39522#5
<mbakke>in short, "it depends" :-)
<mbakke>I don't usually assign copyright for typos or "facts" such as a version change, but then, the combined works might well be copyright-worthy
<drakonis>haw, these guix day notes are very raw.
<nckx>NieDzejkob: Was https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39501#8 what you meant to paste? I don't see a review there nor does that page even have an #8 anchor, which is… odd.
<NieDzejkob>nckx: I blame caching. Link works on my side
<kmicu>[jokin’] Great, I cannot contribute to Guix manual :/
<nckx>NieDzejkob: Hmph, indeed 😒
<vits95>NieDzejkob: I'm, personally, need an raw example: "to do guix pull only once and for all, do <code>x</code>." You just added more text like in that naughty man pages i didn't read.
<vits95>See, most newbies are worry about the installation process enought, to want an examples, not explanations. And some people are ungrammar mostly (like me). And you offer TEXT to us... examples and pictures is better.
<apteryx>actually, I don't have a .xsession file no that machine yet, so it should at least login.
<NieDzejkob>vits95: hmm, I'm not sure I understand your use case. How many users do you have?
<apteryx>is anyone successfully using GDM + ratpoison?
<vits95>NieDzejkob: root + me.
<vits95>one total, 2 virtual
<kmicu>mbakke: wasn’t ‘two spaces after punctuations’ malarkey removed from Texinfo?
<NieDzejkob>apteryx: I'm using slim + ratpoison, don't remember what the issue with GDM was
<kmicu>(or do we use some old Texinfo version which still needs that?)
<mbakke>kmicu: The convention is still heavily used throughout Guix, and I also find it easier to parse text when sentences are better separated.
<NieDzejkob>vits95: then, as your user, do `guix pull'. Then, use `sudo guix system reconfigure'
<apteryx>NieDzejkob: that's usually what I do to, but I like to revisit GDM sometimes, like now ;-)
<vits95>NieDzejkob: but this will leave my root-user vulnerable and not updated? `guix pull #user` `sudo -i guix system reconfigure #root` ?
<NieDzejkob>vits95: do you have any packages in root's profile?
<NieDzejkob>(not the system profile, but the one that's only used by root)
<vits95>no.
<NieDzejkob>oh, your example has -i. Yes, guix pull as user and sudo -i guix will leave you not updated, as far as I understand
<vits95>then `sudo guix #without -i` will update user, but not root-user?
<vits95>Just as i'm get it from Guix manual, somebody should update the system-profile as root from time to time. But this isn't touch the users profiles, if they overlap with the system's defaults.
<vits95>and if i'm installed something as a user, i'm shouldn't download it again to install it system-wide, as root.
<NieDzejkob>"will ... not update root-user". What do you mean? You can't update a user, only some packages it has installed, and as you've said, root doesn't have any
<kmicu>mbakke: Thank you. I can understand that and your motivation. Though that convention is no longer required and there are valid reasons to avoid it (hence removal from Texinfo) especially if that’s not enforced for Guix. For now no patches from me so no issue here.
<vits95>i'm mean that `sudo guix` will update the user's profile (as manual says). And it will be $PATH, that without -i flag will be preserved. So root will have no updates withou -i flag.
<vits95>NieDzejkob: Either way, please, if you propose a patch to the manual, keep wording less and examples more. Rather a separate line examples, not embedded inside a lots of text (because it is a Reference Manual).
<apteryx>NieDzejkob: ah, it worked, by keyboard layout was QWERTY and not dvorak as I thought I had configured. Will double checked.
<apteryx>Confusing though that there's GNOME as the default session choice, even though I didn't install GNOME.
<commanderkeen>how are yall building packages? right now im editing the package definition in emacs and then running guix package -f filename . but from my understanding the file has to return the packge (package ...) and not the definition. Is there a way to do this in geiser repl? i feel like there has to be a better way than what im currently doing
<commanderkeen>it seems like i could load the namespace and then run a build-package function on the definition or something like that from the repl
<commanderkeen>i tried searching a a detailed guide but everything i found lacked details
<mbakke>commanderkeen: see "Building from Git and Running Guix before it is installed" in the manual: https://guix.gnu.org/manual/devel/en/guix.html#Building-from-Git
<mbakke>commanderkeen: in short, once you have built Guix from the repository, you can use './pre-inst-env guix build FOO' to build FOO directly from your git checkout
<commanderkeen>i know i read that but i'll go through it again
<commanderkeen>yeah i think this is worthy to read through again
<commanderkeen>is there a way to run the ./pre-inst-env to get a repl in emacs?
<commanderkeen>like a slime-connect or a cider-connect?
<commanderkeen>but for guile
<janneke>i'm trying to build a libre-linux kernel for the pinebook-pro
<janneke>it's a patched v5.4 or v5.5 kernel, but none of the deblob scripts from https://linux-libre.fsfla.org succeed
<janneke>iow, i would like to patch a deblob script; how to go about that?
<janneke>hmm, i could probably add an optional patches field -- thanks for listening!
<commanderkeen>and i already have guix installed. cant i use the one already checkout?
<NieDzejkob>IIRC the checkouts are kept in the store, so you can't really modify them
<drakonis>so, i'm trying to figure out how to tell my service's udev service extension to use a specific list of udev rules
<drakonis>this is certainly taking a while.
<NieDzejkob>"hmm, OpenSSH is not up-to-date. I wonder what it would take to update it." "WTF? Why are all the mirror 404'ing?" "oh, the release was today"
<apteryx>hmm, ratpoison relies heavily on xterm, but it doesn't propagate/patch its reference to it, so upon installing a fresh system with gdm + ratpoison, the experience seems broken.
<NieDzejkob>"relies heavily" as in provides a keybind to start it?
<nckx>NieDzejkob: Your PGP key mail was still stuck in moderation, should be OK now.
<apteryx>NieDzejkob: well, even getting the help screen would fail without xterm, so yes, heavily ;-)
<janneke>NieDzejkob: in any case, non-tarball origins cannot be patched or modified, or so it seems
<janneke>i'll need to distribute my own, patched deblob script
<janneke>the content of a file moved, or so it seems
<civodul>janneke: i think a git-fetch origin can be patched just fine, no?
<janneke>civodul: that would be nice...but when i add either a snippet or patch to the deblob origin, i get something like
<janneke>@ build-started /gnu/store/01ip9khnzl6d0b1y80cx5336y10adfcg-deblob-5.5.tar.xz.drv - x86_64-linux /var/log/guix/drvs/01//ip9khnzl6d0b1y80cx5336y10adfcg-deblob-5.5.tar.xz.drv.bz2 5979
<janneke>/gnu/store/cnqpra8vr2l5fz00rr4yj4bp3hr00cfw-tar-1.32/bin/tar: This does not look like a tar archive
<civodul>hmm
<janneke>civodul: https://gitlab.com/janneke/guix/-/blob/wip-pinebook-pro0/gnu/packages/linux.scm#L225
<civodul>janneke: a package like "dstat" uses git-fetch with patches
<civodul>janneke: oh but maybe it's the computed-origin thing that breaks with #:patches, no?
<civodul>it's a different mechanism
<janneke>ah, that makes sense (#:patches or #:snippet, either way)
<lfam>I noticed that the Vigra build timed out again
<lfam>Interesting that it succeeds on aarch64 which I'd assume was much slower than x86_64 but, anways, we should increase the max-silent-time
<lfam>Also interesting that the check phase only takes 1047 seconds on i686
<lfam>But ~15000 seconds on aarch64
<NieDzejkob>ahh, damn it. Apparently I don't get to keep my branches named "master" and "upstream-master" without git's bad defaults getting in the way
<kahiru>nckx: ah, thanks. I tried just `modprobe zfs` which obviously did not work because of paths
<kahiru>what if I wanted to have the module loaded on boot, is there some system configuration option for that?
<drakonis>i'm working on figuring these things out
<drakonis>trying to at least.
<drakonis>is the guile repl not supposed to have completions when run through guix repl?
<NieDzejkob>drakonis: what is the Guile version being used by guix repl?
<TempeVolcano>I can't boot guix from a usb
<TempeVolcano>it gives an error on the command line, i don't even get to the menus.
<TempeVolcano>Should I install Ubuntu first?
<TempeVolcano>I have another question. Is it recommended to use Guix as a daily OS for your personal use? I was thinking about doing that to get used to Guix.
<NieDzejkob>I use Guix as my daily OS, it's nice.
<NieDzejkob>"an error" is pretty vague
<NieDzejkob>how would installing Ubuntu first help?
<NieDzejkob>("is pretty vague" is itself pretty vague :P It would help if you could say what the exact error message is, either transcribed or as a photo)
***jonsger1 is now known as jonsger
<NieDzejkob>lfam: I was browsing the patch tracker and noticed that you were quite close to merging #39312 and then dropped it after a round-trip to make the patch apply properly. Any reason, or did you just forget?
<lfam>It's a case of forgetting
<lfam>Are you able to handle it NieDzejkob?