IRC channel logs


back to list of logs

<mark`>Does anyone have experience installing guixsd from the usb image?
<lfam>efraim: Yes, something like that, but I will wait until the libtiff graft is on master. Those merges are a pain, so I'd rather just do one
<lfam>efraim: I missed your message until now, as I read the logs
<paroneayea>I have a 10 year old desktop running as my backup server, but it's falling apart.
<paroneayea>I'm thinking of ordering something lightweight and guixsd compatible
<paroneayea>not really sure what, though.
<paroneayea>anyone doing something they'd recommend?
<paroneayea>I'm sure would be more than enough but costs a bit more than I'd like (the cost isn't so bad, I just don't have a lot of extra cash right now)
<paroneayea>also I'd probably need to switch my backup setup over to guix :)
<paroneayea>currently I'm running dirvish on debian.
<habs>I think there is something gravely wrong with my GuixSD installation. Is there any way to do a more thorough check than 'guix gc --verify=contents'? Any new packages I install always segfault unless I run them with sudo, and I get several hundred "collision encountered" warnings when installing any package.
<alezost>habs: I think you installed different packages in different times which may lead to multiple collisions, try to upgrade your profile: "guix package -u"
<alezost>btw how many packages are in your profile?
<efraim>i wonder if we should add an option to git-fetch for '--depth=1 --shallow-submodules'
<efraim>it always seemed faster for me on my netbook to not have to compute all the tags
<efraim>is lgpl2.1+ and openssl is ok? or should I switch openssl for gnutls
<ng0>Hi. Some of us are at reddit, right? What do you think about doing an "Ask Me Anything I'm a Guix Developer" or something like that? I don't use reddit, but depending on the topic those are good to get some questions solved and to get answers out there which aren't already obvious, and I think there's some percentage of reddit users which are hackernews users...
<efraim>we could try it, i'm not too sure how well I know some of the internals
<ng0>it's not a one person operation, so I assume the "I" could be extended to "we are" and more than just one person answers? I'm not sure how this AmA thing usually works
<efraim>#fsf says lgpl and openssl are incompatable
<ng0>how much resistance would there be if I added tex to the next gnunet version package? It has a feature which uses pdflatex for a very long time now.
<ng0>is texlive-minimal == pdflatex?
<efraim>nows the time to get it in, while its still a new package :)
<ng0>I think i wasn't clear
<efraim>i mean in terms of adding it
<efraim>i'm running find on /gnu/store now for pdflatex
<ng0>The optional dependency is there, it can beadded, i already do so on another system. I'm not sure though if it would require texlive or texlive-minimal or something entirely different, and I know many people here are protective about adding texlive to something
<ng0>ACTION runs e-file to check where pdflatex actually is
<efraim>so far i've found it in texlive and texlive-bin
<ng0>ah. not in -minimal?
<efraim>its still searching, i don't have it on my machine with an ssd so it takes a long time
<ng0>saves me the time to boot a system which boots another system where I could run e-file, which queries a database of filestructures of applications of gentoo :)
<efraim>ah I always forget about locate
<ng0>got it
<efraim>in texlive, didn't see it in texlive-minimal
<ng0>me neither
<efraim>but i might not have texlive-minimal on this machine
<ng0>if it is not in texlive-minimal, I will create a gnunet variation which inehrits this. I'm personally not so happy to add texlive as a dependency
<ng0>s/which inherits this/which provides this
<ng0>this is not my focus atm though, will be for later
<ng0>thanks for searching :)
<snape>hi, is there a way, with GuixSD, to avoid rebuilding the whole kernel when I change a little thing
<snape>I mean, I need to debug something
<snape>is there like a debugging workflow?
<snape>well, I guess I can manually update the linux modules that are in the initrd, and recompress it
<snape>make a fake /var/guix/system that points to the debugged linux/initrd
<snape>and that should boot...
<rekado_>snape: I'm not experienced with this but you can change the list of modules for your initrd in the operating-system declaration.
<rekado_>this doesn't rebuild the kernel afaik
<snape>it does not rebuild the kernel if none of its inputs are updated, but since I'm debugging the kernel, the source of the kernel is updated, and the build-initrd scm function will trigger a whole kernel rebuilding
<snape>(If I understand well)
<efraim>i have the internet again!
<efraim>-lgs links against what?
<habs>I tried running 'guix package -u' and it succeeded but now things are much worse as even the guix binary segfaults when run without sudo. Also I have 120 packages in my profile (GuixSD on a Thinkpad) @alezost. What can be done to further debug / fix this?
<rekado_>efraim: libgs --- ghostscript?
<efraim>rekado_: thats what i'm thinking, didn't seem to be happy with it
<efraim>i might need to add a patch phase after configure to fix something
<efraim>wait, i take it back, apparently it magically built while I wsan't looking
<Petter> the best builds ;)
<efraim>it wasn't happy with ghostscript-gs, did like both, now to try with just ghostscript
<efraim>heh yeah
<efraim>well, with efl gobbling up all the other enlightenment libraries it's time to add another patch making efl the sole enlightenment input on a couple packages
<habs>This is a quick example when I run /run/current-system/profile/bin/ls or any other newly-updated binary: Can anyone help debug why they segfault? Is it a problem with my library paths?
<iyzsong>habs: did you set LD_LIBRARY_PATH? it shouldn't load libcap from ~/.guix-profile
<habs>iyzsong: My LD_LIBRARY_PATH is "/home/habs/.guix-profile/lib:/lib:/lib64"
<iyzsong>LD_LIBRARY_PATH shouldn't be set :-
<habs>iyzsong: Ah, that fixed my problem! Thanks!
<rekado_>LD_LIBRARY_PATH is a hack that tells the dynamic library loader to ignore what the binaries say they want and to load other libraries instead.
<rekado_>in Guix all binaries contain the full paths to libraries they need to load.
<davexunit>it would be cool if 'guix package --search-paths' printed out something that could be eval'd *without* clobbering the previous values of those variables
<davexunit>for example, it will set PATH such that *only* the profile is searched.
<rekado_>davexunit: isn't this the same as entering a sub-shell and then evaling the output?
<rekado_>(or source $profile/etc/profile in a sub-shell)
<davexunit>rekado_: oh yeah, forgot about /etc/profile!
<davexunit>thank you!
<muck>how would you guys log stdout on tty1 within qemu when debugging a wm build ? the wm doesn't start and clearly produces lua related errors (awesome-wm) but i can't read it fast enough thanks to slim restarting and no related logs being in /var/log
<efraim>elementary is now part of efl, same commit to remove the now superfluous elementary or new commit?
<efraim>i mean remove from other package inputs
<efraim>not the package definition
<efraim>hmm, new efl can't find pulseaudio or libsndfile
<efraim>i might need to save the 1.18.0 bump for later, can't find a premade patch in their git tree
<paroneayea>mark_weaver: btw, do you know if the latest update of linux-libre you pulled in fixes that attack we discussed recently?
<paroneayea>ACTION wonders why coreutils is rebuilding on his machine
<paroneayea>new gcc?
<paroneayea>maybe I'm wrong,maybe it wasn't coreutils
<bavier>it might be nice for 'guix build' if one could specify the packages output, e.g. glibc:out, to optimize the case where substitutes are used
<lfam>bavier: You mean, only build the specified outputs?
<lfam>paroneayea: Maybe you were watching the installation process of some other package, which uses the `install` command from coreutils? That prints the store path of coreutils. I've been alarmed by seeing that in the past, until I realized what was happening.
<paroneayea>oh hello
<paroneayea>lfam: I think yes
<paroneayea>let's update guix to have ascii boxes like this ;)
<paroneayea>* UPDATING: glibc *
<paroneayea>I'm sure everyone else will be cool with this :)
<lfam>No, please :)
<paroneayea>okay okay lfam
<nully>paroneayea: needs more *s
<paroneayea>we'll use unicode box drawing characters instead!
<lfam>Only one piece of information will fit on the screen at a time ;)
<nully>one screen full per packag
<davexunit>why not run the titles through figlet?
<davexunit>sounds like the best option here
<lfam>In seriousness, has anyone ever audited the ASCII art programs? I'd rather not add lines of code to our base system for frivolous reasons :)
<lfam>Inkscape makes me nervous already
<davexunit>inkscape is just used when building
<davexunit>and should be able to replaced in the future
<paroneayea>render to PNG using inkscape, then render to PNG, and print to screen using aalib
<paroneayea>PROBLEM SOLVED
<lfam>Speaking of which, inkscape uses libtiff, which has a pending security update on the mailing list :)
<paroneayea>oops, need one less render to png!
<paroneayea>see, it's even easier than I thought
<davexunit>paroneayea: patches welcome
<paroneayea>davexunit: in this case since it's not requiring any major rebuilds, it's safe to just push this change directly to git master without review right?
<davexunit>paroneayea: absolutely.
<paroneayea>and I agree nully, there can always be more *s
<paroneayea>the reaction you really want is "my god it's full of *s"
<davexunit>detect the width of the controlling terminal and print more *s as needed
<lfam>I tried to package this for April Fool's Day:
<lfam>But I couldn't get it to build :/
<paroneayea>so the more the better
<paroneayea>cute lfam
<lfam>Yeah, I was going to present it as some great advance in the state of the art of free software
<paroneayea>man, freshmeat
<lfam>An aquarian interface
<paroneayea>you know, even in its height, I could never find freshmeat manageable to read
<lfam>Much easier to eat
<paroneayea>I'm a vegetarian thank you
<lfam>Pass it over here ;)
<paroneayea>vegetarians don't eat meat, but we do read meat like tea leaves
<paroneayea>as long as it's fresh.
<lfam>Back to serious topics, does anyone know why libarchive triggers 976 rebuilds? Is there some core package which uses it, or is it just an input to many leaf packages?
<lfam>Hm, sees to only appear in a handful of package modules
<bavier>lfam: I meant, when a 'guix build foo' results in downloading all outputs of foo from a substitute server, it would be nice to specify which you're interested in
<bavier>if the 'guix build foo' requires a local build, you just build the whole thing anyhow
<lfam>bavier: Ah, that would be nice
<lfam>Right, I wasn't sure how my interpretation would be possible, but yours seems possible
<bavier>e.g. 'guix build glibc' downloads the debug output, which I'm currently not interested in
<lfam>Some bad problems with libarchive:
<bavier>so something like 'guix package' supports: guix build glibc:out
<lfam>Especially bad for operating systems that *use* libarchive as part of the software update mechanisms
<lfam>But, that's not us, AFAICT
<lfam>Although, I shouldn't be the only person to look at this
<lfam>Cmake uses libarchive :/
<lfam>A handful of other packages, too, but mostly cmake
<efraim>libarchive is used in cmake, not sure what else
<efraim>libcaca is also fun for ascii video watching
<efraim>mpv --vo caca my_video
<efraim>make sure it's a low res video though
<socristotle>Hello. Is there any documentation on installing GuixSD on a libreboot machine with Full Disk Encryption?
<lfam>socristotle: I'm not sure exactly. I think that Petter (in this channel) knows about it. There has also been some discussion on the guix-devel mailing list archives.
<lfam>socristotle: There is also this: <>. I don't know the anything about the quality of the information there
<socristotle>Thank you very much lfam. I'll look into the information to see if it's possible. If not I can always install Parabola.
<lfam>socristotle: I remember hearing that it's possible, but requires some manual steps.
<lfam>Of course, supporting full disk encryption with or without libreboot is something we want for GuixSD
<jmd>What kind of pattern matching applies to guix package -d ?
<lfam>jmd: There's some documentation in the manual, section 3.2 Invoking guix package. Beyond that, I'm in the same boat as you: time to read the code :)
<jmd>lfam: I'll have a look there. Thanks.
<efraim>i normally call it with `guix package -d 1m' for 1 month
<jmd>Anyone looked at getting torbrowser packaged?
<lfam>jmd: Yeah, there's been some discussion here and on guix-devel
<efraim>ng0 has been banging his head against that one for a while now
<jmd>efraim: Why does 1m mean one month. I would have guessed one minute.
<efraim>either ludo clued me in or I checked the code, can't remember
<efraim>its in guix/scripts/package.scm
<efraim>well, i think it is at least
<alezost>efraim: jmd: this string pattern is converted to a time interval by 'string->duration' procedure in (guix ui) module
<alezost>anyway I think a minute would be a too small time interval to be used for --delete-generations
<alezost>ha-ha, but there is "s" for seconds!
<socristotle>Not the right irc channel, but does anyone here own a libreboot x60 tablet?
<sankey>ACTION has an x60 runnin libreboot
<sankey>but it's not a tablet, didn't know those existed
<ngz>Hello. I'm currently struggling with a package definition (at, which throws the following error: ice-9/eval.scm:386:9: Throw to key `match-error' with args `("match" "no matching pattern" ())'. Any help is appreciated, since the package is otherwise finished.
<bavier>ngz: idk about the match error, but the cmake-build-system already builds within a separate "build" directory
<ngz>Ah. I'm going to get rid of that part then.
<bavier>I'm not sure how useful that is to you here, because if you need to override the configure phase, it doesn't expose the location of the source directory
<bavier>but you can usually approximate that with more ".."s
<ngz>I'll add a temporary (display (getcwd)) to see my location, if needed.
<steveeJ>I'm currently stairing at and I'm trying to find out how the guix refresh command guesses the next version to try to download. any hints?
<ngz>bavier: Oddly I was able to successfully build that package at some point, but I'm not anymore.
<bavier>steveeJ: it uses a combination of the gnu-maintenance module and the various import modules
<steveeJ>bavier: I see. the package has to provide specifics for the updater
<bavier>steveeJ: the packages provide only the source urls, which are examined by the updaters
<ngz>Which module provides %current-target-system? I cannot find it grepping the sources.
<bavier>ngz: it's defined in each derivation I believe
<steveeJ>bavier: I'm very impressed that this is a native feature!
<bavier>steveeJ: :)
<bavier>steveeJ: if you're interested in such things, there's more work to be done
<bavier>e.g. it'd be nice to have a generic fallback updater
<steveeJ>bavier: heuristics?
<bavier>steveeJ: right
<bavier>like, maybe it looks for later versions in the same web directory, or scrapes the project homepage for links, etc
<steveeJ>does guile support concurrency?
<steveeJ>I can probably find out somewhere on the webs ;)
<bavier>steveeJ: in a few ways, yes
<bavier>steveeJ: another project, Spack, has an interesting generic updater that we might be able to borrow some ideas from:
<steveeJ>TLS warning :/
<bavier>steveeJ: oh?
<steveeJ>besides that, another interesting project I've never heard of
<muck>ohman finally got awesome 3.5.9 running \\o/
<paroneayea>damn, ledger is failing to build
<paroneayea>ACTION looks at the hydra interface
<bavier>paroneayea: I noticed it too,
<bavier>but it passes on hydra
<paroneayea>bavier: hm
<paroneayea>. o O (Did boost update recently or something?)
<bavier>well, no, the last build was a failure
<bavier>paroneayea: there's a 1.61.0 version that I'm testing
<paroneayea>bavier: let me know if you get it to work
<paroneayea>I need to check a balance for a project :)
<bavier>paroneayea: so far just one package needs patching
<paroneayea>The following tests FAILED:
<paroneayea> 29 - BaselineTest_opt-time-report (Failed)
<paroneayea>181 - BaselineTest_opt-datetime-format (Failed)
<paroneayea>that's why it's failing for me locally
<paroneayea>but that's with the current version
<paroneayea>I'm guessing a boost change, but that's without looking :)
<bavier>both the last two builds on hydra used the same boost version
<paroneayea>looks like it's something different in date/time stuff
<paroneayea>though maybe in the way it's formatted or something, who know
<bavier>mpfr was updated...
<paroneayea>ah, doing the guixsd "install from usb" thing
<bavier>paroneayea: do you use ledger much?
<paroneayea>throws me back to nostalgia for 2003 me
<paroneayea>bavier: yes-ish
<paroneayea> you can find some of my book-keeping here :)
<bavier>oh, cool
<bavier>I just learned of it recently; wondering whether I should transition away from gnucash
<bavier>but I share finances with others in the family, so a text- and cli-heavy approach might not be appreciated
<paroneayea>bavier: maybe look at beancount
<paroneayea>I hear it's ledger like but has a nice web UI
<paroneayea>same with hledger, but I hear beancount is more friendly