IRC channel logs


back to list of logs

***lrvick1 is now known as lrvick
<reepca>... yeah, that'd do it. TEXINPUTS is 165k characters long, which exceeds the per-string limit of 32 pages (128 kB). Now I just need to figure out why in the world we're setting TEXINPUTS to contain *every directory in every input*.
<reepca>rekado_: I see you initially wrote texlive-latex-base, may I ask why we're setting TEXINPUTS the way we are?
<reepca>currently it contains 1433 directories (!)
<apteryx>hello! how can I extract a gunzip'd nar file archive manually? (or with guix command?)
<apteryx>oh, guix archive -x
<reepca>would it work to switch texlive-latex-base to texlive-build-system?
<apteryx>is it known that python2 cannot be built reproducibly?
<bavier`>apteryx: I think so; I'm fuzzy on the details, but I think has more details
<efraim>I made the mistake of looking onto opencv, I have it building on aarch64, now to fix the test suite
***rekado_ is now known as rekado
<rekado>reepca: TEXINPUTS may be a little too large.
<rekado>we can’t use the texlive-build-system there, because that uses texlive-latex-base.
<rekado>reepca: you might be able to reduce TEXINPUTS by filtering the inputs, so that only “texlive-” packages are processed.
*rekado works on a texlive profile hook now
<efraim>I have on my TODO list a telnet service, I want to be able to telnet into a machine and play a terminal game or somesuch
<roptat>hi guix!
<civodul>Hello Guix!
<janneke>hello civodul!
<efraim>hello everyone!
<civodul>bah, evaluation of staging on berlin was broken:
<civodul>i think it should be fixed now
<civodul>in Cuirass the load_path_inputs for "staging" was non-empty, which caused problems down the road as we were mixing two different Guixes
<roptat>ah, so that means the numbers should get higher now, right?
<roptat>maybe we'll be able to merge staging before 1.0 after all
<civodul>roptat: hopefully!
<civodul>though perhaps there were other issues before that
<civodul>we'll see
<civodul>python-numpy-documentation broke on master:
<civodul>gexiv2 on i686 fails on staging:
*civodul is actually a CI bot
<roptat>maybe we need one?
<civodul>we surely do
<civodul>though you wouldn't want it to be the only one speaking on this channel :-)
<roptat>well, if we're annoyed by build failures, there's only one solution: fixing them
<efraim>that's how I ended up looking at opencv
*jonsger wonders if we should add the openSUSE and Arch Linux package to
<roptat>good idea
<civodul>yes, why not
<civodul>though i remember people reporting problems with the AUR (or Gentoo?) package
<civodul>we'd need to double-check
<jonsger>it seem to be maintained and they run into the same gnutls issue then me
<jonsger>but this is not a big thing. And the gnutls maintainer should be now aware of this problem...
<roptat>ha! I totally forgot these ocaml packages that were sent last week
<roptat>I was busy struggling with sbt
<roptat>I wonder how ant finds javac... I'm using a modified version of it that has some classes that are otherwise not present in icedtea, but ant cannot find them, so it's probably using the icedtea version of javac and I wonder why and how
<roptat>this package here:
<roptat>uses #:jdk ,java-jsr308-langtools
<roptat>so it shouldn't bring in icedtea (it does, indirectly though)
<rekado>I wouldn’t add the AUR package, because the AUR isn’t in any way official (it’s also not enabled by default). I’d rather have people install Guix through the binary installation method than via the AUR.
<jonsger>rekado: but I think the entry barrier for most Arch users is lower if it's in the AUR then doing it by "hand" or with the script...
<ng0>iirc the checksum in guix (for a single tarball) was just the base32 of its sha256sum right? I have no guix here right now
<rekado>ng0: no. It’s base32 with a reduced alphabet
***dddddd_ is now known as dddddd
<rekado>jonsger: I’d prefer to see a proper Arch package in the official repositories.
<ng0>Oh, right. Thanks! before I introduce any errors, I have released gnurl-7.63.0 yesterday. I'm about to write the release email. Does anyone of you want to version bump it in guix? I had a series of unfortunate events and need to setup something again for guix.
<jonsger>rekado: me too, but I think we could list it but write some comment like: "This package is not an official Arch package, for a better experience install guix as binary or with the script"
<efraim>ng0: I can bump it
<ng0>thanks :)
<ng0>can you also test run ./gnurl-config ? I'm a bit puzzled by my own build. It runs okay, but I might have introduced a regression in gnurl-config. but maybe it's just on my system.
<ng0>the file (gnurl-config) was behaving strange before, and I'm attempting to fix a bug
<efraim>ng0: it looks ok
<ng0>does it still look like the one described at the original ticket here:
<ng0>wrong question. I mean, can you paste its --libs output?
<efraim>-L/gnu/store/nq4lsyipmfb0q7g26ra45rwwqrh3x8zw-zlib-1.2.11/lib -L/gnu/store/4ig0w6mrmx59pfv56vrq5rh3wx0avrj8-gnutls-dane-3.5.18/lib -L/gnu/store/pyj39jis185kak73rhgn69mkbd8023sy-libidn2-2.0.5/lib -L/gnu/store/rp82m3siqdbz6a1454k7lcl3crncp62i-gnurl-7.63.0/lib -lgnurl -lidn2 -lnettle -lgnutls -lz
<ng0>hm. okay, thanks
<ng0>that means that at least what I applied from netbsd's curl worked as intended
<efraim>here's the static libs: /gnu/store/rp82m3siqdbz6a1454k7lcl3crncp62i-gnurl-7.63.0/lib/libgnurl.a -L/gnu/store/nq4lsyipmfb0q7g26ra45rwwqrh3x8zw-zlib-1.2.11/lib -L/gnu/store/4ig0w6mrmx59pfv56vrq5rh3wx0avrj8-gnutls-dane-3.5.18/lib -L/gnu/store/pyj39jis185kak73rhgn69mkbd8023sy-libidn2-2.0.5/lib -lidn2 -lnettle -lgnutls -lz
<efraim>ng0: so if it looks good i'll go ahead and push it
<ng0>yes, it's okay
<ng0>thanks again
<rekado>I’ve got a working texlive-configuration profile hook.
<rekado>What’s not so nice, though, is that you can’t use texlive-tiny with it.
<rekado>texlive-tiny uses union-build and wraps executables in TEXMFCNF and TEXMF. We set those very same variables in the profile hook.
<rekado>I may need to also change the way texlive-union works.
<rekado>but I’m hopeful that this can be added before 1.0
<ecbrown>i'm curious about process for reporting a package that does not build (in this case, qgpgme). i know there's a build farm, so would filing a bug report for failure to build be superflouous?
<efraim>ecbrown: looks like there's an update, want to try it out?
<ecbrown>efraim: regret to tell you for some reason guix master won't build on my debian stretch system... i sort of have two problems
<ecbrown>s/sort of/certainly
<efraim>ecbrown: did you try running bootstrap again? I found that normally helped me when I was using system libraries
<ecbrown>even made a 1.12.0 branch, eager to compile. :-)
<ecbrown>let me try again
<civodul>rekado: sounds nice!
<efraim>should I merge qt-updates into staging?
<rekado>I’ll push the uncontroversial parts of the tensorflow patches tonight.
<civodul>rekado: cool, i haven't taken the time to review but i'm happy you managed to make progress
<rekado>that’s fine. No review is necessary at this point. It’s a very dirty package definition. It requires a bunch of changes / fixes before it can be merged.
<rekado>the package bundles too many libraries and is more complicated than it should be because of that.
<chewzerita>Just setup guixsd for the second time, but this time I need my vpn to be setup
<chewzerita>... how do I setup openvpn on guixsd?
<chewzerita>the info page is not helpful (sorry), and seemingly no one else online has done it
<efraim>suggestions which file to put a telnet service in?
<roptat>chewzerita, on guixsd you could use the openvpn-client-service
<chewzerita>roptat: yeah, I saw that, but how do I load a username/password? how do I add a previous openvpn config file?
<roptat>ah, maybe that's something we don't support (yet)
<chewzerita>roptat: bummer
<roptat>I know we support public key authentication
<roptat>I'm the original author of the service btw :)
<roptat>I don't know if anyone worked on it since
<chewzerita>roptat: well thanks for your work anyway
<roptat>I'll have a look at the service definition, it shouldn't be too hard to support external config files, other services do it
<roptat>and then maybe I'll had username/password authentication if I find how
<chewzerita>roptat: alright, so should I just run openvpn manually on startup at this point?
<chewzerita>roptat: last time I tried to get it to work with networkmanager it didn't work, so know I just use plain wpa_supplicant
<chewzerita>(the networkmanager openvpn plugin that is)
<ecbrown>efraim: from a fresh guix clone, i ran ./bootstrap and now things are fine, thanks
<efraim>ecbrown: oh good, glad to hear its working now
<ecbrown>efraim: also i was able to compile gpgme 1.12.0 with a small change to gnupg.scm
<apteryx>chewzerita: yes, at this point for username/password authentication you have to run openvpn manually
<apteryx>I'm working on unlocking network-manager-openvpn, but progress is slow
<apteryx>(due to lack of time...)
<roptat>chewzerita, how do you setup password auth in openvpn (in the config file) usually?
<rekado>I’m still having trouble with splitting the guile.scm module…
<efraim>How about check and python-check?
<rekado>I’m also splitting other modules (like perl-compression)
<rekado>but this guile vs guile-xyz split really has me scratching my head.
<rekado>this isn’t supposed to cause problems with finding things like (@ (gnu packages compression) gzip)
<ecbrown>meh, there seems to be a test failure in qgpgme, disabling tests (#:tests? #f) i can get it to compile
<efraim>Is it the same test as 1.11.1?
<ecbrown>i googled that error, and found some discussion about failure due to date being after certificate valid time
<ecbrown>(they were setting date ahead artificially by 15? years, but the cert they were discussing would have gone invalid in early Jan. 2019)
<ecbrown>efraim: yes, same test
<bavier`>1 test failure for mariadb on core-updates
<bavier`>+Note 1588 Event execution time is in the past and ON COMPLETION NOT PRESERVE is set. The event was dropped immediately after creation.
<chewzerita>does guixsd have an equivalent to /usr/share/dict/words
<rekado>interestingly I get the same compilation problem when splitting python and python-xyz.
<rekado>same file, same problem.
<rekado>turns out the new module name was wrong. I couldn’t tell by looking at the error message, though. It’s a completely unrelated error message. Frustrating.
<civodul>chewzerita: there are dictionary packages but i'm not sure what provides "words"
<rekado>some modules contain literal references to the Python version. E.g. ibus.scm refers to “3.6” as the current Python version.
<chewzerita>civodul: okay
<bavier>is it best to just push fixes directly to core-updates, rather than posting to guix-patches first?
<rekado>bavier: if they are not controversial that’s fine.
<efraim>IMO depends on how big the change is, I normally push directly to core-updates
<civodul>rekado: re hard-coded python versions, that's terrible :-/
<rekado>civodul: I’ll fix this later.
<bavier>ok, I'll lean that way. thanks :)
<rekado>I’m going through all modules right now in order to see if they need “python-xyz” or just “python”.
<civodul>woow, thanks for doing it
<bavier>was this shared yet?
<bavier>oh, now I see the author :)
<civodul>bavier: heh, nice
<efraim>does dillo on guix work with https sites?
<efraim>(i'm on another machine ATM)
<bavier>anyone know how to sign with gpg while in an ssh session? I think I'm having issues with pinentry
<efraim>Something about gpg_tty but I don't remember offhand
<pkill9>i'll test efraim
<pkill9>yes it does efraim
<pkill9>i love it's 90s look, lol
<efraim>bavier: I have 'export GPG_TTY=$(tty)'
<bavier>efraim: sweet, that was the final piece; also had to switch the pinentry program and reload the agent
<efraim>pkill9: looks like there might be a bug in debian's dillo then
<efraim>unbelievable! my 2007 macbook has opengl 2.1 and my 2005 ibook G4 has opengl 3.1
<efraim>i bet my 2007 machine has intel graphics
<apteryx>couldn't build groff on core-updates; the doc output is not generated. Is this new?
<apteryx>also it seems Python 2.7.15 is not reproducible.
<bavier>apteryx: I have a patch for groff that I'll be pushing shortly
<efraim>apteryx: there's a patch on guix-patches
<apteryx>(at leas --rounds=2 failed, as well as --check, but didn't get to investigate farther)
<bavier>(and for python-pycodestyle, fwiw)
<apteryx>bavier: OK, good
<davidl>Im having a weird issue with wget and guix download aren
<apteryx>suppose I wanted to build a Guix environment compatible with that of a Ubuntu 12.04 (dockerized) -- I want to force glibc to version 2.15, at least. What would be the best way to accomplish this?
<apteryx>I'd need to define a 2.15 glibc package, for one.
<apteryx>Then, should I go the GUIX_PACKAGE_PATH way or perhaps a Guix channel would be helpful here?
<rekado>apteryx: you would need to rebuild everything.
<rekado>are you sure you need 2.15 and not just a compatible glibc?
<apteryx>I'm trying to build an old OpenWRT (12.09), and m4 currently fails because of the glibc 2.28 from Guix
<apteryx>(openwrt builds its own toolchain, which includes m4)
<rekado>it may be easier to fix this build error than to rebuild *everything* from source because of a glibc downgrade.
<apteryx>right, I tried applying one of our patches but it failed -- I'll give it another shot
<apteryx>and if it fails, I'll review the Docker patches from Danny ;-)
<davidl>Im having a weird issue where wget and guix download arent trusting a site's certificate but the browser does. What could be the reasnon for that?
<davidl>I have checked the SSL_CERT_DIR and SSL_CERT_FILE variables which points to what I believe are the correct paths.
<bavier>davidl: I think icecat has it's own certs, maybe try installing the nss-certs package?
<davidl>bavier: I have installed both nss-certs and le-certs and the issue persists.
<davidl>bavier: Im considering whether I have to install it in the system config.scm instead of a user profile?
***slyfox_ is now known as slyfox
<davidl>bavier: I believe the issue is that I need the le-certs in the system config.scm instead of user profiles, since the site in question that doesn't work to download with wget is using valid LE-certs.
<davidl>but I don't understand why.
<apteryx>rekado: managed to adapt the m4-gnulib-libio.patch so that the gnulib bundled with the old OpenWRT m4 builds :-)
<apteryx>thanks for suggesting
<civodul>switching to guile-json 3.x will be a lot of fun...
<eubarbosa>hey, what about openjdk11? Its exactly what holding me back to install guixOS... not rushing anyone tho!
***hsanjuan_ is now known as hsanjuan
***bmpvieira_ is now known as bmpvieira
<eubarbosa>.*? is now know as .*?1
***sturm is now known as Guest27466
*rekado fixes python-numpy-documentation
<civodul>rekado: great!
<civodul>i'm on a libssh upgrade
***ArneBab_ is now known as ArneBab
<rekado>hmm, now that I’ve spent some time to simplify the texlive profile hook, I think it may not be needed at all :)
<civodul>now *that* is a simplification :-)
<ArneBab>can I use shepherd as a userspace daemon manager (run services once I login)?
<bavier`>ugh, the mariadb test failure on core-updates has a fix, but the bug ticket just says "fixed in <version>, no changelog entry needed." :/
<bavier`>time to go hunting
<civodul>ArneBab: yes, but currently you have to arrange to start it yourself (e.g., via ~/.xsession)
<civodul>rekado: BTW, which machines now have qemu-binfmt set up?
<ArneBab>civodul: when I try to run shepherd, I get "shepherd
<ArneBab>Service root has been started."
<ArneBab>that sounds like an error
<ArneBab>or isn’t it … ?
<ArneBab>… oh, second invocation throws an actual error …
<ArneBab>seems I misunderstood the message
<ArneBab>sorry …
<civodul>it's not an error :-)
<civodul>shepherd manages a service DAG and "root" is the root of that graph
<ArneBab>now I just need to find out how and where to define a service that runs SSH to forward ports …
<ArneBab>I’ll read the info
<rekado>civodul: all of them.
<rekado>except for the big node
<rekado>but the nodes may need to be rebooted to start the service.
<civodul>indeed /proc/sys/fs/binfmt_misc/ seems to be empty there
<civodul>it's surprising the service wasn't started upon reconfigure
<ArneBab>found it: ~/.config/shepherd/init.scm
<ArneBab>when I run shepherd ; herd stop root ; shepherd, the second invocation of shepherd gives an error (address already in use)
<rekado>civodul: this happened before on the first node that I configured with that service.
<rekado>it also had to be rebooted.
<civodul>weird, and i can't run "herd status" there
<ArneBab>civodul: to get rid of the mangling of x86_64 to x86<sub>64</sub> on , you can set #+options: ^:nil
<civodul>ArneBab: thanks, but i'm not the author :_)
<ArneBab>ok :-) — it’s still a great read :-)
<eubarbosa>ambrevar is everywhere hehe
*janneke pushes first fully wip-bootstrap
<janneke>lots of work still to do
<janneke>*fully functional
<Copenhagen_Bram>Hi! I just noticed that Guix's latest version of gzdoom available is 3.3.0. How would I install gzdoom 3.7.1 on GuixSD?