IRC channel logs


back to list of logs

<lfam>thomassgn: Does the error show during the pull or the reconfigure?
<thomassgn>sneek: botsnack
<thomassgn>sneek: later tell lfam: error happens during both, but I found an old mail thread saying to remove the ~/.config/guix/latest symlink; and that did the trick.
<lfam>sneek: I stepped on this botsnack
<sneek>Welcome back lfam, you have 1 message.
<sneek>lfam, thomassgn says: error happens during both, but I found an old mail thread saying to remove the ~/.config/guix/latest symlink; and that did the trick.
<lfam>thomassgn: I'm glad you got past it!
***that_guy_ is now known as nliadm
***Piece_Maker is now known as Acou_Bass
<paroneayea>git failed to build for me
<paroneayea>*** ***
<paroneayea>not ok 3 - git svn branch tests
<paroneayea>ACTION looks at mailing list, guess I'm not the only
<rekado>paroneayea: this seems to be non-deterministic.
***muzik is now known as muzik_
***muzik_ is now known as muzik
<wingo>ACTION packages cloc:
<wingo>what's the state of the art for packaging perl programs? do you make a wrapper program that sets the perl path?
<wingo>seems to work
<ngz>Hello. I'm encountering a problem while trying to package "mgba". In a nutshell it compiles fine, but, during installation, I see "Set runtime path of "/gnu/store/...-mgba-0.6.0/lib64/" to "lib64"" during installation. Then, during "validate runpath" phase, something like "/gnu/store/...-mgba-0.6.0/lib64/ warning: RUNPATH contains bogus entries: ("lib64")" and the whole process ultimately fails wit
<ngz>"/gnu/store/...-mgba-0.6.0/bin/mgba: error: depends on '', which cannot be found in RUNPATH".
<ngz>This is the cmake-build-system, so it might be fixed by using -DCMAKE_INSTALL_LIBDIR appropriately, but it's a bit out of my league, I think.
<wingo>ngz: yeah that sounds about right
<wingo>try #:configure-flags '("-DCMAKE_INSTALL_LIBDIR=lib")
<wingo>just judging by what's in other packages
<wingo>in the arguments block
<ngz>I tried. I get the same error chain, with "lib" instead of "lib64".
<ngz>Fortunately, this is a quickly built package. I can try again. Without much hope.
<ngz>Confirmed. This is the same error.
<ngz>(mutatis mutandis)
<wingo>yeah dunno :/ cmake is a mystery to me
<wingo>rekado: hey i rotated my gpg key (also ssh key). i updated savannah but i don't know what the deal is for guix and its pre-commit hook. i seem to have managed to push but if there is something else i need to do to update the guix keychain lmk
<wingo>i guess this is more of a civodul / mark_weaver thing, dunno
<ngz>wingo: Thank you anyway :)
<wingo>for nothing :)
<balaga>Hi ! I tried to set a build with a publish server. I don't understand why getting a package from my server (as substitute) is by far slower than using hydra as substitute. Is that normal ?
<balaga>to set a build farm*
<ngz>wingo: For the record, it worked with (string-append "-DCMAKE_INSTALL_LIBDIR=" (assoc-ref %outputs "out") "/lib"), i.e., an absolute path. Ah well.
<wingo>yay monkeytyping
<wingo>always be rebuilding webkit
<wingo>so weird thing in guix + gnome: why when i type ~/Documents into the alt-F2 dialog box does it bring up that folder in baobab instead of nautilus?
<wingo>both programs are in the system profile (via the gnome-desktop-service) and neither in my user profile
<wingo>seems it is in Gio somewhere, it's picking the wrong app:
<htgoebel>Hello Guix!
<htgoebel>OT: I just read the thread about "git commit messages" and "automating commit checks." on guix-devel.
<htgoebel>For one of my project I'm looking for checks suppressing whitespace and reformatting-only changes (at github).
<htgoebel>Any pointers?
<wingo>ACTION no idea :)
<wingo>regarding my problem: "gio mime inode/directory" shows that both baobab and nautilus are registered, and that for some reason baobab is the chosen one. need more investigation.
<wingo>and if i add "inode/directory=org.gnome.Nautilus.desktop" to the "default applications" block of my ~/.local/share/applications/mimeapps.list then that fixes it for me.
<wingo>strange that baobab isn't available in the right-click "open with" menu tho from nautilus
<wingo>it could be that nautilus isn't happy with having so many XDG_DATA_DIRS
<wingo>aaaah regarding baobab, apparently the order of dirs in XDG_DATA_DIRS matters!
<wingo>it's preference-ordered
<wingo>and i would guess that the "gnome" package listing baobab as an input before nautilus makes gnome prefer baobab by default for that mime type
<wingo>hum, my XDG_DATA_DIRS has lots of duplicates
<wingo>i guess all is working fine, maybe just that nautilus special-cases directories or something
<balaga>Hi ! I tried to set a build with a publish server. I don't understand why getting a package from my server (as substitute) is by far slower than using hydra as substitute. Is that normal ?
<balaga>I work on 2 VMs : one which is set to be my serveur (substitute), one which is supposed to get packages by my server.
<balaga>My main issue is the subsequent : I build a package like shellcheck,tmux or perl. I make guix publish. The server is available. I make a "guix archive -authorize" and I enter my pub key of my server. I try a "guix package -i shellcheck --substitute-urls="http://myserver...".
<balaga>I can see the different gets on my server, so the connection is ok. Some packages are dowload through the server BUT after a few secs, a phase of compiling is launched and it's very very slow.
<balaga>Is my substitute server not a mean to avoid my second VM to compile ?
<balaga>sorry, i would like to say "build" when I said "compile"
<wingo>does anyone else have the problem that libreoffice doesn't show up in the gnome-shell launcher?
<wingo>and when you double-click on a .doc file for example that libreoffice isn't associated with it
<wingo>i think the .desktop files included with the project aren't being installed, or something like that
<efraim>balaga: did you authorize your machine to get substitutes from your server
<balaga>yes i did
<adfeno>wingo: is XDG_DATA_DIRS set correctly?
<efraim>balaga: I reread your --authorize command, what I normally do is `sudo guix archive --authorize <'
<wingo>adfeno: i think so yes
<wingo>adfeno: you don't have this same problem?
<balaga>I didn't write it entirely when I was explaining. I do exactly your command
<wingo>seems that fedora runs a post-pass over the included desktop files:
<wingo>something we don't do afaiu, and additionally libreoffice is built using gnu-build-system, not glib-or-gtk-build-system so perhaps these files aren't picked up anyway
<wingo>i will give that a go.
<adfeno>wingo: Also, be aware that if you are using Guix in a foreign distro, the distro' Xsession.d scripts might get mistaken if you explicitly set XDG_DATA_DIRS with the usual ${VAR:+:} bashism.
<wingo>adfeno: i am on guixsd on this machine
<balaga>efraim: The downloads from my server seems to be done but there is always a phase of building which length long long time.
<adfeno>wingo: Oh, I see. :)
<wingo>adfeno: if you use libreoffice from guix, can you repro the bug?
<efraim>balaga: there is some grafting that happens which isn't done with the substitutes, what packages are being built?
<adfeno>I don't have LibreOffice installed from Guix. I have been avoiding LibreOffice for almost a year now. It doesn't seem to be non-graphics-acceleration friendly, and requires a X display to edit documents. I have Trisquel's copy of LibreOffice, which seems to work fine.
<adfeno>ACTION misses an ODF (ODT/ODS) mode in GNU Emacs. :(
<balaga>efraim: Sorry if i'm slow, I don't want to tell you something wrong. IMO the first package which is build is guile-2.0.14 or bootstrap-binaries.
<balaga>efraim : the first packages downloaded are all dependancies that I can see with a "guix graph --type=bad-emerged shellcheck". But after there are some packages like "bootstrap-binaries-0", "guile-bootstrap-2.0" which are downloaded. I guess they are dependancies of the precedent packages.
<balaga>efraim: I can understand that there is dependancies to download, but... why are there so many phases of "unpack", "preconfigure", "configure" .... that seem not to happen when using the substitute hydra?
<thomassgn>hmm, I'm getting "system: error: profile contains conflicting entries for coreutils:out" and it lists coreutils-8.26 coreutils-8.27
<thomassgn>that is when I run guix pull.
<thomassgn>first guix pull told me it needed guile-git, which I installed with guix package and then I get this about coreutils ...
<thomassgn>nevermind me, I didn't read the obvious message guix pull was giving me...
<ng0>adfeno: you might want to try antiword or what this cure-all 'can-i-haz-all-formats' application was called
<ng0>no, not antiword.. something else
<tetero>ng0: can-i-haz-all-formats app? what?
<rekado>ng0: did you mean ghc-pandoc?
<ng0>I'll find it again one day.. I didn't start packaging it because it had so many dependencies
<ShalokShalom>hi there
<ShalokShalom>is artifact support provided?
<ShalokShalom>and container?
<rekado>ShalokShalom: what do you mean by artifact support?
<ShalokShalom>single apps, no container
<ShalokShalom>Can apps get isolated, so they dont interference with each other?
<ShalokShalom>Whats about sharing the ressources and rollbacks?
<davexunit>ShalokShalom: you can use 'guix environment --container' to achieve some amount of isolation when you need it.
<davexunit>and IIRC something was introduced awhile back that allows GuixSD system services to run in containers as well
<ShalokShalom>davexunit: which kinds of limitations bring this?
<davexunit>limtations? well, network isolation isn't very mature. you can either give the container full network access or none, there's no virtual network devices bridging containers with the main system like docker has.
<davexunit>but aside from the network access, I was even able to boot full GuixSD systems in a container, which was neat.
<davexunit>sometimes I like to run certain applications in a container as a way to have a scratch pad for some quick hacking
<davexunit>like to get a ruby console or someting 'guix environment --container --ad-hoc ruby -- irb'
<ShalokShalom>ok, fine thanks
<ShalokShalom>and are deltas supported?
<ShalokShalom>so, incredimental updates
<davexunit>I haven't seen any way that it can apply to functional package management
<rekado>it might work outside of the functional package management paradigm, as an optimisation.
<rekado>instead of fetching archives containing identical files and then deduping them in the store one could compute this in advance
<rekado>but it’s not currently feasible
<rekado>this would require support from the build farm and I don’t see a way we can burden it with additional computations for this feature, which would only provide small gains.
<ShalokShalom>davexunit: openSUSE use them since a long?
<ShalokShalom>seems this fair?
<ShalokShalom>is there anything else to mention?
<cbaines>ShalokShalom, well apt supports deltas
<ShalokShalom>just in theory
<cbaines>In practice as well, e.g.
<ShalokShalom>i thought about a yellow there
<pmikkelsen>hi guix!
<pmikkelsen>When writing my MPD service, I noticed that mpd can't find any ALSA device, OSS device or pulse audio device. That made me wonder what the "default" is for guix? And how would I go around starting pulse audio for example, before mpd? I don't see a service for it in the manual
<ng0>pulse audio just starts
<ng0>it starts on demand
<pmikkelsen>Neat, thanks
<ShalokShalom>cbaines: outdated since 2013
<ShalokShalom>no news item since then at least
<ShalokShalom>is it still active?
<ShalokShalom>" refers to deltas for security updates of Debian/stable (currently Squeeze) for the architectures i386,amd64. "
<roptat>hi, I'd like to send a patch to add capstone and radare2 ( and
<roptat>they are tools for reverse engineering binary code, and I'm not sure in which module they should go
<davexunit>roptat: how about gnu/packages/engineering.scm as a start?
<davexunit>someone may suggest something different, but that seems a good place.
<ng0>we had this a long time ago in discussion
<ng0>I have one other software which needs lots of rust
<davexunit>ng0: there is no such module
<ng0>but revrrse-engineering.scm is also not very specific.. so.. maybe just engineering
<ng0>I know
<davexunit>just put it in an existing module so the patches can get sent
<roptat>ok, I'll put it in engineering.scm then
<rekado>ShalokShalom: functional package management is very different from rpm and deb.
<rekado>ShalokShalom: doing differential updates is easier with traditional package management.
<ShalokShalom>differential updates are delta updates?
<ShalokShalom>what are the benefits in practice?
<rekado>delta updates often require less data to be transferred over the network, at the expense of CPU.
<pmikkelsen>hi guix
<bavier`>hellp pmikkelsen
<rekado>pmikkelsen: hello!
<rekado>I’m having a tiny problem with offloading.
<rekado>ssh hydra@ guile -c "'(use-modules (guix))'"
<rekado>this works for exactly one host
<rekado>for other hosts this fails
<rekado>but when I actually log on as the “hydra” user and run the command locally it works.
<rekado>ah, well ssh hydra@ guile -c "'(display (getenv \\"GUILE_LOAD_PATH\\"))'" shows #f on those machines
<rekado>not sure why it works when logged in, though. Not sure at what point this is set.
<janneke>rekado: those are GuixSD machines?
<rekado>but with a twist
<rekado>they used to be Ubuntu
<rekado>and then I ran guix system init
<rekado>I think the problem is some remaining Ubuntu debris
<janneke>i'm sure you know that ssh doesn't give you a login shell
<rekado>I guess I should first clean all the Ubuntu binaries from the servers
<janneke>anyway, you'll need to look at the .bashrc, .profile and what not on that machine
<rekado>yes, they are the same as the account was created by GuixSD.
<janneke>no packages installed/packages missing in the user's profile?
<janneke>if .guix-profile/etc/profile is empty...
<rekado>the profile is indeed empty, but that’s by design
<rekado>the bashrc loads /etc/profile when this condition is true:
<rekado>if [ -n "$SSH_CLIENT" -a -z "`type -P cat`" ]
<rekado>and /etc/profile in turn sources the system profile’s etc/profile
<rekado>oh, another big difference is that on the working machine lsh is used
<rekado>I guess that also affects how the environment is set up
<janneke>i never got offloading to work with ssh on guixsd, fwiw
<janneke>didn't invest any time, just switched back to lsh
<janneke>on ubuntu+guix, ssh is fine
<rekado>I guess I need to figure out how to make it source the system profile’s etc/profile file to set GUILE_LOAD_PATH
<rekado>that’s the last missing bit
<rekado>sshd uses pam
<rekado>and that’s a little more … intransparent
<rekado>I think I got it
***fr33domlover is now known as fr33domlover_
<rekado>sshd does load /etc/environment but it does not set GUILE_LOAD_PATH
<janneke>ah, that makes sense -- where does lsh get that setting from, then?
<rekado>there’s an override in /etc/profile, especially for lsh
<rekado>because lsh cannot use pam_env we source the system profile’s etc/profile instead.
<rekado>so all I need to do is reconfigure the servers with an extended etc/environment service
<janneke>won't we want that for any guixsd+ssh machine?
<rekado>I think so
<rekado>hmm, that environment service is part of “essential-services”, which is a procedure.
<rekado>I can’t seem to just modify-services and extend it
<rekado>hmm, that’s inconvenient
<roptat>hm.. is it possible to specify both an IPv4 and an IPv6 address in static-networking?
<rekado>roptat: doesn’t look like that’s possible at the moment.