IRC channel logs

2017-10-26.log

back to list of logs

***sonnixin_ is now known as sonnixin
<fusion809>Hi folks. I've just booted the official GuixSD 0.13.0 VM image and whenever I do anything package management-related I get the error: "ERROR: In procedure getaddrinfo: Name or service not known"
<sneek>Welcome back fusion809, you have 1 message.
<sneek>fusion809, mbakke says: chsh won't work on Guix, the shell needs to be set through the system configuration: https://www.gnu.org/software/guix/manual/guix.html#User-Accounts
<fusion809>Running it (the VM) per the guide https://www.gnu.org/software/guix/manual/html_node/Running-GuixSD-in-a-VM.html
<fusion809>DuckDuckGo searching the error gets no results. Completely stuck as to how to fix this. Running `ping www.google.com` indicates there's no network
<buenouanq>I know nothing about VM networking.
<buenouanq>If it's the same as when you install from a frash drive I maybe can help.
<buenouanq>Sounds like it might be the same.
<fusion809>Setting it up per the manual (https://www.gnu.org/software/guix/manual/guix.html#Hardware-Considerations) doesn't work. ifconfig -a shows eth0 and lo. Running ifconfig eth0 up doesn't set the net up properly
<fusion809>nor does ifconfig lo up.
<buenouanq>did you run # dhclient -v <interface>
<fusion809>Thanks! It works now after running that. I see the manual does mention it, but the layout made me think that part was for wireless connections.
<buenouanq>the wired/wireless sections are slightly more indented, but yes, at first/quick glance it might be hard to tell
<civodul>BTW, network-manager changed my life
<civodul>it works sooo much better than Wicd!
<civodul>(not that i'm really surprised, but still)
<fusion809>Hey is there a way to expand the GuixSD VM image so I can install extra software on it? I'd like to run GNOME on Guix to make it feel more "homely" to me.
<fusion809>NVM I found something to help me.
<fusion809>I can resize the disk with qemu-img but how to incorporate that extra space into my root partition without deleting my existing partitions. The extra space is at the end of the disk and the first partition is my root partition, second is EFI
<fusion809>Might be simpler just to install GuixSD to a VM using the USB image.
<fusion809>Could use guix system vim-image, but anyone have an example file argument for it? It says it must specify operating system but I haven't the foggiest how to do so for it.
<fusion809>vm-image^
<fusion809>I know of https://www.gnu.org/software/guix/manual/guix.html#operating_002dsystem-Reference but an example file would help
<brendyn>anyone here use openmailbox?
<CharlieBrown>I installed Stow onto Trisquel using Guix. I'm getting a weird error:
<CharlieBrown>ACTION uploaded an image: symlinkerror.png (38KB) <https://matrix.org/_matrix/media/v1/download/matrix.org/fVdcnUCSAyRdBXUfYKvIBsxm>
<CharlieBrown>ACTION posted a file: permprob.txt (2KB) <https://matrix.org/_matrix/media/v1/download/matrix.org/EAWOXdmoDEbzfJRoIvuAOIKp>
<civodul>CharlieBrown: i suspect there are few Stow users here
<civodul>perhaps you should email bug-stow, unless you think the error is Guix-specific?
<CharlieBrown>My solution is simple: `mv ~/Stow/Keep/dotfiles ~/Stow/dotfiles`.
<civodul>good that you found a solution :-)
<CharlieBrown>I hope making my whole home directory a pile of symlinks won't hurt anything.
<clacke[m]>it will only make your life better
<CharlieBrown>clacke: How?
<clacke[m]>there's a reason you did it, right.
<clacke[m]>compartmentalization, easy to .git the subset of HOME you care about, easy to temporarily flip things around
<CharlieBrown>Yeah.
<CharlieBrown>I wanted to separate the things I care about from the things I don't care about.
<civodul>rekado: thanks a lot for going through old patch submissions!
<fusion809>Is it normal for "substitute: updating list of substitutes from 'https://mirror.hydra.gnu.org'...100%" lines to be repeated over and over for over a thousand times during GuixSD setup (with guix system init)?
<fusion809>I as as my system is doing this and it seems to be taking forever with this and I'm wondering is this normal? I've allocated this VM 8GB RAM and 3 CPU cores so it should be going fairly fast.
<fusion809>Aha good that phase has past.
<brendyn>I tried downloading 9k emails via pop with gnus mostly from guix mailing lists, but it used up a few gigs of ram and then crashed. I know many people use gnus but how do they manage to download such a large list?
<efraim>not gnus or emacs related, but I use offlineimap to have a local maildir copy of my mail
<brendyn>Mine is nnml format I think
<brendyn>Gnus uses up all of my 16GiB of RAM and dies with Maximum buffer size exceeded
<brendyn>i added 11GiB of swap, now emacs has enough memory but it's spent half an hour in "Disk Sleep" state burning up my hard drive probably
<brendyn>welp it just used up all the swap too
<efraim>Can you download it in segments?
<jlicht>hi guix!
<brendyn>Possibly. I just solved it by deleting all my emails
<jlicht>Could someone please try to run `guix package -A <anything>'? I am getting some weird ruby-ffi backtraces
<brendyn>oh wow, -A is much more usable than -s
<jlicht>brendyn: I think so too! Also, nvm, I did something weird in one of my custom packages in GUIX_PACKAGE_PATH >.<
<clacke[m]>right, I should remember to use -A more. It's usually what I really mean when I say -s.
<brendyn>I suppose I'll just have to package icedove, since email in emacs is so unimaginably difficult.
<jlicht>brendyn: what issues are you having? I am (for ~80% of my mailing needs) quite happy with mu4e at least
<jlicht>I also concur that setting it up for the first time was a bit intimidating
<brendyn>I mark all my mail as read, close emacs, reopen it then everything is unread again
<brendyn>I can't see any way to resplit mail after ive already downloaded it, so i have these weird folders like 'bogus' and 'mail.misc' with some of my emails in them
<rekado>brendyn: I found gnus to be too difficult to set up, so I went with mu4e.
<brendyn>also i tried to download my mail after a few months of not trying emacs. there was 9000 mail at 100MiB. Emacs used up my 16GiB RAM, 11GiB swap before crashing and still failed to download any mail
<jlicht>brendyn: are you using gmail?
<brendyn>openmailbox.org
<brendyn>but i also have a gmail one too but it doesn't get much mail
<jlicht>because with gmail, the same (duplicate) email is ususally received twice, which leads to some annoyances when trying to mark them as read
<brendyn>i might try mu4e. I'll have to try setting up offlineimap
<jlicht>brendyn: I use isync(mbsync?) for syncing, password-store for managing credentials and then mu (and mu4e) for indexing and reading.
<ng0>hey, does 'NetworkManager' have an 'vpnc' plugin?
<ng0> (vpn-plugins (list network-manager-openvpn)))) -> (vpn-plugins (list network-manager-openvpn network-manager-vpnc)))) then?
<ng0>ACTION tries to reconfigure and hopes nothing explodes (should fix the coreboot… no menu…)
<ng0>ohnoes. right. I need to package that
<ng0>sigh. cursed vpn
<bavier>efraim: do you have a "Processor" field in your aarch64 system's /proc/cpuinfo?
<bavier>efraim: or, does the perl-sys-cpu package build for you?
<efraim>bavier: I'll check when I get back home
<mekeor>which packages do i need in order to compile and run a java source file? as far as i understand, i need "java" and "javac" commands. i installed icedtea but it only gives me "java" command. i think i need "javac", too, right?
<mekeor>which guix-package includes the "javac" java-compiler?
<davexunit>mekeor: I believe icedtea has a "jdk" output that you need to install for that
<mekeor>ah, cool
<ng0>grmph. does the NetworkManager-openvpn work? With my -vpnc package it works but I get a message on org.NetworkManager. … vpnc not being found
<ng0>I know it works because there is a working authentication but it fails in the org.freedesktop part
<ng0>(watching it on tty)
<mekeor>davexunit: works! thank you :)
<davexunit>np!
<ng0>result="fail" reason="The VPN service 'org.freedesktop.NetworkManager.vpnc' was not installed." <- while the file in the dbus directory of the package doesn't look any different than the one NM-openvpn uses
<ng0>happens in xfce and mate
<ng0>actually…
<ng0> /run/current-system/profile/share/dbus-1/serves/ has no networkmanager file
<ng0>and both openvpn and vpnc plugins are in my service list
<ng0>am I supposed to add them to my packages liust aswell?
<ng0>the NM files exist in the other dbus directories, but not the plugin .service files
<ng0>*/dbus-1/services/ I meant
<ng0>well, this (including both packages systemwide) didn't fix it
<ng0>ah!
<ng0>this is because at least nm-vpnc includes the dbus-1 in /etc/
<ng0>*out/etc/
<ng0>iirc the same was trued for -ovpn
<rama_dan>GuixSD uses the "libre" kernel, right? but how does that affect a user's life? -for example, is it not compatible with some software?
<cehteh>hardware rather
<jlicht>rama_dan: I had to get a freedom respecting wifi adapter to make everything work on my laptop
<rama_dan>oh ok
<rama_dan>well, i'm thinking of server use.. and i guess server hardware tends to be quite standard (and thus, supported even by the libre kernel?)
<jlicht>You can check h-node.org to see if your hardware is listed. But usually, it Just Works.
<ng0>rama_dan: even upstream linux has moved away from accepting blobs in the kernel and moved the included blobs to an extra repository
<rama_dan>ng0 jlicht alright, thanks
<jlicht>rama_dan: good luck on your install :-)
<rama_dan>jlicht thanks :P i'm still not sure what to do though.. i'm trying to figure out how to set up server stuff for my SaaS app
<rama_dan>i suppose my top picks atm would be openbsd and guixsd, but it seems i won't be able to use guix on a mac.. maybe ever, but at least not soon
<rama_dan>so i guess the main advantage GuixSD would have over OpenBSD would be Guix itself.. but it's not useable on Macs (and i can't use anything else on the desktop because i need to get shit done :p)
<jlicht>rama_dan: if I recall correctly, there were also some people running a minimal Debian server, with a Guix (not GuixSD) install used for managing most software on the server.
<rama_dan>but OpenBSD is very good too and has built in stuff like httpd and relayd which would probably work really well
<ng0>jlicht: yes
<rama_dan>jlicht yeah, that might work too.. but i'm kind of unhappy with the idea of using any "typical" Linux at all, because they all seem somewhat half-assed to me
<ng0>we don't have httpd (yet) (working on it)
<rama_dan>ng0 you mean porting the openbsd httpd?
<rama_dan>or like.. Apache? :P
<ng0>httpd -> Apache2
<rama_dan>rright :p
<rama_dan>well, i don't want to use apache either
<rama_dan>openbsd has made their own http server which is just called "httpd"
<jlicht>ACTION is off to run
<rama_dan>but should be super solid and easy to configure, much like all the main tools in openbsd as far as i know
<rama_dan>cya :)
<rama_dan>does guix have a package for nginx btw?
<ng0>yes, and service
<ng0>damn… I think the plugins of NetworkManager have no way to configure the dbus-servicedir. I'll try a hack for now and send a patch upstream (though I've never sent anything to GNOME, no idea how open/welcoming they are on build system patches)
<ng0>oh great. they have the service, but they throw it into one big file with other stuff
<ng0>in the wrong dir… hm.
<ng0>I'll figure this out
<rama_dan>ng0 btw, does guix let you configure nginx with a "base path" for its config files? like on <distro X>, the configs need to be in some specific place that's distro-specific, right? i'd like to be able to configure it myself
<ng0>no idea what you mean
<rama_dan>ok, nevermind then :)
<ng0>oh, i skipped a word
<rama_dan>but basically i mean the directory where nginx wants to read its config from
<ng0>the service is documented.. I have no idea rigfht now. you can take a look at the guix-maintenance repo for an example
<ng0>there are the nginx + system configs of bayfront + berlin .guixsd.org
<ng0>you can basically construct what you want
<rama_dan>hmm, ok
<ays>hi! I am just installing guix on my fedora machine. I am following the documentation at https://www.gnu.org/software/guix/manual/html_node/Binary-Installation.html.
<ays>The documentation says mv var/guix /var/
<ays>but my guix tar doesnt have a var/guix, only a guix directory.
<ays>I am using guix 0.13.0
<ays>Are they the same?
<ays>Also, I can't seem to find var/guix/profiles/per-user/root/guix-profile?
<rekado>rama_dan: I’m using GuixSD on different servers.
<rama_dan>rekado oh? how's it been so far?
<rama_dan>you mean like for a SaaS app or other "critical" stuff?
<rekado>rama_dan: on many old Sun Fire X2200 M2 servers, but also on recent Dell servers.
<rekado>with these servers everything works out of the box
<rama_dan>mm
<rama_dan>cool
<rama_dan>so you're hosting the servers yourself?
<rekado>even the on-board network interfaces work fine, which is surprising.
<rekado>yes.
<rama_dan>ok
<rekado>sometimes network interfaces are problematic.
<rekado>depends on the chipset.
<rama_dan>oh :x
<rekado>the Sun Fire servers have NVidia network chips and Broadcom
<rama_dan>and there are problems just because some non-libre stuff is missing?
<rekado>both are known to cause problems often; but in this case they work just fine.
<rama_dan>or just general linux problems? :P
<rekado>I have no problems with these servers at all.
<rama_dan>oh i thought you meant network interfaces you've worked with
<rama_dan>but you meant it in general?
<rekado>but on some servers with bad Broadcom hardware the network card simply won’t be recognized
<rama_dan>ok
<rekado>I also bought 10G Intel network cards, which also work great.
<rama_dan>mm
<rama_dan>well, i'm interested in using Guix(SD).. it seems like it might end up being extremely productive for my (future :p) business
<rama_dan>but ofc there might be some pitfalls i have no clue about
<rekado>when you’re shopping for hardware it pays off to make sure that it will work without non-free software
<rekado>because nothing is more annoying than hardware that stops working because the vendor decided to stop supplying you with the binary blob you need to use it.
<rama_dan>i'll be using a VPS or some "cloud" provider, at least for the foreseeable future
<rama_dan>and ofc, i have no idea if those tend to have compatible hardware
<rama_dan>i mean.. generally any typical linux would run just fine, right? but maybe there's a difference with the libre kernel
<rekado>VPS have virtualised hardware
<rekado>so that would usually not be a problem at all.
<rama_dan>ah, right
<rekado>it’s different for hosted servers, where you have direct access to the actual hardware.
<rama_dan>i have almost no clue about virtualization either, but .. i take it GuixSD would be a "good citizen" there?
<rama_dan>yeah
<ng0>depends on the virtualization type
<rekado>but with a VPS that’s usually abstracted away so the virtual OS doesn’t need to know anything about the underlying hardware.
<rama_dan>yeah, i forgot about that
<rekado>there’s the virtio class of paravirtualisation drivers, which are very simple to implement, and Linux has support for them.
<rama_dan>ok
<rekado>(they are little more than buffers)
<rama_dan>mm
<rama_dan>so that's what would be used with GuixSD under a host OS?
<rama_dan>and everything just works?
<rekado>yes, probably
<rekado>I also use GuixSD under KVM and VMware.
<rekado>both work without problems.
<rama_dan>ok
<rama_dan>sweet :)
<efraim>bavier: perl-sys-cpu fails one of the tests
<bavier>efraim: ok, thanks for checking. same for me.
<bavier>there are some patches at https://rt.cpan.org/Public/Bug/Display.html?id=95400
<bavier>but I don't know if they're "good", i.e. they might have bit-rotted a bit already
<bavier>e.g. my system doesn't have a "Processor" field in /proc/cpuinfo, so the first patch doesn't do any good.
<ng0>isn't it "processor", not "Processor"? That's what I have
<efraim>i also have 'processor' with a small 'p'
<efraim>i think debian runs smoke tests instead
<bavier>ng0: apparently at one point there were both "Processor" and "processor"
<ng0>ah
<bavier>efraim: oh smoke-tests.
<bavier>sure, I like their solution better.
<db48x>I'm playing with guixsd and I've hit a snag
<mekeor>db48x: yeay
<mekeor>db48x: what's the snag?
<db48x>I installed rsync into my user profile, but then discovered that while my user profile is activated when I ssh in interactively, it's not activated when I ssh in non-interactively
<db48x>obviously I could install rsync into the system profile and it'd work (and probably should), but that won't help me for everything
<db48x>how do I tell it what profile to activate in this case?
<mb[m]1>db48x: If you open ~/.bashrc, there's a line testing for `cat`, mentioning noninteractive use. You can change that to test for rsync, and it'll work.
<mekeor>db48x: what does "ssh in interactively" mean?
<db48x>/etc/profile has a bunch of stuff for activating profiles, but the default .bashrc only calls it if SSH_CLIENT is set (it is), and if cat is not found in the current PATH (but cat is found, because it's in the system profile)
<db48x>mekeor: sshing in normally, without a specific command to execute
<db48x>mb[m]1: that makes sense, but it seems like a hack
<mb[m]1>Yes indeed. It would be great to have a better solution. Maybe just source the profile unconditionally? That's what I do anyway.
<db48x>I guess this is an area that needs some thought
<mb[m]1>db48x: feel free to send a patch or start a discussion on guix-devel :)
<db48x>:)
<db48x>another snag is that I can't use Emacs's TRAMP mode to edit files on the guixsd machine; it says that it can't find a working ls
<db48x>I haven't investigated this one as closely yet
<jlicht>db48x: Maybe https://lists.gnu.org/archive/html/help-guix/2016-10/msg00042.html can help for TRAMP support on (remote) GuixSD machines.
<db48x>hmm
<db48x>jlicht: that does indeed work; thanks!
<db48x>I guess I searched for the wrong thing
<ng0>what is the difference between let's say <busconfig><policy user="root"><allow own_prefix="org.freedesktop.NetworkManager.vpnc"/> <allow send_destination="org.freedesktop.NetworkManager.vpnc"/></policy>[…] and the same thing with "allow own" instead of "own_prefix"?
<jlicht>db48x: yw. I can definitely recommend downloading the guix-devel ML archives, as most problems I run into using Guix(SD) have already been mentioned at least once.
<ng0>I'll try to replace the content and see if the policy is then accepted as installed
<ays>I seem to be getting this error when installing guix, can someone help me please?
<ays>
<ays>
<ays>ice-9/psyntax.scm:1558:32: In procedure expand-macro:
<db48x>jlicht: that's also a nice idea :)
<ays>ice-9/psyntax.scm:1558:32: Syntax error:
<ays>unknown location: error: no 'ungexp' substitution (ungexp %utils-module)
<jlicht>ays: Could you please post the command you entered, as well as the complete backtrace to paste.lisp.org?
<db48x>so here's a question that I don't think has been asked before: in a guix import script I'd like to call read to parse some text containing s-expressions
<db48x>but it looks like I need to change some options for read and then put them back afterwards
<db48x>is there a nice way to do this already written, or do I have to figure it out myself?
<db48x>I'm guessing no, because I don't see anyone calling read-set!
<bavier>db48x: where is the text coming from?
<bavier>db48x: anyhow, no, I don't know of that being done before
<db48x>bavier: fetched from a url?
<db48x>looks like this: http://paste.lisp.org/display/359540
<rekado>db48x: what options do you mean?
<db48x>rekado: specifically the keyword option, because the keywords in this sexp are "wrong"
<rekado>do you need to read them as keywords or just as atoms?
<rekado>(with-input-from-string "(:hello world)" read )
***sonnixin_ is now known as sonnixin
<rekado>that gives me the list (:hello world)
<db48x>it feels wrong to not use keywords
<db48x>but technically
<db48x>I suppose even if they get interned into my package or whatever it could still work
<db48x>ah, cve.scm does do something like this
<mlin>Hello, I was wondering if it is at all possible to run GuixSD in gnome-boxes. If not, I need some help with getting qemu working properly.
<rekado>mlin: I haven’t tried running GuixSD in gnome-boxes, so I don’t know if it is possible. Is boxes a frontend for qemu?
<mlin>Yes it is
<db48x>mlin: the qemu directions in the manual worked for me
<db48x>I think I added --enable-kvm
<db48x>so that I could use -cpu host
<mb[m]1>Huh, the wget-1.19.2 release tarball seems to be missing the CVE-2017-6508 patch, even though it is in the git tag.
<mb[m]1>Never mind, somehow 'patch' managed to apply it, even though it was there from before.