IRC channel logs

2017-10-11.log

back to list of logs

<TaoHansen[m]>as a regular user, things are more difficult for me. as an example, from a USB session i was able to invoke `herd start ssh-daemon` without a problem, now as a regular user, `guix package -i openssh` followed by `herd start ssh-daemon` results in `ssh-daemon could not be found`. additionally, configuring anything in /etc/config.scm only results in changes, packages and services available to the root user. when i used
<TaoHansen[m]>NixOS, changes and packages and services were available to all users.
<TaoHansen[m]>how do i resolve whatever cognitive dissonance i'm experiencing here?
<TaoHansen[m]>what parts of the manual should i familiarize myself with? is there a config.scm that applies strictly to nonprivileged users?
<mb[m]2>Packages specified in config.scm are available to all users. Services must also be defined in config.scm, just installing the package won't work. The example I sent earlier should have an OpenSSH service defined.
<TaoHansen[m]>mb: okay got it. sorry for the salvo of questions!
<mb[m]2>np. IIRC on NixOS root's profile was also available to all users, which may have been the source for at least some of the confusion :)
<nee`>I got a murmur service (the mumble server) running on guixSD.
<TaoHansen[m]>packages and use-package-modules: how do i discover what module must be specified for a desired package? ntfs-3g for this example.
<TaoHansen[m]>for all of my packages i've had to look up other's configs because i don't know how to search for a package's required modules.
<mb[m]2>Tao Hansen: You can use specification->package to avoid importing the modules up-front. See https://www.gnu.org/software/guix/manual/guix.html#Globally_002dVisible-Packages
<TaoHansen[m]>mb: is it possible to nest append within my pre-existing (cons* list?
<TaoHansen[m]>looks like it is from looking at others' configs but i continue to get an extraneous argument error.
<TaoHansen[m]>got it to build but this is way over my head. i should probably learn some actual Guile before going any further.
<efraim>wow, qt-5.9 built second try
<janneke>\\o/
<rekado>TaoHansen[m]: when you do “guix package -A openssh” it will not only show you the matching package but also the module where it is defined.
<rekado>TaoHansen[m]: for services, however, you would use something like “guix system search ssh”
<rekado>it returns a list of matching services and the location where they are defined.
<civodul>Hello Guix!
<sneek>Welcome back civodul, you have 1 message.
<sneek>civodul, dustyweb says: https://github.com/WebOfTrustInfo/rebooting-the-web-of-trust-fall2017/blob/master/draft-documents/smarm.md
<wigust>civodul: Hello!
<civodul>nice
<efraim>do I have to add openssh to the global list of packages for the openssh service to work?
<rekado>efraim: no
<rekado>you just need it for scp (on the target machine)
<civodul>it might be a good reason for openssh-service to add it to the global profile, no?
***fr33domlover is now known as fr33domlover_
***fr33domlover_ is now known as fr33domlover__
***fr33domlover__ is now known as fr33domlover
<ng0>hi. is anyone working (or keeping off-master) on an apache service?
<civodul>ACTION isn't
<civodul>nginx FTW! :-)
<ng0>nginx is an option, but we'd prefer apache. if no one is working on it, I'll add it to my job list.
<roptat>I don't think there's any work on apache, at least none that I've heard of
<ng0>ok
<ng0>upcoming (as far as I remembered the list): Services for murmurd, apache, mantis, buildbot, finishing my tlsdate service, and packages for mantis + buildbot. there's more I think.
<nee`>did you already work on murmurd?
<ng0>in the next months that is. possibly into 2018, depending on how documentation etc goes
<ng0>murmurd as a package exists
<nee`>I made a service for it yesterday. It works I just have to write docs and test it some more.
<ng0>nice!
<ng0>that's one upstream documentation less to read for me
<rekado>I have an itch wrt verbosity of Guix.
<rekado>I’d like to hide all build output by default and only print something like “building… (phase 2/7)”
<rekado>this can be disabled with “--verbose”. By default all output would go to log files instead.
<ng0>when the log file location is displayed at the end of the build, that would be a very good idea
<rekado>Another thing I find annoying is that the daemon and guix client both print the same message, e.g a hash mismatch error.
<ng0>python.scm is large. I need to add buildbot and its dependencies, where the one I am working on right now is autobahn and its dependencies. cbor seems like a thing that belongs in serialization.scm ?
<ng0>buildbot of course goes into ci.scm
<ng0>to answer my own question: yes.
<efraim>I was just gifted another PPC macbook
<civodul>efraim: oh, the good'ol days (if we assume there were "good" days in Apple land)
<civodul>i had a G3 at work ~12 years ago with GNU/Linux
<civodul>that was fun
<efraim>I feel like a PPC port of guix would benefit to offloading to a PPC qemu image
<ng0>does python-build-system have a concept of make-flags, or a better way to replicate this: SODIUM_INSTALL=system pip install pynacl
<ng0>maybe just setenv
<efraim>can I separate off /include from Qt? headers should only be needed at compile time, right?
<thomasd>hello Guix!
<thomasd>efraim: yes, I suppose you can.
<thomasd>are you trying to reduce the footprint of packages using Qt?
<wigust>How to get a path to a build directory?
<wigust>I need to replace ./SConscript with BUILD-DIR/SConscript
<wigust>Could I just use current working directory?
<civodul>wigust: yes, that should work
<thomasd>anybody have an idea how to have 'make check' skip certain tests, in an automake-based test suite?
<thomasd>(related to http://lists.gnu.org/archive/html/guix-patches/2017-10/msg00391.html)
<civodul>thomasd: i think you can do "make check XFAIL_TESTS='foo bar'"
<civodul>it will consider that foo and bar are expected to fail
<wigust>civodul: thanks!
<thomasd>civodul: that's what we did, but the success/failure is unpredictable (it seems to be a timing issue), so when we use XFAIL, we sometimes have failure due to 'unexpected success' :-)
<bavier>good day, guix
<wigust>bavier: o/
<davexunit>detailed guide on disable intel me: https://wiki.gentoo.org/wiki/Sakaki%27s_EFI_Install_Guide/Disabling_the_Intel_Management_Engine
<davexunit>probably a few additional packages needed to be able to do this with guix
<mekeor>ACTION waves to bavier and the other guix
<civodul>thomasd: then you can simply remove it from Makefile.in
<civodul>heya bavier, davexunit, mekeor :-)
<thomasd>civodul: thanks. Feeling a bit dense now :)
<civodul>davexunit: interesting, i'd love to run me_cleaner is this laptop i got several months ago
<davexunit>same
<davexunit>I have an old thinkpad x220 that I refuse to let go. would be nice to install coreboot and disable ME on it.
<rekado>I’m supposed to pick a laptop for the office, and it’s such a drag!
<rekado>I find it really hard to pick new hardware :(
<bavier>same
<mb[m]2>rekado: have you looked at the puri.sm laptops?
<mb[m]2>thomasd: will you look at disabling this test on 'master' for armhf (only)?
<jonsger>same here. if you have to pick one, I would recommend you to wait at least until begin of next year for the AMD Ryzen Mobile
<jonsger>I expect to see more power for the same money (like with Ryzen, Threadripper and Epyc + the reactions of Intel ^^)
<mb[m]2>Oh no, ceph CVEs. I may have to disable a lot of tests to build the later versions :/
<rekado>mb[m]2: I had not, because I was put off by the misleading advertisement during the crowdfunding campaign.
<rekado>mb[m]2: but looking at the laptops now, I think they might be good for an office purchase.
<rekado>I’m quite attached to the X200 keyboard and the laptop’s durability
<jonsger>rekado: yeah x230+ keyboard isn't as nice as the older ones :(
<thomasd>mb[m]2: Perhaps we should disable (skip) it for every architecture, no? Since it's failing/succeeding in an unpredictable manner (at least on x86_64), XFAIL is not the way to go (should have thought of that before).
<mb[m]2>Yes, but that will cause a lot of rebuilds. So we need a fix for armhf ASAP, then we can fix it properly on core-updates.
<mb[m]2>thomasd: ^
<thomasd>mb[m]2: ah yes, this is on master, got it. Still haven't found a really satisfactory way to do it though.
<mb[m]2>Thanks for working on it! Hopefully it can be easily substituted out.
<bavier>would be nice if 'guix environment' could take a derivation as an argument
<bavier>uff da, trying to debug a test hang on the bootstrap libunistring, but 'guix environment' seems to need that same package built in order to do its thing
<bavier>ahhh, --bootstrap :)
<bavier>nope, even that wants to build libunistring...
<bavier>argh, trying to instead create a profile from a manifest containing only the build inputs of guile-final's libunistring, guix wants to build a bunch of xorg packages. what?
<bavier>seems like I'm hitting this bug: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=26441
<bavier>which was apparently fixed, maybe I'll just let the test run to see if it ever finishes...
<ng0>would someone know if the intel I350 ethernet card works with linux-libre?
<civodul>bavier: i vaguely remember someone saying that the fix didn't fix it
<civodul>or maybe it just takes a lot of time to run
<bavier>civodul: it's going on 30 minutes now
<bavier>civodul: but yeah, reading some of the gnulib ml discussion, it seems there are additional issues, perhaps in the kernel or related to glibc
<bavier>disabling the test-lock tests locally might be my best bet at this point.
<civodul>yeah
<civodul>we should find a proper solution in core-updates
<civodul>or maybe just skip the offending tests
<civodul>can you reopen the bug and send the link to the discussion you found?
<TaoHansen[m]>civodul: i can't change date time in a standard GNOME session. i found another user who had the same issue but it doesn't look like he followed up. no error messages on the GUI. https://lists.gnu.org/archive/html/help-guix/2016-09/msg00060.html
<TaoHansen[m]>let me know if there's any info i can provide.
<TaoHansen[m]>the switches for Automatic Date & Time and Automatic Time Zone don't work at all but my understanding is that %desktop-services comes with everything GNOME needs to make use of ntp.
<Steap>./pre-inst-env guix weather --substitute-urls=https://hydra.gnu.org -m gnu/packages/python.scm <- Is this how guix weather is supposed to be used?
<civodul>TaoHansen[m]: i think i have a fix for that
<civodul>Steap: no, -m is followed by a "manifest" as for 'guix package'
<Steap>I see
<Steap>can we see the jobs currently running?
<civodul>running on what?
<Steap>rebuilding packages, basically
<civodul>one trick is to do "ls -d /tmp/guix-build*" :-0
<Steap>nah I mean
<Steap>on hydra
<rekado>Steap: you can check queued builds
<Steap>rekado: how?
<rekado>with the /api/queue?nr=10 URL
<rekado>but I think it’s better not to stress hydra.gnu.org with additional requests.
<rekado>unfortunately, this API endpoint is not yet available for berlin.guixsd.org.
<Steap>I see
<Steap>it literally takes days to rebuild stuff :/
<rekado>hydra only consists of a handful of servers.
<rekado>berlin has 18 servers at this point and might be a bit quicker
<rekado>on berlin the limitation is cuirass, which doesn’t yet perform as well as the hydra software.
<Steap>rekado: ok
<Steap>rekado: can berlin be used? I don't see the public key in the Guix repo
<rekado>Steap: yes, it can be used.
<rekado>Steap: the public key is available through the guix-maintenance repository
<rekado>I will add it to the web page shown at https://berlin.guixsd.org
<Steap>rekado: ok thanks
<jherrlin>hey guix, i (still) have some problem with OpenVPN. I just tried out the patch from the bug that i reported, http://git.savannah.gnu.org/cgit/guix.git/commit/?id=4b8b4418e609b5e0bfb6efbc11ac28deaa437e80
<jherrlin>if i am trying to start it with =nmcli connection up vpn-profile= it give me a *Connection activation failed: The VPN service 'org.freedesktop.NetworkManager.openvpn' was not installed.*
<jherrlin>and when i try to start the vpn-client service it's just telling me it cant start
<jherrlin>i dont get much from the logfiles, anyone have ideas on how to start debugging this?