IRC channel logs

2018-02-27.log

back to list of logs

<axg>Is it the case for anyone else that open tramp sessions in emacs, make the rest of it slow down considerably, most notable with input lag?
<atw>sneek: later tell axg let me test tomorrow at work, where I can TRAMP into a few servers.
<sneek>Got it.
<ng0>on the new build server I have, my connection to it times out with the build of chromium. is there something I must do with one/both of the guix-daemons? like increasing timeout?
<ng0>when the build server was in the LAN, this was no issue at all. it's just that my internet connection is terrible
<ng0>okay, also 2 vCPUs are slower than a real quadcore .. but timeouts, I mean of course it is the option --timeout
<ng0>I just forgot its default.. 8 hours?
<ng0>the default is "0".. then it seems like my connection is the problem?
<efraim>does anyone have experience using nfs shares in kodi on guixsd?
<civodul>Hello Guix!
<civodul>hey g_bor
<civodul>how was that Outreachy chat?
<g_bor>Hello!
<g_bor>It went fine.
<civodul>awesome
<g_bor>It seems that I got 3 people interested.
<civodul>nice, let's see how it goes!
<g_bor>I pointed them to this channel, if they have questions.
<g_bor>I've also mentioned the mailing list, and they know my personal accounts also...
<civodul>sounds good
<g_bor>I will have a look at my outreachy account in a few minutes to see if I have any notifications.
<civodul>do we get email notifications as well?
<g_bor>I have received email notifications so far, so I guess yes.
<civodul>ok
<civodul>you said that conversation was on Twitter, right?
<g_bor>Yes, that is right.
<civodul>mbakke: hi! python-hacking fails to run its tests: https://berlin.guixsd.org/build/3845769/log/raw
<civodul>g_bor: bah, too bad :-/
<g_bor>I've also recieved a reminder from the Outreachy Organizers.
<civodul>ok
<civodul>i'm a bit swamped so i hope i'm not missing anything important
<civodul>please do ping rekado & myself if you see we're missing something :-)
<g_bor>My problem with this was, that sometimes it is hard to express your opinion in such a short frame...
<g_bor>Ok, I will.
<civodul>you have 140 chars and as many seconds to express your opinion :-)
<efraim>ready...GO!
<g_bor>It was nice that they provided the questions in advance, so one could prepare.
<rekado>g_bor: thanks for posting answers to the twitter chat!
<g_bor>rekado: you are welcome.
<rekado>I couldn’t do it, because I don’t have a personal twitter account.
<g_bor>I see.
<g_bor>I have one as an accident, I once met a project with no other means of communication at all :)
<g_bor>It was sleeping for years :)
<ng0>actually 280 chars and you can now chain tweets.. but still not enough
<civodul>efraim: could you take care of https://bugs.gnu.org/29483 ?
<civodul>it would be sad to let it bitrot
<efraim>yeah, i'll look at it again
<efraim> https://man.openbsd.org/ntpd.conf I love how well they document it
<groffer>How can I specify keepassxc as dependency so that I have access to core/Database.h at build time? keepassxc has only the "out" output which yields bin/ lib/ and share/ directories only.
<civodul>hello groffer
<civodul>groffer: as a dependency of a package you're defining?
<civodul>just add it to the 'inputs' field
<snape>civodul: core/Database.h isn't installed in the store, if I understanc clearly
<civodul>oh i see
<civodul>i guess the keepassxc package would need to be modified then?
<snape>yes :)
<groffer>ok, I presume it should then define an "include" output?
<groffer>so I'll look at bash and blast+ for inspiration?
<rekado>groffer: it’s probably not a good idea to install internal headers. If the headers describe a public interface then they can be installed.
<rekado>otherwise it might be better to add the *sources* of keepassxc as an input.
<groffer_>rekado, understood. (How) can I refer to the source of a package without stupidly copying the (source ...) from it?
<rekado>groffer_: you can find some examples by searching for package-source
<groffer_>rekado, thanks
<Apteryx>I installed libreoffice, but it crashes :/
<Apteryx>oh: (soffice:11591): GLib-GIO-ERROR **: Settings schema 'org.gtk.Settings.FileChooser' is not installed
<Apteryx>I'm not using Gnome.
<groffer_>Apteryx, you could try "guix environment --ad-hoc libreoffice -- lowriter" (for Writer) or similar
<Apteryx>maybe something about an icon theme: Gtk:ERROR:gtkiconhelper.c:493:ensure_surface_for_gicon: assertion failed: (destination) (when ran in --pure env with --backtrace)
<civodul>mine works fine, installed from yesterday's master
<civodul>could it be that you're missing XDG_* env vars?
<civodul>is it on GuixSD?
<Apteryx>I had icon problems before, I had toyed with a theme configurator (lxappearance) and it might have left some things behind in my home it seems.
<Apteryx>yes it is.
<civodul>hmm!
<civodul>i have adwaita-icon-theme in my profile, dunno if that makes a difference
<Apteryx>no I have XDG envs. I think somethings related to icons not found because I once installed some icon themes.
<civodul>(i'm not using GNOME either)
<Apteryx>I still have arc-icon-theme installed
<groffer_>maybe check if you have GSETTINGS_SCHEMA_DIR set to a non-existing path?
<Apteryx>and it seems I 'require' it now IIRC (if I uninstall I lack icons everywhere)
<Apteryx>yep if I uninstall my arc-icon-theme package I now get warnings like: Gtk-WARNING **: Could not find the icon 'list-remove-symbolic-ltr'. The 'hicolor' theme
<Apteryx>was not found either, perhaps you need to install it.
<Apteryx>But the fatal one is still: GLib-GIO-ERROR **: Settings schema 'org.gtk.Settings.FileChooser' is not installed. Hmm.
<Apteryx>I'm on a somewhat old (3 weeks) guix: 127dc7c7be0559b8683dc90e65b277059ac986bb
<Apteryx>here's my repro: http://paste.debian.net/1012221/
<roptat>civodul: the disclaimer for translations was already removed
<Apteryx>i have to run, but will be connected as apteryx_ ;)
<rekado>apteryx_: there are two packages that explicitly set GSETTINGS_SCHEMA_DIR for GtkFileChooserDialog.
<civodul>roptat: was it? i didn't get a reply to my message, so i thought it got lost
<civodul>well that's good news, then
<civodul>roptat: the French translation is at 100%, i think we ought to thank you :-)
<roptat>I think I spotted a mistake yesterday, I need to review it, but yw
<roptat>actually I only took the work since 0.14
<roptat>also, it's at 11% for package synopsis/description
<nkaretnikov>ACTION waves to civodul
<civodul>hey nkaretnikov!
<civodul>the real nkaretnikov?
<civodul>long time no see :-)
<civodul>how's everything?
<rekado>oh, a name from before my time!
<civodul>yup
<civodul>one had to be really really brave at the time to hack on this thing :-)
<mbakke>civodul: python2-hacking has been broken for a long time: https://hydra.gnu.org/job/gnu/master/python2-hacking-0.13.0.x86_64-linux
<civodul>mbakke: oh indeed
<civodul>it seems that "python setup.py test" doesn't do the right thing in this case
<civodul>perhaps we just need to replace the 'check' phase?
<mbakke>civodul: I'll give it a go :)
<civodul>got you :-)
<mbakke> https://www.xkcd.com/356/
<nkaretnikov>civodul: things are good. Just wanted to say hi :) hope you’re doing well too.
<nkaretnikov>civodul: does Mark still contribute?
<civodul>nkaretnikov: yes, Mark is still around, although not on IRC
<civodul>i'm fine too, good to hear from you :-)
<civodul>you're still doing functional programming these days?
<g_bor>Hello guix!
<g_bor>It seems that I have a fix upstream for the java-jeromq test bug.
<roptat>cool
<civodul>heya g_bor, nice
<g_bor>I would like to ask, what should be done here? Get that commit, or the latest release, and the fix as a patch?
<roptat>the second option is closer to what we generally do, right?
<g_bor>Yes, it is.
<civodul>+1
<g_bor>Ok, I will do that then.
<g_bor>It is a bug in the test in fact, so it does not affect the actual binary in any way.
<atw>sneek: later tell avg I opened a few Dired buffers to remote servers via TRAMP. Emacs seems responsive but autocompletions in ERC and CIDER hang with a message like "Pinging example.com (Commercial)..."
<sneek>Will do.
<mbakke>Phew, I tried updating python-subunit to address the python2-hacking test failure, which sent me down a deep rabbit hole that requires a new bootstrap path for python-pbr :/
<mbakke>I could use cbaines ' circular dependency fix.
<nckx>mbakke: ...yeah. I too've been down that hole before.
<nckx>(Not with the fix, tho'.)
<nkaretnikov>civodul: yeah, after taking a small break, I’m back to writing Haskell and (hopefully) Agda. Though, I have some issues with implementations of functional programming languages. I’d like them to do less useless stuff, be more efficient.
<nkaretnikov>Reading a bit on garbage collection, for instance, with the hope to improve ghc, but it’s going slow because I don’t have much free time :)
<atw>nkaretnikov: I recently packaged Agda. I hope to have the emacs mode packaged soon :)
<nkaretnikov>Good job! (I use vim, though :p)
<efraim>Yay vim
<mbakke>Did someone say evil-mode?
<mbakke>I find the name "python build reasonableness" somewhat ironic.
<civodul>nkaretnikov: nice, quite an endeavor though!
<bavier`>without thinking too much about it, I typed 'guix package -i shellcheck' on an aarch64 system, which isn't in the package's supported system, but guix is happily downloading all the source for ghc. is it going to eventually fail somewhere?
***pkill9_ is now known as pkill9
<g_bor>:( it seems that the jeromq fix is not complete. It breaks other tests...
<rekado>g_bor: could we just disable the failing test?
<hooverville>noob question: is it possible to install a particular build of a package (for example, a current build of a package fails and I'd like an older build that passes)
<rekado>hooverville: you need an older version of guix for that.
<bavier`>or the derivation files?
<hooverville> https://hydra.gnu.org/job/gnu/master/racket-6.11.x86_64-linux/all the latest failed, is there way to specify build 2403143?
<hooverville>rekado: ah
<civodul>ghc-uuid build failure on master due to wrong QuickCheck version: https://berlin.guixsd.org/build/3907242/log/raw
<civodul>ISTR you wrote that it usually works to just change the version requirements rekado, right?
<mange>Is there an easy way to reconfigure a GuixSD system, but using custom sources for a package throughout? I had expected --with-source to do it, but apparently system reconfigure doesn't recognise it.
<mange>I can modify my guix checkout to have the necessary change, but I was hoping to get away without needing to do that.
<civodul>mange: there's no easy way i'm afraid
<civodul>what's your use case?
<civodul>we might be able to come up with a less generic solution that would do the job
<mange>I'm writing patches for shepherd, so I want to make sure a system still boots and behaves properly with my modified version.
<civodul>i see
<civodul>for that i think you basically have to modify shepherd-root-service-type currently
<g_bor>rekado: it seems that these tests indicate actual problems. Looking at the diff in details revealed, that in fact there is a modification in handshake. But the current fix seems to be breaking the specification, so now various tests fail...
<mange>Well, I could just modify the package definition (which should then propagate to the service), couldn't I?
<mange>I was hoping that --with-source would do that transparently for me.
<civodul>right, but that'd lead to more rebuilds
<mange>What depends on shepherd?
<civodul>network-manager, things like that
<civodul>maybe it doesn't matter for testing
<civodul>i have to go, but i'm happy to resume the conversation later on :-)
<rekado>civodul: yes, in some cases it’s valid to tell the build system that newer versions are fine.
<jsierles>rekado: Hey - are you using nfs-utils from guix to run your NFS server?
<mange>On my GuixSD system there are two services in shepherd that are always stopped. "term-auto" and "user-homes". Does anyone know why they cannot be started?
<mange>I checked the code, on start they do things, but then return #f so shepherd thinks that they failed to start. I just don't understand why they do that.
<verisimilitude>Have any of you had luck running the Tor Browser under GuixSD; I've been playing around with it and handled several errors, but now I'm at something of a loss.
<verisimilitude>If anyone's interested, I can describe the issues I've handled, so far.
<g_bor>Actually it seems that the jeromq testsuite is quite fragile. I wonder if this is because something is really broken, or is it created in a way, that it works most of the time....
<g_bor>I'm also testing this one with guix build --rounds=30, and it seems that it does not manage to run to completion.
<apteryx_>Strange. Using make from Guix on a foreign system lead to Python 3 crashes (guix's make calling system installed Python program which somehow was sourcing Python 3 stuff under /gnu/store and failing on something). I unset PYTHONPATH but it kept happenning, I had to uninstall make from Guix.
<apteryx_>I find using Guix on a foreigh distro a risky business. I don't think we should remove GuixSD from the spotlight on the website. ;)
<efraim>i was looking through some image creation scripts for aarch64 boards, seems many of them use extlinux over grub-efi
<mbakke>Oh, we have a real mess of sphinx packages.
<efraim>if you're feeling particularly adventurous, there's always openstack.scm
<mbakke>efraim: That's what I'm working on! :D
<efraim>I open that from time to time since I added to that forever ago and I almost wanted to remove everything and start it fresh
<mbakke>I've solved the circular dependencies in check.scm for testtools and friends, which helped a little.
<axg>So I install font-fira-mono package and for some reason, fc-list does not list it and emacs isn't able to find it. I ran fc-cache which is supposed to update the font cache and the fc-list still has no fira fonts. Previous I installed the font-hack package and that has installed properly and is usable. Is there something wrong with those packages?
<sneek>Welcome back axg, you have 1 message.
<sneek>axg, atw says: let me test tomorrow at work, where I can TRAMP into a few servers.
<efraim>when I last installed fonts I ran `fc-cache -fv`
<axg>efraim: that did it! I should have read the man page a big closer, thanks
<verisimilitude>So, unfortunately, I've had no luck with my wifi card compatible with Free Software; it occasionally works and occasionally ceases to be recognized.
<verisimilitude>Is there a recommended method for installing drivers without using guix?
<verisimilitude>I also want to configure one of my packages and it seems the way to do this is directly modifying the configuration in the private /etc/, but is there a preferred method for this, as well? I lightly scanned the documentation, but didn't find anything about this in particular.
<atw>sneek: later tell axg I opened a few Dired buffers to remote servers via TRAMP. Emacs seems pretty responsive but autocompletions in ERC and CIDER sometimes hang with a message like "Pinging example.com (Commercial)...". Completion seems slower in general.
<sneek>Will do.
<efraim>guix build: error: build failed: cannot close compressed log file (BZip2 error = -6)
<efraim>looks like its trying to do '-6' compression instead of actual 6
<efraim>or however the flag is passed
<efraim>.. or thats the error message
<snape>verisimilitude: what package is it? It doesn't offer an home configuration?
<verisimilitude>It's tor, snape; I've only ever configured it by editing the torrc in /etc.
<verisimilitude>So, the general solution is to use a local configuration, then?
<snape>verisimilitude: there is tor -f <path/to/file>
<bavier`>verisimilitude: the tor-service procedure can be passed a config-file
<bavier`>which is a "file-like object"
<verisimilitude>Alright then; it's appreciated, bavier`.
<snape>Hu. didn't know you were using the service
<verisimilitude>I'm not using it here.
<verisimilitude>Freenode doesn't allow that.
<snape>I meant the Guix service :) as opposed to using only the package
<verisimilitude>Oh.
<verisimilitude>Returning to the Tor Browser, I seem to have ironed out all of the sh issues, yet it still refuses to run.
<snape>by the way, there is no file-like? procedure is it?
<snape>is *there
<verisimilitude>Without consulting the documentation, I'd assume a string containing the filepath would suffice.
<verisimilitude>That's how it is for the rest of the system.
<verisimilitude>The Firefox bundled with the Tor Browser is a shared library and I simply see all manner of issues, including it apparently not existing. I've taken a look at some issues Debian had with packaging it, but I'm not certain they necessarily applied to this situation.
<snape>yes, but I was talking about file-like objects, which are a specific Guix thing :)
<snape>and they aren't plain strings
<snape>Anyway, good luck with Tor!
<verisimilitude>Alright; well, I wouldn't know, then, but a cursory glance at the info manual doesn't reveal a file-like? procedure.
<snape>neither does the code!
<snape>I'm going to create it then
<verisimilitude>It's mostly the Tor Browser giving me issues, currently; I can use IceCat in the interim, but the Tor Browser has many convenient features.
<verisimilitude>Good luck, snape.
<snape>Thanks! I hope you have a powerful machine to build browsers...
<verisimilitude>The Tor Browser is distributed in compiled form, which is what I'm trying to use.
<bavier`>verisimilitude: there has been some work off and on in packaging the Tor Browser; you may be able to find some of the discussion on the mailing lists
<verisimilitude>Alright; I'm supposing it's been unsuccessful, so far, then.
<bavier`>verisimilitude: it's complicated, yes :/
<bavier`>but packaging the bundled version may be easier
<verisimilitude>Well darn.
<verisimilitude>Oh, you meant with building.
<bavier`>right
<bavier`>we'd want to unbundle as much as possible before making it an "official" package
<verisimilitude>That's understandable.
<verisimilitude>It's unfortunate that I'm asking so many questions, but I suppose I'll contribute by packaging some programs I've written soon.
<bavier`>verisimilitude: np, question are certainly welcome; and we always appreciate patches ;)
<verisimilitude>I'd primarily be interested in packaging some Common Lisp libraries, so it wouldn't be particularly in-depth, especially considering the other Common Lisp packages that could be looked at.
<verisimilitude>I'm supposing I could simply announce it here, when that's actually ready.
<snape>does it make sense to create gnu/services/utils.scm, or should I reuse guix/utils.scm?
<verisimilitude>I'd reuse, if I were you.
<verisimilitude>I'm assuming this is related to aesthetics.
<verisimilitude>Now, say, that does bring up another important question; is it permissible to allow anonymous contributions to Guix?
<snape>Yes it is
<snape>some contributors are anonymous
<verisimilitude>Well, alright; I've had issues getting my Common Lisp programs into another repository, specifically because of this.
<verisimilitude>Interestingly, I was the first person who'd tried to do this.
<verisimilitude>I'd expect this policy from a GNU project, though, which is nice.
<snape>You'll find friends within the Guix community who do the same :)
<civodul>snape: if it's service-specific, (gnu services utils) may make sense
<civodul>otherwise less so
<civodul>i find that (guix utils) and (guix build utils) have become way too broad, as one could expect
<civodul>so it's even better if you can find a more specific name :-)
<snape>civodul: thank you; I'll try to find something less broad, and worst case will be (gnu services utils)
<mange>civodul: To follow up from earlier, I got shepherd into my system by writing a new package definition and grafting it in. Seems to be working fine so far.
<snape>(gnu services serialization) sounds good, but a bit long
<civodul>mange: cool
<snape>or (gnu services serial)
<civodul>mange: are you the one discussing these things on the ML as well? :-)
<civodul>snape: doesn't it exist already?
<civodul>ah no, it's (gnu services configuration)
<mange>Yep!
<civodul>ooh, nice :-)
<civodul>good to map people to nicknames :-)
<snape>there is (gnu services configuration), but it's specific to the define-configuration stuff
<mange>Yeah, I sometimes use czan, which is much easier to associate.
<civodul>heheh
<mange>But on IRC I'm pretty much always mange.
<mbakke>Most of the Openstack test suites fail about 50% of the time after the testtools upgrades.
<mbakke>On a 16-core machine. They are much more reliable on my laptop.
<mbakke>Looks like they've forked testrepository into 'os-testr', maybe that works better.
<civodul>perhaps Steap can provide guidance for these things
<mbakke>Looks like os-testr is already obsoleted by another testrepository fork, 'stestr' :P
<mbakke>I'll try it out.
<mbakke>Oh, it's not really a drop-in replacement. Let's see how Hydra fares first.
<civodul>do they have an 8-char limit on project names?
<mbakke>Hehe :-)
<bavier`>gotta support those legacy mainframes
<verisimilitude>So, is there a preferred way to install firmware outside of Guix?
<verisimilitude>I'm thinking I should be able to take the firmware to use from Debian.
<bavier`>verisimilitude: which firmware? is it free?
<verisimilitude>My AR9271 dongle doesn't work, so I was simply going to get the intel wireless firmware.
<verisimilitude>Supposedly, writing "options ath9k_htc nohwcrypt=1" to /etc/modprobe.d/ath9k.conf is a fix, but GuixSD doesn't have this.
<verisimilitude>From what I recall, this didn't work anyway on Debian, years ago.