IRC channel logs

2017-11-24.log

back to list of logs

<mb[m]1>And `git am --scissors`.
<rekado>yeah, scissors is cute
<mb[m]1>I was checking if `git format-patch` could automatically add a link to the upstream gitweb interface. It can call `git-notes`, but those can apparently only store static text, i.e. not refer to the current object.
***pksadiq_ is now known as pksadiq
***pksadiq_ is now known as pksadiq
<brendyn>does changing an input to a propagated-input make any difference when the package is being built
<efraim>not while it's being built
<brendyn>Anyone familiar with Perl? I'm packaging File::Mimeinfo but it's failing to find it's own sub module Filee::Mimeinfo::Magic.
<brendyn>It searches in the perl package for it instead of it's own package, for some reason
<DusXMT>rekado: Yes, I am indeed running guix from a git checkout
<DusXMT>ACTION should probably just get a virtual machine for running guixsd
<DusXMT>(instead of trying to run it on a different distribution)
<brendyn>I think it's good to have people testing on on different distributions
<DusXMT>Okay, an `autoreconf -vfi && ./configure --prefix= && make' later, and it works :)
<DusXMT>ACTION thought make by itself was enough, as it seemed to update the automake files
<brendyn>It works until it doesn't!
<rekado>brendyn: maybe try adding the current directory to PERL5LIB
<brendyn>rekado: I just tried that, but it turns out moving a few inputs to propagated-inputs fixed it. propagated-inputs is still something that I find confusing
<efraim>sneek: later tell lfam: utox got check as an input, should it be a native-input?
<sneek>Got it.
<efraim>sneek: botsnack
<sneek>:)
<rekado>propagation only affects dependent packages
<rekado>Perl, Ruby, Python, R, Guile, etc – they all have a search path where they look for packages by name.
<rekado>they don’t retain full references to their dependencies
<brendyn>i moved an input to propagated inputs, and the result was the package contained this extra .packlist file
<rekado>so we use propagation to make sure that their dependencies are installed alongside the package itself
<efraim>all the qt 5.9.3 modules almost done building!
<ng0>hi efraim! I have some people working on this hardware: https://www.olimex.com/Products/OLinuXino/A20/A20-OLinuXIno-LIME2/open-source-hardware and I don't have any details of their work or experience. Do you know if this hardware is supported at all by our current attemot to get GuixSD on arm?
<ng0>*attempt - not attemot
<efraim>ng0: bases on the status matrix of the A20 here https://linux-sunxi.org/Linux_mainlining_effort and https://www.armbian.com/olimex-lime-2/ , it should work great
<efraim>Of course other than the Mali graphics and the 1GB of RAM
<ng0>cool
<ng0>thanks!
<CharlieBrown>Where is the GuixSD source code?
<amz3>savannah
<amz3>git://git.savannah.gnu.org/guix.git
<amz3>CharlieBrown: ^
<rekado>you can browse it here http://git.savannah.gnu.org/cgit/guix.git
<CharlieBrown>amz3: I guess what Im asking is... Is the Guix src different from the GuixSD source?
<rekado>no
<rekado>it’s the same
<rekado>parts of Guix are only used on GuixSD, but it’s all in the same repository
<amz3>monorepo for the win :3
<CharlieBrown>rekado: cool
<brendyn>ok now for some reason mimeopen works fine when i call it in the terminal, but when xdg-open calls it, it fails to find it's submodule (only searches the perl package, not it's own directories)
***pksadiq_ is now known as pksadiq
<civodul>Hello Guix!
<iyzsong>Hello :-)
<roptat_>rekado: did you stop reviewing my patches because you're waiting for something from me?
<civodul>hey iyzsong, long time no see :-)
<iyzsong>yeah, i'm just hide in the corner...
<CharlieBrown>Bonjour, Guix.
<civodul>bonjour !
<CharlieBrown>Everytime I hear the word "bonjour", I think of a certain Futurama clip. https://www.youtube.com/watch?v=gO_1dEOkb_U
<apteryx>The progress is reporting 0.0% for every downloaded substitutes. Is there still a bug there?
<rekado>roptat_: are yours the java patches?
<rekado>I just didn’t get around to continuing the review.
<rekado>ACTION is currently sitting in a noisy, hot, stinky data centre to fix a couple of broken servers
<rekado>apteryx: what substitute server are you using?
<civodul>rekado: and it's so comfortable that you had the idea to chat a bit while sitting there? :-)
<rekado>exactly
<rekado>:)
<rekado>I’m waiting for a progress bar to creep to the right.
<rekado>I replaced two of the build hosts for berlin. One of them just died and the other signalled CPU errors.
<civodul>oh
<civodul>how did you notice?
<civodul>we still lack a monitoring tool on these
<rekado>a blinking LED
<civodul>i see :-)
<rekado>the dead one had no blinking LEDs, of course. Just dark.
<civodul>until we have a Nagios service or something, you could leave a webcam in front of the rack :-)
<rekado>hah
<rekado>it will take a little longer to set up the new head node
<rekado>I’d like to add a second server for fail-over.
<cehteh> https://github.com/firehol/netdata .. btw is that packaged?
<rekado>I’m rebuilding the systems for all build hosts from berlin using install-berlin.scm
<civodul>what's install-berlin.scm?
<rekado>oh, it’s a Guile script on berlin.
<civodul>ok
<rekado>it generates os configs for numbered nodes, builds them, and then copies them to the target.
<civodul>awesome
<rekado>remote reconfiguration doesn’t work yet, so I need to do that manually on the remote.
<civodul>right, we really need 'guix system reconfigure --remote'
<rekado>I’m reluctant to enable root login over ssh
<civodul>it's weird, i feel like i have to work on the more urgent things like 'guix pull'
<civodul>ok
<rekado>I prefer the ansible way of specifying a sudo user, and then prompt for a password on the remote.
<rekado>re guix pull: do you have something I could test?
<civodul>not yet
<civodul>it's terrible
<civodul>it's a real time sink
<rekado>I really appreciate your work on this!
<rekado>I find it hard to wrap my head around the problems surrounding guix pull
<civodul>it boils down to making sure the running Guix can "jump" into the new Guix
<DusXMT>An interesting thing I found when a derp-up caused substitutes to no longer work for me: tar's test suite fails on btrfs :)
<civodul>for instance we want to use (gnu packages ...) from the new checkout
<civodul>DusXMT: many things fail on Btrfs...
<stefanc_diff>DusXMT, if you can move away from btrfs because it still has rough edges.
<DusXMT>stefanc_diff: I like checksums and automatic data redundancy :)
<stefanc_diff>a power cut(laptop running out of power) for example could result in an inconsistent filesystem
<DusXMT>I keep backups
<roptat_>rekado: yes, the java patches. ok I'll wait the rest of your review :)
<rekado>roptat_: sorry about the delay. Maybe tomorrow I can resume.
<DusXMT>(also, most of the time I'm on a laptop, which's battery sort of serves as a UPS)
<stefanc_diff>play with btrfs as you wish DusXMT , I was pointing out that a trivial thing could cause major headaches.
<DusXMT>I'm aware of the risks, but I'm also aware of the benefits :)
<stefanc_diff>I've just fixed someone else's laptop running btrfs last weekend , inconsistent btrfs journal
<rekado>mb[m]1: your glibc patches mail bounced because it is too large.
<DusXMT>Speaking of backups, I just updated my backup :) (so even if my computer completely fails and falls apart, at most I'll lose a week of work (annoying, but not disastrous), and the time putting the thing back together)
<brendyn>I'm still stuck on this issue with a perl program not being able to find it's submodules. I saw on the mailing list someone made a perl wrapper but it's not in master. What ever happened to that?
<mb[m]1>rekado: ugh, thanks for the heads-up. Strange that debbugs didn't let me know.
<rekado>brendyn: what perl wrapper do you mean?
<mb[m]1>The patch is 30k lines, so probably 1-200k in size. Can try compressing it later.
<rekado>brendyn: what’s on PERL5LIB? Are the directories for the dependencies all listed there?
<brendyn>rekado: https://lists.gnu.org/archive/html/guix-devel/2016-11/msg00609.html
<rekado>wrappers are used for applications, not for modules or libraries.
<brendyn>nothing is in PERL5LIB. if is set it manually it will work but shouldn't it work automatically?
<rekado>we already have a wrap phase.
<rekado>for an application you need to add a phase that wraps the executables in PERL5LIb
<rekado>we do this in a couple of packages. Just search for “wrap-prog”
<brendyn>this module comes with an cli "application" with it
<brendyn>If I understand what that means
<brendyn>ok so since im adding a perl dep to xdg-utils, i need to add a wrapper script to xdg-utils
<rekado>right
<rekado>bleh, glusterfs doesn’t work :(
<rekado>the mount tool changes the PATH so “cat” and “sed” are not found. Causes the mount script to fail.
<civodul>rekado: 'mount' from util-linux, or mount.glusterfs or whatever it's called?
<civodul>mb[m]1: thinking about it, it's a case where it's tempting to grab a generated tarball
<civodul>esp. since the glibc repo includes the Autoconf byproducts
<civodul>so no need for a bootstrap phase
<brendyn>But I can also just install perl-file-mimeinfo directly and use mimeopen, so that needs either native-search-paths or a wrapper too. if i wrap it, then i might be able to about wrapping xdg-open?
<mb[m]1>civodul: that's a good idea. Much better than a 200kB superpatch.
<mb[m]1>I'll do that later tonight and push. Also start the "core" job.
<brendyn>The annoying thing about wrapping is it changes the name of the programe to .foo-real. when i type `mimeopen' it says Try '.mimeopen-real --help' for more information.
<brendyn>and in i3 window manager config, and things like killall, the program name is used
<brendyn>is there a better solution? maybe foo -> real/foo would be better
<rekado>I already posted a solution to the mailing list
<rekado>for many programmes we could prepend a guile script as a wrapper, which calls itself
<rekado>wouldn’t work for binaries, though
<rekado>civodul: it’s mount.glusterfs
<rekado>but that’s not the only problem.
<rekado>it contains many hard-coded paths.
<rekado>nixpkgs has a few patches for glusterfs, so I’ll apply those first.
<brendyn>i have two binaries on this program to wrap, so i copy-pasted two wrap-progam clauses and it worked, but then i reduced it to a (map (lambda (p) (wrap-program ... and now it errors
<brendyn>ACTION typo
<civodul>rekado: ok, sounds like it should be tractable
<civodul>mb[m]1: there are always the usual risks with gitweb-generated tarballs, though perhaps we can assume they're low
<civodul>and as a last resort we can host a mirror on alpha.gnu.org/gnu/guix/mirror
<rekado>brendyn: without the error message we can’t help you.
<amz3>thanks for accepting guile-wiredtiger
<civodul>amz3: yw!
<civodul>that was some time ago already :-)
<amz3>civodul: I don't see in the package definition how does it deal with the unittests
<civodul>amz3: they are disabled, which isn't great
<civodul>an oversight of the review process i guess
<civodul>what would you suggest
<civodul>?
<amz3>I was looking for https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/databases.scm?id=v0.13.0-4854-g17c3f7764#n1523
<amz3>nevermind, it works ;)
<amz3>I must fix the tests to work, but I need first to do some progress about some other parts
<amz3>nice website!
<amz3>there is even links in the source
<civodul>anyone has suggestions re https://bugs.gnu.org/29255 ?
<apteryx_>Maybe it could include a ref to a node of the manual which would explain how to troubleshoot and fix those problems?
<rsriniva>just downloaded the 0.13 xz file and tried an install on my thinkpad
<rsriniva>when i uncompress the xz file, i see a 1.3 GB file ready to be copied to USB, which I did
<rekado>rsriniva: are you following the steps in the manual?
<rsriniva>however during install, it went out to the intrwebs to fetch and install gazillion packages
<rsriniva>my quetion is: what's in the installer? I assumed it will work fully offline and have the packages for the desktop profile?
<vagrantc>no
<vagrantc>it's just a very minimal system
<vagrantc>you *might* be able to do a very bare-bones install without downloading new stuff, but it's pretty much a network installer ... at least, that's been my experience the past couple weeks
<rsriniva>what's the split on this channel bw those who run GuixSD baremetal vs running just guix on top of debian or something similar?
<rekado>I run GuixSD on ~30 machines; I use it as a package manager on ~100 machines, maybe more.
<rekado>but that’s only because Guix can talk to a remote daemon over the network, so the 100+ machines are really just talking to one daemon.
<rekado>so I guess I use GuixSD more than Guix + foreign distro.
<rsriniva>where can i find a comprehensive desktop config scm file with all the bells and whistles for a typical dev workstation?
<rsriniva>a nice starter template that I can further customize rather than use generic ones
<rekado>rsriniva: I recommend starting with a simple config for the initial installation.
<rekado>it’s easy to reconfigure the system afterward (and roll back in case of errors)
<rekado>but for the initial install I’d play it safe.
<rsriniva>yeah i finished the install
<rekado>oh
<rekado>good
<rekado>there’s an example desktop configuration in the manual
<rekado>it’s also part of the installer image
<kmicu>stefanc_diff: what kernel version did you have on that laptop?
<kmicu>DusXMT: you can safely use btrfs. I do not remember any issues reported on #nixos during last years. No problems on server farms, no problem on personal workstations (unless you are using some outdated kernel).
<DusXMT>kmicu: I'm using 4.9.30
<DusXMT>(the one that comes with debian stretch / devuan ascii)
<kmicu>4.9 is LTS so it is good. Bugfixes are backported.
<CharlieBrown>to get gnome helloworld.js to run, someone says i
<CharlieBrown>need the package that contains */girepository-1.0/Gtk-3.0.typelib
<CharlieBrown>Where is it?