IRC channel logs

2018-10-26.log

back to list of logs

<janneke>mirror.hydra.gnu.org timed out, bzip2 got built from source (why?!, are 0.15.0 substitutes no longer available maybe?, did we have to do a guix pull before guix system init?) but the source tarball is no longer served...wow
<LucidDreamZzZzZ>yeah just do a pull
<LucidDreamZzZzZ>use --fallback
<LucidDreamZzZzZ>cuz i dunno
***LucidDreamZzZzZ is now known as luciddreamzzz
<civodul>janneke: problem is that hydra.gnu.org is off-line as i wrote on guix-devel
<civodul>so mirror. still has things in cache, but not everything, and it gets worse :-/
<civodul>i strongly recommend using berlin.guixsd.org
<luciddreamzzz>ooh
<luciddreamzzz>hey anyone know how i get 'bui'
<luciddreamzzz> https://gitlab.com/alezost-emacs/bui
<luciddreamzzz>package search no find
<nckx->luciddreamzzz: 'emacs-bui'?
<nckx->It was added to Guix in 2016 so that can't be the problem.
<luciddreamzzz>nckx-, yes i did not know i needed emacs-
<luciddreamzzz>it works :)
<nckx->👍
<luciddreamzzz>OK and it is done installing woo hoo
***nckx- is now known as nckx
<luciddreamzzz>thank you!
<luciddreamzzz>trying to get Emacs-Guix working
<nckx>It Just Works here, and I'm not much for debugging Emacs.
<luciddreamzzz>nckx, https://melpa.org/#/guix
<luciddreamzzz>its OK I am getting it setup
<nckx>luciddreamzzz: ...I'm not sure what to make of that link?
<nckx>I just 'guix package -i emacs-guix'd it.
<nckx>OK. Good!
<luciddreamzzz>it provides featureful visual interface for
<luciddreamzzz>the GNU Guix package manager
<luciddreamzzz>not sure that will help me
<nckx>I know what it is, I've just never had any problems installing it. That's what I meant by Just Works. :-)
<luciddreamzzz>ahhh
<luciddreamzzz>ok maybe i switch mirror as well
<luciddreamzzz>haha that was Ludovic
<nckx>?
<luciddreamzzz>well I can't be sure :)
<luciddreamzzz><civodul> i strongly recommend using berlin.guixsd.org
<nckx>civodul is Ludovic, yes. :-)
***joehillen_ is now known as joehillen
<Laalf>is there a place where i can look which substitutes are available, maybe even with logs? i cant build python 3.6.5 at the moment because it gets killed and i dont know why. tested it on 2 machinees
<luciddreamzzz>Laalf, http://lists.gnu.org/archive/html/guix-devel/2018-10/msg00434.html
<Laalf>luciddreamzzz: alright. thanks. still the question if i am the only one unable to build python 3.6.5 right now
<luciddreamzzz>np
<nand1>do you think it could be because it runs out of memory
<nand1>the tests for python 3 take up a lot of memory; I think run in parallel
<nand1>on my machine, building python 3 locks the computer because the memory usage goes to 100% during the test phase
<Laalf>i ran it with 16gb of ram and -c 1
<nand1>the thing is, I looked into it a bit
<nand1>this could be a red herring but
<nand1>the tests are not affected by -c (I might be wrong)
<Laalf>sux.
<nand1>one second, let me find a source for you
<Laalf>I can't even put more ram into this notebook because proprietary BIOSes and I don't have a flasher and the nerves for core boot. I do have more ram though.
<snape>it's weird because mirror.hydra.gnu.org doesn't seem to work either
<Laalf>berlin.guixsd.org doesnt have python build as well
<Laalf>when i let python compile on my desktop which has 32gb ram but 8 cores... will i see the same issue?
<nand1>most likely
<nand1>if my guess that it runs stuff in parallel and needs an obscene amount of memory is correct (might be wrong)
<nand1>then compiling it on a computer with one core will work
<nand1>of course the root of the problem is just to modify the build script to pass the magic words to say how many threads to use for the testing phase
<snape> https://berlin.guixsd.org doesn't seem to work at all
<snape>rekado_: ^
<Laalf>so turn off hyperthreading and multi core processing.... not now
<nand1>I hope it's not a red herring though
<nand1>at the very least it's worth a shot trying to modify the makefile
<nand1>that would involve investigating how python 3 is built
<nand1>I found the thread: https://stackoverflow.com/questions/49793880/compiling-python-from-source-multiple-threads-for-tests
<nand1>in any case, try to compile it and monitor memory usage during the test phase to see if you can replicate what I saw
<nand1>maybe compiling python 3 just requires lots of ram
<snape>nand1, Laalf: if you change the python3 receipe, everything that depends on python3 will be built again...
<nand1>trying to limit it with -c 1 still used all my threads during the test phase, where memory usage went to 100%
<nand1>to be fair, I only have 8 GB of memory and tried 12 threads
<nand1>good point
<Laalf>i tried it with -c 1, it uses all 8 threads and more than 16gb ram
<nand1>I want to try to build it on my dual core laptop to see if it will build
<Laalf>i can confirm that -M 1 doesnt help either. i will turn off all but one core and turn off hyperthreading and i will try again.
<roptat>Laalf: snape: I have a substitute for python3: https://berlin.guixsd.org/nar/gzip/h4xlrhvzg5v19s0kg8qmbd99310b21vi-python-3.6.5
<mbakke>There is a bug in Python 3 which causes the test suite to run out of memory on recent kernels: https://bugs.python.org/issue34587
<mbakke>Unfortunately the only workaround I can think of is removing "/tmp/guix-build-python-3.6.5.drv-0/Lib/test/test_socket.py" some time during the build (but before the check phase).
***astronavt_ is now known as astronavt
<pkill9>why doesn't the curl package produce libcurl-gnutls.so.4 even though it is compiled with the '--with-gnutls' flag?
<Laalf>is there anyone who has acpi_call and/or tp_smapi packaged?
<civodul>hi there!
<Laalf>hi!
***astronavt__ is now known as astronavt
<janneke>civodul: o/
<jlicht>hey guix!
<ng0> https://lists.x.org/archives/xorg-announce/2018-October/002927.html
<nckx>Hmm.
<Laalf>isnt that bug reported for years?
<pkill9>who had that package definition generator written in scheme?
<civodul>pkill9: are you thinking of 'guix import'?
<apteryx>pkill9: maybe look at the various importers in guix?
<civodul>bingo :-)
<apteryx>eh
<pkill9>ok
<daviid>bavier: hello! any news on the guile-cv guix package battle :)?
<bavier>daviid: howdy
<bavier>nope, got hung up on some other things
<bavier>and haven't had much time for packaging work lately
<daviid>ok. I know it's actually a tex splitting packages battle more then a guile-cv package battle ...
<bavier>in a way, yeah
<daviid>it's unfortunatel gule-cv itself is probably 'dead easy' to gux package ... but ok, let's see how it goes
<Laalf>LC_ALL=C guix package -p ~/.config/guix/current --roll-back ; switched from generation 0 to 0.... ayyy have no idea what i have done. when hydra is back up i should reinstall
<civodul>yay berlin is busy again! https://berlin.guixsd.org/jobset/guix-master
<civodul>Laalf: "generation 0" is the empty generation, so that doesn't sound good :-)
<civodul>can you do --switch-generations=1 for instance?
<Laalf>civodul: guix package: error: cannot switch to generation '1'. or +1/-1.
<civodul>Laalf: can you use --list-generations and then pick one that's actually available?
<Laalf>civodul: guix package -p ~/.config/guix/current -l gives nothing.
<civodul>so you deleted all the generations i guess?
<civodul>but you do have a 'guix' command around apparently, so there's still hope :-)
<Laalf>civodul: i cant even guix pull -l because of some error. but i will not copy those lines
<civodul>prolly you can just run 'guix pull' and everything will be fine
<Laalf>guix pull will compile python
<civodul>it could be that 'guix pull -l' doesn't properly handle zero generations
<Laalf>and compiling python doesnt work with my mease 16gb ram
<civodul>Laalf: did you see https://lists.gnu.org/archive/html/guix-devel/2018-10/msg00434.html ?
<civodul>that may explain some of the problems you encounter
<Laalf>that is why i plan to reinstall when hydra is back uo
<Laalf>i know guix creates a manifest somewhere. so nothing will be lost
<civodul>you shouldn't need to reinstall
<Laalf>i should just run a sudo guix gc --verify=contents,repair && sudo guix gc --delete --optimize --clear-failure && guix pull && sudo -E guix system reconfigure config.scm when hydra is back.
<civodul>if the problem is that python fails to build, running 'guix gc' won't solve it
<Laalf>that is why i wait for hydra to get back up. i can build it but i cant check it.
<pkill9>does the code that queries for substitutes not handle timeouts?
<civodul>pkill9: not sure, why?
<pkill9>sometimes it hangs when updating substitutes