IRC channel logs

2018-04-10.log

back to list of logs

<Cthulhux>"electronic payment" because what could go wrong
<buenouanq>So I woke up and my computer just says GRUB _
<buenouanq>What did I do? Is there any way to get back to what I had?
<buenouanq>this was GuixSD
<buenouanq>booting to a flashdrive I can mount and access the old boot and root drive, so I don't think it's a dead harddrive
<zybell_>did you try <enter>?
<caffe>so emacs works on 64-bit guixsd but not 32-bit guixsd (unless you use emacs-no-x)
<caffe>has anyone managed to get emacs working with X11 support on i686?
<buenouanq>zybell_: yes, it just say `Reboot and Select proper Boot device'
<buenouanq>everything in the BIOS looks correct
<zybell_>I think grub.conf is missing
<buenouanq>how do I unmissing it?
<zybell_>normally it is renamed to something like grub.conf.old before overwrite (which didnt happen for some reason). cat /<tab><tab> on the GRUB command line should provide you with filenames. Running cat on a file should help you find if its right. If its right running config on it should load its menu/boot directly.
<buenouanq>it's not a command prompt
<buenouanq>pressing anything just repeats the error message
<zybell_>even e?
<buenouanq>even e
<buenouanq>it just says `GRUB
<buenouanq>Reboot and Select proper Boot device
<buenouanq>or Insert Boot Media in select Boot device and press a key'
<zybell_>GRUB must be reinstalled. That you can do from any GRUB enabled OS.
<buenouanq>so, boot to GuixSD install flashdrive, then..
<zybell_>y
<buenouanq>can I just install it the normal guix way? or what
<zybell_>What do you mean?
<buenouanq>from the install usb can I just run # guix package -i grub
<zybell_>I think you can even reinstall guix, if you find the used .scm
<buenouanq>I plan on doing that - The issue is I have a postgres database I would like to recover first (;-___-)
<zybell_>but guix package -i grub wont work
<buenouanq>that's sort of what I thought
<zybell_>more likely is guix reconfigure grub.
<zybell_>But i think you should check the config you would write.
<zybell_>And look in /boot for grub.conf before overwriting!
<buenouanq>there is a /boot/grub/grub.cfg which looks like it should be what I want
<zybell_>By default the config points to your currently running system. Do not forget that!!
<zybell_>SAVE IT
<buenouanq>i'm backing up everything now
<buenouanq>but what I really need is for it to just boot ;_;
<zybell_>After reconfigure you can put grub.cfg back.
<buenouanq>you're saying I can run # guix system reconfigure /dev/sda2/etc/config.scm
<buenouanq>from the install usb?
<buenouanq>and it will be just like i ran that from the normal os?
<zybell_>If you have enough bw and time, it could work by recompiling everything.
<buenouanq>something about that doesn't seem right to me
<buenouanq>if I'm booted to the install media, it will try to reconfigure that and not what I actually want...
<zybell_>you only need to reconfigure grub
<zybell_>not the whole system
<buenouanq>oh oh, so I throw it a config.scm that only contains the grub declaration
<buenouanq>then run system reconfigure
<buenouanq>then copy over that grub.cfg file
<zybell_>y
<buenouanq>ok
<buenouanq>I'ma wait for this to finish copying then try.
<zybell_>You can make it faster by mounting /gnu/store
<zybell_>I mean the reconfig
<zybell_>And more accurate too
<buenouanq>where do I mount it?
<buenouanq>and, how exactly...
<buenouanq>if /dev/sda2 isn't mounted, how do mount /dev/sda2/gnu/store ?
<buenouanq>can I run like # mount /dev/sda2 /mnt/ ; mount /mnt/gnu/store /gnu/store
<zybell_>Now I'm slightly confused. How can I explain?
<zybell_>Youre very near
<zybell_>do you have store on the root partition or extra? Do you have /gnu/var?
<buenouanq>everything is on /dev/sda2 I didn't do anything fancy for the install
<buenouanq>no /gnu/var
<buenouanq>just /gnu/store
<zybell_>Then you need to look in /var. There must be a 'database' which goes with the store. Should be mounted too. Could you give me a ls /mnt/var?
<buenouanq>cache/ db/ empty/ guix/ lib/ lock/ log/ run/ tmp/
<buenouanq>question though, rather than installing and reconfiguring from the installation media, could I just replace it"s grub.cfg with my old one?
<buenouanq>so I'd boot to the USB"s grub and select my computer from there
<zybell_>I would suspect var/guix/. If there isn't anything in the current guix needs I would mount it too.
<zybell_>Replacing grub.cfg: If it doesn't work you don't have a repair shop.
<zybell_>I would know how to merge them. That would be an option.
<buenouanq>yeah but I can always just reinstall it and start over in a few minutes
<buenouanq>from my laptop
<zybell_>Then do it.
<buenouanq>s/reinstall/reburn/
<buenouanq>cause that would be easiest and quickest if it works
<zybell_>Then do it.
<buenouanq>this backup is going to take a while yet, so don't hold your breath
<zybell_>nf
<PotentialUser-21>From a no bios boot / efi partition table to a barebones config.scm install.
<PotentialUser-21>I created a partition and mounted, I compile and it succeeds, however I'm brought to the grub with unbound variables, scm-error, and a no luks partition message.
<PotentialUser-21>So I redo everything and add an efi type partition, format fat 32, get a v fat partition. Redo everything, mount the efi partition to /mnt/boot and have the bootloader target sda.
<PotentialUser-21>Still getting the same exact thing on the last compile
<PotentialUser-21>And ideas?
***firewall is now known as Guest5323
<caffe>I guess I'm just out of luck for emacs on guixsd i686, huh?
<caffe>time to retire some hardware, i guess.
<rekado_>caffe: you may need to build Emacs from source on i686.
<rekado_>caffe: I think there’s a problem with either the dumping or some post-build processing (stripping?) that leaves Emacs on i686 in an unusable state.
<caffe>rekado_: build with, or without guix?
<rekado_>with Guix
<rekado_>you could also try modifying the package expression to disable stripping and see if that helps.
<caffe>i'll get the relevant reading out of the way and have a go at it. thank you, rekado_ .
<civodul>Hello Guix!
<Sleep_Walker>how can I properly send patchset to guix-patches? send some intro mail and wait for bug number assignment?
<civodul>Sleep_Walker: exactly
<civodul>a bit annoying, but it does the job
<Sleep_Walker>I sent first and rest will be just added to the bug seqentially
<efraim>sometimes I send a summary instead of 0/X and attach the patches
<civodul>rekado_: i'm still investigating this 'guix pull' issue with 2.0
<civodul>turns out i uncovered a compilation issue with 2.0
<buenouanq>I'm getting `guix system: error: failed to load '/mnt/etc/config.scm': No such file or directory' yet it's right there...
<buenouanq>Any ideas?
<civodul>buenouanq: can you try "guix system --on-error=debug ..."?
<mbakke>I wonder how to best factorize the "set-network-interface-up" logic from dhcp-client-service.
<efraim>I have a partially working enlightenment desktop service
<efraim>setuid stuff doesn't fully work, but I think the destdir is coded into enlightenment
<mbakke>Considering a 'network-interface-up' service, but if I make it extensible, it could introduce circular dependencies (e.g. virtual network interface depending on an already configured one).
<mbakke>Maybe a "start-all-interfaces" boot service, and a companion service for virtual interfaces.
<mbakke>Hmmm. Maybe we could tell udev to simply do "ip link set $if up" unconditionally.
<ajjlmau>how do I install programs from git on guixSD?
<castilma>hi, I think the right way is to make a package definiton for the program, add it to the path in GUIX_PACKAGES and then guix package -i prog --with-source=/path/to/prog.git/
<ajjlmau>ok thanks
<zybell_>How about a post-down script?
<castilma>can someone check if building 7e4f6f131ed from git fails if LANG=de_DE.UTF-8 with sth like GUILEC gnu/packages/tor.go
<castilma>/bin/bash: Zeile 7: 1051 Speicherzugriffsfehler XDG_CACHE_HOME=/nowhere ...
<castilma>I'd translate Speicherzugriffsfehler with Segfault, but dmesg doesn't mention anything related
<zybell_>castilma:Its SIGBUS or something like that, caused by a Segfault although technically linux(kernel) doesnt do segfaults anymore,they are all part of the page fault handler now.
<civodul>castilma: that sounds bad
<rekado_>castilma: check your environment variables. This is on a foreign distro, I assume, so you have to make sure that your environment settings don’t lead to Guix things loading plugins or libraries from the host system.
<civodul>could you "ulimit -c unlimited" and re-run "make"?
<civodul>oh right, that could be the issue
<castilma>well, LANG=en seems to work.
<rekado_>Our draft for the “Reproducible genomics analysis pipelines with GNU Guix” paper has just been submitted to biorxiv.
<civodul>woohoo!
<civodul>URL?
<rekado_>will be available by the end of the day or tomorrow.
<civodul>cool
<rekado_>someone somewhere needs to confirm the submission and then we’ll get a DOI
<rekado_>a grant application for reproducibility work with Guix and Cuirass will be submitted on Thursday
<rekado_>as I browsed the biorxiv pages I found a paper on Bioconda that compares the number of bioinfo packages in Guix with the complete Bioconda collection.
<rekado_>they restricted the packages to whatever is in (gnu packages bioinformatics), which isn’t a fair comparison in my opinion.
<rekado_>oh well, better things to worry about
<jboy>congrats!
<civodul>rekado_: heheh :-)
<civodul>bioconda must be bigger though, no?
<rekado_>civodul: it’s bigger because it retains recipes for previous versions, but it also includes CRAN packages or bioconductor things (which in Guix are in separate modules).
<civodul>oh i see
<civodul>another rebuttal that needs to be written ;-)
<rekado_>heh :)
<rekado_>I asked on the bazel mailing list if there’s a minimal set of Java dependencies that are sufficient to build a minimal version of Bazel; the response is: these 60+ jar already *are* the minimal set.
<buenouanq>lol
<civodul>rekado_: awesome :-)
<civodul>do you have a link to the thread?
<civodul>these people are somewhat involved in reproducible builds, so they should care
<civodul>(or pretend they do)
<rekado_>here it is: https://groups.google.com/forum/#!topic/bazel-discuss/Tc09B9ivKTU
<civodul>very clear reply, heh
<rekado_>yeah, no wiggle room
<rekado_>but I’m not convinced that this is actually true.
<rekado_>does bazel *really* need the Android SDK for all common operations?
<rekado_>As it is I don’t see a way to package this within the next two years.
<rekado_>each one of these directories is a mine that could explode into hundreds of dependencies.
<rekado_>the innocent “java” sub-directory, for example, contains the Java 9 JDK.
<rekado_>There is no icedtea variant for Java 9 yet.
<rekado_>“guava” 24 has bootstrapping problems (because it needs groovy which needs guava)
<civodul>bah
<civodul>but does it just Java 9?
<civodul>if it were only Java dependencies and if we had a recursive Maven importer that works well, perhaps we could get close to that
<rekado_>a maven importer would help, but we’d also need a good maven build system.
<rekado_>doing this with the ant-build-system is hopeless.
<rekado_>the maven build system would need to create a temporary local maven repository in the build environment; this needs a way to map Guix package names to maven identifiers/paths.
<rekado_>it’s not very complicated, but we need to figure out the best way to make this work, e.g. by adding a property to all java packages.
<zybell_>If I had that bazel problem I would do: Get a working bazel somewhere (not necessarily guix) build tensorflow with it while recording the executed programs by whatever means(strace -e exec,wrapper scripts,...)necessary,sort the executed programs by pattern,rewrite patterns to make-rules,write a makefile from it,test it for same order of executed progs,debug kinks.
<zybell_>Then add that makefile to the tensorflow package(as patch if needed)and package.
<rekado_>oh, it may be easier: https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/makefile
<zybell_>If it works.
<bavier`>seems like a red flag if your build tool exhausts the RAM of your build machine
<zybell_>The idea of that workflow was to be able to prove equivalence bazel<->makefile.
<rekado_>zybell_: with this very large set of dependencies I don’t think that observing the build with strace is going to be feasible.
<zybell_>logging
<rekado_>I’d rather use my time for other things than analyzing bazel calls.
<rekado_>ACTION updates gnucash
<zybell_>only execs
<efraim>if you do manage to update gnucash then we can drop some old dependencies
<bavier`>would be nice to drop gtk2
<efraim>that would be nice but I'm not hoping that far
<efraim>we can start with webkit1 and qt4
<rekado_>the new gnucash works with gtk3
<rekado_>the new aqbanking stuff also.
<rekado_>the gtk2 variant for webkitgtk is also no longer needed.
<efraim>looks like gnucash is/was the only user of webkitgtk@2.4 and webkitgtk-gtk2
<rekado_>it also supports Guile 2.2 now
<efraim>i can see why they bumped to 3.0
<jonsger[m]>guix only needs some excotic guile libraries and some bootstrap binaries...
<snape>ACTION will probably update Gajim
<jsoo> how should a build with `cargo-build-system` work? i am trying to build alacritty and it fails looking for ~/.cargo/registry.
<catonano>rekado_: I just read about the submission of your article (reproducible genomycs analisys pipelines). Congrats !!
<Sleep_Walker>has anyone tried curl using SPNEGO together with mit-krb5?
<Sleep_Walker>hm, something is fishy about our curl or mit-krb5 or path configuration
<efraim>does anyone know if 'chattr +C' is saved across reboots?
<ajjlmau>why does my /etc/hosts file reset when I reboot?
<zybell_>+C should be saved across, but depends on the filesystem.
<efraim>I'm running on btrfs, I applied +C to /tmp because it seems it should speed up compiling, as a side effect syncthing seems to have built with no problems
<zybell_>+C on dir on btrfs set +C on new files
<zybell_>But /tmp in many dists is special:`rm -rf /tmp;mkdir /tmp`on reboot
<efraim>that much I knew, just not sure if I'd have to reapply on each reboot on GuixSD
<zybell_>idk if GuixSD does that;look in stage-1 script.
<ajjlmau>why doesn't my changes in /etc/hosts stay?
<zybell_>I would guess there is a .scm which makes /etc/hosts. Normally that should be explained in a comment at the start of hosts.
<daviid>is master the guix 'stable' branch?
<agaric>ajjlmau: https://www.gnu.org/software/guix/manual/html_node/operating_002dsystem-Reference.html , appears that guix creates a default one if no custom file is specified
<ajjlmau>should my changes stay if I specify custom file?
<agaric>i'm pretty sure that's how it works - that's how i get my custom sudoers working
<ajjlmau>ok thanks, I'll try
<agaric>:-)
<ajjlmau>I have this other problem where guix tells me to include xorg in the config file and when I do it, guix complains that xorg-server is provided more than once
<ajjlmau>I'm trying to configure slim
<efraim>sneek: later tell lfam on my GuixSD system with btrfs i ran `sudo chattr -R +C /tmp' and syncthing built successfully
<sneek>Okay.
<efraim>sneek: botsnack
<sneek>:)
<agaric>ajjlmau: could it be that you're configuring slim-service on top of %desktop-services? (i'm pretty new to this as well)
<agaric>ajjlmau: seems relevant: https://lists.gnu.org/archive/html/help-guix/2016-05/msg00012.html
<ajjlmau>agaric: oh, yes :D
<ajjlmau>I got it to work
<ajjlmau>I'm not very good with guile
<agaric>congrats. same here. i installed guixsd to start getting familiar with it.
<civodul>icedtea fails on i686: https://berlin.guixsd.org/log/0wzm6cnlvn310sknvrbz7pazj8hrsbr1-icedtea-2.6.12
<rekado_>civodul: how weird.
<civodul>rekado_: maybe a race condition in creating that libjvm.so?
<rekado_>I’m reb-building it now on berlin to see what happens.
<rekado_>This happens in a separate phase “install-libjvm”. I don’t see how there could be a race-condition.
<rekado_>I wonder if on i686 there actually is a directory lib/amd64…
<civodul>oh, good point :-)
<rekado_>but that commit is from Oct 2017. Why didn’t we notice this earlier…?
<rekado_>the new gnucash also needs boost with ICU, so it can only be built on core-updates.