IRC channel logs

<temp_trisquel>/sys/firmware/efi is available.

<temp_trisquel>> Could not prepare Boot variable: No such file or directory

<temp_trisquel>I decided to reinstall OS.

<temp_trisquel>wigust: Hi. Which game did you recommended instead of NetHack? I forgot.

<temp_trisquel>wigust: Also, please, link their website.

<temp_trisquel>Could not prepare Boot variable: No such file or directory in GRUB installation

<buenouanq>We've done everything we can for you in here temp_trisquel - Have you tried the actual help mailing list yet?

<temp_trisquel>buenouanq: guix-devel?

<buenouanq>no, help-guix

<temp_trisquel>Last GuixSD question.

<temp_trisquel>So, do I need mapped-devices dependency for partition which mount-point is on encrypted device?\\

<temp_trisquel>*device?

<ng0>anyone got an idea what's going on with the website generation?

<ng0>this bug: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=27925

<ng0>when I just try to build the website in an environment with "guix guile-json haunt guile-commonmark guile-syntax-highlight" I get a different problem, which is "no code for module (commonmark)". Which has nothing to do with the guix build -f guix.scm problem.

<ng0>strange enough the second problem is fixed once I no longer use environment but install the packages, which is what I wanted to avoid

<ng0>so something with env variables on my side is leaking into it, obviously

<htgoebel>Hi guix!

<htgoebel>I'm ATM packaging "breeze-gtk", a "GTK+ theme matching KDE's Breeze theme".

<htgoebel>Should I keep the original name or change it to "gtk3-engines-breeze" (fedora) or "metatheme-breeze" (suse)?

<efraim>whats the upstream name?

<htgoebel>IMHO "gtk3-engines-breeze" would fit best since it is used for GTK.

<htgoebel>efaim: "breeze-gtk"

<efraim>i don't know what the significance of "engine" in the name is, so I'd go with breeze-gtk

<htgoebel>okay.

<efraim>it seems a number of the ^r-* packages hardcode '-I/usr/local/{lib,include}'

<temp_trisquel>WTF Could not prepare Boot variable: No space left on device

<ng0>when I talk and write about Guile, is it "Guile Scheme" or just "Guile"?

<temp_trisquel>Guile is Guile Scheme interpreter.

<ng0>I've seen both

<ng0>I know.. But for a text intended easy to understand, just writing "Blabla is written in Guile" would still be okay I think

<ng0>I think I could use Guile Scheme and point to explanations.

<ng0>this is for a text not intended to become very long or in depth and not primarily for people with developer experience

<temp_trisquel>I think, in Scheme.

<temp_trisquel>Guile is interpreter, not language.

<ng0>?

<ng0>no, in general. at all

<ng0>my sentence did not lack a ".", i just did not put an ending dot

<ng0>Guile is an implementation of the Scheme language. so to write just Guile should be accurate if I'm not aiming for academic accuracy. I want to simplify explanations.

<ng0>"… written in Guile (a Scheme implementation)" would produce a sentence with nested parens, but I think this should work

<ng0>grammatically I still need a major review of all the texts, so fixing the grammar in that line in the context of the sentence isn't necessary right now

<ng0>and nested parens give a preview on Guile ;D

<temp_trisquel>Can somebody help with EFI?

<temp_trisquel>"Could not prepare Boot variable: No space left on device"

<temp_trisquel>SOMEBODY HELP ME!

<temp_trisquel>HELP ME TO INSTALL EFI GUIXSD!

<efraim>Have you sent your config.scm to help-guix for us to help you?

<temp_trisquel>No yet.

<temp_trisquel>It looks like I have known problem.

<rekado_>efraim: is it a problem that the R packages hard-code -I/usr/local? At build time this won’t be found anyway.

<rekado_>I just unbundled all the minified JS that’s included in r-shiny, so I guess we can finally add this package to Guix.

<rekado_>It’s needed for the next big upgrade of all CRAN packages that I’m working on.

<rekado_>temp_trisquel: please send you config.scm to help-guix. Otherwise there’s very little we can do for you.

<sirgazil>sneek: later tell ng0: regarding Guix website, have you tried building the wip-website-update branch? The "guix.scm" file was renamed to "build.scm". Maybe that change is related to the errors you're seeing?

<sneek>Got it.

<sirgazil>sneek: later tell ng0: FWIW, while working on the website update, I couldn't create an environment either, and had to install packages by hand. Also had to rename the "guix.scm" file to something else; otherwise, importing the guix package installed with guix would not work well.

<sneek>Got it.

<temp_trisquel>rekado_: I think reason isn't config.

<temp_trisquel>rekado_: Reason is /sys/firmware/efi/efivars no space left.

<temp_trisquel>rekado_: How can I free space there.

<temp_trisquel>rekado: *?

<temp_trisquel>*rekado_

<efraim>rekado_: the only problem I can think of with that would be if someone were building without sandboxing

<rekado_>temp_trisquel: I don’t think anything under /sys/ is actually backed by disk.

<rekado_>temp_trisquel: who’s trying to write there and what?

<temp_trisquel>rekado_: What can I do?

<rekado_>temp_trisquel: as I wrote before, I have no idea about EFI.

<temp_trisquel>rekado_: GRUB is trying to write EFI variable here.

<rekado_>temp_trisquel: have you been able to write EFI variables there before?

<temp_trisquel>rekado_: No.

<temp_trisquel>It looks like they made EFI just for fun if OS cannot be installed.

<rekado_>this is a shot in the dark, but you may need to add iomem=relaxed to the kernel arguments to write there. It’s just a guess.

<temp_trisquel>It says "No space left on device", not "Read-only file system"

<temp_trisquel>*filesystem

<rekado_>temp_trisquel: it’s not a real file system, so things are a bit different here.

<quiliro>hello guix!

<quiliro>mi estas bone

<quiliro>mi uzas guixsd

<quiliro>mi legas manualon por learnas guixon kaj konstrui manualon

<quiliro>de fari paketon

<temp_trisquel>So what can I do?

<quiliro>temp_trisquel: about what?

<rekado_>temp_trisquel: have you tried my suggestion yet?

<temp_trisquel>rekado_: Upload config?

<temp_trisquel>quiliro: No space left on efivars.

<rekado_><rekado_> this is a shot in the dark, but you may need to add iomem=relaxed to the kernel arguments to write there. It’s just a guess.

<rekado_>

<temp_trisquel>rekado_: I will try.

<quiliro>temp_trisquel: I suggest you place your config on a server so you and we can see the files

<quiliro>the config i mean

<quiliro>then...you can make versions with svn or git

<quiliro>and see your progress

<quiliro>and learn from it

<quiliro>how to use mumble?

<quiliro>just installed it

<quiliro>must connect a dev with rms in 2h30

<temp_trisquel> http://termbin.com/ww80

<temp_trisquel>quiliro: If you need to contact RMS, write on rms@gnu.org

<quiliro>temp_trisquel: cool....did you diff with the original template?

<temp_trisquel>quiliro: / I don't know.

<quiliro>temp_trisquel: i did...we agreed to mumble in 2h30

<temp_trisquel>quiliro: No.

<quiliro>temp_trisquel: with diff you can check the differences

<temp_trisquel>I changed only GRUB and FS settings.

<temp_trisquel>Nothing more.

<temp_trisquel>You can see config on termbin.

<temp_trisquel>(of course, I'm changing username, UUID, etc.)

<temp_trisquel>CONFIG http://termbin.com/ww80

<temp_trisquel>rekado_

<quiliro>temp_trisquel: i don't know about encrypted partitions

<quiliro>but the rest i have almost the same except for the desktop

<quiliro>why don't you test without encryption first?

<quiliro>i have a macbook air

<quiliro>with EFI

<quiliro>and works perfectly

<rekado_>temp_trisquel: your “mapped-device” has target “the-root-device”, but “/” is on “my-root”.

<rekado_>temp_trisquel: that’s wrong.

<quiliro>did you mount your efi partition on /boot/efi when init or reconfigure?

<quiliro>oh....true

<quiliro>rekado_: experience!

<temp_trisquel>rekado_: No. mapped-device is LUKS. my-root is inside LUKS.

<temp_trisquel>quiliro: /mnt/boot/efi

<quiliro>ok

<temp_trisquel>quiliro: When init.

<rekado_>temp_trisquel: but how would Guix know how to mount it?

<temp_trisquel>rekado_: ESP or LUKS?

<rekado_>temp_trisquel: the mapped device maps the encrypted LUKS thing to "the-root-device"

<rekado_>so you’d get /dev/mapper/the-root-device

<temp_trisquel>rekado_: Yes.

<rekado_>and that’s where the root device is on

<rekado_>so where does the partition with the label “my-root” come from?

<temp_trisquel>rekado_: But /dev/mapper/the-root-device has my-root partition.

<rekado_>okay

<temp_trisquel>mount LABEL=my-root /mnt

<temp_trisquel>=

<rekado_>why do you mount /dev/sda1 as /boot/efi?

<temp_trisquel>mount /dev/mapper/the-root-device /mnt

<temp_trisquel>rekado_: /dev/sda2 is root. /dev/sda1 is EFI System Partition.

<rekado_>yeah, but why do you mount the EFI system partition?

<temp_trisquel>rekado_: I don't know. That's written it manual.

<temp_trisquel>rekado_: Do I need to mount it?

<rekado_>that’s only during installation.

<temp_trisquel>rekado_: Do I need to include this partition in config at all?

<rekado_>I don’t think so

<rekado_>you just need to mount it when you install GuixSD

<temp_trisquel>rekado_: Where? Will GRUB find it then?

<temp_trisquel>rekado_: Or /boot/efi is default path?

<solene>hello, I would like to run a static binary I downloaded on the web. When I run it by doing ./bin_file I get the error "./bin_file : no such file or directory"

<solene>is there some magic trick with guix environment maybe ?

<temp_trisquel>solene: It looks like it's 32-bit and you're using 64-bit OS?

<lfam>solene: Does the file have the execute bit set? (`chmod 500`, for example)

<solene>temp_trisquel: it's a 64 bit binary

<solene>file returns this : " ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.18,"

<solene>lfam: yes, it has the execute bit. I didn't think about it, I checked but it's there

<rekado_>solene: it’s the interpreter

<rekado_>solene: on GuixSD we don’t have /lib64/ld-linux-x86-64.so.2

<wigust>cbaines: Hello, sorry, I forget to change commit message in “* doc/guix.texi”.

<bavier`>you might be able to use patchelf to reset the interpreter

<rekado_>solene: it’s part of the glibc package

<rekado_>solene: it might be possible to just link the current system’s glibc package’s ld-linux-x86-64.so.2 to the global location at /lib64.

<solene>rekado_: is there a way to use guix glibc to run it ?

<solene>i'll try LD_PRELOAD

<quiliro>temp_trisquel: when booting with the installed mount boot/efi on /mnt/boot/efi

<wigust>temp_trisquel: Hello, it's in http://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/games.scm#n4621

<quiliro>temp_trisquel: and then install

<rekado_>solene: this won’t help when you have a statically linked binary.

<quiliro>temp_trisquel: but you must be sure root is correct as noticed by rekado_gi

<rekado_>solene: all you need is to either provide the interpreter at the expected location or override the interpreter.

<solene>rekado_: who is the interpreter in that case ?

<solene>I don't know much about how binaries work

<solene>I copied the library in /lib64/ and now I get error while loading shared libraries: libstdc++.so.6: cannot open shared object file: No such file or directory

<solene>I think, if I really want it to work, I'll need to copy every library somewhere to get it working

<solene>thanks for your help rekado_ !

<lfam>What is the name of this construct that I see used within substitute-keyword-arguments, where the value passed to #:tests? seems to be repeated: https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/java.scm#n2530

<rekado_>solene: you can fix that with preloading.

<lfam>Sometimes it's like ((#:tests? _) #f), other times it is ((#:tests? _ #f) #f)

<rekado_>solene: i.e. add the actual location of libstdc++.so.6 to LD_PRELOAD

<rekado_>solene: are you *sure* this is a statically linked binary…?

<rekado_>lfam: I don’t know! The latter looks wrong to me.

<lfam>rekado_: I noticed the latter in (gnu packages qt)

<solene>rekado_: I need to go, I'll try later ;)

<bavier`>lfam: the latter uses the default-value semantics

<efraim>lfam: when you want to overwrite the tests? flag you have to pass it a new default also

<bavier`>in case the #:tests? keyword isn't present in the inputs

<rekado_>ah, that’s good to know

<efraim>I think its suppsed to be the second one, but I could be wrong

<efraim>We also have one in python I think

<efraim>solene: you might need patchelf to change it to link to where guix has the librarirs

<quiliro>temp_trisquel: did you have a problem with efivars?

<temp_trisquel>quiliro: Yes, no space left on device.

<quiliro>Did you make a search about it?

<quiliro>perhaps will bring some light

<temp_trisquel>quiliro: They're saying dump-* files must be removed but there are no such files.

<temp_trisquel>quiliro: What can I do?

<quiliro>temp_trisquel: change the-root-device to my-root

<quiliro>and test

<temp_trisquel>quiliro: It's EFI issue.

<quiliro>i see your efi config the same as mine

<quiliro>and mine works

<quiliro>your efi is in /dev/sda1 , right?

<temp_trisquel>EFI System Partition is /dev/sda1

<quiliro>but still your root definition is incorrect

<temp_trisquel>/sys/firmware/efi/efivars is not HDD, I think.

<lfam>From 'gnu/packages/libunwind.scm': FIXME: As of glibc 2.17, we get 3 out of 34 test failures. Report them upstream.

<lfam>`git blame` shows the date: 2013-02-11 22:44:42 +0100

<lfam>;)

<quiliro>temp_trisquel: wouldn't you try fixing that first? is there a reason not to fix it?

<quiliro>temp_trisquel: you think it is correct?

<temp_trisquel>quiliro: Yes, it's correct.

<quiliro>why?

<temp_trisquel>my-root is partition label.

<temp_trisquel>the-root-device is not partition label, it's used to be in /dev/mapper

<lfam>Can anyone give advice about Python pip? Our Python 3 package includes a pip binary. Is that one "good enough"? I ask because I tried installing both the python-wrapper and python-pip packages, but there are conflicts when building the profile

<lfam>warning: collision encountered: /gnu/store/s31njsb8cplmly202gjsxk9165cyb672-python-3.5.3/lib/python3.5/site-packages/pip/vcs/__pycache__/git.cpython-35.pyc /gnu/store/4mbfwmjjjs1rckp93gp0xywnwnnm7bai-python-pip-9.0.1/lib/python3.5/site-packages/pip/vcs/__pycache__/git.cpython-35.pyc

<lfam>And others

<temp_trisquel>Can you help?

<quiliro>temp_trisquel: i got confused with the comment about target and my-root on the template

<temp_trisquel> https://www.gnu.org/software/guix/manual/html_node/Mapped-Devices.html

<temp_trisquel> https://www.gnu.org/software/guix/manual/html_node/File-Systems.html

<temp_trisquel>Help!

<solene>does someone know if webrtc is disabled on icecat ? All I can find is a 2015 commit that activates it

<solene>I need it :(

<quiliro>temp_trisquel: i found your problem

<quiliro>temp_trisquel: your uuid is wrong

<quiliro>temp_trisquel: you should find the one for your own encrypted partition

<quiliro>temp_trisquel: i do not know how to find it...but i am sure that is why your partition is not found...your uuid is the same as the one on the sample in the manual...that is impossible

<quiliro>temp_trisquel: the rest is correct assuming the manual is fine

<temp_trisquel>quiliro: My UUID is classified.

<quiliro>ot

<quiliro>ok

<quiliro>so i dont know what else to test

<temp_trisquel>quiliro: I'm using correct UUID in config.scm on my laptop.

<quiliro>try a basic config without encryption first

<quiliro>then see what the problem is

<quiliro>i have no idea what else to do

<temp_trisquel>Legacy BIOS will work.

<quiliro>perhaps guix has a bug with efi with encryption

<quiliro>but you can be sure if you test just efi

<quiliro>without encryption

<rekado_>quiliro: it doesn’t make much sense to test without encryption

<rekado_>quiliro: this cannot easily be changed without a reinstallation.

<quiliro>he is installing

<rekado_>encryption should not be a problem here.

<rekado_>*they

<quiliro>so he can test with a bare-bones firsr

<quiliro>t

<quiliro>so what is the problem

<temp_trisquel>Is it possible to resize LUKS partition later?

<quiliro>he followed the manual exactly

<temp_trisquel>If I will decide to replace BIOS with UEFI later?

<rekado_>yes, as I’m saying: they want to use encryption, so let’s set it up like that from the start.

<temp_trisquel>*later.

<quiliro>i have not been able to use mumble

<quiliro>it does not detect my input or output audio devices

<rekado_>temp_trisquel: I’m not sure if you can do that easily. You’d have to change the partition table.

<rekado_>temp_trisquel: I cannot give advice on switching to EFI later.

<temp_trisquel>I think, I just need to replace BIOS Boot Partition with EFI System Partition and reinstall bootloader.

<rekado_>temp_trisquel: AFAIK you also need to use GPT.

<rekado_>but if that’s all then it should not be hard to switch at a later point.

<rekado_>GuixSD’s “guix system reconfigure” will install the bootloader to the new target.

<rekado_>so maybe you can just try to install with legacy BIOS

<quiliro>upon starting mumble, I get pulseaudio: connection failure

<lfam>rekado_: At the institute, do you still use the setup where the store and daemon are exposed from a central server over NFS? If so, do your users find the profile hooks terribly slow?

<quiliro>but pulseaudio is installed

<lfam>quiliro: Is pulseaudio running? Try `ps aux | grep pulseaudio`

<lfam>I'm not familiar with pulseaudio, however, so I don't have any advice beyond making sure it's running

<rekado_>lfam: yes, we’re still using this setup

<rekado_>lfam: since we’ve moved to a dedicated NFS server for Guix it’s no longer as bad as it used to be.

<rekado_>lfam: the profile hooks are still slow, but users seem fine with it now.

<lfam>rekado_: Unfortunately I am seeing the manual-database hook take ~2 minutes :/

<rekado_>oh, that’s excessive

<kristianpaul>quiliro: o/

<lfam>Maybe I can add a knob to just change the list of profile hooks. In this context, the only one that's useful is the ca-certificate hook

<wigust>Could Guix preserve “build-tree” if build is succeeded?

<wigust>Like “guix build -K” but for succeeded builds?

<lfam>wigust: We don't have it currently AFAIK, but when I need it I set it up by hand by adding a failing build phase after the compress-documentation phase and using -K.

<lfam>A phase such as: (add-after 'compress-documentation 'stop (lambda _ #f))

<wigust>lfam: Thanks for tips.

<lfam>rekado_: I'm curious, how can I find more information about the source libunwind-for-julia? I'd like to check if there is a newer release but I don't know where to look

<efraim>rekado_: what happened with r-crosstalk? also bioinformatics needs the (gnu packages cran) module i think

<rekado_>efraim: I don’t see r-crosstalk used in the bioinformatics module.

<efraim>rekado_: r-plotly adds it as a propagated input

<rekado_>lfam: libunwind for julia comes from here: https://github.com/JuliaLang/julia/blob/1a43098cf7868045e64f051146484dbb2d132633/deps/unwind.mk

<rekado_>efraim: I’m on commit 2267e4d3f64c9224997f8d48d69c505fa91c9aa6 and I don’t see it :-/

<rekado_>efraim: I updated a lot of R packages today

<efraim>it was in f9b376063643ae4b7c4403d0debad590fba5942d

<rekado_>but that’s in statistics, not bioinformatics.

<rekado_>lfam: the latest version of libunwind for julia is 1.1-julia2 according to https://github.com/JuliaLang/julia/blob/5eb41125ac717e56ae425fb441d87574212456cf/deps/Versions.make

<efraim>yeah, but that was from the r-rcpp move

<rekado_>ah!

<rekado_>how could I have missed that?

<rekado_>I built all R packages

<rekado_>maybe I screwed up the rebase

<rekado_>thanks! I’ll fix it in a minute.

<efraim>looks like machine-learning.scm also needs (gnu packages cran)

<rekado_>sorry about that!

<efraim>i got that one

<rekado_>just pushed the one to add it to machine-learning. Thanks for catching this!

<efraim>oh, looks like you got it first

<lfam>Thanks rekado_!

<lfam>I filed a bug report with julia to discuss failing tests, and they suggested it could be related to libunwind. But I'm not sure...

<lfam> https://github.com/JuliaLang/julia/issues/23115

<rekado_>lfam: we are not stripping the julia binaries for this very reason. But other binaries are indeed stripped.

<rekado_>maybe we should add #:strip-binaries? #f to libunwind-for-julia?

<rekado_>a comment in 'disable-broken-tests says:

<rekado_>;; These tests fail, probably because some of the input

<rekado_> ;; binaries have been stripped and thus backtraces don't look

<rekado_> ;; as expected.

<rekado_>it looks like “libunwind” is in fact stripped and thus the inheriting libunwind-for-julia is stripped as well.

<lfam>rekado_: Are you sure libunwind-for-julia is stripped? I only checked the binaries with `file` but it says they are "not stripped"

<rekado_>I’m not sure but we do have a strip phase in the gnu-build-system.

<rekado_>unfortunately, I cannot test this right now because a disk error prevents me from building this.

<rekado_>guix build: error: build failed: cannot link `/gnu/store/.links/07pb5sx365w6ysg21z9s1vz97jhjmacmx2adi9226m04pax77jrl' to `/gnu/store/2n78drw2qmkyxz6ijpvajhkalc5yq5zf-lapack-3.7.1/lib/cmake/lapack-3.7.1/lapack-targets.cmake': Input/output error

<rekado_>:-/

<rekado_>maybe I should replace this disk

<lfam>Hm, the build log does indeed say that it is stripping the binaries

<lfam>I'll poke around some more

<rekado_>it might be enough to add (arguments '(#:strip-binaries? #f)) to libunwind-for-julia

<lfam>Yes, I'm trying this

IRC channel logs

2017-08-03.log