IRC channel logs

<pmikkelsen>I think I solved the erlang man-page problem, and I have sent a patch to guix-patches. That is the right list, isn't it?

<lfam>Yes, that's the right list

<pmikkelsen>good

***quasisan1 is now known as quasisane_

***quasisane_ is now known as quasisane

<jlicht>hello guix

<lfam>Hi jlicht

<jlicht>hi lfam!

<jlicht>I am ... quite perplexed currently xD

<jlicht>my libreboot T400 decides to go to sleep every 30 seconds, like clockwork

<lfam>Hm, sounds annoying ;)

<jlicht>so to prevent that, I issued a `chmod 444 /sys/power/state´ as root

<jlicht>which works, but now I am getting several ¨Failed to open /sys/power/state: permission denied¨ messages + screen flickering XD

<lfam>Hm...

<jlicht>To Be Continued, I guess

<lfam>What could it be?!

<lfam>This is GuixSD?

<lfam>If so, can you share your OS configuration file?

<jlicht>it is! My guts are telling me it is some weird interaction between the libreboot T400 + elogind

<lfam>It doesn't make sense to me that libreboot could cause some user-space application to start echo-ing 'mem' to /sys/power/state

<lfam>But, I have no other hypothesis to test

<jlicht> http://paste.lisp.org/+7JGI

<jlicht>it only ¨starts happening¨ once I have a config with xorg enalbed

<lfam>The configuration seems straightforward...

<jlicht>at first I had my own frankenstein list of services + %base-services, so I tried to go back to a more factory-setting ;-)

<jlicht>lfam: I could also post dmesg, (Does that contain any sensitive info?)

<lfam>It could, it really depends on what your system is doing and what you consider sensitive.

<jlicht>It is a fresh system, with a dummy password and no info on there ;-)

<lfam>I think it will be fine then, but you should skim it to be sure

<jlicht>hmm, ¨paste too large¨

<lfam>You could compress it with `xz` and send it to bug-guix@gnu.org

<lfam>It's definitely a bug for the system to sleep uncontrollably

<jlicht>lfam: Good idea :-)

<lfam>There's only a few places in our elogind's source code that contain the string 'sys/power/state'. So it might not be too hard to figure out what's going wrong by reading that code

<lfam>s/There's/There're

<lfam>Recently I noticed that elogind has a life of its own: https://github.com/elogind/elogind/

<lfam>It's possible they fixed a bug that could cause this. At least, I'd read their bug tracker if I got stumped

<jlicht>lfam: let´s hope so :-). I would prefer to do my thesis work on a system with only free software :D

<jlicht>But I have to admit the experience is again a plus for free software; now I can at least ask people for help (or perhaps try to fix it myself ...)

<jlicht>It reminds me a bit of the xkcd about yak shaving

<reepca>Hm, apparently one of libunistring's tests has been hanging (and causing "test-lock" to use 382% CPU) for quite awhile. Any guesses what might cause that?

<wigust>Hello Guix

<solene>It seems there is no doc about a service for mpd in the manual https://www.gnu.org/software/guix/manual/guix.html

<rekado>sneek later tell jlicht I have an update for elogind, but there’s a small problem with it, so it will take me a little longer to push it.

<civodul>Hello Guix!

<rekado>Hi!

<civodul>rekado: i think you can move to Guix proper the Cuirass fix from berlin.scm, WDYT?

<jsierles>what does this do? looking at gcc-toolchain's inputs. ("ld-wrapper" ,(car (assoc-ref %final-inputs "ld-wrapper")))

<efraim>Its using whatever 'ld-wrapper' is the first ld-wrapper from %final-inputs

<jsierles>ok. i don't understand what that's for

<reepca>"Final derivations used as implicit inputs by 'gnu-build-system'. We still use 'package-with-bootstrap-guile' so that the bootstrap tools are used for origins that have patches, thereby avoiding circular dependencies." - comment from (gnu packages commencement) about %final-inputs

<jsierles>alright

<reepca>still pretty weird though - as far as I can see, %final-inputs should be an alist mapping names to packages. Taking the car of a package seems a bit odd.

<reepca>ahhh never mind, I had it wrong, it's not an alist in the usual sense

<reepca>it's not a list of pairs, it's a list of 2-element lists

<reepca>so assoc-ref would give the cdr of the first pair, which is the second pair, (some-package . '())

<reepca>so that explains that. But it doesn't explain why it doesn't just use the variable ld-wrapper directly. Perhaps which one is in %final-inputs may vary.

<jsierles>yeah, ah well, just curious

<jsierles>anyone know the status of the conda packaging? I saw on the devel ML that it might be close to ready

<ryanwatkins>Hey guys, how might one install a terminfo file for a given terminal like st?

<ryanwatkins>How do I install c99 aswell:(

<catonano>I am reconfiguring and it's building inkscape

<catonano>why ? How is inkscape a dependency of anything ? The Gnome desktop ?

<catonano>bah

<nee``````> catonano: I think it's somehwere in the gtk-doc dependencies, along with latex which is like 1gb big. It also annoied me some weeks ago.

<nee``````>I think it only happens when the server didn't build the docs before.

<catonano>nee``````: sigh :-/

<catonano>nee``````: thanks anyway for confirming this happes

<catonano>happens

<reepca>catonano: gtk-doc -> dblatex -> inkscape. dblatex is also used by hydra and lilypond if you have any of those as well.

<reepca>that'd actually be a pretty useful tool, "guix why" or something like that

<nee``````>It would be useful to have a guix pull channel that only gets updated when the server has built the binaries for everything.

<catonano>reepca: thanks. I don't know what dblatex is but its name doesn't suggest me any reason why it shhhould depend on Inkscape. As for the build farm lagging, I know rekado is working on that

<reepca>"for svg conversion" says a comment

<rekado>jsierles: the conda package has been merged already.

<jlicht>hello #gux

<jlicht>*guix

<catonano>hi jlicht

<ryanwatkins>Any idea why ncurses terminfo files are not working for me?:D

<rekado>jlicht: hi!

<rekado>jlicht: I have an update for elogind, but there’s a small problem with it, so it will take me a little longer to push it.

<jlicht>o/

<jsierles>rekado oh! well that's the best answer i could expect

<rekado>ACTION doing sneek’s work here

<jlicht>rekado: wait, what?

<jlicht>ah, I see. Is there a problem with elogind related to my help-guix mail?

<rekado>earlier you wrote that your laptop suspends all the time.

<rekado>it might be related to elogind.

<jlicht>rekado: Yeah, I elaborated a bit on the ML: https://lists.gnu.org/archive/html/help-guix/2017-07/msg00094.html

<jlicht>but thanks for the update either way. AFAIK, elogind does not track systemd development?

<rekado>there is another upstream with active development.

<rekado>I don’t know if they follow systemd.

<jlicht>rekado: thanks for informing me either way. Is there any way in which I can help test the update before you push it?

<efraim>How's our GNOME Wayland support going? I think our gedit is built with x11 only

<rekado>jlicht: the only problem is with the check phase in the “po” directory. It’s hardly fatal.

<rekado>jlicht: I could just disable it and you could test the patch.

<efraim>jsierles rekado: conda also builds on aarch64

<efraim>I don't know what it does, but it built without any problems

<jsierles>heh

<rekado>efraim: it’s a package manager :)

<jsierles>can I generate a profile from a manifest without 'entering' that profile?

<jsierles>just want to generate it for others to use.

<jlicht>rekado: I do not see the po-issue as a show-stopper, so I'd be very happy with a link to your patch :-)

<rekado>jsierles: doesn’t “guix package -p /some/where -m the-manifest.scm” do just that?

<jsierles>rekado: i don't know. the docs says "Use profile instead of the user’s default profile" which doesn't sound like that

<efraim>Wouldn't that 'enter' the profile? Does 'guix build' take a manifest?

<jlicht>You only 'enter' the profile by setting GUIX_PROFILE and sourcing /some/where/etc/profile in my experience

<jsierles>my goal is not to provide a command to others to use. it's to generate a profile without 'losing' access to the guix command line. since guix is not part of that profile.

<rekado>jsierles: it just builds the profile and creates a link at /some/where

<rekado>that’s what “-p” does. It won’t override the default profile.

<jsierles> ok - so what path is supposed to be passed to -p?

<rekado>jlicht: I’m building elogind with tests disabled now. If that builds fine you can give it a try.

<rekado>jlicht: I won’t have time to test the new elogind myself, though, so I cannot guarantee it actually works.

<jsierles>this happened because i ran 'guix package -m manifest.scm' and lost access to guix, as it's no longer in my path

<rekado>jsierles: anything you want

<rekado>jsierles: it just has to exist

<jsierles>thanks

<jsierles>so now i don't have the guix command, is it possible to roll back to the previous profile somehow?

<jsierles>besides manually

<rekado>I’d do it manually.

<jsierles>ok, done :)

<rekado>Or you can reference guix from your previous profile generation directly (/var/guix/profile/…/guix-profile-link-42/bin)

<rekado>but that’s more difficult than just changing the link yourself.

<rekado>jsierles: I’m sure you can also just build a manifest, but that’s not exposed on the command line. You can do it with Guile, though.

<jsierles>rekado what do you mean? i am using a manifest already

<solene>is guile-dbi package usable with guile version 2.2.2 ?

<rekado>jsierles: I mean “just build”, not “build and link to a target” as “guix package -m” does.

<jsierles>rekado: ah, so guix build generates a profile too

<jsierles>is there a way, without using guile, to see the target profile hash?

<rekado>jlicht: The install failed… I’ll let you know when it’s ready to test.

<jlicht>rekado: If you don't have time atm, don't bother: I am in no hurry.

<jsierles>does 'guix import pypi' require a requirements.txt file? I found a package (cufflinks) that doesn't have it in the tarball

<rekado>cufflinks is already in (gnu packages bioinformatics)

<jsierles>yeah, i was just using it as an example, as i've seen a couple packages like that.

<jsierles>i guess in that case I just define the deps manually

<rekado>jlicht: I’ve sent you a patch for elogind. It builds for me.

<jlicht>thanks rekado. I will try to apply it once home. Mind you, how do I add a patched version of elogind to my GuixSD OS configuration?

<jlicht>ah, I can just apply the patch to guix, and then use that guix to reconfigure. nvm my previous question

<jsierles>i'm seeing this error: guix build: error: python-feather-format: unknown package

<jsierles>despite having the correct GUIX_PACKAGE_PATH

<jsierles>i have another package in that path (different file) which builds fine.

<jsierles>ok, there was a syntax error.

<ng0>python-graphene fails, missing test input pytest>=2.9

<ng0>what's strange is that even if I provide it it iisn#t found

<ng0>maybe -3.0

<civodul>jsierles: yeah files in GUIX_PACKAGE_PATH are skipped when they contain errors

<jsierles>it was a warning, that's what threw me off i guess

<ng0>I don't have the time to solve the graphene issue, I reported it as a bug.

<civodul>ACTION has 'guix pull' using Guile-Git \\o/

<rekado>civodul: woo!

<civodul>running a check now and will submit in a moment

<civodul>pretty cool!

<jsierles>what does 'out' mean in the list of packages from 'guix package -A'?

<rekado>jsierles: packages can have multiple outputs.

<rekado>jsierles: the default output is called “out”

<rekado>jsierles: the “git” package, for example, has multiple outputs for additional features

<rekado>(such as send-email or git-svn)

<jsierles>i see. how would you build one of the other outputs?

<jsierles>nevermind, i found it.

<rekado>by appending “:” followed by the name of the output

<wigust>Hello Guix

<jackhill>hi wigust!

***Introoter is now known as MooMachine

***MooMachine is now known as Introoter

<jackhill>hmmm, I'm running guix on top of Debian. guix import gem foo fail with guix/build/download.scm:424:6: X.509 certificate of 'rubygems.org' could not be verified. I assume that this is because I don't have the needed certificate stores installed. Is it that, or something else.

<jackhill>full error at: http://lpaste.net/357257

<lfam>jackhill: There are two issues: 'insecure-algorithm' and 'signer-not-found'

<jackhill>ah, so it's tring http://… ?

<jackhill>or sha1?

<lfam>I don't know exactly

<ng0>sneek: later tell adfeno: Is it really just this? https://git.parabola.nu/abslibre.git/tree/nonprism/icedove because this can be dealt with.. it's better to point patches and issues at me and not sites for resources.

<rekado>jackhill: have you set the certificate related environment variables?

<jackhill>for what it's worth, I see the same thing with hackage.

<jackhill>rekado: probably not, I don't remember setting them.

<ng0>ah, sneek isn't here

<jackhill>rekado: thying now…

<jackhill>rekado: that was it.

<jackhill>rekado, lfam: sorry I missed that, thanks for the help.

<rekado>you’re welcome! Glad it’s fixed.

<jsierles>import should warn about that - bet it happens a lot

<ng0>what's up with the python failures? I've got another one, for pyopenssl now

<ng0>or maybe my clock is wrong

<ng0>2 minutes off.. oh

<ng0>ok

<jsierles>i'm seeing: guix import: error: no source release for pypi package pyarrow 0.5.0

<jsierles>what should i do in this case?

<jsierles>i'm trying to build python-feather-format which depends on pyarrow in it's setup.py: install_requires=['pyarrow>=0.4.0'],

<jsierles>i see we have python-arrow, but it's expecting the 'pyarrow' package I suppose.

<jsierles>ah, nevermind. python-arrow is an unrelated package. so I do need pyarrow

<jsierles>guess I have to use git-fetch - does that allow building in subdirs?

<jsierles>for example, the pyarrow source is located here: https://github.com/apache/arrow/blob/master/python

<ng0>hm. does someone have an short mcron run every 5 minutes every day for me?

<ng0>there's this 2:15 example in mcron docs, I think I could just substitute the 2 with a 0 and the 15 with a 5 and it should work, right

<lfam>jsierles: It needs to change to the subdirectory before building?

<jsierles>lfam: yeah, since 'setup.py' is in a subdirectory of the git repository

<lfam>ng0: I think mcron can take the regular cron language too

<jsierles>worse case I guess I can create a repo just for that package.

<lfam>jsierles: What about copying this? https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/kerberos.scm#n70

<jsierles>or try to build a source release and upload that somewhere

<jsierles>hmm

<jsierles>what would i 'add-after' for a python build?

<jsierles>ah, i guess it has an unpack phase too

<jsierles>or add-before build maybe.

<ng0>lfam: https://gitlab.com/ng0/systems/blob/master/guixsd/servers/shadownet/shadownet.scm#L35 should work

<lfam>I usually do those things after the unpack phase because I think it's less likely to break if we add new pre-build phases in the future

<jsierles>makes sense. trying

<jsierles>how do you get a sha for a git reference?

<jsierles>ah, found it. guix hash -rx

<jsierles>what might cause: extraneous field initializers (synposis)

<rekado>a typo

<jsierles>dang

<jsierles>if i remove that line, it says synopsis is mssing

<rekado>well, you spelled it “synposis”, not “synopsis”

<rekado>“npo” vs “nop”

<jsierles>doh

<jsierles>what does extraneous field mean? an unsupported field?

<jsierles>guess so. thanks !

<jsierles>synposis should mean something in english

<rekado>it stands for the post-conference events: “syn” meaning “with”, “posis” meaning “drink” :)

<rekado>we had good “synposis” after FOSDEM :)

<jsierles>gotta get that into OED

<catonano>today I went away, leaving the computer compiling

<catonano>now I came back. It was in sleep mode. I woke it up. It started compiling as if it had never stopped

<catonano>some webcore thing. It seems like a hard bone to chew

<rekado>that’s part of webkitgtk, I think

<jsierles>OK, got python-pyarrow definition working!

<jsierles>except, it needs apache arrow which is not in guix :S

<jsierles>so time to package that

<jsierles>rekado: which file would be appropriate for apache arrow?

<rekado>what does it do?

<rekado>ACTION runs hardware diagnostics on a server right now

<jsierles>Apache Arrow is a columnar in-memory analytics layer

<rekado>don’t know what this means :)

<jsierles>basically for passing data like arrays or dataframes between languages

<jsierles>among other things

<rekado>“serialization”, maybe?

<jsierles>we use it for exporting and importing data structures across languages in our execution model

<catonano>yes, webkitgtk 2.16.3

<jsierles>i guess that would work

<rekado>catonano: maybe a substitute is already available on berlin.guixsd.org?

<catonano>rekado: maybe but now it's grinding. How would I inform it ?

<catonano>wouldn't I loose whatever it has compiled until now ?

<rekado>you would indeed lose the progress of the compilation

<rekado>but maybe you could do C-z to stop the process and then run “guix build --substitute-urls=https://berlin.guixsd.org -n webkitgtk” to see if it will download the substitute.

<rekado>and if not: just “fg” to resume the compilation

<catonano>I'm gonna try

<catonano>Ok, instead of C-z I pressed C-c by habit. So it's all lost now

<rekado>:(

<catonano>rekado: eh, it happens

<rekado>if that doesn’t work you can also try https://hydra.gnu.org

<catonano>no worries, it was an experiment anyway

<rekado>it might already have the package and the mirror is behind

<catonano>it says:

<catonano>substitute: updating list of substitutes from 'https://hydra.gnu.org'... 100.0%

<catonano>The following derivation would be built:

<catonano> /gnu/store/xvm1jsbmy409n87mhc3xx73snh8jn1f5-webkitgtk-2.16.3.drv