IRC channel logs


back to list of logs

<lfam>sneek: later tell efraim: mesa from core-updates built for me on x86_64
<sneek>Will do.
<efraim>lfam: thanks. I'll have to figure out why its failing to find libdrm on aarch64
<sneek>Welcome back efraim, you have 1 message.
<sneek>efraim, lfam says: mesa from core-updates built for me on x86_64
<efraim>sneek: botsnack
<efraim>I removed i915 from the gallium drivers and configure passed, so its only libdrm_Intel that's causing my proplem
<efraim>BSD ports to the rescue!
<efraim>lfam: does libdrm on x86_64 build libdrm_freedreno?
<lfam>efraim: no
<lfam>See line 619
<efraim>didn't think so
<efraim>So aarch has freedreno and no Intel, Intel platforms the opposite
<efraim>I'll check what people have and prepare a patch today
<ryanwatk`>Hi guys, I would really like to define an 'emacs server' service for a guixsd machine of mine, I presume it is doable somehow in the OS definition, does anybody have any idea how it would look? Should I look to the shepherd manual?
<ryanwatk`>Likewise, I figured some screen sort of daemon service would also be nice. My home environment involves many systems so I wanted to offload a lot of system resources to the non embedded devices :D
<rekado>efraim: is there a way to build the mesa drivers in a separate package?
<efraim>rekado: possibly, but I think (match (%current-system) ... is the best option, this way everyone gets the drivers that they can actually use and its all in one package
<efraim>currently we already drop the i915 dri-driver for non-intel systems, this just extends it for all the architectures and for the gallium-drivers too
<efraim>on aarch64 libdrm provides amdgpu, freedreno, nouveau, radeon, vc4 and kms
<snape>ryanwatk`: you should search for "normal user" in the shepherd manual
<sneek>snape, you have 1 message.
<sneek>snape, wingo says: please do!
<wingo>heh, i think that was about pushing your patches
<wingo>good morning snape :)
<rd161616>Hi guys
<snape>hi wingo :)
<rd161616>Previously I used qemu with the kvm option to run really fast and performant virtual machines, do you know how I can do it in guixsd please ?
<snape>rd161616: --enable-kvm doesn't work?
<rd161616>Well I have my barebones system with a few packages installed so far and IAM happy about it but i don't know which package i should install for qemu kvm snape
<snape>rd161616: "qemu" package should be enough, according to its documentation
<rd161616>Thanks mate, u think I'll be able to use the internet on it without too much problems ?
<snape>I think you won't have more problems than with standard distros :p
<rd161616>Lol sûre I understand
<rd161616>Also I don't have X11 installed is it a problem if I need windows 7 virtually
<snape>But I don't know for sure, because I never used qemu on GuixSD directly
<snape>I don't think X11 is a problem
<rd161616>Or does it use the frame buffer directly
<rd161616>Ok thx mate
<rd161616>Have a nice day
<snape>you too!
<efraim>rekado: I haven't built it yet, but my current mesa patch is at
<efraim>.. or not, let me correct that
<efraim>rekado: big oh, not zero
<efraim>I can probably combine mips and _ since mips i believe can build i915
<ryanwatk`>snape: thanks, I will take a look, I am just watching the fosdem talk now about composing system services
<ryanwatk`>another problem I seem to be getting is geiser and guile/guix integration, it keeps saying I should M-x run-geiser and when I do, no documentation is available from the original buffer
<snape>ryanwatk`: maybe geiser can't guess your scheme implementation
<snape>you can eval geiser-impl--implementation in a scheme buffer, to check it is guile
<snape>I had a similar issue: I had to set geiser-active-implementations to '(guile)
<snape>so to force geiser to use "guile" implementation
<snape>Actually, without (setq geiser-active-implementations '(guile)), every scheme file with the word "use" is considered a Chicken file
<snape>(Chicken is another scheme implementation)
<snape>even though I run "run-guile"
<snape>it might be a geiser bug, don't know
<snape>I'll report it.
<jmd>Is there a way one can persuade the guix daemon to continue a build which has aborted for some reason?
<ryanwatk`>snape: setting scheme to guile gave me some documentation but I am seeing ludo has an emacs mode of Scheme Guix Guile, is that an additional emacs mode for guix devs?
<ryanwatk`>snape: could very well be a geiser bug :D
<thomasd>ryanwatk`: could be “guix-devel-mode”, part of package emacs-guix
<civodul>Hello Guix!
<thomasd>bonjour civodul
<civodul>so we have a python 2.7 failure on core-updates:
<jmd>civodul: From which checkout was that build done?
<civodul>be81133a13932bc83b0697d73581a47674ef6987 per
<jmd>That one doesn't include lfams latest python fixes.
<civodul>ACTION triggers a new evaluation
<jmd>ACTION wishes that he understood how to use and interpret the website.
<efraim>civodul: I'm thinking of this for mesa core-updates, aarch64 doesn't have i915 at all
<efraim>i'm also going to trigger an nss rebuild on core-updates, aarch64 also needs the USE_64 flag
<efraim>6 hours for nss on aarch64
<civodul>efraim: the mesa thing sounds good, it might not even trigger a rebuild on x86_64/i686, does it/
<efraim>it moves the dri drivers higher up so it'll cause a rebuild
<efraim>but its on core updates so its not really a big deal
<ryanwatk`>any way I can change guix package mirror? I am getting a 404 on wireless-regdb-2017.03.07 from upon a guix system reconfigure
<rekado>ryanwatk`: if you have a different URL for that package you could download it with Guix, which adds it to your store.
<rekado>then Guix wouldn’t have to try to find it at the upstream URL.
<ryanwatk`>rekado: oh, that makes sense :P
<ryanwatk`>rekado: unfortunately, guix package -i wireless-regdb is still fetching this url, I presume I can get the tar from elsewhere and install from file right?
<ryanwatk`>rekado: so I changed the url manuall and did a guix package -f but somehow it still thinks I do not have the package :(
<rekado>ryanwatk`: have you tried to do “guix download the-new-url”?
<rekado>if the file has the same hash and the same name I suppose it would just work.
<ryanwatk`>rekado: I did not know the command, I got it working by making a temp scheme file with the url replaced it seems, guix download sounds much easier though, thanks
<jmd>What's the next big thing for Guix?
<rekado>jmd: didn’t you ask this question just a few days ago?
<rekado>I think it hasn’t changed since then.
<jmd>Did I?
<rekado><jmd> [20:08:22] What's the next big thing in Guix?
<rekado><lfam> [21:28:17] jmd: I'd say a better mechanism for updating Guix, and then
<rekado> the fabled "channels"
<jmd>rekado: You keep a searchable index of everything said on #guix?
<rekado>no, it’s just the ERC buffer on my workstation
<jmd>Wow. It's bigger than mine.
<thomasd>about the update mechanism, an option to update to a version for which substitutes are available, would be nice (at least some point, maybe we'd need more build infrastructure to make that work well?)
<thomasd>this is probably obvious, but I haven't read the remark anywhere else, so thought I'd mention it anyway :)
<thomasd>I wonder how others deal with substitutes and updates? I look which commits have been evaluated on, checkout one of those in git, and upgrade from that.
<efraim>rekado: civodul: I believe llvm is only really supported on x86/x86_64
<rekado>I just upgrade to the latest version unless a kernel update had just been pushed.
<rekado>efraim: oh.
<efraim>er, only really supoprted *on mesa* on x86/x86_64
<roelj>Damn.. how long do these nss tests take? It's been running for 2 hours now..
<efraim>it took 3.5 hours on my 2007 laptop, 6 hours on aarch64
<roelj>Thanks for an estimate :). Can we do something about this (reduce the repetitiveness of the tests)?
<roelj>Or is there a specific reason it needs to do things about a million times?
<efraim>i checked the source, parallel builds aren't supported, we override 'check so I don't think its multithreaded ATM
<catonano>I just reviewed a patc that had been reviewed already. Can anyone point me to an introduction about how to use tags in the Debbugs-Emacs thing ?
<catonano>I mean
<efraim>do we have a package with a conditional input that I can use as an example?
<catonano>if a patch ahs been reviewed and is waiting for a reaction by its author. Is there a flag for that ?
<rekado>catonano: I’d also like to know the answer to this question.
<rekado>I think it may be something for which we have to find an accepted procedure.
<catonano>rekado: I'm sorry, I made myself a fool
<civodul>efraim: in that case we could make the dependency on llvm conditional on x86
<civodul>most likely people on arm/mips don't care that much about mesa anyway :-)
<efraim>civodul: that's my plan, i forgot the syntax though :)
<rekado>civodul, efraim: doing so would also address Mark’s concerns.
<efraim>thats what I was thinking
<efraim>and I was worried about being able to build llvm
<civodul>rekado: right
<efraim>on a similar vein, mysql failed after 7 hours to massive test failure due to lack of CPU time & RAM shortage
<efraim>ok, i have mesa using llvm on for x86_64-linux and i686-linux, and not using it otherwise
<catonano>could be "pending" ?
<catonano>in a normal debugger "pending" would mean that the buug has not been triaged yet
<catonano>or a patch hasn't been posted yet
<catonano>but ib gix-patchhes
<catonano>it could mean that it is waiting for the author to react to thhe review
<catonano>can bugs be ordered by tag, anyway ?
<civodul>not sure
<catonano>in the debbugs web interface I see there are 5 bugs markes with "More information needed"
<catonano>were they manually markked by someone ? Or are they automagically classified by an algorithm run by Debbugs ?
<wingo>manually tagged "moreinfo" i think
<rekado>do you know how to display only those bugs that were assigned to me?
<rekado>I’m still struggling with debbugs-gnu-search
***chatter is now known as Guest9305
***Guest9305 is now known as chatter29
<civodul>catonano: most likely i added the "moreinfo" tag
<civodul>rekado: i don't think bugs can be assigned to people, can they?
<catonano>civodul: so when you mark them with "more info" do you mean that you asked the author to please send a reviewed patch ?
<catonano>so it seems
<rekado>civodul: they can. I just hit “C” and then “owner”, which sends a control message to assign the bug to myself.
<civodul>oh, interesting
<civodul>catonano: yeah, that or, if it's a bug, means we need more info to act on it
<rekado>oof, fixing R reproducibility issues is hard :(
<catonano>civodul: thanks
<rekado>actually, fixing *any* reproducibility issues is hard, as a rule
<rekado>in R timestamps and mtimes end up in generated files, but the Makefiles are pretty hard to follow.
<civodul>ACTION sympathizes
<rekado>I just don’t know what rule executes a piece of code.
<rekado>it’s an R snippet that’s run at build time; if I remove handling of timestamps there I might actually break runtime code, so I’m trying to find a way to just reset the mtime of generated files right before the code is run.
<rekado>it’s an unfortunate downside to code reuse.
<rekado>Oh, sweet! —>
<rekado>could this make bootstrapping scala easy?
<bavier`>is the java-based scala difficult to bootstrap? I've probably missed the discussion.
<civodul>rekado: not as exciting as Guile 2.2 :-)
<civodul>though that fixes a serious shortcoming of Scala IMO
<wingo>civodul: how long will it take guix to switch to 2.2, do you think?
<wingo>i guess it can do either when you build from source
<civodul>wingo: i'm using it on 2.2 since a few days
<civodul>there are still a couple of test suite failures that i need to investigate
<civodul>but we're almost there
<civodul>as for using 2.2 in derivations, that could happen on the next core-updates
<civodul>it should be transparent
<civodul>switching packages to 2.2 by default will be easy, most of them already work with 2.2
<wingo>i will try to upload a 2.2.0 tarball somewhere tonight
<civodul>i'll have 'guix pack' things for you to test
<wingo>should give us enough time to try "guix pack" before blessing the tarball on thursday
<wingo>yeah great
<wingo>ACTION updating guix on laptop
<wingo>nss sure takes a long time to build
<wingo>or test, rather
<rekado>bavier`: the JVM-based Scala cannot be bootstrapped.
<rekado>to build a release from source you need to have a pre-built development version of Scala.
<rekado>this goes all the way back to the origins of Scala.
<snape>wingo: after nss, there is icecat...
<bavier`>rekado: I see.
<bavier`>so now we might be able to use the native compiler to bootstrap the jvm compiler
<rekado>bavier`: yes, that’s the plan.
<[df]_>dependencies on the java standard libs might be a problem
<rekado>[df]_: indeed.
<rekado>the native compiler comes with implementations for some of the libs, but I suspect it isn’t yet comprehensive enough.
<bavier`>and scala-native uses sbt for its build
<rekado>argh, really?
<bavier`>yup, I don't see any other build system files
<rekado>I find this sad: “*To contribute to this code, it is strictly forbidden to even look at the source code of the Oracle JDK or OpenJDK!*”
<rekado>“This is for license considerations: these JDKs are under a GPL-based license, which is not compatible with our BSD 3-clause license.”
<civodul>sad, indeed
<bavier`>why not prohibit contributors from reading any gpl'd code? /s
<civodul>why not ask them to live in a cave?
<rekado>hmm, the thing is written in Scala, so I think one would really need to have the JVM Scala first.
<ryanwatkins>I love how it is 'very important' :D
<wingo>snape: yay indeed, there is icecat :)
<ryanwatkins>but wnat about if they think about the gpl licence source code, I think it is high treason
<lfam>High treason is usually a capital offense. But, at least they'll give you a cup of Java first... ;)
<civodul>there'll be 2 repro build talks at LibrePlanet!
<civodul>pretty cool
<ryanwatkins>lfam: ;)
<lfam>ACTION basks in the thunderous laughter and applause
<civodul>wingo: so now you can do things like: "guix pack guile-next --symlink=/opt/gnu/bin/guile=bin/guile"
<civodul>let me know how it goes!
<efraim>wingo: guile-2.1.8 built in about 6 hours on my aarch64 board
<bavier`>rekado: I just found your writeup on bootstrapping haskell. good read.
<bavier`>rekado: btw, do you know whether the "logo" image on is truncated on the right, or is that how it's supposed to be?
<wingo>efraim: cool. hopefully with that recent gc patch that went in after 2.1.8, that time should go down too
<jmd`>It seems that bash or one of it's dependencies has recently changed in master. Shouldn't that have been pushed to staging/core-updates ?
<lfam>What do you see that makes it look like bash was changed?
<efraim>there was the CVE that required rebuilding an early bash version for grafting
<jmd`>efraim: Oh maybe it is only a graft.
***jmd` is now known as jmd
<jmd>What is the stdin of the guix daemon?
<jmd>It looks like the python entropy fix didn't work :(
<efraim>mesa and nss fixes for aarch64 pushed to core-updates!
<jmd>Where can I get a aarch64 machine to try it?
<efraim>i'm using the odroid-c2
<efraim>its about $50 for the board, just needs a microSD card, power from usb and an ethernet cord
<buenouanq>have you preordered your eoma68 libre tea boards yet #guix?
<buenouanq>first ever singleboard with an fsf ryf cert
<buenouanq>ships in may
<jmd>efraim: I've heard it'll fit the same enclosure as a RAspberry pi is that correct?
<efraim>dunno, I have mine just sitting out on top of an external harddrive
<efraim>i don't know if the ports would line up
<joshuaBPMan_>Hello, is there way I can make simple guile games? maybe something that'll run in the browser?
<bavier`>buenouanq: put in my order a long time ago
<buenouanq>I'm so excited for this :3
<buenouanq>joshuaBPMan_: watch the SICP lectures and get hacking
<joshuaBPMan_>thanks buenouanq
<buenouanq>browser things are going to be a level removed, but there is a guile web server
<joshuaBPMan_>guile web server? that sounds cool.
<buenouanq>you should maybe find a guile specific room for this though
<buenouanq>someone who can into services, am I getting the syntax on this right?
<buenouanq>And I can add multiple instances of this (with different configs, ports, etc) to my OS config yeah?
<efraim>i notice our ibus package doesn't have wayland support, or emoji support
<lfam>Hm, python@3.5.3 failed on core-updates for i686-linux:
<lfam>I'm going to try building it locally. Hopefully it's a transient error
<civodul>lfam: it definitely looks like a transient error
<civodul>those seem to happen frequently these days :-/
<civodul>not sure why
<lfam>Okay, I've restarted it on Hydra. I'm still building it here, too
<civodul>this build exited with code 100, which is normally for permanent build failures (per offload.scm)
<lfam>I can build python@3.5.5 on x86_64 for the i686-linux system
<civodul>sounds like we need to restart dependency-failed builds
<lfam>Well... doing. Hopefully the HTTP request goes through
<rekado>bavier`: I can’t reproduce the logo truncation error in icecat :(
<thomassgn>my laptop keeps crashing during some of the heavier builds due to overheating. I've added max-jobs=1 and cores=2. The laptop is an hp elitebook i7 4G ram and I have been running both nixos and gentoo (everything compiled) fine earlier. I find MCE messages in my logfiles, but can't make much out of them. Any tips for checking build daemons current settings or getting more info from MCE/Kernel?
<civodul>thomassgn: i have the same laptop running GuixSD but i haven't seen this
<civodul>are you using GuixSD?
<civodul>to check the daemon's settings, just just "ps aux | grep guix-daemon" and look at the command-line optinos
<lfam>optinos are the atomic configuration particles that compose the daemon's configuration files ;)
<luser1>civodul: I was the person who had the corrupted ssd yesterday you helped. Wanted to say thanks :)
<thomassgn>civodul: aye, guixSD. Will check, thanks
<thomassgn>so guix daemon starts 2 instances with 1 core each. as expected. Don't understand why this makes it overheat...
***the_ktosiek is now known as ktosiek
<civodul>luser1: good to know it works :-)
<civodul>thomassgn: BTW, i'm curious how you managed to install GuixSD on that laptop
<civodul>it requires UEFI, which is undocumented :-)
<thomassgn>civodul: mine is an old one (I think 6 or 7 years old by now). so UEFI is optional, I've turned it off. (:
<thomassgn>elitebook 2540p
<lfam>Does it crash if you just load the processors with useless work?
<lfam>Like, running `yes > /dev/null &` per-core?
<thomassgn>lfam: I'll try, let's hope not :P
<lfam>thomassgn: I guess there's no need to try if the MCE messages specify an overheating condition.
<lfam>Anyways, that's how I warm up cold laptops ;)
<civodul>thomassgn: lucky you ;-)
<thomassgn>lfam: haha, nice. I think it's overheating, but I haven't been able to pinpoint much, I don't even know what happens at boot in my logs so the messages from MCE may be unrelated. But I just started the MCE daemon manually (didn't look like it was running from 'sudo herd status'). Am retrying to build system; as there now is a new mcelog file in /var/log/ ....
<buenouanq>man, just as I mentioned the EOMA68, it gets pushed back a couple months (;-___-)
<buenouanq>efraim: I've yet to successfully get ibus working on GuixSD at all...
<buenouanq>can someone who knows about services verify that I'm doing this correctly
<lfam>buenouanq: Does it work when you try it?
<buenouanq>I haven't gotten that far yet - I'm really just asking about the service declaration syntax I think.
<buenouanq>the #:... "..."
<lfam>I haven't used that service but presumably you want those fields to have a value besides ""
<buenouanq>manual has some with and some without quotes, but can they all be quoted?
<buenouanq>yeah, just place holders
<buenouanq>I'll fill them in when I better know what I'm actually doing
<buenouanq>I think some of these too I might want to leave off and just let the postgres config deal with them
<thomassgn>heh, system just went down again, if anyones interested theres mcelog at
<lfam>Maybe you'll have to use --cores=1
<civodul>thomassgn: this is from dmesg?
<thomassgn>civodul: mcelog --daemon creates this file /var/log/mcelog
<buenouanq>So, I have a postgres service running, but this obviously didn't install all the tools (things every tutorial mentions) I need to access it from whatever user. In addition to the service, am I supposed to install the postgres package under said user?