<mik_>is it meant to be used with sudo? <mark_weaver>the exceptions are "guix system reconfigure", "guix system init", and guix-daemon itself, which should be run as root. <mark_weaver>to run those commands directly out of the source tree, you'd do: "sudo ./pre-inst-env guix-daemon ..." or "sudo ./pre-inst-env guix system ..." <NiAsterisk>GNUdists sounds akwardly close to nudists. not that I judge one or the other, it just sounds too similar <sprang>any guix-related stuff happening around libreplanet this weekend? <sprang>I'll be in Cambridge tomorrow morning <sprang>oh, yeah, I know about the session. I was just wondering if there were any informal gatherings <sprang>well, I'll be around tomorrow and Friday if anybody wants to grab a meal and/or a beer <sprang>flying out soon, I'll be on IRC tomorrow <kristofer>dumb question, but do I need to be root to reconfigure the system? <kristofer>any pointers on creating a system configuration with nginx-service? <lfam>I'm glad to hear the xfce4-power-manager is working for people <Jookia>Hmm. I have a derivation to build but I get errors when running "(run-with-store %store-monad (base-initrd '()))" <Jookia>Oh, I have to open a connection. My bad. <Jookia>Does anyone have an example of using guix build -e ? <mark_weaver>guix build -e '(@ (gnu packages commencement) gcc-toolchain-5)' <mark_weaver>to access a private variable (not exported), use @@ instead of @ <mark_weaver>btw, this syntax is not specific to 'guix build', it is guile syntax for accessing a variable in another module without relying on importing it first <mark_weaver>any scheme expression that evaluates to a package object can go there <lfam>So, is anybody fetching substitutes over https? I still can't get that to work <dvn>new to guix. using it on a foreign os. I installed glibc locales, and set the path but I'm still getting this error: warning: failed to install locale: Invalid argument <Digit>sat looking at my unupdated gentoo, realising it's only going to get scarier to update the longer i leave it... and gets me thinking of guix again... and how safe one feels when performing system upgrades with the knowledge of clean roll-back. <jmarciano>dvn: if you as user get the error, you can put this in environment: export GUIX_LOCPATH=$HOME/.guix-profile/lib/locale <jmarciano>if one of guixbuilderXX get the error, you can ignore. I had the same problem. <Jookia>mpv and youtube-dl are pretty fantastic tools <roelj>Has anyone else run into "error: libgcrypt version mismatch" when trying to run the guix-daemon? I cannot seem to solve this.. <efraim>are you on guixsd or on a foreign distro? <roelj>A foreign distro, Fedora 23. <roelj>It has version 1.6.4 of libgcrypt, and Guix has 1.6.5. <efraim>so it could be referencing the wrong one for the daemon <roelj>I tried compiling guix with --with-libgcrypt-dir=/gnu/store/...-libgcrypt-1.6.5/ as well as with --with-libgcrypt-dir=/lib64 <efraim>I actually have libgcrypt in my profile to ease some pains I had <roelj>Both resulted in the same error. <efraim>from config.status its referencing the libgcrypt in my profile <roelj>I checked with ldd whether it really linked to the store version or the system version, and that seemed OK. <roelj>Now this is weird.. I don't have a directory /var/guix/ anymore. <roelj>Is it normal to have it in /usr/local/var/guix/ instead, when compiled from git? <efraim>it depends if you passed --localstatedir=var to configure <efraim>good call on the /var and not just var <roelj>So I've run "./configure --with-libgcrypt-libdir=/gnu/store/4iqlcz069gw3qnx2iry9yifwg9mc5k51-libgcrypt-1.6.5/lib/" <roelj>Then make and sudo make install <roelj>config.status says: S["LIBGCRYPT_LIBDIR"]="/gnu/store/4iqlcz069gw3qnx2iry9yifwg9mc5k51-libgcrypt-1.6.5/lib/" <roelj>Yet, ldd /usr/local/bin/guix-daemon says: libgcrypt.so.20 => /lib64/libgcrypt.so.20 <roelj>What should I configure differently, so that guix-daemon uses the libgcrypt from the store instead? <alezost>roelj: try "./configure --localstatedir=/var --with-libgcrypt-prefix=/gnu/store/4iqlcz069gw3qnx2iry9yifwg9mc5k51-libgcrypt-1.6.5" <alezost>although I don't know why you have this problem, I don't specify libgcrypt explicitly ***kelsoo1 is now known as kelsoo
<rekado>roelj: I never run "sudo make install". <rekado>and I also didn't need to pass "--with-libgcrypt" <efraim>i did the binary install on debian <efraim>but did ./configure --localstatedir=/var with my git repo <efraim>vlc fail: configure: error: "You cannot build VLC with Qt-5.5.0. You need to backport I78ef29975181ee22429c9bd4b11d96d9e68b7a9c" <wingo>i think i might have fixed that gnome suspend-forever bug. fingers crossed <efraim>2.2.2 will have to wait for us to package qt-5.6 <jmarciano>somebody knows if emacs-guile can be extended with guile? <wingo>yay, fixed it! anyone using gnome and having the suspend problem, or who are not using gnome because of it, you can pull and things will work. <wingo>also last night i was checking something <wingo>and while the brightness keys and other media keys don't seem to be working yet <wingo>you can adjust screen brightness using the control panel <wingo>which means the pkexec / elogind / pam / sudo dance is all working as it should, it seems <janneke>wingo: do we have a list/checklist for potential guixsd gnome users? <wingo>janneke: users or developers? :) <wingo>there are things to do but i don't know of a list <wingo>adding gnome-backgrounds to the metapackage is an easy one <janneke>hints like these can help people to switch <wingo>ah, you mean for users then ;) <wingo>it's fairly straightforward i think. <wingo>you just cons (gnome-desktop-service) onto your %desktop-services <wingo>and then you can log in in gnome <janneke>yes--well developers could pick stuff that is problematic and fix it <wingo>and things work as well as they do in xfce, more or less. except that gnome-shell wants to talk to networkmanager instead of wicd <wingo>i don't know what the right solution is there -- do we make gnome-shell talk to wicd? or do we switch to networkmanager? and if so, how? <janneke>if i have networkmanager/wifi and openvpn, i'm good <wingo>janneke: what do you do now? something wicd-specific? <janneke>i tried gnome 6weeks ago and reverted to minimal debian+guix <wingo>ah, so you are not on guixsd yet then <janneke>i have migrated all my work to guix; "only" use debian+gnome-shell <janneke>no emacs/guile/gcc/bison/whatever from debian only network+gnome <jmarciano>janneke: thanks yes. I tried it, now I know it can be used like elisp. Only it is slow yet. <janneke>jmarciano: great...i was hoping guile would speed things up... <rekado>wingo: yay! So, this might mean that the shutdown/reboot bug in Xfce would be fixed too. <jmarciano>I think it will be, only at this moment is slow. <janneke>wingo: i've resurrected my gnu standard error-messages patch for guile <janneke>but i cannot get guile master's debugger to ,break or ,break-at? <wingo>i don't know if xfce talks to logind by default <wingo>janneke: that could be a bug, i rarely test the debugger <wingo>a report for how to reproduce the bug would be welcome :) <jmarciano>may I ask... if developers use Amazon AWS? Or is Amazon rather rejected? <jmarciano>I mean, if Amazon is rejected due to freedom principles, then packages relating to S3, <jmarciano>Route53, etc. cannot be included. Like some file systems on S3, and such <jmarciano>like s3-fuse file system for example. I just wonder. <jmarciano>openstack is I guess alternative to Amazon tools, when GuixSD starts running, openstack could be good solution. <jmarciano>I mean when GuixSD starts invading the Internet. <efraim>it also works with amazon-compatable services <jmarciano>and is Amazon AWS rejected for some principle of freedom? <jmarciano>of course they don't run services on free software <jmarciano>GuixSD could be new method of quick startup of web services. <wingo>one day i want to be able to create vm's to run on virtual hosting providers, using guix. <jmarciano>Something like Docker... can be implemented through GuixSD <wingo>so to that extent i think libraries that integrate with various hosting provider's facilities are welcome. dunno though, there are many opinions :) <jmarciano>I guess hosting providers shall integrate with GuixSD, not other way around. <wingo>this discussion is a bit abstract :) <dptd>Hello there! Anyone have a minute to help newbie with GuixSD installation? <wingo>i have something to do now and so i can't, but i wish you luck dptd and that you find someone to help you! <wingo>also, please keep notes as to what is a problem so that we can improve the docs :) <dptd>I have downloaded USB installation image. Converted it to the VirtualBox VDI. Mounted and booted. I follow the official instructions (started ethernet, started cow-store, modified sample system config and run guix system init). However tests/store tests keeps failing. I cannot find logs anywhere. :\\ <dptd>wingo: Thanks. Actually when I will be done I will probably post somewhere step-by-step instructions how to install GuixSD in VirtualBox. <Jookia>Is it blasphemy to use Guix to build an initramfs that doesn't use Guile? <Jookia>civodul: I'm trying to build something that's supposed to fit under 3M, but Guile alone in an initramfs goes waay past that size :( <civodul>right, 7MiB for /run/current-system/initrd here <civodul>thing is, i'm not sure you can do much better without Guile <civodul>because you need klibc, Bash, etc. etc. <Jookia>well, i could try and get it down to busybox-land <civodul>yeah, but i suspect we could make the same space savings with Guile <dptd>Anyone knows where I can find test logs from guix system init? Store tests failed and I cannot find why... <Jookia>I could aim for 8MB instead but I'd still have to write a small networking service <Jookia>Yes, I want to be able to get ethernet and wi-fi in this initramfs <civodul>in the current framework, you'd "just" need to add the relevant kernel modules, and give an expression that sets things up <Jookia>Probably not- but won't I need something like wpa_supplicant? <civodul>rekado: r-dnacopy-1.44.0 has a wrong 'home-page' <Jookia>civodul: Is there any low hanging fruit that could be shaved off guile-static-stripped? <civodul>Jookia: no idea, i haven't checked, but i know i haven't really tried to make it small <Jookia>I wonder if replacing glibc could help. Looks like I have some fun things to test <civodul>grr hydra is dying and the mirror doesn't have everything in cache <dptd>Anyone tried to install GuixSD in VirtualBox? <rekado>dptd: someone did and they kept notes, but I cannot find them right now. <Jookia>dptd: Have you tried using qemu instead? <rekado>civodul: oh, you're right about r-dnacopy. Sorry! Slipped by the review. <dptd>rekado: I found one or two pages and I did exactly what was there. Still no luck though. <rekado>dptd: which guide did you follow? <dptd>Jookia: No, I did not. Actually I just went to the project website, never heard about it before. <dptd>And official Guix manual of course. <dptd>The problem is that after guix system init <blabla> store tests fails. I cannot find any logs anywhere and I am stuck. <dptd>Jookia: I will check it today. How it compares to VirtualBox? Have you used both? <Jookia>dptd: I mainly just use qemu, it's just as fast :) <rekado>a graphical frontend for qemu is provided by virt-manager. <rekado>roelj: I restarted the guix-daemon and now also get "error: libgcrypt version mismatch" :) <dptd>I see. However my issue is probably not related to VirtualBox itself. No idea why this test fails. I cannot find any logs anywhere. Should it be that way? <Jookia>should you be running tests? aren't there substitutes? <dptd>I believe tests are run automatically after system init. <dptd>After preparing the partitions and config file all I have done was to run guix system init /path/to/config <rekado>roelj: "make clean" and "make" fixed it for me. <dptd>Jookia: Yeap, before I have run system init. <mkoenig>substitute download is terribly slow (1% per hour) and "--no-substitutes" also fails (downloading pkg-config) :/ <rekado>hydra doesn't work right now and I don't really want to wait for substitutes to create an environment, but even with "--no-substitutes" it seems to wait for hydra. <rekado>when I switch to mirror.guixsd.org it wants to download the bootstrap binaries --- even though I have them already. <rekado>"guix environment --no-substitutes --pure -l guix.scm" <rekado>unpacking bootstrap Guile to '/gnu/store/aiz8db2gni401wc9fgidmcggxyb1czis-guile-bootstrap-2.0'... <mkoenig>civodul: seems to be working, but now download speed is 1% per 5 minutes <rekado>without --no-substitutes (and using mirror.guixsd.org) it does this: "Downloading cib4xy...-diffutils-boot0-3.3 (368KiB installed)..." and then fails with 504. <civodul>rekado: i'm currently tuning mirror.guixsd.org to be less tolerant of delays on hydra.gnu.org <jmarciano>I just wonder, as I have realized, nobody I know uses KDE... <mkoenig>jmarciano: not together with guixsd but yes <jmarciano>mkoenig: aha yes, because of German connection? <jmarciano>yes, sure. I tried it many times, I was thinking for it to become my default. But that was back in 1999-2001 <jmarciano>After testing the speed and all usabiliy, I stayed with simple window managers. <mkoenig>i prefer xmonad but i don't want to use it on multimedia systems <NiAsterisk>do we even have ruby/rails ? i send an email on the recent CVEs of it as I had 3 results for guix package -s rails <jmarciano>aha tiling window manager, well I tried it back in time. But if I tile, my windows would get smaller and smaller. <NiAsterisk>I used kde with kde3 or 2 or somewhere around year 99/2000 and now I use kde plasma 5.5 when I can or awesome-wm.. <rekado>NiAsterisk: we have ruby but no rails. <jmarciano>maybe guile-wm could become something like that... <NiAsterisk>idk, I just use gentoo's kde overlay. 5.5.5 is the newest plasma <NiAsterisk>what else do I need for gnome-desktop-services to be valid? <NiAsterisk>"In procedure module-lookup: Unbound variable: gnome-desktop-services" <mkoenig>i wonder how you can offer gnome without systemd <rekado>you need to load the module in which it is defined. <NiAsterisk>which is (use-package-modules gnome xfce wicd avahi xorg ratpoison certs emacs tor) where I assumed gnome is the right name <rekado>it's a part that has been carved out from systemd (mostly through the efforts of wingo) <NiAsterisk>oh, it did not work because i wrote services and not service <janneke>mkoenig2: it has a tiling.scm module... <janneke>i succeeded in running guile-wm for the first time on guixsd <janneke>but it crashes on my and i haven't found a way to make a bug report <janneke>i have high hopes for guile-wm, it needs some love... <mkoenig2>what's the best way to install the proprietary nvidia driver <Jookia>mkoenig2: you'd have to find someone's branch of guix that allows it <Jookia>yeah, all GPU vendors are unfirnedly to OSS hardware <efraim>mkoenig2: AFAIK linux-libre actively blocks loading non-free firmware <NiAsterisk>but, not advocating this, you can rewrite the linux-libre package according to your needs. <jmarciano>efraim: good, when I found actively blocking, I have installed it. <jmarciano>mkoenig2: I guess GuixSD is free software distribution. Non-free will never belong to GuixSD neither will be sponsored, or otherwise supported or referenced to. <redshelled>Hey everyone, I'm getting started with GuixSD, but I can seem to get it to install. libarchive always fails one test. Any ideas? <redshelled>is anyone else able to build libarchive on amd64? <redshelled>I did last night, would it be worth trying to pull again? <redshelled>for the sake of being useful, the test that is failing is test_format_newc <mkoenig2>efraim: oh, didn't know. this will be a problem.. ***mkoenig2 is now known as mkoenig
<dptd>redshelled: I have also some strange problems. I am still trying to install GuixSD on the VirtualBox. After guix system init <blabla> test/store.scm is failing. <davexunit>dptd: I don't understand. running 'guix system init' doesn't run the Guix test suite. <davexunit>and even so, a test failure isn't necessarily indicative of a broken installation because the tests are run in isolation from the main system. <dptd>davexunit: I don't understand either. However after running this command I had some tests output printed out. <mkoenig>does guix install required nouveau automatically? <davexunit>dptd: sorry, we need more information in order to debug this. <davexunit>what is your OS config? are you using substitutes? <davexunit>have you run 'guix pull' before installation? <dptd>Okay, let me write it step-by-step. :) <davexunit>you must run 'guix pull' beforehand, otherwise you *will not* receive pre-built binaries <davexunit>because 0.9.0's binaries have long since been garbage collected. <davexunit>our next release will make it more clear that you need to run 'guix pull' immediately <dptd>I have done everything according to the official instructions. I am trying to install GuixSD in VirtualBox. I downloaded newest usb installation image, converted it to the VirtualBox VDI and mounted. I successfuly booted this image. Created partitions (boot and root) and labeled root. Mounted it, started cow-store and copied sample desktop config to /mnt/etc. Edited the sdX in the config. After that I actually have <dptd>run guix pull and after that guix init /mnt/etc/config.scm /mnt --fallback. <dptd>And - as I wrote before - it failed. Last prints in buffor were tests run summary (2 skipped, 1 failed). <dptd>I will. However it will take a while to go back where I was yesterday. I will let you know when I will be done. <davexunit>dptd: did you take note of what guix said it was going to download vs. what it was going to build? <davexunit>guix is building guix instead of downloading it <davexunit>and the test suite fails on a virtual box machine <dptd>Okay, will check without --fallback. Makes sense what you are writing. <davexunit>so there's a bug report there if you can run 'guix build -K guix' and capture the test suite log <dptd>However at the very beginning I was not using this command. During the debugging process I read (or someone told me to) to try with this flag. ;) <davexunit>note that we do not condone or support virtualbox specfically because it requires nonfree software. <davexunit>but I suspect the test suite failure may be a bigger issue. <mkoenig>guix pull always fails, but guix pull --bootstrap seems to work!? <dptd>davexunit: Roger that, thanks for the help. <Jookia>Oh, I knew there was a few btrfs bugs a while ago <davexunit>wingo: so happy to hear about the infinite suspend loop being fixed! reconfiguring my laptop now. <mkoenig>still fetching.. but now with a realistic speed <davexunit>mkoenig: hydra is frequently under heavy load <mkoenig>this is not a valid option for guix pull <davexunit>mkoenig: oh sorry, I thought you were doing something else. <ajgrf>how do i get the correct hash for git-fetch? <dptd>davexunit: Should I also use substitute-urls during system init? Currently I see some logs like "system is somehow slow" related to hydra. <civodul>so, with my latest nginx changes, hydra.gnu.org will *feel* faster <civodul>that is, hydra.gnu.org and mirror.guixsd.org will give up quickly instead of waiting forever <civodul>so effectively, if hydra.gnu.org is overloaded, you'll get fewer substitutes <civodul>it turn, that will probably help reduce load on hydra.gnu.org <civodul>yeah, i think it will make the UX less bad <civodul>assuming it feels better to build things locally than to wait forever <davexunit>I'm currently building an awful lot of GNOME <kristofer>it's difficult to walk away from an installation because of the 502 errors <davexunit>"I really like Guix's ideas, but I think that project made a very unwise choice when it picked Scheme rather than Common Lisp." <jmarciano>I guess Guile is used because of future, to be very extensible <redshelled>plus there is an advantage that scheme is taught at a lot of uni's <jmarciano>There is application on Android FOSDEM Companion, one can many new talks about Guix. Get it on www.f-droid.org <paroneayea>but if you can use a lisp, that's better than most of the world still :) <jmarciano>"Your Distro is a Scheme Library"... thanks Ludovic <davexunit>paroneayea: I'm currently writing "Scheme Considered Unwise" to be published by the ACM <piyo>Countdown to april fools in 3 2 1... <jmarciano>GuixSD will blur distinction between users and developers. It is teaching users to become developers. <civodul>davexunit: that's because you got an awful lot of 504/502 without noticing :-) <davexunit>I think it may still be in the grafting phase <davexunit>because it's just printing 'updating list of substitutes' a million times :) <davexunit>oh except I still haven't figured out why 'guix environment' breaks SIGTERM <davexunit>I need to figure that out... probably something really silly. <jmarciano>I repeat it from Fosdem video, I tried: ,use (gnu): ERROR: no code for module (gnu), and I have guix. What to do? <davexunit>you need to make sure that the guix modules are on GUILE_LOAD_PATH <davexunit>I have no idea how you have installed Guix so I can't give specific advice <jmarciano>explain what you mean above with (and GUILE.. ? <davexunit>jmarciano: just because those variables are defined doesn't mean that they are correct <davexunit>otherwise ,use (gnu) would work fine in your REPL <davexunit>paroneayea: maybe "considered unwise" can be the new "considered harmful" <jmarciano>I have it, but not finding, can you check the name of gnu file? <jmarciano>maybe I can find it and load to path (gnu), but I need some clue <davexunit>the first thing you should do is identify where the Guile modules are for Guix <jmarciano>/home/data1/protected/.guix-profile/share/guile/site/2.0/ that is LOAD PATH <jmarciano>I see shepherd, I can ,use (shepherd), but I don't see inside gnu <jmarciano>and I have gnu somewhere in /gnu/store/...guix.../site/gnu.scm like that <jmarciano>I wonder why that... I could try loading with hadn <davexunit>it's because you don't have guix installed to your profile <jmarciano>hmm, it seems like it magically disappeared... <davexunit>'guix package -i guix' will install a version of guix there, but it won't be the same version you are using. <davexunit>if you run 'guix pull', you could add $HOME/.config/guix/latest to GUILE_LOAD_PATH and GUILE_LOAD_COMPILED_PATH <davexunit>then you'll have the latest guix modules available <jmarciano>aha, let me try that too (I did it few times) <jmarciano>504, "Gateway Time-out" on mirror.guixsd.org... but OK, let me try that <jmarciano>and I have now latest, but no, it does not find it... <davexunit>we *really* need details here otherwise we can only guess and keep guessing wrong. <jmarciano>ok I removed double path definition... now I get better error: ERROR: no code for module (guix config) <jmarciano>GUILE_LOAD_COMPILED_PATH=/home/data1/protected/.guix-profile/lib/guile/2.0/ccache:/home/data1/protected/.guix-profile/share/guile/site/2.0:/home/data1/protected/.config/guix/latest <jmarciano>GUILE_LOAD_PATH=/home/data1/protected/.guix-profile/share/guile/site/2.0:/home/data1/protected/.config/guix/latest <jmarciano>and in latest, I can see: guix.scm guix directory, etc. gnu.scm, so it loads. but maybe "config" is not loading <davexunit>does /home/data1/protected/.config/guix/latest/guix/config.scm exist? <jmarciano>inside is definition for config package, but why is extension .in ? <davexunit>jmarciano: because config.scm is a programmatically generated file <davexunit>it's generated as part of the configure phase of the build process <davexunit>civodul: does 'guix pull' do something special such that guix/config.scm isn't generated? <davexunit>jmarciano: that's unlikely to change anything <dptd>davexunit: Okay, I am done. This time it failed on "guix substitute: error: host name look up error Name or service not known". It is the result of running system init with --substitute-urls set to mirror. <Jookia>dptd: are you having internet issues? <Jookia>i recall some VMs have internet issues <dptd>No, I keep pinging google. Solid ~20ms. <davexunit>dptd: what is the exact command you are running? <davexunit>dptd: did this failure happen immediately or did it successfully download some things? <ajgrf>if i'm packaging something from git instead of a tarball, how do i know which hash to put in the package definition? <Jookia>ajgrf: usually i just build then get the new hash from the failure <davexunit>dptd: if it was a spontaneous error, all I can recommend is to try again. our servers are simply overloaded, unfortunately, and it sucks. we're actively working on getting new servers but it's going to us some time. <davexunit>ajgrf: the correct way to calculate the hash of a git repo is to create a clone of that repo that you can safely throw away, checkout the commit you are interested in, *delete* the .git directory, and run 'guix hash -r .' <rain1>is there a reason to do it that harder more complex way? <dptd>davexunit: Yeah, I can see that. People are writing that here all the time. Not a problem though, thanks for the help. Answering your question - it is spontaneous error. Some staff is being successfully downloaded. <jmarciano>after guix pull, no change... ERROR: no code for module (guix config) <dptd>I will keep trying though. Love lisp and really want to contribute. <davexunit>dptd: :/ sorry about that. hang tight everyone! this UX really sucks right now, but it *WILL* be fixed. <rain1>davexunit, why "don't recommend this."? <davexunit>don't just assume that the thing that was downloaded was correct <davexunit>I don't want to recommend "easy" things that lead to bad habits. <jmarciano>exactly, that is opposed to why guix is there <davexunit>the proper easy way would be to extend 'guix download' to handle git repos. <rain1>well don't but you shouldn't tell other people to stop helping people do their task easily <jmarciano>rain1: such things oppose proper usage of package manager.. right? <rain1>I don't know what you mean by thta <dptd>kristofer: Random packages. Each run stops on different one. <jmarciano>that is like when you do on Debian apt-get package, the PGP keys are not trusted, so what do you do? Proceed for insecurity? <Jookia>fwiw i didn't know that was how you did it because it wasn't documented in my guix manual, or maybe i missed it <davexunit>I care about promoting proper packaging practices. I also care about building tools to make that process the easiest possible path. <rain1>davexunit, your instructions didn't mention auditing the source code <rain1>there's no difference in security it's just extra work <davexunit>but I know that this isn't the first time that this topic has been brought up. <mkoenig>gateway tome-out; package "pciutils" <pizzaiolo>pinging here, also getting 505 timeout error when connecting to guixsd.org or hydra.gnu.org <civodul>pizzaiolo: when installing substitutes? <Jookia>Is there a 'smart' way of making a closure of a staticly linked program not drag in glibc when running "populate-store", without just creating a derivation that copies glibc? <civodul>you need to find out why the statically-linked thing drags glibc <civodul>Jookia: it could be because it actually uses it, or because of debugging symbols, etc. <pizzaiolo>civodul: I'm trying guix package -u and guix reconfigure, both are throwing the 505 error <civodul>while downloading /gnu/store/foobar, right? <pizzaiolo>hydra.gnu.org seems to be working momentarily <Jookia>civodul: Is there a graph command I could use? I'm also finding wrapping my package in "glibc-for-bootstrap" i causing two glibcs to be copied too <civodul>pizzaiolo: thanks; there was a cache misconfigurations so for a while 504s were getting cached for one hour on guixsd.org <civodul>so i think this particular item will remain 504 on mirror.guixsd.org for a few more minutes... <Jookia>civodul: I see. Is there documentation or code on how this is determined? <civodul>Jookia: references are determined through "conservative scanning", aka. 'grep -r' <davexunit>civodul: I think we scared off that one person at my talk in January that asked about the conservative scanning on the mailing list. <davexunit>he proposed that packages should explicitly encode their runtime dependencies. <davexunit>wingo: I've managed to crash gnome-shell by activating the "hot corner" and attempting to drag a window to a new workspace <davexunit>try to move a window once, some weird graphical stuff will happen and it will kick you out of that overlay menu <davexunit>try a second time and it will crash with the "Oh no! Something has gone wrong." screen and you have to kill X <davexunit>this is one of those "I have no idea where to even start debugging" things <wingo>it's not elogind at least ^-^ <Jookia>I wish I wasn't such a newb when it came to object code <wingo>davexunit: i think a start would be getting a .xsession-errors file going <wingo>right now we don't have one afaiu <davexunit>wingo: ooh but maybe this is! ;) asking GNOME to shut down doesn't work. <davexunit>wingo: are we missing a service that writes that log file? <wingo>there are a bunch of errors on tty1 it seems <wingo>i guess logs are going there <wingo>that is a shepherd problem afaiu <wingo>so i think the first thing to fix is that gnome-session's warning that says WARNING: Using null backend for session tracking <wingo>probably gnome-session needs to be compiled with logind support <wingo>needs same s/// over it as polkit, to change some instances of systemd to elogind <Digit>ACTION notices, besides the guix talk schduled at libreplanet, there's also a debian-centric "Beyond reproducible builds" <wingo>and needs removal of integration with systemd's journal <wingo>afaiu that is the extent to which systemd has percolated into gnome-session <Digit>yup, the 10am session block, on sun. you're in luck <Jookia>It seems no amount of stripping will stop glibc paths being put in busybox's program <mkoenig>do u use the original hydra by nix? or a modified version <davexunit>civodul: the mirror for /gnu/store/cnhqk5cwyrfaq7py2l1sq2ddcxxjh7x6-gcc-cross-boot0-4.9.3-lib may be corrupt on mirror.guixsd.org <davexunit>I'm getting a consistent failure trying to 'guix environment --ad-hoc gcc-toolchain' using that mirror <wingo>davexunit: i'm on gnome-session fwiw <wingo>davexunit: just sent gnome-session change to list; look ok to you? <NiAsterisk>not related to your comment. but I just launched it. cool <wingo>haven't had a chance to try it properly yet tho <wingo>rebuilding the system, i mean <wingo>will push once i give it a go :) <NiAsterisk>some akward settings like you described, screen brightness needs root pw, etc. and I had to start rxvt-unicode through some other terminal application <wingo>for me screen brightness does not need root pw NiAsterisk <wingo>i assume you installed gnome-desktop-service? <wingo>i woudl remove gnome and xfce packages from your packages list and instead add (gnome-desktop-service) and (xfce-desktop-service) to your services <NiAsterisk>errors are thrown this way, I have not logged them <NiAsterisk>rephrasing: errors appear when I have them in packages list and add those two to the services. <NiAsterisk>mkay. I just got spam from a school district in downingtown. <jmarciano>NiAsterisk: how do you know it is school district? <davexunit>ACTION found a problem with 'guix environment --container --network' on ubuntu (possibly other distros) <jmarciano>you can find last "received" IP address, and whois IP to find where to complain <NiAsterisk>because it goes back to www.dasd.org/Domain/1414 <jmarciano>if you see domain name, it need not be true, look for the IP before it came to your server <NiAsterisk>maybe I find an admin / staff email address. not that I care much, but maybe it helps them to fix it. <wingo>seems gnome-shell needs gdk-pixbuf+svg <redshelled>is there a reason running guix system init would be trying to build guix and not using the substitute? <wingo>oh god it's gtk+ that needs gdk-pixbuf+svg :P <NiAsterisk>wingo: no, with only the gnome package I get popups of Administrator Authentication Dialogues each time I change the brightness, but I can just ESC that <wingo>NiAsterisk: I am getting that too fwiw <redshelled>dptd: did you fix the issue of the test failing on store.scm <NiAsterisk>also: Mar 17 17:49:51 localhost gnome-keyring-daemon[452]: Unsupported or unknown SSH key algorithm: ssh-ed25519 <NiAsterisk>no idea how keyring daemon is build or if this is intended <wingo>i had some old gnomey packages installed in my profile <wingo>after having removed those, my backlight works and i can power off from the menu <rain1>NiAsterisk, im using ed25519 and its good but i don't think ihave keyring daemon <wingo>first time since i've had this computer that the backlight keys work :-) <NiAsterisk>a .desktop or whatever for xfce, gnome and others to "see" rxvt-unicode would be nice. <davexunit>NiAsterisk: icecat also is in need of a .desktop file <wingo>i guess we need to add an accountservice or something <wingo>i.e. a service in addition to the package <wingo>i seem to have determined though that you shouldn't have "system"-type things in your user profile <wingo>at least, i believe that was the source of some problems i was having <wingo>davexunit: so when i tried to drag something from the gnome shell activities view just now <wingo>it did kill gnome-shell but it restarted fine <wingo>i wonder if having made gnome-session have a non-null backend "fixed" that too <wingo>and the error that causes it to fail is that it can't load close-window.svg <davexunit>wingo: gnome-shell crashes still if you try to drag the window a second time. <davexunit>the close button icon is definitely not showing up so perhaps that is part of the issue <NiAsterisk>this sucks much battery... I wonder if this battery is broken. I can spot nothing in powertop <efraim>NiAsterisk: Debian has a package called laptop-mode-tools <efraim>and a couple of others it looks like <dptd>Strange. I am during the guix system init and after transferring python from hydra it is stuck with such message. "killing process 20719" and the line below "killing process 20719: no such process". <dptd>I am waiting more than 10 minutes now - nothing changes. Anyone have seen something like that before? <efraim>I've seen it occasionally in a terminal window where I _wasnt_ building software <efraim>I ignored it and waited to see if anything broke but it seemed ok <efraim>so it's likely a bug but not a fatal one <dptd>But my installation process is stuck now. Not sure if it will progress any further right now. <raphmaster>I have seen it me too after or during the download of python-2 <raphmaster>rerun guix system init /mnt/etc/config.scm /mnt and it always stop there by killing process <NiAsterisk>efraim: yes, but with GuixSD there's no such thing (yet) <davexunit>another triumph for 'guix environment' today <davexunit>not a huge deal, but I was able to make a guix environment container on my ubuntu machine at work that can build and run one of our web applications <davexunit>it's a good interim solution while we don't have all the *hundreds* of gems that this application needs <davexunit>I just include ruby, bundler, a gcc toolchain, and the C libs I need <davexunit>I'm quite happy. this would *not* have worked without containers. <davexunit>because guix libs and ubuntu libs would have mixed and blown up <davexunit>and for the folks complaining about not having /usr/bin/env: I just run 'mkdir -p /usr/bin && ln -s $(which env) /usr/bin/env' in my container <davexunit>I have a patch or two that I want to make to 'guix environment' now to improve things further. <davexunit> guix environment --container --network --ad-hoc \\ <davexunit> --share=$GEM_HOME --share=$HOME/.ssh --share=/run/mysqld \\ <davexunit> ruby bundler mysql imagemagick libxml2 libxslt gcc-toolchain \\ <davexunit> make git coreutils openssh libffi pkg-config which sed gawk openssl <efraim>shouldn't it be echo "TAKE THAT DOCKER" <davexunit>guix environment --container -- echo TAKE THAT DOCKER <rain1>I need to rememeber these commands <davexunit>I think this is a good example of a middle-ground state with guix. it doesn't *have* to be all-or-nothing. <NiAsterisk>do we have an example.scm for a system with luks encryption? <rain1>NiAsterisk, for encrypted /home only, I think its not perfect for encrypted root yet <rain1>it might already be working well on libreboot? I'm not sure <rain1>there was an issue about secret derivations <rain1>I would like to understand it better <rain1>was the problem that with the current setup initrd would be readable by everyone? and that lets something .. bad happen? <mark_weaver>on a libreboot machine, you can have GuixSD with a *fully* encrypted disk, including /boot <mark_weaver>rain1: that's only an issue if you try to put the encryption key in the grub.cfg, to avoid having to type in the password <rain1>so putting a key on a separate USB should work well right now <rain1>maybe I should try that in a VM <rain1>and if its a keyfile rather that password entry it cant ask you twice :) <mark_weaver>normally with that setup, you need to type in your disk password twice: once for grub and once for linux <mark_weaver>hmm, libotr-4.1.1 is consistently failing one of its tests on my yeeloong <tg>fbut why would anyone put their encryption keys in grubcfg? <mark_weaver>tg: actually, I think it was probably the initramfs, not the grub.cfg <mark_weaver>but there is one reason: if the purpose of the encryption is to protect against a compromised HDD firmware <mark_weaver>but setting that aside: on a Libreboot/GuixSD fully-encrypted disk setup, since the initramfs is stored on the encrypted root partition, in theory it could contain the disk key to avoid typing the password a second time. <mark_weaver>however, on GuixSD the initramfs is in /gnu/store, and everything there is world-readable <mark_weaver>Guix does not currently have any provision for secret data within the stuff that it builds. <jmarciano>secret data could be on the USB, or it should be left to user how and from where to slurp secret data. <jmarciano>it would be insecure to fix it in GuixSD configurations. <jmarciano>my password is sha256 password from a plain text password... I like it that way. <rain1>jmarciano, how do you input it? <rain1>it must be too hard to type it? <redshelled>Is there a way to force a package to install with the substitution and not try to compile from source <redshelled>or maybe a way to delete a previously downloaded source so that I can install guix and not try to compile it during install <jmarciano>but maybe I input it on X, that is why is easy for me. <jmarciano>I do like: echo (with or without -n, you must remember it) PASSWORD | sha256sum <jmarciano>but all in one line, few commands, then: cryptsetup create protected /dev/sdX <jmarciano>on the other hand, the backup disk, has password on my mounted encrypted partition, as file. So that is solved. <jmarciano>there are various ways, look: man cryptsetup <jmarciano>I did not try, but I guess, that something like: echo | something | sha256sum | cryptsetup --key-file=- should work. Maybe gawk is required too. <jmarciano>if those tools are available in guixSD start <jmarciano>look --hash option, I guess that makes it more secure, you enter wee, and you get long hash as passwor <rain1>that doesn't increase keyspace though <jmarciano>I have no idea what is "keyspace". I have the space, I put key inside. That? <rain1>if there are 100 small passwords [keyspace=100] <rain1>if you hash them all, its true you have big complicated hashes <rain1>but there's still only 100 of them <jmarciano>the point is: how is somebody going to know how did you hash it yourself? <jmarciano>not by tool, but yourself, before the tool is invoked. <rain1>rather than say 2^256, if you chose a random hash instead of basing it on a password <rain1>yes thats true, if they don't know that... :) <jmarciano>random hash is for /tmp, I don't need to know it <jmarciano>ok I give you beer if you decrypt my hard disk <jmarciano>sometimes I use md5, sometimes sha512, 224, etc. I forget it. But I try it multiple times, and I get it. <jmarciano>or I forget, did I use echo or echo -n, newline or not... <jmarciano>I am not sure if keeping a key on USB stick is "secure". <rain1>it has different positives and negatives <jmarciano>when bank laptop is stolen... international spying agencies open their eyes and ears. <jmarciano>If I am asked, /home encryption should be by default. Secondary option would be "No encryption". <jmarciano>sensitive data in /var - should not be there. Databases shall keep data in /home/var or something <jmarciano>idea: cryptsetup can be easily used with files, to be read. Keep temporary file on memory, if there is such in Guix. Like /run/shm <NiAsterisk>that's a good idea, however it is highly discussable. if it were in email/bm thread form I would go into detail with my positive/negative opinions on this. <NiAsterisk>jmarciano: one option later could be to ask users in a non-specified graphical setup assistant dialogue to set encryption options, like some *buntu do now or whatever distros these days do.. I haven't looked outside of Tails, Gentoo and others the last years on desktop systems. some started including encryption I heard. <jmarciano>I guess extX file system was not made to be slow, but rather fast. Encryption slows things down. <jmarciano>/tmp, swap, /var, /home - sensitive data should be. <jmarciano>and encryption shall be set up quickly, even before the X <Gamayun>They probably all have encryption. But perhaps not as an easy option in a graphical installer. <jmarciano>aha like that. I know that Debian has it easy, but I never tried it. <jmarciano>GuixSD will give more control to the user. I can create system.scm or something, and always get what I want and nobody else. <suitsmeveryfine>Hmm, is there a delay before emails are delivered to the devel-list? <civodul>suitsmeveryfine: the first message can take an hour or two; subsequent messages directly get in <suitsmeveryfine>Thanks civodul. I'll then just send this warning to any MacBook2,1 users here that might be thinking of building from master: DO NOT try to suspend to ram in Gnome 3. It just corrupted my file system. <suitsmeveryfine>which is totally fine in my case. I've just been using that system to do testing on behalf of the Guix SD project. <suitsmeveryfine>But I'll now give up and switch to Parabola on that particular computer :) <mark_weaver>I've been investigating the libotr test failure on mips, and so far it seems very likely to be a code generation bug in gcc :-( <mark_weaver>adding print statements to investigate the failed check fixes it <mark_weaver>suitsmeveryfine: the suspend-to-ram problem with Gnome 3 on GuixSD seems to happen to all of us, but this is the first I've heard of filesystem corruption from it :-( <mark_weaver>actually, that bug is the reason I switched back to Xfce for now <suitsmeveryfine>OK. Well, I'm not reinstalling GuixSD at this time also because I can't get the touchpad to work. <mark_weaver>suitsmeveryfine: that's very surprising. are you sure you didn't end up with an old version again? <mark_weaver>the problem was fixed very recently, maybe you didn't try the version that fixed it <suitsmeveryfine>Yes, I wrote that mainly to pizzaiolo so that he would't try it in Xfce. <suitsmeveryfine>pizzaiolo: resuming from suspend in Xfce used to worked for me but then it stopped <pizzaiolo>suitsmeveryfine: my only issue with suspend to ram so far has been an unresponsive touchpad <suitsmeveryfine>pizzaiolo: yes, but there have been many new changes related to power management in the more recents commits. <civodul>lfam: great work on the Biber review, once again! <lfam>civodul: Thanks! And indeed, I am glad you replied as well <jmarciano>I was suspending like this: sudo sh -c '/bin/echo mem > /sys/power/state' and it worked well with my encrypted partition. <lfam>I'm still having trouble with https substitutes on my "foreign distro" system. I think that I must have done something to root's copy of guix (used by all users) that is preventing it from upgrading <mark_weaver>closing the lid hasn't worked for me on the Libreboot X60, but choosing "logout" from the Xfce menu and then clicking on "suspend" works for me now <lfam>I see that people are talking about power management on xfce. Is there a consensus that xfce4-power-manager has improve the situation? Or degraded ti? <jmarciano>It is GUI, how it could improve the situation... ;-) <mark_weaver>so, if closing the lid simply doesn't do anything for you, what I described above may work <lfam>civodul: I get the backtrace from #22937 <mark_weaver>lfam: at some point recently, things improved dramatically for me with Xfce, but I'm not sure whether it was xfce4-power-manager or wingo's (xfce-desktop-service) or some combination. <suitsmeveryfine>mark_weaver: closing the lid worked before but not any more (in Xfce on the macbook) <lfam>In root's profile, I have /gnu/store/...-guix-0.9.0.f888c0b, which is the daemon I'm using <suitsmeveryfine>mark_weaver: closing the lid also doesn't work on my libreboot+T60 on any GNU/Linux distro. The screen then stops working. <civodul>lfam: oh so that means guix-daemon calls the "old" 'guix substitute' <civodul>lfam: maybe /root/.config/guix/latest points to an older Guix? <lfam>Okay, I did do `guix pull` and `guix package -u .` yesterday <mark_weaver>suitsmeveryfine: Libreboot on the X60 seems not to provide the usual ACPI events to the OS, or something along those lines. <lfam>It points to /gnu/store/is65rfjl6iakmif60ybyzii7rxagj5n6-guix-latest <mark_weaver>suitsmeveryfine: the situation might be similar on other computers of a similar generation, e.g. the T60 and Macbook2,1 <lfam>Which appears to be 47 years old ;) <civodul>lfam: :-) could you compare substitute.scm in that directory to what we have currently in master? <suitsmeveryfine>mark_weaver: yes, one might think so but the T60 and MacBook2,1 are quite different <rain1>mark_weaver, that otr thing is really interesting <rain1>I wonder what it will take to make reproducible builds <suitsmeveryfine>mark_weaver: the brightness controls and touchpad work differently also. <mark_weaver>rain1: another possibility, I suppose, is that the test code depends on undefined behavior in C <rain1>I think we need a boringc like djb brought up at some point <mark_weaver>rain1: well, compiling it with -O0 would probably fix problems like this <lfam>civodul: The substitute.scm in root's 'latest' is identical to that of current master <lfam>I bet that, a few months ago, I did `guix package -i guix` as root. I wonder if that could cause this issue? <civodul>maybe you could 'strace guix substitute' as root and see which substitute.scm is being used? <lfam>I've never run `guix substitute` before :o <mark_weaver>jmarciano: for the code that arranges to use it in our slim service, see %default-slim-theme in gnu/services/xorg.scm <rain1>(midori4:20870): GLib-GIO-ERROR **: Settings schema 'org.gnome.system.proxy' is not installed <rain1>so i can't use midori, but it's odd because I do have gsettings-desktop-schemas in the propagated-inputs <rain1>if i recompile midori (using package -i) it works again <lfam>rain1: If it works after rebuilding midori, perhaps some dependency of midori changed in a way that broke things? <rain1>it might be from a system reconfigure <mark_weaver>rain1: iiuc, midori uses an old version of webkit that has large numbers of security flaws <lfam>Yeah, the "live reconfiguring" feature is not 100% perfect yet :) <rain1>im getting firefox as we speak <rain1>i blame gschema to be honest <NiAsterisk>wild. So I wrote a new package, and it used to work in the first tests, but now pre-inst-env guix build can't even locate it. <lfam>Perhaps the code in the package definition is malformed <NiAsterisk>maybe.. but then lint would be able to find it. and it is just unknown <lfam>No, I've had lint be unable to find package definitions due to malformed code <lfam>Good luck. Feel free to share it if you get really stuck <mark_weaver>I think I found the problem in the libotr test suite <mark_weaver>test_auth_clear in test_auth.c never initializes the context. there's a missing call to otrl_auth_new(&ctx); <mark_weaver>it works by accident on some platforms, presumably because of the way those static test functions get inlined together into the same parent function <lfam>civodul: I confirmed with strace that `guix substitute` is using the same code as that of the current master branch. However, I find that when I run the daemon from a root shell, I can fetch over https! It's only when running the daemon with systemd that things go awry. But, the systemd service file executes this file: /root/.guix-profile/bin/guix-daemon, which is what I am running in that root shell. <lfam>Also, when running the daemon from root's shell, if my regular user does not ask for an https substitutes-url, I fetch from http://hydra.gnu.org, rather than some https URL. I'm not sure if that is expected or not. 
