IRC channel logs


back to list of logs

<PotentialUser-85>Hi, I tried installing Guix in a VirtualBox VM. It boots to a login screen now, and my password gets accepted, but then the screen goes black for a second, before just throwing me back to the login screen. If i go to tty i can log in, but I don't know how to start XFCE4 from there?
***Uncle is now known as elais
<Krafter_>I may have accidentally deleted all files in ${HOME} and now guix pull wont work. Can I reset it?
<Krafter_>PotentialUser-79: I am almost at the same spot as you, I may be able to help you in a bit.
<samplet>Krafter_: Try using “/var/guix/profiles/per-user/YOUR-USER-NAME/current-guix/bin/guix”.
<Krafter_>samplet: "error: file exists: "var/guix/profiles/per-user/YOUR-USER-NAME/current-guix/""
<samplet>Krafter_: You have to replace “YOUR-USER-NAME” with your username.
<Krafter_>samplet: I did.
<samplet>Krafter_: And you ran the “guix” executable there?
<Krafter_>yes that binary with pull at the end.
<samplet>Krafter_: Were you in the directory when you ran it?
<Krafter_>It also says: "Migrating profile generations to 'var/guix/profiles/per-user/MY-USER-NAME'...
<Krafter_>samplet: no.
<Krafter_>samplet: Do you think I can safely delete /var/guix/profiles/per-user/MY-USER-NAME/* ?
<samplet>Krafter_: Maybe try recreating the symlinks from your home directory to the profiles there.
<Krafter_>Recreate to what?
<samplet>As in “~/.config/guix/current” points to the “current-guix” one, and “~./guix-profile” points to the latest “guix-profile” one.
<samplet>Which should be “guix-profile” with no numbers.
<Krafter_>samplet: That did it. It turns out my ~/.config/guix/current was missing.
<Krafter_>I seem to strike every odd bug imaginable.
<samplet>This is the part where I unhelpfully recommend that you don’t accidentally delete all your files again. ;)
<samplet>That Swedish language bug you found was pretty striking, true!
<Krafter_>Past me would have been very grateful.
<noobly>what is guix doing during install when it says 'substitute: updating substitutes from
<noobly>i have a screenful of this message during every install, and never get past it
<apteryx>I've got a "python" package, named typeshed, which only consists of data (python stubs, *.pyi). It comes without a file. Should this data be installed to a out/share/typeshed location?
<apteryx>or is there a more suitable location for a package which only installs data?
<dftxbs3e>samplet: hi there!
<samplet>dftxbs3e: Hey.
<dftxbs3e>it's a video about the whole float128 issue lol
<dftxbs3e>an hour for this
<dftxbs3e>looks quite serious
<dftxbs3e>samplet: should we try again now? do you think cross compilation works better on core-updates?
<samplet>dftxbs3e: I don’t think there’s been much progress. Everyone was working on 1.0.0 and then immediately 1.0.1.
<samplet>I had to give up on my build. I was making progress, but there were just so many little issues I lost the will for it.
<dftxbs3e>samplet: I understand.. I'm thankful for everything you've done so far!
<dftxbs3e>I wish there was a clear roadmap on the progress that was made and what's left to do on core-updates
<dftxbs3e>doesnt Ludovic hang on in here?
<apteryx>dftxbs3e: he does
<samplet>dftxbs3e: Now that version 1 is out and semi-stable, I imagine everyone will relax for a bit, work on some fun stuff, and then eventually decide that “it’s time to merge core-updates!”
<samplet>At that point, everyone will put together things that can and should be done on core-updates.
<dftxbs3e>samplet: aha okay - sure, take some time off
<dftxbs3e>apteryx: well that's good. I have yet to see him! What's his nick?
<dftxbs3e>ok! thanks :D
<apteryx>he's there almost everyday of the week (european time)
<dftxbs3e>well then probably I just didnt catch on his nick
<dftxbs3e>uhm and also
<dftxbs3e>I'm in Europe but I live by night these days
<dftxbs3e>So I probably missed him due to that..
<dftxbs3e>samplet: and there's GSoC too!
<tenXsandwich>I'm having some trouble building some projects with CMake in a guix environment, are there some particular metapackages that depend on things like libstdc++ etc?
<Marlin1113>for some reason, i don't seem to have a .guix-profile
<vagrantc>Marlin1113: is there /var/guix/profiles/per-user/vagrant/guix-profile ? that's what it should symlink to
<vagrantc>Marlin1113: where "vagrant" is your username :)
<Marlin[m]>there was, but it was empty for some reason
<Marlin[m]>so i just reinstalled it
<r0sebush>anybody get the xorg service to dispaly on windows 10 hyper-v install of guix system? i;m seeing session relovale messages for uder gdm.. no gfx uiser inter login screen shows up
<kabo>is there a cheat sheet for guix for people comming from other distros?
<kabo>like sudo apt update -> guix pull
<kabo>or something
<str1ngs>guix package -u -> apt upgrade. or guix system reconfigure for system
<str1ngs>some concepts in guix though do not apply to other distro's
<kabo>ok, but as long as the concepts in other distros map to guix it should be fine, right?
<kabo>basically the stuff I do most is apt update, apt full-upgrade, apt install, and apt remove
<kabo>should I do guix install, or sudo guix install, or update the config file and run guix pull?
<str1ngs>apt install -> guix package -i . apt remove -> guix package -r .
<kabo>guix upgrade: warning: consider running 'guix pull'
<kabo>should I run guix pull with sudo or no?
<str1ngs>I'm kinda selective about guix pull myself
<str1ngs>are you using GuixSD or hosted guix?
<kabo>guix sd
<str1ngs>if you just installed I might do a guix pull yet.
<kabo>ok, running it without sudo now
<str1ngs>then a reconfigure
<str1ngs>run with -n on reconfigure just encase it decides to build the kernel or something silly :P
<kabo>ah found this
<str1ngs>seems about right :P
<kabo>so guix pull and sudo guix system reconfigure /etc/config.scm is the same as sudo apt update and sudo apt full-upgrade then
<str1ngs>yes, kinda
<str1ngs>though you can't compare upgrade to reconfigure.
<kabo>what does -n do?
<str1ngs>-n is dry run
<str1ngs>it will give you a list of what it will do.
<str1ngs>kabo: are you familiar with stow?
<str1ngs>aka GNU stow
<str1ngs>well package and reconfigure. basially symplinks programs etc into either a user profile, or the system profile
<str1ngs>package are actually in /gnu/store so unlike upgrade-full they might not get remove. also the profiles can be rolled back or selected on boot
<str1ngs>once root nodes, are no longer used by a user or system profile. you can garbaged collect them
<str1ngs>so there is not really an one to one distro cheat sheet. just simular goals I guess
<kabo>those are details I'd love to dive into eventually. first I just wanna figure out how to make sure my laptop has the latest info on what versions are available for all the available packages.
<kabo>and how to upgrade all upgradeable packages
<str1ngs>guix pull, guix system reconfigure, and guix package -u .will acheive that
<kabo>sweet :)
<kabo>guix pull is taking a lot longer than apt update...
<str1ngs>it use to take longer lol
<str1ngs>it should not be to bad if you are using substitutes?
<kabo>I just did a fresh install, I assume it's using substitutes?
<kabo>guix pull just finished
<kabo>so took 10 minutes
<str1ngs>that's not to bad
<kabo>hmm, according to the manual: Note that sudo guix runs your user’s guix command and not root’s, because sudo leaves PATH unchanged
<kabo>I ran sudo guix system reconfigure /etc/config.scm
<kabo>and got warnings
<kabo>/root/.config/guix/current not found
<kabo>guix pull was never run
<str1ngs>sudo guix describe
<str1ngs>what does it output for hash?
<str1ngs>looks good
<kabo>scarry warning though
<kabo>that's the warning
<str1ngs>maybe switch to root and run . guix pull --commmit=7a890145e3bc9d7aebd753ab2ccc3ba1aecfaf06
<str1ngs>that will sync user and root guix profiles
<dftxbs3e>samplet: I think I'll be able to get Guix to work on powerpc64 big endian. The float128 issue isnt present there.
<dftxbs3e>and I can run a big endian VM on a little endian host, so that's fine by me
<str1ngs>why notify-send you no work!
<samplet>dftxbs3e: That sounds like it might be easier, sure.
<str1ngs>I have access to a power8 and power9 machine. but I don't have root access.
<dftxbs3e>guix doesnt need root AFAICT
<str1ngs>I've beening meaning to use a cgroup container to do some porting
<str1ngs>it needs root to install
<dftxbs3e>you don't need to install
<str1ngs>and it makes assumptions like groups etc
<dftxbs3e>I have a Talos II machine at home
<dftxbs3e>so I'm trying to port it for myself
<dftxbs3e>I can give you access to a VM if you need one
<str1ngs>you stil need the store eventually
<dftxbs3e>well you can compile guix to not use /gnu/store as store
<str1ngs>that's what I did, but then it assumed build users
<str1ngs>I was actually going to look at seeing if I could get guix to use rootless containers
<str1ngs>I've was able to do that in go. maybe I could do that with guix aswell
<str1ngs>using cgroups
<str1ngs>then you can run guix, as none root in a container. and still add groups, and use the /gnu/store path
<dftxbs3e>yeah that sounds like good improvement, to not require root at all for guix
<dftxbs3e>some packages will still need root but you arent force to use those
<str1ngs>I have the mental process to do that. and most distro's support it out of the box
<str1ngs>I just need to figure out if guile can make the syscalls. and the forks. I think it's possible
<str1ngs>worst case that little bit can probably be done in C. though I need to understand guix better.
<str1ngs>alternatively runc could be used for machines that dont have root access temperally
<str1ngs>I was going to create a guile script that basically setups and runs a rootless container. that might be a good starting point
<sirmacik>had to install debian again. Having this ghostscript bug presist and need to recompile chromium so often was too much for my workflow to allow me to work normally :/
<civodul>Hello Guix!
<user_oreloznog>Hello !
<user_oreloznog>The principal Gimp developper asked something, at the end of this page:
<user_oreloznog>I'm not able to answer myself :)
<user_oreloznog>the article is highlighted on the first page
<civodul>hello user_oreloznog!
<user_oreloznog>o/ :)
<dftxbs3e>hi civodul, I was told you're Ludovic, glad to meet you, I really like GNU Guix!
<civodul>hi dftxbs3e, glad you like it!
<civodul>user_oreloznog: i don't actually have an account, but maybe roptat or someone else can answer
<civodul>(or i could create an account, but...)
<user_oreloznog>yes, maybe sneek can tell to him...
<dftxbs3e>civodul: I have a Talos II POWER9 machine and I've been working with samplet on bootstrapping powerpc64le, but it's dependent on gcc 6.2+ migration for everything because glibc 2.26+ requires ieee float128, I'm currently about to try powerpc64 big endian that doesnt suffer from this double format transition
<user_oreloznog>sneek: later, say to roptat if he can have a look at:
<user_oreloznog>sneek: later tell roptat if he can have a look at
<user_oreloznog>Hmm, i can't say it to sneek...
<jonsger>dftxbs3e: nice, I'll get my Blackbird hopefully in the next weeks
<jonsger>dftxbs3e: on which branch are you working? I think on the core-updates branch I didn't had this float128 problem anymore, there was just a different build failure...
<civodul>we should starting working on merging core-updates
<civodul> <- 1.0.1
<dftxbs3e>jonsger: I went for the shortest path and built it on v0.13.0 + some cherry picked patches at
<jonsger>so dftxbs3e you're going for v0.13.0 because it has a glibc < 2.26?
<dftxbs3e>jonsger, yes, and I think I can bootstrap upwards to 1.0.1 afterwards
<dftxbs3e>jonsger: I actually havent tried yet. But as powerpc64 big endian doesnt have the issue, I'm going to try to get 1.0.1 running directly with the big endian bootstrap binaries.
<dftxbs3e>jonsger: the issues in core-updates about bootstrapping are not powerpc specific, so that's why I gave up on that route
<dftxbs3e>I'm not experienced enough with GNU Guix to start doing main maintainer work
<jonsger>dftxbs3e: oke, I'm not really interested in big endian. I think efraim got powerpc32 big endian working just out-of-the-box
<dftxbs3e>jonsger: personally, I'm just trying to have it run on my machine. You can run a big endian VM on a little endian host, and that's what I'll be doing to start hacking on GNU Guix on my Talos II if that works first.
<dftxbs3e>Having to use my crappy x86 laptop to do so or rent powerful cloud servers has showed quite annoying, so the earlier I get running completely on ppc, the better it is!
<dftxbs3e>rekado: are you okay?
<rekado>it’s me!
*rekado should move the yubikey to a different USB port…
<dftxbs3e>rekado: really! those two strings looked similar
<jonsger>dftxbs3e: fully understand, doing work on bootstrapping guix on Laptop is not fun. It even kind of kills my workstation...
<rekado>my Polish colleague said it looks like Polish with typos.
<jonsger>dftxbs3e: maybe you could send a mail to with your current status and CC, he has also a TalosII
<dftxbs3e>jonsger: well it's not quite finished, neither I have done a lot of work, I spent time understanding things and testing.
<jonsger>dftxbs3e: I doesn't have to be "finished". Just describing what you undestanded, what your plans are etc. This also can reduces duplicated work...
<civodul>rekado: :-)
<dftxbs3e>jonsger: okay well, I'll do this some time later
<roptat>hi guix!
<sneek_>roptat, you have 1 message.
<sneek_>roptat, user_oreloznog says: if he can have a look at
<roptat>I'll take a look
<dftxbs3e>that's so nice!
<civodul>howdy roptat :-)
<dftxbs3e>that bot is, genius, really, simple solutions to simple problems :D
<roptat>well, I also have backlogs :p
<dftxbs3e>it would be nice if it sent some context along with it
<roptat>there's now also an hpc question I can't answer
<dftxbs3e>and sent PMs instead maybe
<roptat>I like it like that, you just have to tell it the context so it repeats it back later :)
<dftxbs3e>roptat: right. well that means people must know about the bot!
<mfg>Hey :), when running `guix environment python-numpy python-pandas python-plotly python-ipython' ipython is missing in the environment.
<mfg>How do I resolve this ?
<str1ngs>mfg: maybe you want --ad-hoc instead?
<mfg>Ahhhhhhh ..... I knew i forgot smth ....
<str1ngs>guix environment is not intuitive . should be guix build environment :P
<dftxbs3e>or rather, guix build --environment
<str1ngs>that would work too
<g_bor[m]>hello guix!
<g_bor[m]>I've just notice something strange with a freshly installed vm.
<g_bor[m]>Everything was working just fine for a while, but then the root password was reset to empty, and the user password was locked. Any idea how that might happen?
<str1ngs>g_bor[m]: passwords might not survive reconfigures
<str1ngs>if you want password to presist you need to declare them in your config. but that has security draw backs
<str1ngs>like passwords in clear text. in the store
<roptat>dftxbs3e, I wrote answers to the two new comments :)
<dftxbs3e>roptat: on linuxfr you mean?
<dftxbs3e>I wasnt the one who asked for it
<roptat>oh yes, you're right, sorry ^^'
<dftxbs3e>roptat: the person who did is offline so..
<g_bor[m]>str1ngs: Is there a reason for that?
<dftxbs3e>but thanks for helping convincing people to run GNU Guix aha
<g_bor[m]>I was under the assumtion that if you don't set in the config, then it is handled as state, and not touched by reconfigure.
<dftxbs3e>roptat: je suis aussi content que la communaute francaise prennent part importante dans GNU Guix! ;)
<roptat>dftxbs3e, bah, civodul est aussi français ;)
<roptat>à part peut-être les allemands, qui pourrait dire que sa communauté a une part plus importante que la notre ? :p
<dftxbs3e>roptat: j'avais compris, j'ai vu ses travaux a cette universite francaise de bordeaux dont je ne me rappelle plus le nom
<roptat>inria ?
<dftxbs3e>c'est un activiste libriste qui m'a parle de GNU Guix dans un centre d'animation sur Paris, je m'attendais pas a y voir autant de francais! :D
<str1ngs>g_bor[m]: normally the are preserved. there are rare cases where they are not.
<dftxbs3e>c'est surement pas une competition mais ca me rend heureux, j'avais un ressenti de nullite sur la france recemment
<str1ngs>g_bor[m]: I don't know which case specifically maybe someone more knowlegable does
<g_bor[m]>ok, thanks
<rekado>civodul: now supports “advanced” queries:
<rvgn>Do following syntax 1 and syntax 2 have any functional differences?
<rvgn>Syntax 1: (users (cons (user-account [...] ) ℅base-user-accounts)). Syntax 2: (users (append (list (user-account [...] )) ℅base-user-accounts)).
<rekado>rvgn: no difference.
<rekado>both result in a list
<rvgn>rekado Thank you!
<rekado>%base-user-accounts is a list and we want to add something to it.
<rekado>“cons” puts *one* item to the front of the list.
<rekado>“append” takes two lists and glues them together
<civodul>rekado: oh!
<rekado>for more than one item you can also use “cons*”
<civodul>rekado: you're a hero!
<rvgn>Thanks a lot rekado. Now I understood the difference.
<rekado>civodul: I wonder if we’re doing something wrong with fibers. The fact that the web interface doesn’t respond all the time is worrying.
<rvgn>rekado What would be the difference between "cons" and "cons*" ??
<rekado>see a few lines above :)
<rekado>“cons” can only work on *one* item and a list.
<rekado>“cons*” works on one or more items … and a list.
<rekado>(cons 1 (cons 2 (cons 3 (list 9 8 7)))) ==> (1 2 3 9 8 7)
<rvgn>Oh I missed that. Thanks!
<rekado>(cons* 1 2 3 (list 9 8 7)) ==> (cons* 1 2 3 (list 9 8 7))
<civodul>rekado: it can be hard to debug, but the likely issue is that a fiber is blocking on something
<rvgn>Got it now. :)
<rekado>(cons* 1 2 3 (list 9 8 7)) ==> (1 2 3 9 8 7)
<rekado>civodul: I think we should just have a pool of web workers instead of trying to be smart.
<rvgn>rekado When append glues lists, which list is used first?
<roptat>the first: (append (list 1 2 3) (list 7 8 9)) ==> (1 2 3 7 8 9)
<rvgn>Gotcha! Thanks.
<roptat>you can experiment with that by yourself with the guile repl ;)
<rekado>rvgn: when in doubt you can run “guile” and …
<roptat>(just run "guile")
<dftxbs3e>mhm, 504, "Gateway Time-out" on
<rekado>yeah, what roptap wrote
<civodul>thanks, efraim! :-)
*kmicu finally has a master password for Guix PetaBytes Data and High Performence Computing Center. [evil laugh]
<dftxbs3e>kmicu: aa what is this? the finalization of the CI infrastructure?
*kmicu nmaps the Whole Known Internet and hits it with user:admin pass:jjjjjjhit.itnutnujjeekhhecptjeyynyddykjjixtd credentials. [evil laugh intensifies]
<rekado>kmicu: changed it already. Now its jjjjjjhit.itxkceuuybjuk.pgpu.hpgkujky.kxgpdh
<rekado>efraim: this is a great blog post! Thanks!
<dftxbs3e>question: I'm getting a cannot open shared library error, is there a way to tell Guix to add a folder to load shared objects from?
<rekado>dftxbs3e: only the usual way: LD_LIBRARY_PATH (or LD_PRELOAD). But maybe you want to add the directory to the RUNPATH at compile time instead?
<dftxbs3e>somehow version 0.13.0 is trying to use my system's file binary
<rekado>… is LD_LIBRARY_PATH set? Generally it won’t use your system’s libraries unless you force it.
<dftxbs3e>rekado: oh ok, I see, but that library already is in /lib64 ..?
<civodul>dustyweb, davexunit: what would you think of having a short blog post like to introduce our new GSoC intern and their project?
<rekado>dftxbs3e: LD_LIBRARY_PATH generally should not be set, especially not to /lib64 of the host system, because it’s so full of libraries that you just don’t have much control over the effects.
<dftxbs3e>rekado: well, GNU Guix isnt trying to compile that binary, it uses my system's "file" utility instead. Not sure why, but it does, I tried LD_LIBRARY_PATH and it's pending to see if it worked, what would you recommend here?
<dftxbs3e>it did not work
<mekare>hi guys. Trying to do a ping inside an environment (--container --pure --ad-hoc inetutils -- ping but getting an error about icmp being an unknown protocol. Indeed /etc/protocols is not there. I thought that net-base would do the trick but not really. I tried to fill /etc/protocols then I got another error about not getting allowed to do raw socket (as ping is not setuid in env afaik). Any pointers
<mekare>? :)
<rekado>dftxbs3e: I’m not sure I understand. In the build environment your system’s “file” utility doesn’t exist. In what context does this happen?
<dftxbs3e>rekado: it does exist, but when ran within the environment, the wont load because it can't find it, even though it's in /lib64 folder on my system
<rekado>by “build environment” I mean what “guix build” sets up.
<rekado>this is in “guix environment”?
<dftxbs3e>ldd says that is only referenced with that relative path and not the full /lib64/
<dftxbs3e>rekado: just build, my system is fedora 29
<rekado>sorry, can you send email to with more details?
<dftxbs3e>rekado: I don't have more details to give, what do you not understand?
<rekado>when using “guix build” a container/chroot is set up that doesn’t have access to /lib64
<dftxbs3e>oh ok
<dftxbs3e>I don't know how GNU Guix works, I'm mostly a user
<rekado>I don’t know what you’re building, for example, what exact command you used, etc
<dftxbs3e>I'm building the "hello" executable on version v0.13.0
<dftxbs3e>for powerpc
<dftxbs3e>on powerpc
<dftxbs3e>a patched version of, v0.13.0
<dftxbs3e>well I don't think I want to bother anyone with that if the solution isnt obvious
<dftxbs3e>So I'll just work it out, eventually, thanks]
<efraim>Yay! My post is live
<kabo>hmm, I need node 10 for work, any idea how soon that'll be available?
<efraim>Maybe powerpc64 needs the /lib64 -> /lib change aarch64 has in gcc
<rekado>kabo: would you like to try to upgrade it?
<kabo>hmm, tempting, but I'm a js dev, not a C dev...
<kabo>I wouldn't have a clue where to start
<rekado>kabo: I’m also not a C dev.
<rekado>luckily you don’t need to be one to upgrade a Scheme package definition.
<kabo>I have to go to bed now anyway. could you please message me a few links to get me started and I'll think about it?
<rekado>kabo: try the Contrtibuting section in the Guix manual. It’s got all you need to set up an environment to contribute to Guix.
<rekado>(the actual upgrade of the package definition might only take a handful of lines to be changed)
<roptat>mekare, have you tried --network? :)
<roptat>(although I don't think you can have suid binaries in environments)
<pkill9>why does chromium need mesa wrapped with LD_LIBRARY_PATH?
<pkill9>well, need itself wrapped with LD_LIBRARY_PATH pointing to mesa i mean
<rekado>hmm, someone accesses build logs on the cuirass web interface, but my requests aren’t processed at all.
<rekado>it looks like it’s handling just a handful of clients at a time.
<dftxbs3e>rekado: well I disabled usage of zlib in file
<dftxbs3e>and it's working so
<mfg>it's me again with another question: How do I mount NFS? i installed the nfs-utils package and tried to mount -t nfs, but it doesn't work due to "/home/mfg/mnt: bad option"
<mfg>do i need to activate some services ?
<civodul>rekado: does Cuirass have a backlog of requests?
<civodul>that could happen if it's too slow
<rekado>I don’t know. How would I see that?
<rekado>I attached to the process with strace and saw requests to build logs once every few seconds.
<rekado>but my request (made while cuirass was waiting in epoll) never showed up
<rekado>even with a very long nginx timeout I don’t get through.
<civodul>you can tail -f /var/log/cuirass.log
<civodul>and in another terminal, you wget something
<civodul>you'll see if your request is displayed right away in the log or not
<civodul>also, you can judge by the amount of EPIPEs whether it's lagging behind
<civodul>(EPIPE means the client disconnected; IOW, the request was processed too late, or took too long)
<rekado>my request does appear in /var/log/cuirass.log
<rekado>quite a few requests for build logs. I wonder what’s so interesting about them :)
<civodul>i like to read build logs at night, always inspiring
<civodul>but anyway, your request is processed on time?
<civodul>you could "strace -f" the thing to see whether there's an sqlite query blocking the whole thing
<rekado>civodul: my request is never processed.
<rekado>nginx eventually times out.
<rekado>strace doesn’t tell me anything about sqlite queries blocking things.
<rekado>oh, just saw this: In procedure accept: Too many open files
<rekado>also: 2019-05-21T14:02:11 heap: 62.06 MiB; threads: 81; file descriptors: 1023
<rekado>that’s not good
<civodul>das ist nicht gut!
<civodul>i guess you have to restart it
<dftxbs3e>increase kernel limits?
<roptat>do you forget to close some files?
<civodul>i suspect Guile-Git is the culprit:
<civodul>though it could be due to the way Cuirass uses Guile-Git, not sure
<janas>The hash for emacs-idris-mode appears to be changing every day or so, however when I inspect the expanded directories with diff it doesn't report any differences
<janas>Any other tools that could help me figure out the problem?
<rekado>janas: huh, you’re right. The package definitions don’t follow our guidelines.
<rekado>(or do they? I don’t think we should ever get tarballs from melpa.)
<janas>One way to solve it would be to switch to a git checkout, although the github repo doesn't seem to have release tags
<janas>But I'm curious about he hashes, I thought the NAR hashes were only supposed to depend on the file contents?
<jlicht>janas: have you run `diffoscope' on the directories?
<jlicht>it might be some timestamp or permissions issue
<janas>jlicht: They have different modify times, although that could be from the diff tool. But I thought that the modify time was stripped from the nar file before the hash was computed? Or am I remembering that wrong
<janas>No, the modify times seem to be from melpa
<roptat>janas, are you talking about the source of the package, or binary substitutes?
<janas>roptat: The source
<janas>ahh okay the archives are only converted to the nar format for the substitutes?
<roptat>sources are kept as is
<roptat>so if it's a tarball whose format preserves modification time, every version of it will be different even if the file contents are the same
<roptat>guix doesn't modify the source
*janas is learning
<Krafter_>Is it necessary to install a wm globally to be able to use Xorg?
<mekare>roptat: tx, I missed that one !
<Krafter_>Or any graphical desktop.
<roptat>no, I think you only need a login manager
<roptat>mh... unless the login manager is not able to find a desktop environment / window manager installed in the user profile (but it should be able to read .xsession)
<Krafter_>roptat: I am intending to use i3. And in my .xsession I source .xinitrc which loads i3. But when I try to log in with gdm I get booted right back out.
<roptat>could it actually be the bug from 1.0.0 (related to %base-packages)? or maybe you need to specify i3 with its absolute path
<atw>hello #guix!
<Krafter_>I just saw something. I had forgot to set the startup files as executables.
<Krafter_>Hello atw!
<ArneBab>minifeedback: if guix tells me that there’s a conflict between two packages, I wish it would just tell me all conflicts and then ask whether I want to update all of them together.
<ArneBab>as it is now, I just re-run guix install several times with more and more packages, until guix finally installs what I need
<jonsger>dftxbs3e: the bootstrap-tarballs for x86_64 just build fine on core-updates, so I don't share your assumption that the bootstrap-tarballs are general broken on core-updates
<PirBoazo>Do you know if a channel exist in french?
<civodul>PirBoazo: there's no French-speaking IRC channel but i'd say it's ok if when speak a bit of French here
<civodul>once French speakers have become too noisy, we can create #guix-fr ;-)
<PirBoazo>Thank civodul ;-)
<civodul>si tu as des questions en français, on est plusieurs à pouvoir y répondre
<PirBoazo>I have install the guix 1.0.0 with too many dificulties.
<PirBoazo>Et j'ai eu la surprise de trouver la doc en français , c'est super pour moi.
<PirBoazo>Je cherche à comprendre un peu mieux comment tout cela fonctionne, dans l'objectif de créer un paquet pour un logiciel .
<PirBoazo>existe t il des tutoriaux expliquant la démarche ?
***Server sets mode: +cnt
<jonsger>PirBoazo: Il y a une introduction dans notre blog: Mais il est en anglais...
<PirBoazo>Ce sera un bon début
***notnotda1` is now known as notnotdan
***nckx- is now known as nckx
<zimoun>Are some servers down?
<PirBoazo>Merci civodul & jonsger A+
***jfred3 is now known as jfred
<civodul>hi zimoun!
<Marlin[m]>Awesome, you can install nix inside of guix
<civodul>zimoun: which servers?
<davexunit>can you install guix with nix? that way you can use guix to install nix to install guix to install nix to install...
<Marlin[m]>Funny thing
<Marlin[m]>Nix's install script never works for me
<Marlin[m]>But with guix it worked
<Chaekyung>are the guix anime girls used to shill it everywhere official artwork available in high resolutions somewhere?
<rain1>guix anime? i never saw this
<Chaekyung>I am not entirely sure I want to post the links
<zimoun>civodul: roughly speaking
<Marlin[m]>Can i use gnu shepherd on devuan?
<civodul>zimoun: it seems to work, except through Tor, which has been unreliable for me lately
<zimoun>civodul: Thank you. To me, it is unreliable; sometime off, sometime hanging on. If no one notices that, it should come from my side...
<roptat>Chaekyung, but now we're curious :)
<donofrio_>why does guix install inxi cause it to build freedesktop and many others
<roptat>probably dependencies?
<roptat>so you managed to run the daemon after all :)
<zimoun>civodul: and for example, I am failing to reach, or refresh
<efraim>donofrio_: try inxi-minimal
<Marlin1113>seems like installing nix in guix doesn't set the NIX_PATH enviroment
<rvgn>Why Guix 1.0.1 installer is rfkill-ing the Wireless by default?
<roptat>zimoun, I don't have issues with
<roptat>zimoun, also you might want to give a look at there are new comments :)
<zimoun>roptat: thank you for the test. Weird, I have huge lag or not found for almost all the Hum?
<zimoun>roptat: trolling... mais on n'est pas encore vendredi ;-) I am giving a look.
<Marlin1113>yeah, nix through guix lacks the channels
<Marlin1113>i added a channel and now it works
<kmicu>rvgn: do you have a Lenovo (Ideapad) laptop?
<rvgn>kmicu Yes, X200T.
<rvgn>kmicu Lenevo Thinkpad X200 Tablet.
<kmicu>Hmm X200T, then I’m not sure whether you are affected by or that’s a pure Guix issue 🤷
<rvgn>kmicu I was not affected with Guix 0.16 Installer.
<rvgn>Took me a while to figure out "rfkill unblock all" command. xD
<kmicu>X200T is not Yoga so that’s not that kernel change so maybe a regression from ‘The wpa-supplicant service would sometimes fail to start in the installation image, thereby breaking network access; this is now fixed.’.
<kmicu>Or maybe you’ve pressed hardware rfkill by accident? 🤷
<rvgn>kmicu I am sure that did not do that ;-)
<rvgn>kmicu That bug was found in Guix 1.0.0 and fixed in Guix 1.0.1 right? I shouldn't have encountered that issue.
<kmicu>It depends. That bug was about wpa_supplicant not rfkill.
<kmicu>Maybe someone here could reproduce the issue?
<rvgn>kmicu Okay
*kmicu sees that 1.0.0 is on kernel 5.0 and 1.0.1 is on 5.1 so maybe the issue is really related to that rfkill kernel change.
<rvgn>Any way, I managed to finish my fresh re-install of guix system. Just wanted to know what happened.
<kmicu>Could you test rfkill behavior on Guix 1.0.0 installer? If that’s a kernel change then there is no fix in Guix.
<rvgn>kmicu Oh shoot! If what you see is true then its gonna be another critical bug which prevents installer to connect to wifi on some thinkpads.
<rvgn>kmicu Sure, I can try that.
<kmicu>Thank you!
<donofrio_>eek up to 3gb+ in gnu directory.....
<donofrio_>just for inxi.....and hello build
<kmicu>donofrio_: inxi-minimal or inxi?
<donofrio_>normal inxi
<donofrio_>how do I reclaim/rollback installed dependencies and the like (aka how do I go back to 2gb or less in gnu directory with inxi-minimal (would be testing apache already but it's still building...and I enjoy this project, what makes it better than homebrew and others?)
<bavier>donofrio_: 'guix gc'
<kmicu>Uninstall inxi, install inxi-minimal and then garbage collect to reclaim disk space.
<kmicu>What makes it better depends on your needs. Maybe nothing is better, maybe a lot. Do you really need reproducible and declarative environments? If what you use right now works for you then you shouldn’t switch.
***str1ngs_ is now known as str1ngs
<davidl>Can I install packages via pip by temporarily change /gnu/store to a read-write filesystem or would that be very ill advised?
<str1ngs>davidl: do not write to the stop manually
<davidl>str1ngs: I know that's the general advice but what happens if I use pip install csvkit after having remounted the /gnu/store as read-write and then remount as read-only? What is the actual risk of what might go wrong?
<str1ngs>davidl: if you need to use pip use pip install <foo> --user
<str1ngs>that will use ~/.local iirc . the guix way though is to write a package declaration that installs to the store. maybe guix import pypi will help with that.
<davidl>str1ngs: ok thx, Ill try that. I tried guix import pypi csvkit, then on the propagated inputs etc, but it quickly became a project I don't currently have the time for :/
<str1ngs>maybe pip install --user is the quick and dirty for now
<davidl>str1ngs: yep, it was :)
<str1ngs>you did not hear, this tip from me. I disavow all knowledge of this conversation :P
<davidl>str1ngs: haha ok, I did not hear that
<kmicu>davidl: store consists not only from files but also a database. Modifying store not via Guix tools is the easiest way to break it beyond any repair.
*kmicu pastes One Does Not Simply Read-Write Store meme.
<donofrio_>can I stop remove and gc in the middle
<donofrio_>of a build running
<donofrio_>"- 'build' phase"
<donofrio_>I love that dependancies are in check within this project
<donofrio_>just wow lol
<donofrio_>I'm taking up 3.6 gb in gnu directory now
<donofrio_>I bet I'd be able to startxfce4 after it's done if I went down that road....but what I want is to have apache, pcre, apr, mod_jk, mod_cluster, jboss eap, tomcat all guix installable, without installing all the thousands of packages it doing for inix full.....
<Marlin[m]>My next pc will be built with Guix Gnu+Hurd and libreboot in mind
<donofrio_>I already do that (have for last two years now) ;)
<b0f0>Hello, (I dont have Guix installed) 1. What happens if I want a package that is not in the repository, do I have to write a package for the program that I want to use ? 2. I have nvidia graphics, I think the property driver is not supported here with this distro, but free and open source driver probably is available in GuixSD? 3. When you need to d
<b0f0>o something with linux kernel, like config something and the then compile the kernel, do you write the config in guile script or how do you set your kernel ?
<Marlin[m]>nice botnet
<civodul>hi b0f0
<civodul>b0f0: there are 10k packages in Guix proper
<civodul>if what you want is not there, maybe there's a third-party "channel" somewhere that provides it
<civodul>if not you'd have to write it, which is often rather easy
<civodul>re (2), the "nouveau" driver is indeed part of linux-libre
<civodul>re (3), efraim published a great post just today on this topic :-)
<Marlin[m]>you can get nix inside of guix too
<Marlin[m]>i have it for some stuff guix doesn't have, like smplayer
<civodul>b0f0: this one:
<Marlin[m]>and it also doesn't allow proprietary stuff by default
<civodul>Marlin[m]: this is not the place to discuss proprietary stuff, though :-)
<civodul>also i'd encourage you to contribute missing packages to Guix
<b0f0>if I find an interesting Vim plugin, do you then write a package for it, or can you just use one of Vim s plugin managers ?
<civodul>i don't use Vim so i can't tell for sure, but i suppose its plugin manager works
<civodul>you may find it more convenient to have a Guix package though
<civodul>well, hopefully ;-)
<Marlin1113>i'm not talking about proprietary stuff*
<Marlin1113>nix is free software
<Marlin1113>and is on guix's repos
<civodul>i know i know, and that's not what i wrote
<b0f0>thank you for the effort in answering fully my 3 question. I also appreciate telling me that there is a way to use nix packages in guix, that could be usefull for me too. Also like the civodul answer about the custum linux kernel blog post. Great ! I have to try it.
<Marlin1113>any tips for an fsdg compliant distro that doesn't use systemd?
<Marlin1113>i'd use guixsd on my laptop, but it failed the last couple of times
<Marlin1113>dunno why
<civodul>b0f0: glad that answered your questions :-)
<ng0>it gets easier when you don't exclude systemd... but if you have to, there are lists of such distros out there.
<Marlin1113>i'm better off using sysvinit
<Marlin1113>bash scripts i can understand
<Marlin1113>anything else and i get confused af
<donofrio_>will I kill my packages if I break out of this build?
<civodul>donofrio_: you can fearlessly ctrl-c a build, nothing bad will happen
<mekare>hi, does anyone succeeded in getting ping working in environment --container --pure --network --ad-hoc X Y Z ?
<b0f0>in the "About GNU Guix" text its written that "unprivileged package management" is a feature, does this mean that I can clone a github project in my home directory or does with "unprivileged package management" mean something completly different ?
<donofrio_>civodul, k cause I've broken things on other distro's that required --fix
<b0f0>it could be that what is meant by that is that I dont need root acceess to install packages, but I am not sure
<vagrantcish>you don't need root access to install packages in guix
<vagrantcish>you can even install custom packages as a user, using guix as a package manager
<b0f0>Thats Nice ! thx vagrantcish.
<vagrantcish>which is really useful when updating a package or building a new set of packages ... the tools basically work the same weather you download the packages from guix's substitute servers or you build them locally
<vagrantcish>(just a matter of how much time your machine spends building vs. downloading)
<b0f0>are the packages then source code and then it is compilled, its not a binary package ?
<b0f0>or both ?
<vagrantcish>weather you build the source code or the substitute server builds the source code, it's all the same.
<vagrantcish>pretty much
<vagrantcish>guix package --install PACKAGE ... will check your configured substitute servers and download from them, and if none of the substitutes have it, it will build it locally
<b0f0>good to know, nice !
<vagrantcish>you can also choose not to trust any substitute servers and it will always build locally
<donofrio_>nice that gc is great....only 400mb used right now
<donofrio_>ok how do I do 'apt-cache search' things?
<mekare>guix search or guix package -s
<krafter>The fonts in icecat is all F***ed up. Why?
<krafter>Sorry, only the fonts of the UI elements consist of black boxes. All other pages work fine.
<civodul>krafter: is it on a foreign distro?
<krafter>civodul: No.
<civodul>did you install any font-* packages in your profile?
<krafter>civodul: Yes.
<civodul>and then run "fc-cache -f"?
<krafter>All the fonts in webpages work fine.
<krafter>civodul: Your command worked. What does it do?
<rekado>hmm, email’s acting up again. More of my mails don’t make it through to the list :-/
<civodul>krafter: it updates Fontconfig's font cache, thereby making new fonts that you installed "known" to graphical programs
<civodul>rekado: bah
<civodul>email has become so terrible
*nckx bites their tongue.
<nckx>civodul: I ended up reimaging openSuse on the Overdrives, then re-installing Guix from scratch :-/
<nckx>Now it fails to compile Python like expected. ☺
<nckx>But at least that ‘expected’, right?
<civodul>hey nckx!
<civodul>so Python doesn't build on AArch64?
<nckx>civodul: Not with 8 gigs of RAM.
*civodul found , which looks interesting
<civodul>nckx: and we don't provide substitutes?
<civodul>"guix build python2 python -s aarch64-linux -n" says we do :-)
<nckx>Hmm. That's weird. Which farm (=key)? This is Guix System…
*nckx goes to look for the exact derivation.
<jonsger>nckx: what do you mean by reimaging?
<nckx>jonsger: unxz < original.overdrive.hdd.xz > /dev/sda ☺
<jonsger>oke, that's a quite old openSUSE on it. Something like 42.2?
<nckx>jonsger: Sure, but it has a bootloader etc. for a platform I know nothing about.
<nckx>I also don't know enough about openSuse to upgrade the system first & to be honest didn't want to waste even more time on that.
<nckx>It does the job of running ‘guix init’.
<jonsger>yes, Leap 15.0 brings guix-0.14 and Leap 15.1 has guix-0.16. But 42.x doesn't have guix :(
<nckx>jonsger: I used the binary tarball. It works quite well!
<jonsger>for sure...
<nckx>jonsger: Are you Suse-related?
<jonsger>nckx: I work for SUSE :P
<nckx>I guess that qualifies.
<nckx>Does Tumbleweed (← this is me pretending to know something about openSuse) bring newer Guix?
<jonsger>yes. Tumbleweed has guix 0.16, but 1.0.1 is already submitted, but still in review queue :(
<jonsger>I only use Tumbleweed, because Leap and SLES only have Guile 2.0.x, so no modern Guix on it anymore...
<nckx>Hm. Yes. Those are very enterprisey versions.
<jonsger>I drove the update to Guile 2.2 on tumbleweed, but sadly I was to late for the SLES/Leap train...
<jonsger>this discussion remembers me, that I want to add a section to with openSUSE and Arch packages...
<str1ngs>hello, how do I add a input runpath to the dynamic section of an elf. via a package declaration.
<str1ngs>I though inputs where automatically added?