IRC channel logs


back to list of logs

***emyles``` is now known as emyles
<anon987321>hi guix
<anon987321>has anybody managed to get vterm working on guix's emacs package?
<anon987321>i currently get "Loading file /gnu/store/b38pn0gnj4jsrf79lg4kr80rn5kaim0q-emacs-26.2/share/emacs/26.2/lisp/term.elc failed to provide feature ‘vterm-module’"
***sneek_ is now known as sneek
<emacsomancer>anon987321: I've run into the same issue
<anon987321>i have sent a bug report to the mailing list
<anon987321>maybe i should try packaging it to guix
<anon987321>hmm, getting errors when importing vterm from melpa
<anon987321>any ideas?
<anon987321>seems like it has to do with gnutls
<jsatx>hello :) super basic question, how to specify pkg-config requirement in package definitions in order to avoid "The pkg-config script could not be found" configure errors?
<P64>hi everyone
<P64>i have a x64 customized guixsd and i have to instantiate it with such below command for embedded board:
<P64>guix system disk-image --system=armhf-linux -e "(@ (gnu system install) beaglebone-black-installation-os)"
<P64>but above command just work on armhf architecture.
<P64>therefore how i could do that on original system (x64 customized guixsd)?
<samplet>P64: You should be able to do it with “qemu-binfmt-service-type”.
<samplet>Check out the manual for a brief example.
<P64>samplet: Thanks. I'll check it.
<g_bor>hello guix!
<g_bor>I would like to know a bit about the reference scanner
<g_bor>I am implementing different string searching algorithms in guile, and it depends on the workload which one would be optimal for the scanner
<g_bor>My question is, how often the haystack does not conatin the needle...
<g_bor>also, is it possbile to match all the hashes at the same time? If yes, then we might be better of dong that.
<htgoebel>ropat, civodul: Re. debugging-symbols for cmake: The buildtype has to be set to "Debug" (seems to be case-sensitive).
<htgoebel>I was using "debugfull", as teaches.
<htgoebel>And source needs to be available in /tmp/guix-build-….drv-0/, same for other packages.
***jonsger1 is now known as jonsger
<roptat>saluton guix!
<ng0>if you read your news entries on you want to run something on your news entries which replaces "special characters" with html code characters.
<ng0>or at least the atom/rss file
<roptat>or, should we simply declare the right encoding?
<ng0>if you don't use this CDATA thing, you must replace the characters
<ng0>for atom i have 2 examples gnunet used
<ng0>this is what we used to write:
<ng0>this is what gets generated now
<ng0>not the stage one
<fps>hmm, does anyone know what package is needed to get the ecrm1000.tmf latex font?
<ng0>but shows me most of the symbols are already getting replaced
<fps>ok, and authorblk.sty?
<ng0>beats me where the few odd characters are coming from, but I had too little sleep
<fps>oops, authblk
<fps>found the tmf file in texlive-fonts-ec
<fps>could the build farms maybe maintain an index of output files for built packages?
<fps>that would be so suuuuuper helpful :)
<jonsger1>fps: +1 :)
<fps>the build farm would just need mlocate installed and configured i guess and have a REST api call added
<fps>are they still using cuirass?
<jonsger1>fps: yes, it's still cuirass
<grumbel>Is it possible to download git repositories with 'guix download'?
<civodul>Hello Guix!
<numerobis>Hi #guix! Is there a service to create a file? I'd like gitolite to have a given ssh key, in order to enable hooks that mirror my repositories to github at each push. Is extra-special-file a good option for this?
<g_bor>numerobis: I believe you can use extra-special-file just fine.
<g_bor>hello guix!
<g_bor>I am trying to code around epoll.
<numerobis>g_bor: great, thanks!
<g_bor>I found a quite simple implementation, provided that pointers are at max 64 bits, and the architecture is little endian
<g_bor>do we have any supported architecture not fitting these constraints?
<g_bor>and also int is mac 64 bits.
<jonsger1>can someone have a look at It blocks ATM "guix upgrade" on foreign distros
<nly>leoprikler efraim inferiors did it, i have 4.19.69 now. Thanks
***jonsger1 is now known as jonsger
<raingloom>hmmm, keepassxc no longer opens any of my databases after upgrading last night....
<raingloom>previous version still opens them....
<g_bor>hello guix!
<numerobis>g_bor: do you know if there's a way to specify the owner of the file with extra-special-file (I need the gitolite users to have read permission, basically)?
<g_bor>numerobis: I don't think it is possible right now.
<g_bor>I will hav a look at the implementation..
<g_bor>I had a look at 38068, the patch somehow does not apply. I will either rewrite it later, or someone could send an updated patch.
<g_bor>Also, I might be doing something wrong...
<numerobis>g_bor: Thank you!
<nixo_>Hi guix! Any idea on why mount options at boot are wrong? `mount` gives: /dev/mapper/root on / type btrfs (rw,relatime,ssd,space_cache,subvolid=5,subvol=/), while /etc/fstab has different options and `mount -o remount` uses the right one (specified in the file-systems field)
***Server sets mode: +cnt
<g_bor>38068 squashed.
<jonsger>g_bor: thanks for pushing my patch :)
<leoprikler>brendyyn: perhaps you're interested in gitless
<efraim>git worktree is great, i use it in guix for master/staging/core-updates
<nckx>brendyyn: Same. I have ~/guix (for master), ~/guix/.worktrees/{staging,core-updates}.
<str1ngs>sneek: later tell anon987321, it works for me using locally built emacs-libvterm. you need to install cmake, gcc-tools, libvterm. And I used this expression in ~/.emacs .
<sneek>anon987321, you have 1 message.
<sneek>anon987321, str1ngs says: it works for me using locally built emacs-libvterm. you need to install cmake, gcc-tools, libvterm. And I used this expression in ~/.emacs .
<str1ngs>anon987321: also it requires libtool to be install to build.
<g_bor>the date filter on the issuse tracker does not seem to be working corretly for me when specifying numerical dates.
<g_bor>The example works fine
<anon987321>that's what i did tho
<g_bor>hello guix!
<g_bor>I was having a look at wayland
<g_bor>we should update wayland protocols, but it has 1578 dependantss.
<g_bor>I believe this should go to core-updates.
<g_bor>What is the current status of core-updates?
<g_bor>or should I create a wip-wayland-updates branch instead?
<g_bor>It seems that there are two trivial changes involved so far...
<leoprikler>iirc core-updates was merged a while ago, but I'm not an authority on this
<g_bor>leoprikler: IIRC there was a proposal to make a very short core-updates cycle. I believe the next merge is scheduled quite soon.
<nckx>c-u was open last I checked a few days ago.
<g_bor>I would like to know if it is still ok to push this now...
<g_bor>ok, I will try this then.
<nckx>g_bor: Just saying what I know.
<nckx>Don't trust my memory.
<g_bor>Anyone more knowledable about wayland, what are the critical package to check?
<g_bor>nckx: ok, but where can I get this info?
<nckx>g_bor: You ask mbakke and they tell you.
<leoprikler>probably all the WMs and DMs using it
<nckx>We all agreed that it's suboptimal af 🙂
<g_bor>ok, fine.
<nckx>sneek: What is core-updates?
<sneek>I've heard Core-updates is fontconfig, eudev and cups-minimal FTBFS, tcsh's test hangs
<g_bor>nckx: could we somehow make this info publicly availabe, for example on the website
<nckx>I tried teaching sneek that it's ‘open’ but it refuses to learn this new info.
*mbakke is summoned...
<g_bor>mbakke: hello!
<g_bor>Is core-updates open?
<civodul>g_bor: it's wide open!
<mbakke>g_bor: very much so!
<nckx>g_bor: Yes, the problem is everyone agrees that this is needed but nobody does it.
<civodul>free style and all!
<nckx>And my Web skills are limited to <?php ?>.
<g_bor>I believe was we should agree on is how that should work.
<civodul>php is to 90s, Guile is the future!
<g_bor>I don't think that making a static site for this is the best...
<nckx>While we discuss the perfect solution, how about pushing a static $branch is open/closed page to maintenance.git?
<mbakke>dongcarl: I thought LLVM would be rebuilt with that patch; if it truly is just Clang and does not rebuild the Rust toolchain, it should be good for 'master' indeed
<civodul>g_bor: not sure to what extent it's applicable, but GCC has automated tracking of release critical issues with periodic reminders sent to the mailing list
<civodul>nckx: we could do that
<civodul>my impression is that what would work well is to have someone responsible for a branch
<civodul>responsible in the sense of making sure deadlines are honored
<nckx>Isn't that the case in practice?
<civodul>not necessarily someone hacking on it, simply looking at the calendar
<nckx>_o_ mbakke.
<civodul>nckx: in practice, we can't ask mbakke to be responsible for a zillion things :-)
<mbakke>agreed! :D
<civodul>heh, see? :-)
<civodul>plus, when you're hacking on the branch, you're less likely to respect deadlines
*mbakke would love to stay and chat, but has to go back to work for a bit
<civodul>heh, good luck mbakke!
<nckx>Well, hence my gentle suggestion that, while integrating milestones and reminders and Guix Data into Mumi is all really cool, it always gets stuck there, and maybe we just need a <table> (it can have dates & all) to get us 80% out of this mess. And not bother mba_kke every day 🙂
<nckx>And that's something I could do.
<g_bor>ok, then I will update wayland first, then have a look at what can be done about this.
<g_bor>We might even get some server side git hooks in place...
<g_bor>Can we?
<nckx>g_bor: We can.
<civodul>nckx: ah yes, definitely, but we also need to actually pay attention to the data
<civodul>and that's where having a well-identified data meister could help
<fps>civodul: what would you think about cuirass indexing all files in package outputs (using e.g. mlocate) and providing a REST call to search for packages containing files?
<civodul>fps: i think it would be great!
<civodul>perhaps something the Data Service could do, actually
<civodul>but either way, that'd be very useful
<fps>is the data service part of cuirass? i'm only a user in the form of the guix build farms. i have never even looked at the source ;)
<jonsger>mhm, incomplete build logs on CI are a bit bad :(
<nckx>fps: No, it's a separate thing.
<nckx>Wrong link. fps:
<nckx>It's based on CI data though.
<civodul>jonsger: re incomplete build logs, this is at least in part due to a guix-daemon/offloading bug i fixed yesterday
<civodul>i'll deploy the new daemon hopefully today
<jonsger>civodul: nice. But apparently historic build logs can't be fixed...
<civodul>yeah, and there might be another offloading bug leading to truncated logs
<zimoun>I am a bit lost about how the search-paths is declared in the package definition. Other said, I would like to add CPATH and LIBRARY_PATH to the package protobuf@2.6.0. How can I do that?
<nckx>zimoun: That would mean protobuf makes use of CPATH and LIBRARY_PATH at run time, which sounds wrong.
<nckx>I mean, it's definitely wrong, but what is the problem you're trying to solve?
<anon987321>hey guix, i'm getting errors when i try to use guix import
<anon987321> this is what i geth
<nckx>anon987321: Works for me. Do you have the nss-certs package installed?
<nckx>If not, do 🙂
<anon987321>just installed it, same issue
<roptat>make sure your SSL* environment variables are defined
<nckx>Specifically, $SSL_CERT_DIR and $SSL_CERT_FILE
<nckx>On Guix System, they point to /etc/ssl/certs and /etc/ssl/certs/ca-certificates.crt, respectively.
<anon987321>that was the issue
<anon987321>now it works
<anon987321>seems like variables are wonky on foreign distros
<zimoun>nckx: hum? but why include/ is exported otherwise?
<anon987321>just sourcing .guix-profile/etc/profile isn't enough
<zimoun>do you mean I need to create a lib package?
<zimoun>for example `guix install protobuf@2.6 -p /tmp/proto' populates /tmp/proto/include
<nckx>zimoun: I don't understand your last 3 messages or the problem you're trying to solve.
<zimoun>nckx: I do not understand why protobuf populates include/ if I cannot access to the file.
<roptat>maybe explain what protobuf is and what it does?
<nckx>zimoun: Why can't you access <protobuf>/include?
<nckx>Guix packages do this all the time.
<nckx>(string-append (assoc-ref inputs "protobuf") "/include") etc.
<zimoun>nckx: in general etc/profile exports CPATH or LIBRARY_PATH
<zimoun>in this case, they are not exported.
<nckx>And what do search-paths have to do with it?
<nckx>Does protoc require CPATH and LIBRARY_PATH at run time?
<nckx>zimoun: What are you trying to *do*? 🙂
<zimoun>nckx: Ok, I got your point. Other said, I would like to have a protobuf lib package. I guess
<zimoun>I am working on patch 3 if bug#37988
<nckx>Thanks, I'll take a look.
<anon987321>could anybody help me package emacs-libvterm? i'm not very good at this :c
<zimoun>nckx: thank you
<anon987321>it has a server-client model, so it's not just a matter of importing the melpa archive i think
<zimoun>However, I do not understand how Guix knows if it is used or not at runtime. And so why sometime CPATH and LIBRARY_PATH are exported and sometime not.
<nckx>zimoun: So you're writing your own protobuf@2.6 package now, or are you just using the existing @3.10.0?
<roptat>zimoun, that's the role of the search-paths in some package definitions
<roptat>basically, if a package needs an environment variable, it is defined in its search-paths
<roptat>if the package is explicitely installed (or propagated), the variable is defined in etc/profile
<roptat>otherwise, it's not defined
<zimoun>nckx: I am trying to use the existing protobuf@2.6.0 first.
<nckx>zimoun: You mean 2.6.1?
<nckx>Or is this an older copy of Guix? Current master has no 2.6.0 AFAICT.
<zimoun>nckx: yes, sorry. The R package claims 2.6.0 but it should work with 2.6.1. I hope.
<nckx>Seems likely. What roptat says above it correct, but I fear it might be read as ‘…so you need to add CPATH and LIBRARY_PATH to search-paths somewhere’ which isn't.
<roptat>that's not what I meant
<nckx>I know.
<zimoun>roptat: thank you for the explanations. I think I got the point.
<nckx>But it's what zimoun thought.
<zimoun>no no, I am not thinking... I try to understand :-)
<zimoun>roptat: how Guix knows the package needs an environment variable?
<roptat>zimoun, it's up to the guix developers
<roptat>they have to specify the search-paths for package that use them
<nckx>zimoun: CPATH and LIBRARY_PATH are search-paths for the build systems and compilers. Which is why both should just be set (and just work™) in a Guix build environment.
<nckx>When you add ("protobuf" ,protobuf) to the R package's inputs, Guix will automatically set path variables that point to protobuf/lib, protobuf/include etc, because *consumers* like GCC declare things like LIBRARY_PATH and C_INCLUDE_PATH in *their* search-path. Not protobuf (or the thousands of other packages with /libs and /includes).
<nckx>zimoun: Does that make sense?
<zimoun>roptat, nckx: I am still confused. For example, reading the definition of the packages gmsh (maths.scm:1741) and protobuf (protobuf.scm) I do not understand why gmsh exports CPATH and LIBRARY_PATH and protobuf not. I understand that gmsh requires CPATH and LIBRARY_PATH at run time (I guess) and not protobuf. But how Guix knows that.
<roptat>zimoun, but that's the case, because gcc has a definition for LIBRARY_PATH and CPATH
<zimoun>nckx: yes it makes sense even I failed this way. The simplest first. I probably did a mistake. I will try again using your explanations.
<nckx>zimoun: Your confusion comes from ‘exports’. gmsh is *not* ‘exporting’ its anything, its asking Guix to create the variable for it to *consume*. The variable will contain all directories matching that pattern from all packages in that (build) environment.
<roptat>right, nicely put :)
<nckx>Guix will see that protobuf has a /lib directory, so it will add it to LIBRARY_PATH. protobuf doesn't have to declare anything. It's elegant but confusing until it clicks 🙂
<nckx>roptat: Thanks. I find it harder to explain than to understand.
<zimoun>roptat, nckx: I have not yet clicked ;-) I need to experiment a bit. Thank you both for the explanations.
<nckx>anon987321: I can't dive in to learn how everything works & help you package it from scratch, but we'll certainly try to help with any more specific questions you may have.
<anon987321>thanks, i'll try to get it working later
<roptat>zimoun, maybe this: if you run "guix environment --ad-hoc ocaml-dose3", you get ocaml-dose3 in the environment, but OCAMLPATH is undefined. If you run "guix environment --ad-hoc ocaml ocaml-dose3", suddenly OCAMLPATH is defined to the ocaml package AND the ocaml-dose3 package
<roptat>because the second environment has a package that declares OCAMLPATH, it tells guix to add OCAMLPATH to etc/profile of that environment, and to point it to any lib/ocaml of any package of that environment
<roptat>so OCAML doesn't point to, say, <ocaml-mccs>/lib/ocaml, because it's not part of the environment
<roptat>but if you run "guix environment --ad-hoc ocaml ocaml-dose3 ocaml-mccs", it would
<roptat>because now it's part of the environment, so guix generates OCAMLPATH accordingly
<roptat>does that make more sense?
<roptat>also, neither ocaml-dose3 nore ocaml-mccs have a search-paths specification, guix only adds a path because it exists in the store path of these packages
<roptat>if you run "guix environment --ad-hoc ocaml ocaml-dose3 guile", then OCAMLPATH only points to ocaml and ocaml-dose3, not to guile, because guile does not have a lib/ocaml directory
<nckx>I thinks that's a dandy explanation.
<nckx>zimoun: Re: your last mail to that bug report: a ‘snippet’ is a (modules '((guix build utils))) (snippet '(begin …)) form. It's like a phase, but it runs before the source code is returned to the user. There are *many* examples in gnu/packages/*scm.
<zimoun>roptat: I am trying... :-) Thank you for your detailed explanations.
<nckx>So you can just convert the phase to something like (snippet '(begin (delete-file-recursively "src/win") #t))) and it should work.
<nckx>…and the result of ‘guix build --source r-rprotobuflib will not have a src/win directory.
<fps>is there some hacky workaround for the ssh-server not starting thing?
<zimoun>nckx: thank you. Yes, already did. I am trying to remove the bundled. As explained in the bug report, the ./configure calls another ./configure && make (to build protobuf); I am trying to remove that. By using the Guix protobuf@2.6.1
<nckx>zimoun: That should work!
<janas>Wow I learned a lot from idling on #guix today
<janas>What's the difference between native-search-paths and search-paths?
<zimoun>nckx: I did a try: patch 3 v2 :-) Thank you for the help.
<nckx>zimoun: Looks good, although I didn't test it. If you'd send a V3 for any reason, please put #:use-module (gnu packages protobuf) alphabetically.
<nckx>Oh, and remove 2.6.0 from the description.
<nckx>zimoun: And add the patch to in the same commit.
<nckx>Dammit, I wasn't going to get sucked into reviewing 😛
<dongcarl>Hi all, wondering what people's email setups look like. I've been copy and pasting into a webUI like a n00b
<zimoun>nckx: ah yes, sorry for the 2.6.0 from the description. Rekado told me. And did not know about the alphabetical order. I will do the changes for a V3.
<nckx>dongcarl: That's one determined n00b though. For patches?
<nckx>zimoun: Thanks!
<nckx>(Patches in gnu/ also alphabetical under dist_patch_DATA).
<dongcarl>nckx: Yeah for patches and for correspondence in general... I've been editing in my scratch buffer in emacs, formatting it to 80col then copying and pasting into a webUI lol
<zimoun>nckx: forgot about V3 because I miss the gnu/ part. Sorry
<nckx>dongcarl: For patches I use git send-email, which pops up an emacs buffer to edit the message but sends the mail itself. It's configured in ~/.gitconfig:
<nckx>For all the rest I use mu4e, that other Emacs MUA (after gnus)).
<nckx>TBH it's a janky mess, but a lovable, extensible janky mess.
<nckx>Oh I already said it was Emacs never mind.
<nckx>zimoun: No sorries needed.
<dongcarl>nckx: that was intensely honest and I love it lol
<dongcarl>I'm gunna explore the options a little, see what's out there right now
<vagrantc>dongcarl: i use notmuch-emacs + offlineimap + msmtp for sending mail
<vagrantc>sending, receiving, reading, etc.
<efraim>i recently switched from offlineimap to mbsync. took a few tries to not lose any mail but it's much faster
<dongcarl>vagrantc: That's very good to know!
<efraim>mutt + mbsync + msmtp
*vagrantc has been meaning to look at offlineimap alternatives as it's future is uncertain
<dongcarl>I heard that mbsync can’t handle IMAP keep-alive and automatic resyncs?
<nckx>Oh, I forgot mbsync/isync which I also use.
<efraim>I miss the built-in sleep timer for offlineimap, it worked well as a user shepherd service
<nckx>dongcarl: mbsync can't handle IDLE, no. It's a primitive call-me-once-every-n-minutes thing.
<efraim>I've been having trouble using mcron to read ~/.config/cron/ to pick up cron jobs
<nckx>dongcarl: What are automatic resyncs?
<dongcarl>nckx: Just read from here:
<efraim>I also heard about interimap from the debconf videos, it sounds very interesting
<nckx>Ah, QRESYNC. Never had a MUA that used it so don't know what I'm missing.
<nckx>IDLE would be nice though.
<efraim>from the demo it looked basically like instantaneous syncing
*efraim can't spell
<nckx>mbsync takes about 5 seconds per sync here.
*efraim apparently can spell but has no confidence
<efraim>yeah, but you don't run it in a loop
<efraim>mbsync -a took 17 seconds here
<nckx>efraim: Well, technically I do, but not a tight loop, that is certainly true.
<nckx>Once a minute or so.
<efraim>nckx: with a cron job?
<efraim>I run mine manually most of the time
<nckx>efraim: Kind of: (setq mu4e-update-interval 60)
<efraim>ah, that'd do it
<nckx>I't about as clever as a cron job; if I manually refresh at T-5s, it'll refresh again at T.
<janneke>i'm still using gnus but have been missing m4 a lot, maybe it's time to accept m4 is not coming back and look into offlineimap and friends
<nckx>janneke: ‘m4’? Seriously?
<nckx>Hm, I can't even search for it.
<janneke>nckx: oh no wait eh [n]mh
<nckx>Now that I've heard of (but never used).
<janneke>it was lovely, each mail in a separate file; but i didn't use emacs all that much then
<samplet>Is it a lot of trouble to cancel <>? I want to push a cleaned-up branch soon-ish, and the results of those pending builds are no longer relevant.
<samplet>The branch is “wip-haskell-updates”, to be clear. :)
<efraim>I'm trying to factor out my guix-configuration from my OS-config so I can share it across my machines and I'm having a hard time :(
<efraim>I guess I could share the servers and keys and not an actual guix-configuration
<g_bor[m]>hello guix!
<g_bor[m]>I am trying to test the wayland upgrade on core-updates, and building guile-2.2.6 failed a test. Anyone else noticed this?
<g_bor[m]>Now rebuilding with --keep-failed to get the testsuite log.
<g_bor[m]>The test summary reported three errors, all related to webserver
<htsr>hi guix!
<pat_h>Hello, I am following the instructions given in the Guix System docs here: , but have encountered a problem. I first ran `guix pull` and then `sudo guix system reconfigure /etc/config.scm` as instructed, and was given the following warning: "/root/.con
<pat_h>fig/guix/current not found: 'guix pull' was never run" ; do I actually need to run `sudo guix pull` instead of `guix pull`?
<htsr>pat_h: your user guix and root guix are not the same
<g_bor[m]>pat_h: I believe that should be sudo -E guix system reconfigure ...
<pat_h>Does this mean the doc page linked is incorrect?
<htsr>look at the note
<g_bor[m]>pat_h: I believe so.
<amz3>hello guix!
<htsr>use sudo -i guix pull
<amz3>what the hot topic this days?
<pat_h>I'll try with `sudo -E `, yes I see the note but I am unsure what the implications of it are for this situation. For context, I have just installed Guix System, am planning to only have a single user, and the docs recommend the reconfiguration step immediately after a new install
<g_bor[m]>I don't know when that was written. I believe it is not uptodate.
<pat_h>Ok I see, I
<pat_h>I'll give `sudo -i guix pull` a shot
<g_bor[m]>amz3: One that I know of is adding a branch status page to the website.
<amz3>g_bor[m]: for each branch?
<amz3>g_bor[m]: based on cuirass build?
<g_bor[m]>amz3: actually at least for core-updates and staging, and not based on builds, but based on maintainer decision about freeze and merge deadlines.
<pat_h>If I were to contribute a patch for the docs would it make sense to change it to recommend new users run `sudo -i guix pull` after install rather than `guix pull`?
<g_bor[m]>This has already came up eariler, so that we know without asking around if it is ok to push to core-updates and staging.
<htsr>some links in the devel manual are broken, e.g. at the Introduction link at the begin is a 404 (
<amz3>g_bor[m]: is there a ticket?
<g_bor[m]>pat_h: depends on the way you use it. I usually do not bother with root guix on guix system, but do guix pull and reconfigure using sudo -E, but doing it the other way around can also work
<g_bor[m]>amz3: I don't think so.
<g_bor[m]>What I am trying to do is to update wayland on core-updates right now.
<g_bor[m]>But I encountered a guile test failure...
<lvmbdv>henlo frens, what user:group should guix-daemon run as? (system-wide, just as a package manager)
<amz3>g_bor[m]: tx
<pat_h>g_bor[m]: Ok perhaps there can be an edit to specify both options and the pros/cons of each, which I am ignorant of but can at least provide a little more clarity perhaps
<amz3>g_bor[m]: did you paste the guile test error?
<g_bor[m]>pat_h: yes, that makes sense...
<htsr>pat_h: another solution is to use the root account with su
<g_bor[m]>amz3: I lost it, I am now rebuilidng with --keep-failed to keep the full test log.
<amz3>no promise. I was just curious. But it seems like something I can do.
<nckx>pat_h: Don't recommend sudo guix pull in the manual, it complicates things for no benefit.
<nckx>Thanks for looking into this. Somebody had the same problem yesterday, seems like something's not quite right with the current docs.
<g_bor[m]>there was some inconsistency in the output. Test summary displayed not failed test tests, but then there was another line stating one test failed...
<g_bor[m]>amz3: It still builds... and it looks like it will take long...
<g_bor[m]>I will paste it as soon as I have it, but that might well be only tomorrow...
<pat_h>nckx: Alright I won't then
<amz3>g_bor[m]: did you check in cuirass that guile already built suceessfully in that branch?
<amz3>or the new tool by mbakke (I am not sure how useful in this context it can be)
<htsr>how can I create a one-shot service? can't find it in the docs of shepherd
<g_bor[m]>no, I did not... Well, it does not have a substitute available...
<amz3>is mbakke tool used by the community?
<amz3>not sure about its name
<htsr>ok found how :)
<amz3>htsr: how? :)
<htsr>just add a (one-shot? #t) to your shepherd-service definition
<g_bor[m]>amz3: it seems it was built on cuirass on oct 20:
<g_bor[m]>there is also
<g_bor[m]>after some review, I believe it is good for core-updates
<htsr>with guix, is there something to search which package provide a file? like dnf whatprovides or apt-file search
<htsr>if I add a shepherd service with (requirement '(networking)) it will start every time the networking service is started, right?
<g_bor[m]>htsr: I believe it is the other way around.
<mange>htsr: That just means that it won't start unless the networking service is running, and starting your service will attempt to start the networking service. You would also need to add "auto-start?" to your Guix configuration to make it automatically start on boot.
<htsr>mange: thanks!
<alextee[m]>hi im trying to install guix, how do you convert the iso.xz file to iso?
<alextee[m]>im trying to burn it to a usb
<htsr>how can I restart not manually some services after a guix deploy?
<pkill9>alextee[m]: you need to extrct it with xz
<htsr>alextee[m]: if you are on linux, it should be installed
<g_bor[m]>htsr: I believe it should be done by guix deploy. It should be like running guix system reconfigure, which does that.
<g_bor[m]>i.e. automatically restarts the services it can
<alextee[m]>pkill9: htsr thanks, worked
<mange>g_bor[m]: I don't think you're right. Reconfiguring does not restart running services (see the patches/discussion at, and thus deploying doesn't either.
<htsr>g_bor[m]: it just use the newest service when it restart
<htsr>but is there a way to force restart? or should I do ssh ... -s 'herd restart httpd' ?
<mange>htsr: That's pretty much your only option at the moment. :(
<htsr>thanks :)
<stikonas>what would be a good guix tutorial for somebody not very familiar with scheme?
<stikonas>or should I learn some scheme first?\
<htsr>maybe the blog
<g_bor[m]>oh, yeah... I was in confusion. It seems it only reloads.
<leoprikler>stikonas: have a look at the cookbook
<stikonas>leoprikler: thanks!
<g_bor[m]>I believ like the commenters on that thread said that a finer grain mechanism should be provided by specifying a reload/restart/manual restart only policy...
<g_bor[m]>also it would need some symlinking to get this done.
<mange>g_bor[m]: The patches in that thread allow for services to provide an arbitrary procedure for upgrading, but then the discussion came up that it's really a user policy decision, not a service author decision, and that's where I got stuck.
<g_bor[m]>Currently the only real solution on Guix System is to restart. For example the config file absoult path is sometimes hardcoded, which will be the old version, making a sighup uneffective.
<g_bor[m]>mange: I believe that policy should be simply exposed as confgiuration.
<g_bor[m]>Does that make sense?
<alextee[m]>when it goes to the partition screen (manual), when i press enter on a partition it says "creation error - primary partitions count exceeded"
<htsr>g_bor[m]: in operating-system?
<mange>g_bor[m]: It does, and I agree, but the difficulty is that it's up to the service writer to expose that configuration. I couldn't come up with a general mechanism to expose this as a user policy.
<alextee[m]>i currently have 4 partitions (swap, /, /home, /boot/efi) and i would like to keep at least the /home since i have my data there
<alextee[m]>what does this error mean?
<g_bor[m]>mange: ok, now I see where the real problem lies.
<leoprikler>alextee[m]: there's this limitation, that you can only have 4 primary partitions
<g_bor[m]>And yes, it is the responsibility of the service writer, but actually everything is...
<bluekeys>Please will someone explain to me in terms I can understand how to submit a patch. I have a few emacs packages I've mashed together from elpa, but I don't understand how to do a git patch
<leoprikler>since you've already filled them up, you can't create a new primary one – you'd have to make one of your existing one extended, so you can subpartition it
<mange>bluekeys: Do you have a git checkout of the Guix repository?
<g_bor[m]>There might also be some quite complex upgrade procedures. One example I can think of is upgrading postgresql.
<bluekeys>@mange, yes I do
<alextee[m]>leoprikler: can't i just mount guixsd on a current partition and have it format it and install there?
<mange>bluekeys: Excellent! And you've made your changes in it already, and have them as commits?
<alextee[m]>i dont mind going in parted and cleaning up my partitions, gonna do it now
<bluekeys>I've done the whole pre-inst bit, and the pacakges seem OK and I've got a commit
<bluekeys>Am I far off?
<mange>bluekeys: The usual Guix policy is to have one commit per package added (see "(guix) Submitting Patches" in the manual, particularly point 11). Then you can run "git format-patch origin/master..HEAD" (or whatever an appropriate range is to specify your commits) to generate .patch files that you can send to the guix-patches mailing list. :)
<leoprikler>alextee[m]: i'm not quite sure what you're trying to do here
<bluekeys>OK, I've borked 11 I think and put 2 packages into a commit. But, I've got everything stored somewhere else as .scm files... Let me tidy up.
<bluekeys>Does anyone know the command to drop all local commits?
<leoprikler>Guix does not impose any particular layout onto you. As long as your config.scm fits what you did in parted, things will work.
<leoprikler>Guix also doesn't do formatting for you (perhaps it could do that for guix system init, but for now you're required to do those things manually)
<nckx>bluekeys: git checkout origin/master (origin is the default, anyway).
<alextee[m]>leoprikler: nevermind, i figured it out
<nckx>Guix doesn't use MBR? Oh.
<g_bor[m]>mange: I've just finished reading that thread. It is really nice.
<leoprikler>I have yet to do a Guix instally (or any OS install) with GPT, but GPT should be supported.
<g_bor[m]>It looks like that there are so many ways to manage restarts, that it really makes sense to allow for customisation.
<g_bor[m]>I believe this could be introduced quite seamlessly, by making the current behaviour (i.e. manual restart only) the default.
<g_bor[m]>I am not sure it is the best default, but seems to be the most workable in the current settings.
<nckx>leoprikler: I've never done one without GPT, I'm just surprised that the installer doesn't just use GPT for everything.
<leoprikler>nckx: nice
<leoprikler>I'm not sure whether GPT is widely enough supported to make it the default, though
<leoprikler>and I have yet to see a system, which does not allow MBR
<nckx>leoprikler: UEFI systems don't.
<lvmbdv>is /gnu supposed to be mod 700
<nckx>lvmbdv: I can't see how that could work. It's 755 here.
<lvmbdv>see, i thought so
<nckx>GPT works on both UEFI and old BIOS boxes, but maybe there's a weird old motherboard somewhere that does a silly.
<bluekeys>Whilst I'm waiting for make check. Is it possible to configure guix pull to pull my git source tree?
<stikonas>nckx: strictly speaking UEFI system can support anything (including MBR), there is just some minimal stuff (GPT, FAT32) that they must support, but they can support more
<lvmbdv>ty nckx
<stikonas>nckx: that said, I always use GPT...
<mange>bluekeys: Yep! You can do this using channels (see "(guix) Channels" in the manual). If you use a "file://" URL then it will fetch from your git repository on your local machine.
<bluekeys>ty mange
<alextee[m]>wow im really surprised to see jack and jalv and friends in the guix packages list. nice
<mange>bluekeys: I'm pretty sure it will only pull in committed code, though, so anything you just have sitting in your worktree won't end up getting loaded.
<nckx>stikonas: And *if* its not outright unsupported, since Windows never supported it at al, it going to be buggy. But hypothetical point taken. The whole thing is a glorious gory mess 🙂
<g_bor[m]>amz3: I have the results.
<g_bor[m]>Failure again.
<g_bor[m]>The test summary shows:
<leoprikler>I must do something horribly wrong, because my installation usually ends up something like: try doing stuff with GPT, it doesn't work, MBR it is, then. If I do consider GPT in the first place, that is.
<g_bor[m]>The errors seem to be the curlpit here...
<g_bor[m]>The cuirass build does not have those
<g_bor[m]>the first three web-server tests fail with:
<bluekeys>omg. I have a patch! Is there an emacs command to submit the patch?
<g_bor[m]>(system-error "connect" "~A" ("Connection refused") (111))
<mange>I usually send patches as an attachment on an email.
<bluekeys> ty mange I'll be back tomorrow to finish this but keep an eye out for emacs-dracula-theme... I'd like to begin by thanking the academy and the guix ;) I couldn't have done it without the guix ;)
<kirisime>For some reason neither nautilus nor gthumb is generating image thumbnails, but pcmanfm-qt does.
<alextee[m]>i got an error during installation: failed to get canonical path of /boot/efi
<alextee[m]>and then "failed to install bootloader"
<alextee[m]>am i supposed to create a separate partition for /boot/efi?
<stikonas>separate partition for /boot/efi is a good idea anyway
<stikonas>then rest of the stuff in /boot can be outside FAT and can use e.g. POSIX permissions
<alextee[m]>leoprikler: i think what i was having before is a bug
<alextee[m]>i am pressing enter on an existing partition to change its properties but it tells me that it exceeded the number of partitions
<alextee[m]>but if i delete another partition first it lets me edit those properties
<kirisime>Hm. Actually it seems like all thumbnails nautilus attempts to generate fail and end up in ~/.cache/thumbnails/fail/gnome-thumbnail-factory.
<nckx>alextee[m]: Yeah, /boot/efi is the mountpoint for the (separate) EFI System partition.
<alextee[m]>thanks nckx just to be sure, i should set the boot flag on that /boot/efi partition right?