IRC channel logs


back to list of logs

<lispmacs[work]>leoprikler: looks familiar. Still broken, then?
<leoprikler>lispmacs[work]: as far as I understand yes
<leoprikler>Why are you running a fringe platforms which can't even run a Rust compiler?
<leoprikler>Whyvn: you'd have to do something about the log file on your own, wait a sec
<lispmacs[work]>leoprikler: uh, well, if Rust can't run on x86, then Rust is fringe platform
<leoprikler>But rust can run on i686 if you use their tRust™ed rustc.
<lispmacs[work]>leoprikler: I haven't played around with rust - was just wondering if I would be able to run Icecat on this old x86 laptop
<leoprikler>you'd have to time-machine your way out of that, sadly
<avalenn>cidovul: I had the same reaction reading the thing on m68k
<leoprikler>but idk what icecat needs rust for (probably something very safe only browsers can do)
<leoprikler>Whyvn: your #:log-file seems sane too
<lispmacs[work]>leoprikler: I suppose there is ungoogled-chromium, assuming that is building
<leoprikler>but it might help to strip it, so that you can observe stuff on stdout, then enable it again afterwards
<avalenn>cidovul: the part I agree the more is that "build systems are mess"
***gurmble is now known as grumble
<lispmacs[work]>Is gnome-shell a dependency of %desktop-services? I keep trying to build non-gnome DEs but get gnome-shell pulled in
<lispmacs[work]>like (service lxqt-desktop-service-type)
<lispmacs[work]>and then build fails because gnome-shell is broken on x86
<lispmacs[work]>maybe something I did wrong in the config file?
<dftxbs3e>lispmacs[work], desktop-services has gdm which may include gnome-shell
<dftxbs3e>lispmacs[work], gnome-shell built for me, how do you see it broken?
<lispmacs[work]>on 32 bit x86, gnome-shell depends on ruby-memory-profiler, which fails to build due to a test-failure
*lispmacs[work] leaves to drive to the store
<hpfr>aha, the matrix bridge wasn't working before but it seems to be now
<raghavgururajan>hpfr, hello
<hpfr>hi :) I think element got confused about this room after I got kicked for inactivity, so I had to fully leave and rejoin and sent something to check the logs
***iyzsong-- is now known as iyzsong-w
<apteryx>go libraries should have their inputs as 'propagated-inputs', right?
<apteryx>sirce they don't link and just carry source code
<pocketroid>Greetings programs
<aidalgol>I just got this while running 'guix system container guix-system.scm'
<aidalgol>Looks like the substitute server is giving me a corrupt file?
<aidalgol>Oh, I just had to specify the --fallback option so it would build locally instead.
<aidalgol>So building the container as a non-root user worked, but I can't run it as a non-root user.
***janneke_ is now known as janneke
<aidalgol>Do'h! Missed this bit: "Currently, the script must be run as root in order to support more than a single user and group."
<aidalgol>Next set of problems:
<civodul>Hello Guix!
<cbaines>morning :)
<civodul>hey cbaines, how's everything? :-)
*civodul tracked an annoying test failure in wip-build-systems-gexp
<cbaines>things are going OK :)
<civodul>is it patch-review Friday?
<cbaines>I haven't had much focus lately, so it feels like I've not got a lot done
<cbaines>I was thinking about trying to review patches today, I've got the day off work at least
<civodul>oh nice
<cbaines>I'll see how things go, I've got so many other things to try and push forward as well
<civodul>i've come up with this strategy, which is to not review anything for several days after a review round
<civodul>otherwise i end up unable to keep up
<civodul>(with the replies coming in)
<cbaines>on the more optimistic side, I've got more leave coming up in March, and I'll probably drop down to just Monday to Thursday at work from April, which should mean I can spend more time working on Guix while also getting more sleep!
<civodul>ah, well done! four-day work weeks are nice
<cbaines>yeah, I'm hoping it's going to be nice :)
<cbaines>right, first thing to do today, I want to deploy another bunch of Hurd VMs for Guix Build Coordinator agents, but this time use the new dynamic auth thing I implemented...
<civodul>oh, what's this dynamic auth thing?
<civodul>i think i saw those words on guix-commits but didn't pay enough attention
<cbaines>Rather than creating individual IDs and passwords for each agent, you create a token, and then agents request a UUID+password from the coordinator using that token
<cbaines>so you can configure a bunch of agents with different names (like the machine hostname) and the same token, and they'll individually login
<cbaines>which should make setting up a bunch of childhurd VMs simpler, and maybe things like running agents on top of a scheduling system (like OpenShift, ...)
<civodul>cbaines: interesting!
<pimi>hi guys
<pimi>I need a little bit of help, I am packing a python package and the sources from pypi does not have a requirements.txt, but I saw a file requires.txt that have the same content as the requirements.txt from github sources inside the fake package(*.egg-info) from the pypi sources
<leoprikler>IOW the importer fails and you have to do stuff manually?
<pimi>is it OK is move requires.txt as requirements.txt? or there is another approach
<leoprikler>Guix itself does not read the requirements.txt other than during import.
<pimi>but python setup give me the error that there is no requirements.txt
<leoprikler>hum, if renaming the file after unpack fixes the package, than go ahead.
<leoprikler>but I fear the build might be more broken than that
<leoprikler>which package is it btw?
<pimi>thanks, I will try this
<leoprikler>hmm, you could try using git as source, no?
<leoprikler> has a normal requirements.txt
<cbaines>success :) I should soon have 16 childhurd VMs trying to build packages
<cbaines>unfortunately, they still regularly hang, so they're not making steady progress
<civodul>another problem is that the build environment on GNU/Hurd still isn't properly defined
<cbaines>yeah, indeed
<sharlatan>Hi all
<sharlatan>I want to assamble mdadm raid for already configured disk which came from other install, I could not find any examples in docs
<cbaines>sharlatan, which docs were you looking at, as that doesn't sound directly related to Guix?
<AllanAdair[m]21>shcv: about your python package. Did you try using a
***ece5 is now known as ece
***ece6 is now known as ece
<balance>Hello! Can anybody help with next issue (running on notebook i686), "guix pull" was successfull and than next command: "guix system reconfigure /etc/config.scm" warning: cannot determine provenance for current system ... build of /gnu/store/v3zgvwa1c0w44skld09iminwrg5j93c9-ruby-memory-profiler-1.0.0.drv failed ... guix system: error: build of `/gnu/store/k25lm8nxsiba262y92ps802n79f8x5dz-gnome-shell-3.34.2.drv' failed
<madage>balance: do you use channels?
<mrprofessor>So I have installed sway in guix. How do I start it?
<balance>madage: seems no, this is just simple installation on crypted partition
<madage>hmm.. this message is telling there is no path from your current guix to the one you are trying to upgrade to
<madage>'guix system describe'
<sharlatan> cbaines: Guix mapped devices - raid-device-mapping. I pazzled with this "it is chosen during the initial creation and formattion of the RAID"
<sharlatan>so I wonder if it destroys data which is currently present on RAID
<mrprofessor>I was following this
<mrprofessor>got the same error, did what the OP did and sway started. But I don't see any display server
<madage>balance: and 'guix describe'
<balance>madage: guix describe: guix 6891f95 repository URL: branch: master commit: 6891f95739c733df217ceaf5d0787cbed380ec1c
<pkill9>mrprofessor: run `sway` in the console to start it
<mrprofessor>pkill9 I am running it in vbox also running `sway` gives me this error
<mrprofessor>“XDG_RUNTIME_DIR is not set in the environment. Aborting.”
<balance>madage: guix system describe: Generation 1 Mar 05 2021 02:29:04 (current) file name: /var/guix/profiles/system-1-link canonical file name: /gnu/store/kgz6rpf9k9s34wg8i0ypid5xxkgp0s2j-system label: GNU with Linux-Libre 5.9.3 bootloader: grub root device: label: "fsname" kernel: /gnu/store/j3jg1j0q7g7kz8j77fp5137hcwxr8hfy-linux-libre-5.9.3/bzImage configuration file: /gnu/store/49dk51n83krpk00agzpmd49x6qym0hmx-configuration.scm
<madage>the guix system describe is missing the 'commit' info, which is what actualy matters
<madage>this is a fresh install, right?
<balance>madage: yes, fresh install, installed from usb, then reboot "guix pull" "export PATH="$HOME/.config/guix/current/bin:$PATH"" "export INFOPATH="$HOME/.config/guix/current/share/info:$INFOPATH"" and then "guix system reconfigure /etc/config.scm"
<balance>madage: all these commands are from root
<madage>balance: why?
<madage>also, nevermind, I mistook the warning for an error
<madage>it looks like your system tried to upgrade even though it couldn't check the provenance
<madage>have you tried it again?
<madage>use and paste the full logs from the reconfigure failure
<balance>madage: yes tried about 3 or more times, results are equivalent
<balance>madage: used this instructions
<roptat>looks like it failed to build ruby-memory-profiler. Are you on x86?
<roptat>(32 bits?)
<balance>roptat: yes x86 32bits (lenovo ThinkPad x60)
<roptat>I've seen a bug report recently about that, not sure it's fixed yet
***pocketroid_ is now known as pocketroid
<madage>nice, kudos raghav!
<cbaines>civodul, I did see that, it's interesting, although it seems to me like continuing with the client side changes here is still a good strategy
<civodul>cbaines: yes, definitely
<cbaines>the tracking aspect is somewhat already supported by the Guix Data Serivce, although that still needs improving
<civodul>yes, that + 'guix challenge'
<civodul>it'd be nice to have a blog post one of these days to summarize where we are
<cbaines>I think we're so close to having a client side implementation... at least a basic one
<cbaines>civodul, I saw you pushed a fix to the backtrace printing issue in the substitute script, that will be really useful
<cbaines>civodul, I'd like to go ahead and merge the (guix substitutes) module creation at some point in the next week, I don't know if you have any more thoughts on that change?
<civodul>cbaines: oh i thought you had done so recently
<civodul>lemme check
<civodul>i hadn't seen you were also looking at those backtraces, apologies :-/
<cbaines>no problem, your approach looks neater, although I'm not sure I understand it yet
<civodul>there were two parts: not unwinding the stack, and sending the stack trace to FD 4
<civodul>the latter was not happening due to the way with-continuation-barrier works
<civodul>kinda obscure
<civodul>cbaines: the (guix substitutes) patch LGTM; i'm not a fan of things like %unreachable-hosts in a library-looking module, but maybe we can improve that later
<civodul>(with the risk that it won't happen :-))
<roptat>ah, camlboot timed out on ci :/
<roptat>is there a way to increase the timeout?
<zimoun>if a package contains #:make-flags (list "foo"), then another package inherit and contains #:make-flags (list "bar"), does the inherited one append or replace?
<nckx>If you want to combine them, you have to tell Guix how, using substitute-keyword-arguments.
<zimoun>nckx, thanks. So I do not understand why commenting the #:make-flags in vim fixes the build in vim-full where vim-full contains a #:make-flags.
<cbaines>civodul, yeah, the unreachable-hosts hash is a little awkward, being that it's a UI thing inside an error handler. Personally I'm tempted to wait until looking at doing retrying better before trying to fix this
<cbaines>roptat, I think Cuirass now supports using the timeout related package properties
<balance>madage: done, thanks,
<balance>roptat: looks like the same error, want to install mate desktop too, but seems this problem is not in mate)
<apteryx>it'd be nice to have cycles auto-detected and diagnostic printed when it occurs
<apteryx>builtin Guile
<cbaines>apteryx, there's a long lost patch which sounds similar
<cbaines>I don't know if it still works
<civodul>cbaines: yes, it makes sense ot keep it that way for now; so, LGTM! :-)
<cbaines>civodul, cool, I'll test it locally some more, and merge if I don't see issues...
<civodul>roptat: you can add a max-silent-time property to the camlboot package
<civodul>assuming max-silent-time is what needs to be changed
<civodul>cbaines: yay! be sure to run "make as-derivation" too
<apteryx>cbaines: oh! thanks for the pointer
<cbaines>apteryx, there's a commit here that might be more up to date
<nckx>balance: You didn't paste a log but it's almost certainly the same failure.
<balance>nckx: log means '/var/log/guix/drvs/v3/zgvwa1c0w44skld09iminwrg5j93c9-ruby-memory-profiler-1.0.0.drv.bz2'?
<zimoun>nckx, in the case of vim-full, for an obscure reason, the list is not replaced at all.
<rekado> contains a link to the cookbook, but it is relative to the /manual URL
<rekado>the URL is, but it should be
<cbaines>zimoun, the arguments for vim-full include the arguments for vim, with a change to the phases
<cbaines>zimoun, if changing the vim #:make-flags has an effect on vim-full, then the arguments from vim are probably taking precedence
<zimoun>cbaines: why, I thought inherit simply copy then anything in the inherited package overwrites.
<cbaines>zimoun, you're right, vim-full is overwriting with a value that incorporates the vim arguments. See the bit where it uses (package-arguments vim)
<zimoun>cbaines: ah right! Thanks, I got it. :-)
<cbaines>this means the vim-full arguments includes #:make-flags twice, with different values, and in that scenario, only one of those values will be used, and I'm not sure which one (although I think you've worked that out experimentally)
<cbaines>maybe packages should be considered invalid if they've got duplicate argument values
<roptat>balance, nckx looks very similar to
<roptat>it might be an issue in ruby itself: "guix build --with-input=ruby@2.6=ruby@2.7 -s i686-linux ruby-memory-profiler" passes all its tests successfully
<zimoun>cbaines: in this case of vim-full, it is the one from vim, which leads to build failure.
<roptat>(we use 2.6 by default)
<roptat>although they report an issue with 2.7, so I'm not sure
<nckx>balance: (Bit late, but for future) Yes; you cat view it with bzcat.
<nckx>or *your cat can, I guess.
<nckx>zimoun: Doesn't what vim-full does result in a list with two #:make-flags keywords? (Just from a quick glance at Web git, might be wrong)
*nckx AFK.
<madage>balance: no need to thank me, I was leading you astray, I'm sorry.. your problem is probably what nckx and roptat are talking about
*madage suggests bzless, just in case the logs are longer than screen size
<shcv>AllanAdair[m]21: Oh, thanks; I didn't know about I was eventually able solve to it by writing a custom build step in though
<balance>nckx: corrected (with log now), thanks
<Sleeping-Frog>A package on Guix can't be build, and several other packages don't work at all.
<Sleeping-Frog>Where can i report them?
<roptat>can someone restart evaluation of camlboot now?
<roptat>Sleeping-Frog, if it's ruby-memory-profiler, we are working on it, it's already reported at Otherwise, send your report to, specifying architecture and adding the log file for the failed package
<Sleeping-Frog>Thank you!
<civodul>roptat: do you have a link to the failed build?
<roptat>I'm mostly interested in
<roptat>but other architectures need love too :)
<apteryx>has anyone seen that one? 'Wrong type to apply: #<syntax-transformer set-grafting>' while running 'guix graph --path'
<cbaines>roptat, do you know why the build process is silent for so long?
<apteryx>full backtrace to the earlier error:
<roptat>it's taking a lot of time to build a specific file, and need to wait for it before it can continue, so it cannot do anything else in parallel
<roptat>apteryx, something similar happened to me once in another project, and it was because I had an old .go cached
<apteryx>so a possible fix could be 'rm -rf ~/.cache/guile/ccache' ?
<apteryx>tried, but it persists
<roptat>in my case it was in a ./pre-inst-env, so 'find -name '*.go' -delete'
<apteryx>ah, or 'make clean-go'
<shcv>is there a way to pull just one channel?
<roptat>if you don't pull the guix channel at least, you're going to be in trouble :)
<roptat>if you mean, update only one channel, then I think you can pin all other channels to a specific commit in your channels.scm
<shcv>hmm; I was just thinking it might save time / reduce load on server if I only pulled the channel I'm working on
<roptat>then, pin all other channels to their current commit
<roptat>then it won't need to rebuild all the stuff for them, it'll only rebuild your channel
<roptat>if you only specify one channel and it's not guix, the result will be incomplete (it won't be able to use guix packages, it won't have the guix command, etc)
<shcv>an option for guix pull that temporarily pins selected branches to their current commit sounds like it might be useful though
<nckx>A ‘guix pull --channels="..."’ UI WBN.
<nckx>Hah :)
<roptat>although, it's already used to specify a channels.scm file
<shcv>yeah; maybe "--update" or "--only"?
<shcv>or go the other way and do "--pin"
<roptat>we could have both
<nckx>Oh, whatever. My first thought was ‘guix pull --only’, but that's a bit too clever.
<apteryx>shcv: looking at this another way, is there no other flow when working on a module than pulling?
<nckx>--channels is a really misnamed squattypants :(
<apteryx>on a channel*
<shcv>apteryx: what do you mean?
<apteryx>by using adding the local channel repository to the load path directly, via -L for example?
<shcv>so... test my package with "guix environment --ad-hoc <package> -L <channel path>"?
<roptat>that works most of the time, but sometimes it doesn't catch issues that appear when compiling the channel in guix pull
<roptat>but it works for testing packages
<nckx>...oh? “'--url', '--commit', and '--branch' [apply only] to the 'guix' channel.” Might want to mention that wart in the help message, then... (I'll add it.)
<nckx>We can always extend it to support --commit=[CHAN=]commit later; I was just surprised.
<civodul>roptat: did you try setting a max-silent-time property?
<civodul>that would trigger a rebuild
<apteryx>roptat: you were correct about needing make-cleango; I don't have the error anymore (only a hang due to the cycle I'm trying to detect :-D)
<roptat>civodul, I did
<edgar-vincent>Hi all.
<nckx>Hullo edgar-vincent.
<roptat>civodul, I noticed that some changes I made were not taken into account, for instance, I made at least two changes to php that are not present on ci
<edgar-vincent>I am running Guix System. Unless I'm mistaken, despite what is written in the 'Getting Started' page of the manual, I do get the `Consider setting the necessary environment variables` hint after running `guix install`. Should I set the env vars as suggested?
<civodul>roptat: i've restarted from the web UI (first time i click on this button :-))
<roptat>civodul, for instance, this is the commit where I added max-silent-time:
<roptat>it says nothing to build
<civodul>ah, maybe it doesn't work the way i thought it works
<civodul>yeah prolly it's still the same derivation
<edgar-vincent>(BTW, does anyone use Matrix on Guix? If so, which client do you use?)
<nckx>edgar-vincent: If you want to set the new variables in the current shell (there's no way for Guix to do that for you) because you want to run new software that relies on them.
<nckx>They'll be present automatically in any new terminals you start.
<shcv>edgar-vincent: I had been using the emacs matrix-client.el, but it's not compatible with emacs-native-comp yet
<roptat>same here, I removed a failing test on arm in php: but it says nothing to build
<roptat>civodul, so for php, it should not be the same derivation, yet it was not taken into account
<edgar-vincent>nckx: Thanks. I would like, actually, to use matrix-client.el, as shcv mentioned. However, I share my Emacs init file with other, non-guix systems, and I was thinking of running the guix-specific `require` only if GUIX_PROFILE is set (presumably in .bashrc?). Is it the right way to do this?
<shcv>edgar-vincent: it looks like the only currently packaged clients are quaternion, emacs-matrix-client, and nheko
<cbaines>roptat, I believe doesn't compute armhf derivations currently
<edgar-vincent>shcv: Yes - I tried nheko, but it won't log in. I think someone is working on packaging Fractal, but I don't know if it's still being worked on.
<roptat>cbaines, sure, but it should have changed all other architectures too
<roptat>at least aarch64
<shcv>I don't think you need to have any guix-specific requires if you don't want to
<cbaines>roptat, the Guix Data Service suggests it didn't
<roptat>I mean, I changed the guile code inside a phase, so the derivation should be different
<nckx>edgar-vincent: I don't think GUIX_PROFILE is set on Guix Systems.
<nckx>It's not on mine
<edgar-vincent>nckx: It isn't here either.
<shcv>It looks like guix sets EMACSLOADPATH though
<nckx>It's an implementation detail anyway. That said (although I'm not an emacs wizard), I think shcv is right that you shouldn't need to sniff out Guix in your emacsrc.
<nckx>If that's needed it might be a bug?
<roptat>cbaines, oh!
<roptat>I see, thanks :)
<cbaines>roptat, I think the lines you changed were within a if block that looks at the system
<roptat>yes, and it's not part of the phase for other architectures
<roptat>now I understand
<edgar-vincent>shcv, nckx: I've added `(require 'matrix-client)` in my init file, but it fails.
<edgar-vincent>(and the package is installed, obviously)
<nckx>civodul: Any chance of the Guile/Guix REPL supporting C-o? It's a bash habit of mine when working from history.
<shcv>edgar-vincent: if you install it via guix instead of some other emacs package manager, you probably want to call (guix-emacs-autoload-packages)
<edgar-vincent>shcv: That's was I'm worried about. I don't think I should be running that on a non-guix system.
<nckx>edgar-vincent: Does one of the directories in $EMACSLOADPATH contain the matrix-client .el file?
<civodul>nckx: yup that'd be nice!
<civodul>we'd have to check the (Guile-)Readline to see how this is implemented
<nckx>edgar-vincent: I don't use autoloads but my .emacs is full of (require 'foo)s that work fine on Guix System.
<nckx>mu4e etc.
<edgar-vincent>nckx: $EMACSLOADPATH contain: /run/current-system/profile/share/emacs/site-lisp:/run/current-system/profile/share/emacs/27.1/lisp
<nckx>Same as mine.
<nckx>civodul: Sweet.
<edgar-vincent>and emacs-client.el is in ~/.guix-profile/share/emacs/site-lisp/
<shcv>edgar-vincent: you can use (fboundp ...) to check if a symbol has a function binding, and (autoloadp (symbol-function 'function)) to check if it's an autoload object; I don't know if you need the latter given the former
*nckx AFK.
<edgar-vincent>sry matrix-client.el*
<edgar-vincent>shcv: Thanks, I'll try that.
<edgar-vincent>`require` still fails after evaluating (guix-emacs-autoload-packages). There's something very basic I must be missing.
<edgar-vincent>hi cage_
<shcv>are you sure you installed emacs-matrix-client? I just installed it, ran emacs -Q, and got it to load with (require 'guix-emacs) (guix-emacs-autoload-packages) (require 'matrix-client)
<edgar-vincent>shcv: same thing here :(
<edgar-vincent>oh wait
<shcv>do you set load-path in your init? maybe you need to (add-to-list 'load-path "~/.guix-profile/share/emacs/site-lisp")
<edgar-vincent>shcv: Yes, that works.
<edgar-vincent>I'll do the (add-to-list 'load-path ...) only if (fboundp guix-emacs-autoload-packages) is t. Thank you!
<edgar-vincent>Do you use emacs as a service (daemon)?
<edgar-vincent>(I have an awful lot of silly questions, installed guix only yesterday...)
<shcv>I would like to, but I haven't figured that part out; plus I break it too often to make much of a difference anyway
<shcv>partly because my current main setup is in WSL2 (shhh)
<shcv>that and I was having trouble getting emacs --daemon to work with EXWM
<shcv>it worked on my previous setup though...
<shcv>it'll be nice when emacs can be a wayland compositor instead
<edgar-vincent>Do you think it is wise to 'define' a new channel in /etc/config.scm by getting it to write to /etc/guix/channels.scm?
<edgar-vincent>I'd like (if that's wise or even doable) to be able to define my whole system from config.scm. Instances of my system should have to same extra channels.
<shcv>sounds reasonable to me
<shcv>but others with more experience might know better
<shcv>e.g., there may be a more direct way to specify channels
<edgar-vincent>Also, related to that, I read on Ambrevar's blog that it is possible to to something similar with GPG keys, etc. Ideally, I'd like to be able make instances of my system where no extra setup is required.
<edgar-vincent>Ok, thanks :)
<edgar-vincent>to do*
<civodul>nckx: BTW, shell-mode lacks C-o too; do you have a trick for that?
<lederstrumpf>hi, I'm wetting my feet with guix system and set up a server with encrypted root & boot partitions. to decrypt it remotely, how can I enable enable preboot ssh, say via openssh or dropbear - I couldn't find kernel modules to add to the initrd image for either with linux-libre.
<civodul>hi lederstrumpf! i'm not familiar with pre-boot ssh, it could be that it's not supported yet
<civodul>now, you can customize the code that runs in the initrd
<civodul> has some info
<civodul>rekado: the build farm has never been this busy: :-)
<roptat>civodul, I suppose we don't see all the workers there, do we? for instance camlboot is being built, but not present on that page
<lispmacs[work]>hi, I'm having some difficulty and confusion about something: I've got guix installed on an old 32 bit x86 laptop. I adjusted my system config to include mate-desktop-service-type. But for some reason the laptop tries to build gnome-shell
<lispmacs[work]>I switched mate-desktop-service-type to lxqt-desktop-service-type, but system reconfigure still tries to build gnome-shell
<lispmacs[work]>I was wondering if maybe gnome was a dependency of %desktop-services, which I still have in the file, but that doesn't make much sense
<roptat>lispmacs[work], maybe it's gdm?
<lispmacs[work]>roptat: I'm open minded. does lxqt use gdm? or is that a dependency of %desktop-services?
<lispmacs[work]>here is my config file:
<roptat>it's part of %desktop-services
<roptat>it's the login screen, it's not part of any desktop
<lispmacs[work]>roptat: does gdm run in gnome-shell?
<lispmacs[work]>probably does
<roptat>mh... maybe not, gnome-shell is not part of the dependency graph
<roptat>it uses gnome-session though
<roptat>well, you should see a path in the dependency graph that explains why you need gnome-shell
<lispmacs[work]>roptat: heh, I wonder if it has something to do with the set-xorg-configuration
<roptat>when your reconfigure fails, it should print every intermediate package between your system and the failed package
<roptat>ah indeed,
<roptat>gdm-shepherd-service does use gnome-shell
<nckx>lispmacs[work]: Yeah, GDM needs GNOME shell to run.
<lispmacs[work]>ricklepick[m]: nckx: here is the finally output of the reconfigure failure:
<nckx>civodul: ...not use shell-mode? I'm an emacs traditionalist: it's a surprisingly good text editor for a mail client.
<lispmacs[work]>I opened a bug report on the build failure, but was trying to figure out why I needed to build gnome-shell at all
<nckx>I live outside of it.
<roptat>lispmacs[work], right, ruby-memory-profiler, you're on x86 (32 bits), right? :)
<lispmacs[work]>roptat: yep
<nckx>lispmacs[work]: Because you can't run GDM ‘outside’ of GNOME shell, they are integrated. The r-m-p failure on i686 is, unfortunately, known ☹
<nckx>You're the 3rd person to report it today alone, that's pretty bad. Sorry.
<mrprofessor>Hey people!
<mrprofessor>I have installed guix without a DE and was hoping if I can get help running i3 or sway.
<lispmacs[work]>nckx: I guess the question is now what is the easy way forward to getting lxqt running on this old laptop. I'm not quite sure how to replace gdm with something else in my config
<roptat>lispmacs[work], can you try to reconfigure with --with-input=ruby@2.6=ruby@2.7 ?
<zimoun>what is the current blocking for merging core-updates? I mean the last merge seems from May 2020. If all the branch is not in a state to be mergeable, maybe we could simply merge a part of it. WDYT?
<lispmacs[work]>roptat: if I try adding that command line flag to a system reconfigure call, it says that is not a valid option
<lispmacs[work]>I can successfully do that with a guix build call
<roptat>oh no :/
<lispmacs[work]>so, that transformation works, but I'm not sure how to get that into my reconfigure
<nckx>mrprofessor: You've come to the right place! No need to ask to ask; just ask.
<zimoun>if someone could explains me the reason of the backtrace I get in <>.
<mrprofessor>I am pretty new to guix and love the concepts behind it. Also I am new to setting up DE all by myself. So I followed this and tried when that didn't work.
<nckx>zimoun: Oh dear... so my guess was right?
<Noisytoot>The issue with the formatting in the mistletoe patch is that Parinfer's paren mode automatically changes it
<Noisytoot>So I'll have to disable it
<nckx>zimoun: The original vim-full looked dodgy, your patch looks correct. Your backtrace is simply because you're calling the string "LDFLAGS=-lexpat" as a procedure.
<zimoun>nckx: yeah the duplication of #:make-flags as cbaines pointed out is not good. But I miss why `(append '("foo") '()) works and simple `("foo") fails.
<zimoun>nckx, but why it is called?
<nckx>`(cons "LDFLAGS=..." ,flags) works, `("LDFLAGS=...") won't. You seem to be one level of quotation ‘off’, mentally :)
<zimoun>ah it makes sense! Thanks
<nckx>zimoun: <I miss why `(append '("foo") '()) works and simple `("foo") fails> For the exact same reason that, stripping one level of ` from each, (append ...) works and ("foo") doesn't.
<nckx>zimoun: OK
<zimoun>nckx, yeah I got it. :-)
<nckx>I'm glad it makes sense because the more I wrote the less it made sense to me 😉
<zimoun>it should be `(list "foo").
<zimoun>well, it would be simple to just write `(list "LDFLAGS=-lexpat") instead of the append+delete.
<lispmacs[work]>is it possible to install lxqt into the system without gdm?
<roptat>sure, you can use another dm
<mrprofessor>So I am ran `WLR_DRM_DEVICES=/dev/dri/card0 XDG_RUNTIME_DIR=~/.tmp sway -d 2> ~/.tmp/sway.log` and looks like sway is running. How do I switch to it?
<roptat>like slim or sddm
<lispmacs[work]>roptat: is there any instructions on that in the manual?
<roptat>lispmacs[work], yes, let me find it
<nckx>zimoun: Both work. `(list "LDFLAGS=-lexpat") is simpler but means vim-full won't inherit any new fixes from vim. It's your judgment call which approach will reduce maintenance/mistakes.
<roptat>lispmacs[work], this example for slim
<roptat>and you can use something similar for sddm
<Noisytoot>I have sent the fixed patch to, is that correct?
<nckx>mrprofessor: So you want to run Sway manually (like old ‘startx’) without using a login manager?
<Noisytoot>It isn't shown on or yet
<mrprofessor>@nckx yeah I thought so. How else can I run it?
<roptat>Noisytoot, it'll show up eventually, don't worry
<lispmacs[work]>roptat: okay, thanks
<nckx>mrprofessor: I expect most people to launch it with a graphical log-in screen, not by manually typing stuff. This is how I start Sway:
<lispmacs[work]>roptat: will I still include lxqt-desktop-service-type in the normal way?
<roptat>lispmacs[work], yes
<mrprofessor>@nckx could you please share your config?
<mrprofessor>It goes in config.scm right?
<nckx>It goes into your system configuration's (services ...) field, yes.
<nckx>mrprofessor: I don't use %desktop-services, so I don't know what that cookbook patch is trying to say.
<nckx>I don't share my complete configuration, too much trouble to redact and it wouldn't be as useful as you think. Far too much custom weirdness.
*nckx AFK.
<mrprofessor>np @nckx. Thank you though!
<mrprofessor>So after editing config.scm how do I reload it? @roptat @nckx
<nckx>mrprofessor: To reload/apply your configuration: sudo guix system reconfigure /path/to/your/configuration.scm; you'll use that command a lot.
<nckx>Sometimes followed by a reboot.
<nckx>(Side note, not for now: to update your OS you'll use ‘guix pull && sudo guix system reconfigure ...’, ‘guix pull’ alone won't update anything.)
*nckx AFK again, but good luck. If you'd get stuck here there's also help-guix at
<pkill9>how feasible would it be to use gpg-encrypted text files uploaded to a public server as private messages, where the client tries to decrypt every uploaded file (except those it's already read, maybe deletes those) until maybe it finds a message that it can decrypt, which means it for them? I guess I've probably just reinvented an inefficient form of email maybe
<pkill9>actually no it's not email, because it doesn't send anything to any servers, it just lets clients find the messages
<mrprofessor>nckx: makes sense
<pkill9>and obviously only people 'following' you would possibly 'receive' the message
<mrprofessor>@nckx I am getting errors here
<tricon>@pkill9 1. Great name. 2. Seems pretty inefficient, but it's not a bad idea at small scale.
<sneek>Welcome back tricon, you have 1 message!
<sneek>tricon, nckx says: No, I wouldn't put wpa_supplicant.conf in a package. I'd use ‘plain-file’ to include the text in my system.scm (or your own module if you like that kind of thing) and ‘etc-service-type’ to add it to /etc.
<tricon>@pkill9 I began running my own mail server for a somewhat similar situation: I can GPG encrypt incoming email with my public key, decrypting it locally on the client after pull. Email is obviously inherently insecure, but it at least reasonably protects my email at rest.
<tricon>(Or that's the goal. Still need to tweak a few things.)
<mrprofessor> /msg NickServ VERIFY REGISTER mrprofessor quntfqhvivym
<tricon>Oops indeed.
<nckx>> mrprofessor has quit (Quit: #guix)
<nckx>Poor fellow.
<tricon>He just got educated.
<nckx>* hunter2 has joined.
<pkill9>tricon: is it easy selfhosting an email server?
<pkill9>including maintenance
<emacsen>pkill9, medium, and even someone like me, who has been doing it for ~20 years has started to use an out of the box solution
<pkill9>this article i'm reading says it's the configuration that is hard
<lispmacs[work]>pkill9: i did it a few years ago. Setting up the server was not too painful, but the real nightmare is keeping your email server off and on all the whitelists and blacklists
<pkill9>that's what i heard lispmacs[work]
<emacsen>pkill9, it's mostly the changes to DNS and things that is hard.
<tricon>@pkill9 I wouldn't say it is trivial, but I don't find it too complicated either, in part due to the fact that I have been able to discover copious documentation for my particular setup.
<pkill9>how do you do that lispmacs[work]?
<pkill9>i would think guix makes it much easier to host a mail server
<nckx>It does.
<pkill9>do you host your own mail server nckx?
<lispmacs[work]>pkill9: years ago I ran a mail server on Gentoo, I think it was.
<tricon>@pkill9 I admit: My mailserver is running on Debian in Amsterdam, but I would very much enjoy converting it to Guix.
<lispmacs[work]>more properly, there was an imap server and an smtp server and an authentication server
<nckx>pkill9: Yeah, on Guix System.
<tricon>Yes. I use Postfix + Dovecot + Sieve
<nckx>I don't understand why would throw: error: (xorg-configuration (xorg-configuration (keyboard-layout keyboard-layout))): extraneous field initializers (xorg-configuration)
<lispmacs[work]>dovecot + postfix + cyrus I think, or something like that
*nckx same as tricon, with Roundcube for when I'm without sane computing.
<pkill9>nice nckx, i don't suppose you'd be willing to share the config?
<nckx>Eh, no, not same, idiot, I use OpenSMTPd.
<pkill9>nckx: do you run into many problems?
<nckx>I have to go again (sorry about that) but will respond later!
<lispmacs[work]>I got to the place where I could send and receive emails, but was constantly dealing with issues were such and such big name email services was convinced I was a spammer server because some dns detail wasn't perfect or I wasn't on some whitelist I had never heard of
<tricon>@pkill9 Seems to me like you could fulfill your messaging idea with a private mail server. I admit that may be like swatting a fly with a nuke.
<tricon>lispmacs[work]: I imagine DKIM and SPF would go a long way towards resolving much of that; but yes, that is extra work.
<tricon>Email was great before everyone "else" got on the Internet ^_^
<lispmacs[work]>there are some protocols you have to have turned on as well where the servers involved talk to each other a few times back and forth to be sure nobody is a spammer
<lispmacs[work]>and be sure your provider has your reverse dns correct, not that they will care if it isn't
<tricon>This ^
<tricon>@rudr7942 Greetings.
***rudr7942 is now known as mrprofessor
<tricon>Oh it's mrprofessor!
<tricon>I've been duped.
<mrprofessor>Ahh..sorry for spamming. Setting up my irc client.
*lfam enjoys the variety of commit messages about removing grafts
<lfam>I like efraim's the best
<lfam>"Absorb replacement"
<lispmacs[work]>i got lxqt and slim installed on this 32 bit Panasonic CF-30
<tricon>I like old hardware.
<lispmacs[work]>i asked IT dept for a an x86-64 laptop with 8 GB of ram. I got a CF-30 from the XP era
<lispmacs[work]>but the sturdy case is cool
<tricon>That granted me a chuckle.
<lispmacs[work]>not to mention the awesome DB-9 connector
<mrprofessor>haha that's a tough one!
<tricon>Toughbooks are tough. We have employees here that want them, but my IT director made a good point that we can replace a current laptop 3x for the price of 1x Toughbook.
<tricon>@mrprofessor Well played.
<lispmacs[work]>QTerminal. That sounds fancy
<mrprofessor>I personally love thinkpads but I wish the base models had a better screeen. The mac screens are quite good and less pricey now-a-days.
<tricon>We use Lenovos/Thinkpads. They're solid. I'm endeared that they still have the mouse-stick (or whatever that is called).
<tricon>I very much agree re: Mac displays.
<mrprofessor>I know people who use that red mouse stick without moving from home row. That's pretty cool IMO.
<simonsouth>It's called a TrackPoint and yes, once you get used to it it's wonderful.
<tricon>@mrprofessor I never considered that.
<tricon>@simonsouth Gratsi for the correction.
***zimoun` is now known as zimoun
<Noisytoot>Could someone merge
<Noisytoot>I have sent the improved patch
*lfam looks
<mrprofessor>Can someone help me with my config? I have added @nkcx's bit to my config and there are some errors.
<mrprofessor>I can learn scheme tomorrow:(
<lispmacs[work]>roptat: hi, I got slim and lxqt installed, but can you help me with a follow up question? to get it to work, I had to comment out my set-xorg-configuration line, gives me an error about xorg being defined or configured twice.
<lfam>Noisytoot: Your patch doesn't apply to the current master branch. Can you rebase it and send a revised patch?
<lispmacs[work]>roptat: do you know how I could reintegrate that in?
<lfam>mrprofessor: What errors do you get?
*Noisytoot git pulls
<lfam>And what are the "bits" you added?
<Noisytoot>What "bits"?
<Noisytoot>If you mean the changes to the variable above, I sent a reply, which removes that
<Noisytoot>But I'm not sure how to link to a reply
<lfam>Sorry Noisytoot. That question was for mrprofessor
<mrprofessor>this was his bit.
*Noisytoot compiles the latest Guix
<mrprofessor>I was trying to set up sway
<cbaines>Noisytoot, one other comment is that it python-mistletoe might better fit in the markup module. python-xyz is rather large and if a more appropriate module exists, it would be good to use it.
<lfam>mrprofessor: And the errors?
<mrprofessor>one sec
<lfam>cbaines: From your build farm, do you have any sense of the status of core-updates? Like, is it a "totally broken collection of patches"? Or something better?
<mrprofessor>(xorg-configuration (xorg-configuration
<mrprofessor> (keyboard-layout keyboard-layout))) extraneous field initializers (xorg-configuration) @lfam
<nckx>tricon, mrprofessor: TrackPoint is the one true pointing device. I'll never use anything else again.
<nckx>☝ weird thing is that's a verbatim copy from my config and it ‘guix system build’s fine.
<nckx>I don't *think* it relies on my aforementioned custom weirdness...
<lfam>It's weird that it does wayland and X, but I don't have experience with graphical Guix System
<mrprofessor>So don't use gui at all @lfam, or use guix as a env manager?
<lfam>I use Guix on other distros, or headless
<nckx>lfam: What's weird about that?
<lfam>nckx: It's not weird if it works for you!
<nckx>Is there a new way?
<lfam>I thought it was like, one or the other
<lfam>You should probably take over debugging here. I was just guessing
<cbaines>lfam, regarding core-updates, it's possible to process the branch at least, but it looks like there are quite a few Python and Rust packages that have broken compared to master
<nckx>I feel obligated to help but don't really have time to dive deep :(
<lfam>Well, there's no obligation. Got to keep that in mind
<lfam>One can help best when they want to
<lfam>cbaines: Alright, thanks
<lfam>People are wondering... when will core-updates happen?
<lfam>From my point of view, we couldn't have possibly begun before January, and at that point staging was in progress
<mrprofessor>np @nckx. Thank you for your help though.
<cbaines>lfam, I can perhaps try merging master in to core-updates at some point in the next few days
<mrprofessor> /JOIN #go-nuts
<mrprofessor>This irc client sucks
<lfam>cbaines: I think it should depend on the plan for the upcoming release, and if we decide to include core-updates in that release
<lfam>If we decide to release on April 18, I think that we will have to exclude core-updates
<cbaines>yeah, I'd like another release, and I don't think it's helpful to rush core-updates
<pkill9>what irc client are you using mrprofessor?
<lfam>It really depends on the amount of helpers available
<cbaines>I still want to make process improvements so things in general are easier
<lfam>There's a lot of room for improvements :)
<mrprofessor>pkill9 circe for emacsen
<mrprofessor>for emacs
<cbaines>lfam, if you have things on your mind, I'd love to know :)
<lfam>Not really
<lfam>zimoun and I are discussing the plan on guix-devel, in the "release on april 18?" thread
<cbaines>ah, well, problems or ideas, I can also try and do some thinking
<nckx>mrprofessor, lfam: Oh, I didn't mean I didn't *want* to; I have to force myself not to :)
<lfam>Boy can I relate
<mrprofessor>@nckx Understood. I just realised that y'all are the maintainers.
<tricon>mrprofessor I have yet to use Circe. Just been using ERC.
<nckx>mrprofessor: I wanted to try something more 1337 than my trusty HexChat, so I tried circe. It tried to load my 500-line, 100+-channel ZNC scrollback and promptly (well...) took so long that the connection timed out again.
<nckx>Some emacs software is just not good.
<Noisytoot>Why does Emacs need so many IRC clients
<tricon>I rather enjoyed playing with jj, but in the end I decided to favor something within Emacs.
<mrprofessor>I agree. @nckx @tricon circe seems a bit lousy. And I am using irc after years tbh. My computer is full with shitty electorn apps(discord/slack/teams)
<nckx>...and to prove that I have absolutely no self-discipline I debugged mrprofessor's system .scm anyway 😛 The error message is very misleading: the problem is a missing ‘sddm’ service module import.
<lfam>Addicted to helping
<mrprofessor>nckx: Please start a patreon channel.
<nckx>Then it returns the more reasonable: service 'xorg-server' requires 'elogind', which is not provided by any service
<lispmacs[work]>can somebody look at my .scm? I had to comment out my set-xorg-configuration sexp:
<nckx>‘nckx is procastri^W helping people on #guix, please donate today’.
<tricon>mrprofessor Electron wretches me. We use MS Teams at my place of work, and the other head IT guy and I just use Slack (myself via emacs-slack). I was constantly complaining how utterly slow MS' Electron-based MS Teams "app" is.
<lispmacs[work]>but wanted to reintegrate that back in somehow
<mrprofessor>This year my company decided to migrate to teams. It's a piece of shit and needs intune to be installed in phone. That gives the employer the ability to wipe my
<mrprofessor>@nckx ye!
<Noisytoot>mrprofessor: Refuse to use it
<Noisytoot>Or get them to provide you with a phone
<mrprofessor>Yep already did and I got a work phone. Now I have to carry 2 phones.
<mrprofessor>and an extra cable for charging.
<mrprofessor>@tricon since you have already looked into it, is there something I can do to fix it?
<Noisytoot>" error: version mismatch. This is Automake 1.16.2,
<Noisytoot> but the definition used by this AM_INIT_AUTOMAKE
<Noisytoot> comes from Automake 1.16.3. You should recreate
<Noisytoot> aclocal.m4 with aclocal and run automake again.
<mrprofessor>@nckx sorry for you ^
<mrprofessor>I meant @nckx not @tricon
*Noisytoot guix pulls
<pkill9>that sucks mrprofessor
<pkill9>my course uses microsfot teams, which isn't so bad because it's really just used for conferences calls, sharing files, and emailing/messaging
<pkill9>but i wouldn't want to use it for work
<tricon>@mrprofessor I carry a work phone now as well. I have Teams on it, though I rarely use Teams from the phone.
<mrprofessor>It's free with office subscription, slack is pricer that office subscription. Do the math.
<tricon>The more I utilize Teams the less impressed I am with it. Of course, this goes without addressing the fact that it is proprietary through-and-through.
<tricon>mrprofessor Yes, it's a "bargain" to the enterprise.
<tricon>Teams feels like a bunch of half-baked features that all work very slowly. It does voice/video meetings pretty well, though.
<mrprofessor>yeah! It's an utterly lousy app tbh. I hate it with passion.
<cbaines>I probably updated Emacs earlier today, and now output to the Emacs shell is really slow... has anyone else seen this?
<mrprofessor>So all my team mates
<tricon>cbaines: I hate to say it, but I ended up getting used to the slowness of Eshell. If you find a solution, I'd love to hear it.
<mrprofessor>@tricon yeah the video meetings are good. The office integrations are good. and that's about it.
<tricon>mrprofessor: Agreed.
<pkill9>i wonder how well jitsi + nextcloud would be as a replacement
<cbaines>tricon, I'm using shell mode, rather than eshell, and it wasn't this slow before (as far as I remember)
<tricon>pkill9: You have me curious as well.
<tricon>cbaines: Roger.
<tricon>I've really enjoyed Jitsi. Been promoting it as "much easier to use than Zoom."
<pkill9>is signal the best messaging application to use?
<pkill9>for security etc
<pkill9>plus ease of use
<pkill9>afaik it is
<tricon>Wouldn't it be funny if Gemini/Gopher end up being The Solution to deal with all this fingerprinting/spyware/tracking/Internet-suckage?
<tricon>pkill9: I really like Telegram; but it's not E2E by default.
<tricon>So I'd probably vote in favor of Signal as the the best buy default.
<pkill9>I'm thinking of asking soeone I talk to to use that to talk to me
<nckx>mrprofessor: Just adding (service elogind-service-type) to (services ...) makes it build. I didn't try running it.
<pkill9>since it basically functions like whatsapp, except it's FOSS and audited and not owned by facebook
<shcv>tricon: don't forget matrix
<tricon>shcv: I did! You just reminded me of that, thank you.
<shcv>pkill9: reading the backlog, your encrypted messaging idea sounds very similar to
<shcv>tricon: it looks like it's picking up some steam; have you seen the FOSDEM presentation? Their new work on Spaces sounds really cool
<tricon>I really like federated services.
<tricon>shcv: I have not. I shall watch that this weekend.
<shcv>It's probably on peertube somewhere, but here's the yt link:
<mrprofessor>@nckx still the same error. I feel helpless since I can't read scheme. I would try installing with i3 tomorrow.
<nckx>pkill9: I don't really share configs, sorry, but I found all the horror stories about mail overblown, at least for personal mail. I spend much more time fixing software upgrades than fixing mail problems. OTOH if (say) setting up things like FCrDNS is already intimidating, and not the first trivial formality you take care of when setting up a server, it's probably not for you. That's fine too.
<nckx>mrprofessor: You didn't add (use-service-modules desktop networking ssh xorg sddm) <-- yet.
<nckx>That will give you the elogind error, which adding the line above will fix.
<shcv>pkill9: BitMessage is built on bitcoin tech, and basically gossips encrypted + PoW stamped blocks to all nodes; if you can decrypt it, it's for you
<lfam>Gosh that sounds inefficient
<nckx>pkill9: That said I have had some fun problems, because e-mail is a rigged monopoly, and not all blacklists are actually benevolent.
<mrprofessor>@nckx Thank you! It's building now.
<nckx>mrprofessor: `/gnu/store/` pops up a Sway desktop!
<nckx>That should have been: `guix system vm /tmp/mrprofessor.scm` (including the `s)
<nckx>(Copy/paste is badly broken here and I don't know whether to blame Sway or Guix)
<nckx>(Or Alacritty.)
<nckx>It hasn't led to embarrassing pastoes but it's only a matter of time.
<nckx>Anyway, that command is a great way to test your configuration without having to bother with rebooting or roll-backs.
<shcv>where is the use-setuptools? configuration parameter for the python-build-system default value set?
<roptat>lispmacs, sorry for the lack of response, my computer crashed and then I had a meeting ^^'
<mrprofessor>makes sense @nckx
<lfam>shcv: In guix/build-system/python.scm
<nckx>shcv: (build-system python).
<nckx>lfam: I don't think so?
<mrprofessor>Since y'all are here, how do you find time to work on guix when you have a job/exams?
<shcv>weird; I was looking in guix/build/python-build-system - what's the difference?
<roptat>lispmacs, since you manually specify the slim service, you can add a configuration to it: (service slim-service-type (slim-configuration (xorg-configuration (xorg-configuration ...))))
<lfam>shcv: It's the difference between so-called host-side and build-side code
<tricon>mrprofessor: I can justify it as part of my job. Major blessing.
<lfam>What don't you think, nckx?
<roptat>lispmacs, (the first xorg-configuration is the name of the field, the second one is the name of the record that you had in set-xorg-configuration)
<shcv>mrprofessor: I'm also trying to use it to make my experiments reproducible
<nckx>shcv: That's where it's handled, but not (AFAIK) where the default is set, which is line 168 of (build-system python).
<nckx>lfam: ☝
<nckx>Because :)
<lfam>We gave the same answer nckx
<lfam>I wrote it out with filesystem paths, and you use Guile module paths
<roptat>civodul, "building of `/gnu/store/vrsyk998wcyxs1m1h51m2nramgag8ibg-camlboot-0.0.0-0.506280c.drv' timed out after 3600 seconds of silence" did I make a typo?
<nckx>lfam: Yours would be (guix build-system python).
<mrprofessor>@tricon how do you mean?
<lfam>nckx: Yes, your answer made a typo
<roptat>ah yes I did
<lfam>I was going to ignore it ;)
<nckx>lfam: No typo, I was in the wrong directory 🤦
<nckx>lfam: Don't ignore typoes!
<nckx>They won't ignore you.
<Christoph[m]>Have you heard about Nix working towards a content-addressed store?
<lfam>Yes, we have heard about it
<Christoph[m]>Is that relevant to guix, too?
<nckx>Many a time.
<tricon>mrprofessor: I deploy and manage servers as a part of my job, which gives me an opportunity to dive into Guix as the OS host.
<lfam>It could be, Christoph[m]
<nckx>Christoph[m]: Absolutely, at least as an interesting test case, maybe more (depending on the results).
<lfam>It would represent a fundamental overhaul of Guix, so it's not something that is answered simply
<mrprofessor>@tricon cool. yeah that makes sense. :)
<nckx>Christoph[m]: It's not clear yet that CAS is an improvement.
<lfam>Nix is obviously bigger and more established, but as an outsider I'm not convinced that Guix can learn from it
<lfam>Beyond what we already took, of course
<lfam>I mean, just because they did something, it doesn't mean we should also do it
<Christoph[m]>lfam I once heard civodul say that the store is a monad. That gave me the impression that the store is well separated by a nice api. Is that wrong?
<nckx>‘Copy’ not what I meant by ‘learn‘.
<lfam>Each project has its own mistakes and advantages
<civodul>roptat: yes, there's a typo: 'properties' must be an alist
<nckx>If it fails horribly we can learn from that.
<civodul>so: '((max-silent-time . 43234))
<roptat>civodul, yes, fixed it just now
<roptat>sorry for the noise ^^'
<civodul>did evaluations succeed meanwhile?
<roptat>nope, after 3600 seconds of silence
<civodul>i believe things break if 'properties' is not a properties
<civodul>er, is not a proper list
<roptat>looks like I didn't break anything
<civodul>good :-)
<lfam>Christoph[m]: I'm the wrong person to ask :) I have little experience working at that level
<roptat>you can still assoc-ref a list that's not an alist
<roptat>it just returns nothing
<roptat>er, #f
<nckx>Christoph[m]: Nice is subjective :) but an API would still need to be adapted to a radically different addressing method. I don't know if Nix's CAS is so radically different; another thing we can learn.
<Christoph[m]>I assume that guix has grafts as an answer to some of the problems Nix tries to solve with content-addressed store, like unnecessary rebuilds in certain circumstances?
<lfam>More generally, I'd say that Guix somehow avoids the problems that made this change attractive to Nix
<roptat>CAS would have been nice for camlboot :)
<civodul>Christoph[m]: i don't think a CA store solves the same problems as grafts
<lfam>Like you said, we have grafts
<nckx>Christoph[m]: I still don't really understand how that's supposed to work. You still need to build it at least once, and you still need to unconditionally trust whoever tells you you don't have to bother.
<nckx>That doesn't require CAS at all.
<pkill9>is there much benefit of icecat over chromium?
<lfam>The example given in the Nix article, that changing a build dependency of Go should cause rebuilds of Go software, seems to miss the point of Nix entirely :)
<lfam>I mean, that it *shouldn't* cause rebuilds
<lfam>But, I must be the one who misses the point
<dftxbs3e>how does CAS wok?
<lfam>That's the article
<mrprofessor>@nckx I definitely don't think it's a disk space issue
<dftxbs3e>lfam, I read the RFC but I don't understand
<civodul>also, there's content-addressibility, and there's the CA store and corresponding model
<nckx>lfam: Perhaps, but we're in the same boat then.
*nckx offers rum.
<lfam>I think that, basically, we don't really understand the POV of the Nix team, and they might not understand our POV, if they needed to
<nckx>mrprofessor: Can you share the contents of that bz2 file?
<lfam>Guix has some advantages that make other choices less attractive
<nckx>E.g. with ‘bzcat FILE | xclip’, or ‘bzless’ and manual copying, or...
<dftxbs3e>how can you depend on the output of something..?
<dftxbs3e>how do you get back to source then?
<lfam>But, I suppose the very special case of "changed inputs" but "bit-identical outputs" would be nice to address cleanly
<dftxbs3e>it says CAS will not be everywhere because it couldnt
<cbaines>so, somehow Emacs shell is unusably slow on my desktop now after pulling in some updates, but using previous Emacs outputs also exhibits the same behaviour :(
<Christoph[m]>Well, if you build one package and notice that it didn't change, every rebuild a change in this would have caused will be skipped.
<lfam>Right, just seems like something that happens rarely, given the work required to implement a completely new model
<lfam>But, maybe I've tried to avoid that case for so long that I am blind to the use cases
<nckx>pkill9: Less evil upstream, less participation in the proprietary Web that Google is working towards, also (a teensy bit) faster IME.
<lfam>I'm more interested in the adjacent blog post How Nix grew a marketing team
<nckx>mrprofessor: back in about 15m.
<Christoph[m]>civodul Are you suggesting a combination of input- and content-addressed? Maybe links with the content-addressed name onto the input-addressed store item?
<roptat>this is the extensional model described in the thesis, right?
<civodul>a more complex variant thereof to allow coexisting with the intensional model
<civodul>Christoph[m]: i'm not suggesting anything :-), just saying that we need to clarify what the goal is
<roptat>Christoph[m], not exactly, I think there's an issue if you build two packages with just the output input-addressed store name changed, you might get different outputs that differ only by that name
<civodul>often it seems that "CA store" is a goal in itself
<civodul>but i don't think that should be the case
<roptat>need to afk :)
<rekado>one thing that annoys me about our store is embedded file names and I think we’d gain much more from separating store file names than we would from a CAS store.
<rekado>I idly wonder, for example, if it were feasible to stash store reference in a separate file and wrap executables so that they magically use those separate references
<rekado>we have related magic in “guix pack -RR”
<rekado>if two outputs only differ in their references then extracting those references would allow us to reuse the same output for both builds. Grafts would become trivial, too
<nckx>mrprofessor: ‘No space left on device’ -- OK, and why do you disagree?
<nckx>‘Space’ can be more than mere bytes. Check ‘df -i’ for example.
<nckx>There's also a bug in ext4 (although civodul disagrees -- I'd love to know more) that Guix tends to evoke, although it's less likely.
<pkill9>what' sthe purpose of a content-addressed path?
<mrprofessor>It's a new vm with auto resizable disk
<Christoph[m]>rekado What are embedded file names?
<nckx>mrprofessor: There's no such thing as auto-resizable from the perspective of the guest, but it shouldn't matter.
<nckx>If you have enough free space (also in /tmp, but that's part of / for you) and inodes I'm not sure what to recommend anymore.
<mrprofessor>hmm? What's the `dynamically allocated` disk in vbox?
<nckx>Christoph[m]: That built software (e.g., binaries) contain embedded file name strings.
<mrprofessor>I thought it's auto resizable
<nckx>mrprofessor: That just means that VirtualBox simulates a (say) 50G hard drive but the image won't actually be 50G at the start. Sectors will only be written on the host when they are first used by the guest. But the guest always sees a 50G drive, and the ext4 file system on it will in no way change its size on the fly.
<mrprofessor>np nckx. I will start with a DE. However this seems like a cool place to hangout and I will be here.
<nckx>It's a technique to save space on your host OS but not relevant inside the VM.
<mrprofessor>'t know that.
<nckx>mrprofessor: Are you using ext4?
<nckx>I recommend trying again with btrfs (you can keep the same system configuration, no need to start from scratch).
<mrprofessor>I used the graphical installation and it defaults to ext4 I guess?
<nckx>It does.
<mrprofessor>how do I use btrfs?
<nckx>I'm considering wether we should change that.
<lfam>We need to hire a PR department first
<nckx>mrprofessor: I'm not too familiar with the GUI installer; is there no option to select the fs type?
<mrprofessor>AFAIK there was ext and msdos with guided partiton.
<nckx>lfam: We actually should (not hire, but grow).
<nckx>But as good hackers we look down on the people who do so, whilst they overtake us at high speed and become $relevant. 😉
<lfam>I use btrfs but there are always some comments when I mention it. I can't imagine the FUD-storm if we made it the defualt
<lfam>I don't look down on those other distros. But I think I'm a minority here, seeing the value of commercializing things
<nckx>mrprofessor: I think you're confusing two things but it's understandable. I'd try the installer again and see. btrfs should be an option, or it's definitely something that should be fixed.
<pkill9>looking at it seems signal is indeed the best easy-to-use messaging application in terms of privacy and security
<nckx>lfam: I didn't mean you personally lfam. I said it because I knew you agree.
<lfam>I know :)
<mrprofessor>nckx thanks for being patient tho.
<mrprofessor>Signal is pretty good pkill9
<nckx>mrprofessor: I'm not, but I try to work on it, so thanks for mentioning it.
<pkill9>it was even commended by the NSA for it's security: "In December 2014, Der Spiegel leaked slides from an internal NSA presentation dating to June 2012 in which the NSA deemed Signal's encrypted voice calling component (RedPhone) on its own as a "major threat" to its mission of accessing users' private data"
<lfam>Some people dislike Signal for what I consider to be niche complaints. The goal of Signal is not to help, like, North Korean dissidents, but to make total surveillance of communications at the ISP level impossible
<mrprofessor>They are too powerful for plebs like us.
<lfam>It's a first step to making the internet "private by default". It's not a total solution to all threats
<lfam>Similar to HTTPS
<lfam>It's for people who already communicate with smartphones
<lfam>If one finds that unacceptable, they won't like Signal
<mrprofessor>People get prosecuted because their chats are accessible to the govt in diff countries.
<lfam>SMS / MMS is recorded by default by competent governments. Signal is an alternative to that
<mrprofessor>One got arrested because she contacted greta thunberg to spread awareness for the farmer's protest in India.
<lfam>That's surprising. Whatsapp uses the same cryptosystem as Signal
<mrprofessor>for fucking sedition
<lfam>The police must have accessed this person's phone and read the chats there. Signal wouldn't help with that
<mrprofessor>yeah ^
<lfam>I tell people that "you never know what will be illegal in the future". Under Trump it was an easy sell
<mrprofessor>encrypt the fucking phone, don't use face/fingerprint scanner etc
<mrprofessor>or don't irk the govt.
<lfam>I mean, if the government is out to get you, your phone chats are not going to be relevant
<mrprofessor> True.
*lfam afk
<pkill9>signal has a self-destructing message feature though
<pkill9>that could be helpful
<mrprofessor>yeah. Don't keep chat backups too.(sensitive)
<pkill9>but it requires you to use it ofcourse
<mrprofessor>protonmail is pretty interesting too.
<tricon>mrprofessor: ProtonMail is what I switched from. Don't get me wrong: I am grateful for them and their service; but I wanted something a little more extensible, and piping everything through the bridge (via mbsync) for local, non-web-based delivery became slow and cumbersome.
<tricon>There are also some concerns about the browser as an attack surface, as local storage in the browser is utilized for the key that is used to decrypt messages on-demand.
<dftxbs3e>Protonmail's bit a of a lie, their stuff wasnt Free Software for a while (some parts still arent), also JavaScript cryptography in the browser is insecure in general because code control is still at the server's
<tricon>And: I have heard it stated that law enforcement can leverage who you are talking to and when enough that the content of the messages become less relevant (this is debatable depending upon circumstance, of course); and the headers are not encrypted, only the body.
<tricon>dftxbs3e: Agreed.
<shcv>and remember, being able to read your messages only matters if they either haven't decidid you're a target yet, or they follow some kind of due process
<shcv>is it possible to make more stripped down files with guix pack? Like with ad-hoc environments
<shcv>because my python package is producing a pack tar of 400MB because it's pulling in GCC etc