IRC channel logs
2022-12-14.log
back to list of logs
<mitchell>Hello guix, Does anyone know how I can do package substitution on g-expressions? Specifically I want to modify the cryptsetup package used by the luks-device-mapping. <mekeor[m]>TIL polkit rules are written in javascript (and evaluated with mozillas javascript-engine "spidermonkey"). how ugly to use javascript for auth :( <mekeor[m]>mitchell: as a scheme/guile noob, i wonder, if you could simply wrap your (operating system) into a (let ((cryptsetup MY_CUSTOM_CRYPTSETUP))). i guess, it won't work. <mekeor[m]>i just tried that scenario in a guile repl. it did not work indeed :) <lechner>apteryx / hi, i reconfigure that node all the time, but i don't reboot often because the static networking messes up the consuming services. should i restart mcron? <rekado>I just realized that 389-ds-base provides no way to update a directory server instance’s configuration <rekado>you can only create it from scratch <rekado>or update individual values one by one <lechner>mitchell / you can try something like (set! cryptsetup ...) but that's never worked for me. <rekado>lechner: but an instance is a database; removing it destroys all data stored in it. <rekado>it’s as if we had a postgresql system service that would wipe the database on each reconfigure. <rekado>the whole thing is also stateful <lechner>oh sorry, i made an openldap type distinction between db 1 and 2 (one holds the "configuration" and the other the directory data) <rekado>e.g. all instances are written to /etc/dirsrv/registry <rekado>not sure if I can make this declarative by moving things into /gnu/store <rekado>I think you can add records from ldif files, but not the instance config <rekado>you can print the whole config with “dsconf instance-name config get” <lechner>rekado / how is updating an instance's configuration different from updating records? <rekado>different configuration file formats and property names <rekado>the problem is merely that there’s no built-in way to migrate an instance to a new configuration <rekado>and I don’t feel like we should aim to do stateful operations as part of the system service <rekado>maybe it should just do nothing but print a warning when an instance of the given name already exists <rekado>let the user figure out what to do about it <rekado>(delete it, back up the records, etc) <lechner>mitchell / the cryptsetup variable does not appear to be parameterized in gnu/system/mapped-devices.scm. i face a similar issue with xkb-config for the keyboard configuration <lechner>rekado / ideally, we would have our own directory server, perhaps called Shelter, that rewrites the operating system declaration for future deployment <rekado>we seem to have a disagreemetn about the meaning of the word “ideally” :) <nckx>rekado: Bummer about the SSDs, but at least it makes things simple. <nckx>But with 659% more Unicode. <jlicht>should a package generally look for polkit actions in its own output, or rather in some (services-configured) /etc/... location? <Guest51>if I try to do it manually it will prompt me mount: /mnt/cdrom can't find in /etc/fstab <nckx>That should happen only if you don't specify the device to mount, or didn't add a device to your file-systems field. <nckx>(It sounds like you're just running ‘mount /mnt/cdrom’, say. There's no way the system can know what you want mounted there without you telling it somehow. That's not related to Guix.) <nckx>It's been many years since I've touched a CD drive but I believe they're called /dev/sr0 or so… You can add that to your ‘mount’ invocation, or add something like (file-system (device "/dev/sr0") (type "iso9660") (mount-point "/mnt/cdrom") (mount? #f) (create-mount-point? #t)) [untested] to your file-systems. <jgart[m]>Hi, do home shepherd home services run on foreign distros like debian/ubuntu? <apteryx>lechner: yes you can simply restart mcron <Guest51>@nckx I am kindie newbie in guix. I have solved wıth my solutıon wıth udısk <Guest51>Also a stupid question can I install 'guix' on 'real hardware'. Will binary blobs result in some problem ? <lechner>apteryx / i did, and the spurious entries disappeared. thanks! <lechner>Guest51 / I run Guix only on "real" hardware, and on all of it---except my router <nckx>I don't have any hardware that requires blobs though. <lechner>but do you consume microcode updates? <apteryx>ACTION tries the 1.4.0rc2 guix-install.sh along teh guix-binary-1.4.0rc2.armhf-linux.tar.xz on a buildroot armv7 machine <apteryx>so far I've had to guix pack a bunch of stuff to help it, such as gpg, getent from glibc, groupadd from shadow <apteryx>also GNU tar, because busybox tar was throwing tar: invalid tar magic <apteryx>then hack the install script because I have /gnu and I don't want it to warn me (these are my packs...) <apteryx>currently failing at: [1670993830.%3N]: [ INFO ] Installing /var/guix and /gnu... mv: can't remove '/gnu': Is a directory <apteryx>ah, because of my pre-existing /gnu... hm. <cronoStar>Hey can some one help me out with emacs-guix. If anyone who has it installed, can you M-x guix-command and tell me if you just get shell options or a list of guix arguments. Im only getting shell options, and dont want to spend the time debugging and fixing. <cronoStar>It would be really hard for me since im such a novice programmer and not get at elisp. :( <apteryx>cronoStar: I get: [...]guix/scripts/deploy.scm:177:7: Unknown # object: "#~" <apteryx>It's known that emacs-guix needs more love <cronoStar>Set (guix-repl-use-server nil) then try again. Thats another bug I got on my list. <apteryx>cronoStar: like (setq guix-repl-use-server nil) ? <cronoStar>Yah im trying to get better at elisp and reading alot of the code base. Trying to get emacs-guix up to speed with mainline guix would just be killer. <cronoStar>in eshell or the scratch buffer eval and then try M-x guix-command <apteryx>now I see: r Run in Guix REPL s Run in shell c Copy command line <cronoStar>So theres two bugs, one is the gexp error you get is from the socket repl not being able to read the short hand gexp "#~". But it can read the normal gexp procedure -> https://issues.guix.gnu.org/59256. The guix-command just doesnt do the generation from the code. <cronoStar>No probs :) emacs-guix is so cool and could that one killer feature for an already killer program. <lechner>cronoStar / with that work you could make a lot of people very happy <jgart[m]>seems like python-flask has not been updated in a while but there are a few tickets open to update it <jgart[m]>how should we unstuck that? Should I send new patches? Last time iirc I was told that someone had already submitted an update to python-flask but then the already submitted patch set wasn't merged and is still open <apo11o>cryptoguix - Approved by the FSF, used daily my RMS <apo11o>will be under development until 2050 tho <jgart[m]>We should start the meme coin in order to fund the Hurd <jgart[m]>But that might procrastinate it's development further <apteryx>bah. relocatable guix packs don't seem to work on that box <apteryx>it says -sh: /bootstrap/gnu/store/l6nb9j7gw4f1lk157g3picjmqpybqpaz-gnupg-2.2.32/bin/gpg: not found, but it's there <vagrantc>ACTION actually tests the "graphical" guix installer for the first time <vagrantc>i think most of my previous Guix System installs predate it <vagrantc>either doing it from a foreign distro or just the earlier installer where you manually walk through steps <mekeor[m]>hello. is it possible to define custom polkit rules in guix system? <ssb>hi! Do I understand correctly that using "guix build" inside guix system container is not implemented? <ssb>(in the container that is started by "run-container" script, that is) <rekado>mekeor[m]: it’s generated by “guix system container” <rekado>ssb: you can talk to the host system’s Guix daemon by sharing locations explicitly <mekeor[m]>i see. so the explanation would be "because there is no guix daemon running inside the container"? <ssb>mekeor[m], guix daemon is runnig inside the container, but the /gnu/store mounted read-only by the script, and --share option cannot ovveride this <rekado>ssb: in my experiments with rstudio I --expose=/gnu/store and --expose=/var/guix/profiles, as well as the socket location. <rekado>you cannot have the daemon run inside the container because you don’t want two different daemons to touch the same store <rekado>the container merely shares a subset of /gnu/store from the host, so it should never write to it <rekado>let the host system’s daemon write <ssb>right, I'm experimenting on a foreign system without outer daemon <rekado>no daemon but /gnu/store exists? <ssb>copied from working guix system <ssb>it sort of works, even. After hacking the script to allow read-write bind mount of /gnu/store I was able to build something, but somehow managed to destroy guix binary ) <rekado>does your copied store include a valid /var/guix database? <ssb>rekado, nope, thanks for the hint! <rekado>whenever the daemon sees stuff in /gnu/store that’s *not* also in the database it will delete it as junk <rekado>the contents of the database are not deterministic <jonsger>will there be a merge of master into version-1.4.0 before release? <rekado>for service activation snippets it would be nice to have a prefix for any output that gets printed to the console <rekado>for example, the postgres services runs “initdb” which prints a few lines that can easily be confusing in the output of “guix system reconfigure” <rekado>389-ds-base has output like “Starting installation ...” <rekado>users have no idea that this is output from 389ds <rekado>another example: “Error: root_password must be at least 8 characters long” <rekado>would be useful, I think, to prefix all output with the service name that produced the output <bredge>Please make a proot container for termux 😡 <rekado>that’s just a substring of your previous message. <ArneBab>When I log in two users via GDM on separate X11 sessions, whenever I switch between them, the programs in the inactive one lose internet connectivity. Do you know why that happens — and how I can change it? <rekado>ardon: I think it would be good to have a dconf system service that would allow us to set arbitrary dconf key/value settings. <apteryx>nckx: actually the changes I thought may be responsible are in master, not yet released in 3.0.8 <apteryx>so I'm not sure what ends up doing fstatat <nckx>I did a (very) quick grep of {guile-,}sqlite, but only for statx, not fstatat… <apteryx>right, and fstatat is very old, probably not an issue <ekaitz>hi all!! do we have a guide on making new build-systems? <nckx>No results for either, and this is a very inefficient/unreliable way to find out who calls what, I know. <elevenkb>tried two hard packages and failed (pretty much) so i'm hoping that i'll manage with an easy pure elisp one. <lembrun[m]>unmatched-paren: Note that this machine is a brand new thinkpad, with ssd, powerful cpu etc <nckx>ekaitz: I don't think so. Forking by example, I'm afraid. <elevenkb>so... does one have to add a build phase in order to compile the texinfo documentation of a simple elisp package? <ekaitz>nckx: which one is the best as a reference? i need something that calls a build step, just like `make` does but nothing else <attila_lendvai>will guix deploy send a lot of network traffic to the machine where it's deploying? is it different with build-locally? #false? <nckx>Or minetest-, for its smolness. <nckx>I also chose minetest- because it inherits from gnu-, which is what I'd suggest most new build systems do. <nckx>Which build system is this? <lechner>attila_lendvai / just enough to install everything, i believe. most activity seems to occur on the host from which 'deploy' is called. i think everything goes into the local store first. <unmatched-paren>ekaitz: dub is the Awful Combination Of A Build System And Package Manager that D uses <ekaitz>unmatched-paren: yeah I just read the beginning of the file hehe... looks like I can use it btw! good as a reference <unmatched-paren>ekaitz: dub-build-system is simple because it doesn't handle dependencies yet <ekaitz>should it? like... how do we handle dependencies in the build system? we download them too? <nckx>‘Implicit inputs’ is probably the answer. <nckx>Oh, or that way, if you like pain. <nckx>Please avoid inflammatory opinions like ‘works’. <apteryx>haha, perhaps this is the error message thrown when my local machine hasn't yet authorized the offload machine? <apteryx>it's thrown by the local guix-daemon <nckx>apteryx: The >32-bit one? <nckx>I used to know what that actually means, but I frogot. <lembrun[m]>but the thing is I get the blinking cursor thingy only with seatd <apteryx>nckx: yes, I'm adding the key to confirm <lembrun[m]>with elogind I can atleast get to wlgreet type my user and password but I get thrown back to the login over and over <apteryx>nckx: guix offload: successfully imported <unmatched-paren>ACTION would like to see if their config works by default but doesn't know how to run a graphical guix system vm <lembrun[m]>unmatched-paren: Don't have much in there, only error I get is unable to receive ipc reponse <tricon>lembrun[m]: then your OS is complete. <apteryx>nckx: indeed. didn't expect spending half a day configuring an offload machine. all this is to be able to natively build guix system, hoping it might run on that same target after ;-) <nckx>…and I just got your mail. <nckx>How's it intertwined with the daemon code? <apteryx>it's in a low-level readInt procedure <apteryx>so we have to see in which context it was used and do some error checking before, I guess <nckx>I meant more that it's (probably; total guess) reading garbage as an Int at all. <unmatched-paren>does guix system vm just create a window with the virtual display in it? <apteryx>nckx: perhaps local-store.cc:importPath:1347? <unmatched-paren>lembrun[m]: aha, the problem in the first case might be that greetd doesn't set up XDG_RUNTIME_DIR <unmatched-paren>you'll want to add ``export XDG_RUNTIME_DIR=/run/user/$UID'' to your .profile <lembrun[m]>ok, right now I just have the HOME_ENV variables in it <apteryx>ACTION has gdb ready to break in daemon <VesselWave>Hello, I am trying to package hyprland. How to access a file from input (read only)? <VesselWave>Thank you! Only search or maybe I can specify what input? <VesselWave>Ok, thanks looking forward to try this beatiful WM <lembrun[m]>unmatched-paren: that was a good try but I think guix-home set thats by default <apteryx>nckx: fun, I can't reproduce with gdb and a locally built daemon <apteryx>I get a nice error message telling me to authorize the missing key <apteryx>nckx: correction, can't reproduce anymore at all, at least by simply removing the now authorized key from /etc/guix/acl. odd! <lembrun[m]>wondering if that pam error was when i put elogind back <nckx>apteryx: Fun! Have you already downloaded everything you needed to download? <rekado>ACTION pushed the directory-server-service-type to the master branch. <lembrun[m]>and that's wh yit throws me back to the login prompt <apteryx>I should try with a different package/derivation <elevenkb>has anyone already packaged emacs-eat for guix? <tricon>elevenkb: I used the importer for it. Haven't cleaned it up or submitted it yet. Are you wanting to package it? <elevenkb>yah, i'm busy packaging it. completely forgot about the importer :p <elevenkb>i'm particularly curious about what to do with the "/e" and "/integration" directories? what is the convention with additional files like that? <apteryx>nckx: for the repro, I'm just using 'guix offload test /etc/guix/machines.scm 10.42.0.243'. Deleting the item it uses for the test '/gnu/store/m28xn85i7nnpy1w9amisav7b3fqdl5p2-export-test' doesn't change the result. Odd. <unmatched-paren>lembrun[m]: that's strange. could you consider continuing this in #kennylevinsen so that the greetd developer can see? <elevenkb>tricon: for now i'll just patch the way default values for `eat-term-terminfo-directory` and `eat-term-shell-integration-directory` are defined. <jonscoresby>unmatched-paren: I just sent an updated vim-build-system patch, but I forgot to cc you <unmatched-paren>jonscoresby: sorry, i don't have time to review the whole thing right now <Guest363>i'm thinking of using gnu guix. How can i find out if the drivers of my equipment are a problem? <rekado>Guest363: as a quick test you could try to boot a live system of Trisquel. It too uses the Linux libre, which is the default in Guix. <rekado>you can, however, change the default kernel easily. <lembrun[m]><unmatched-paren> "lembrun: that's strange. could..." <- I'll see that tomorrow <lembrun[m]>I did try the other command but I get xdg_runtime_dir not set <lechner>Guest363 / we may also be able to advise if you share your hardware details <Guest363>Does this mean i can use gnu guix without any problems? <lechner>Guest363 / very likely. what do you use presently? <lechner>the two things to worry about are the integrated i915 graphics and the Atheros wifi card <lechner>nckx may have had some negative experiences with integrated intel hardware recently in this channel, or he may have advised someone who did <lechner>i had that integrated chip once in debian or ubuntu (i cannot remember) but the laptop was stolen soon <lechner>you can easily find out by booting into a debian USB live image on your inspiron <lechner>your downside scenario is to enable the kernel they maintain in #nonguix <lechner>make sure you have an ethernet cable and access to your router, and come back here or write to me privately if you need more help.