IRC channel logs

2024-10-27.log

back to list of logs

<x8dcc>My internet died, sorry if I missed anything
<euleritian>There's a confusing description in the latest blog post: "If it is vulnerable, the last line will contain your system is not vulnerable, otherwise the last line will contain YOUR SYSTEM IS VULNERABLE." — see https://guix.gnu.org/de/blog/2024/build-user-takeover-vulnerability/
<lh>podiki: sorry, by current I meant the one currently on master, 24.04. I believe the bug is https://gitlab.freedesktop.org/drm/amd/-/issues/3497
<lh>civodul: forgive my ignorance but what exactly is a “ci green light” in this context? how do you tell if newly succeeding/failing builds come from changes on master incorporated in rebases rather than actual changes in the branch? browsing through some failures, they certainly seem unrelated to mesa
<podiki>lh: ah, gotcha.
<weary-traveler>reading guix-devel, and i saw the post advocating for distance from FSF. anyone know what's being referenced here: "... FSF made a bad political move that has destroyed its social capital and, as a side-effect ..."
<weary-traveler>what political move did FSF make?
<singpolyma>I assume referring to the RMS board situation
<weary-traveler>his reinstatement?
<hiecaq>Hi friends, I'm planning on moving from Guix on Foreign distro set-up to a full Guix System, so I'm experiencing with that in my VM. One thing I notice is that /tmp is not mounted as an tmpfs. Is there any reason why it is not put in %base-file-systems currently? Or more specifically, will there be any issue if I do it myself as putting it in file-systems config?
<dhoffman>is there still a dedicated bootstrap irc channel or has that discussion consolidated here?
<dhoffman>i'm taking a stab at bootstrapping for RISC-V and it seems possible but very experimental at this stage
<[>reinstating RMS to the board was a good move though
<mccd>Heya, woman in emacs doesn't have all the man pages that should be available when I'm using the exwm desktop environment
<mccd>is there anything I need to set up for it to work?
<jakef>mccd: might need the man-db package in a profile to get the man pages for packages in that profile
<mccd>jakef ah that makes sense, I'll try it
<mccd>thanks
<Rutherther>sneek: later tell x8dcc, I can't really say whether font packages _should_ add XDG_DATA_DIRS, but it's not something they would be doing now. You have to do it yourself, even on Guix System this is handled by adding lines directly to /etc/profile of the whole system. (or for guix home from profile files)There is currently no "Guix" native solution to this that wouldn't be more of a workaround (like the dummy package I proposed)
<sneek>Got it.
<roptat>hi guix!
<Lumie>Morning #guix
<Rutherther>hiecaq: hi, you could potentially run into issues if you make /tmp tmpfs. Guix uses /tmp by default for building packages, and it can happen some builds can be quite large, filling up your ram not only by the programs taking up ram, but also from the files taking it. Depends on how much ram you have and what programs you are going to be building
<mccd>I'm trying to install gsettings, which as I understand is in glib:bin. When trying to install it by adding specification->package+output "glib:bin", I'm getting an error with a conflict with networkmanager
<mccd>saying there is a conflict of entries for glib
<mccd>but they are both the same version
<mccd>I seem to be able to install guix glib:bin with the command line though that's not very reproducible
<Rutherther>sneek: later tell mccd, it is reproducible. The error stems from you trying to install them both in the same profile. As soon as you try guix install you are installing it to different profile, so no error. They are likely not the same version - does the hash match or not?
<sneek>Got it.
<mccd>There's no way really to change grub right? I can't add GRUB_HIDDEN_TIMEOUT_QUIET=true somehow?
<sneek>Welcome back mccd, you have 1 message!
<sneek>mccd, Rutherther says: it is reproducible. The error stems from you trying to install them both in the same profile. As soon as you try guix install you are installing it to different profile, so no error. They are likely not the same version - does the hash match or not?
<mccd>Also Rutherther thanks, I managed to solve it by adding it to my profile instaed
<mccd>of system
<Rutherther>mccd: there is always a way to change stuff, but it's true it's not so easy in this case, you would have to define a new bootloader, and copy some code from the grub one from guix probably
<lilyp>sneek later tell apteryx: I've pushed the glib update now – feel free to submit your patches on top
<sneek>Will do.
<mccd>When using guix with geiser and emacs, is it possible to evaluate, for example, the home config directly?
<mccd>or how can I "interactively" work with emacs and guix?
<intermet>Hi there. I am packing an exectutable "ocaml-app" written in ocaml. I am dynamically loading plugins with Findlib. The issue is that when I do "guix shell ocaml-app" the OCAMLPATH is not set correctly in order to find plugins. Basically I need to do "guix shell ocaml-app ocaml" in order for the OCAMLPATH to include a lib/ocaml/site-lib where the
<intermet>plugins can be looked for.
<noe>Yeah that’s because of load paths, AFAIK you need to include the package that declares the load path for it to update. Same goes with man pages (man-db), and c libraries (gcc-toolchain)
<noe>See (guix) Search Paths
<intermet>Thank you. Indeed my fix was to add ocaml to the propagated inputs of ocaml-app? But I don't find this solution very convenient...
<noe>I think its up to the user to add ocaml to their profile like you do with guix shell, see also #22138
<peanuts>"Search paths of dependencies are not honored" https://issues.guix.gnu.org/22138
<Rutherther>intermet: you shouldn't add it to propagated inputs. Just add it to the shell. If it's an app, then you should wrap it in one of the phases, if it's modules, users add it to the shell when they want to use it
<intermet>At the very end I will be using a guix system vm not a shell. I am a bit annoyed to add ocaml as a package just to fix the search path.
<Rutherther>intermet: how do you intend to use the modules if you don't add ocaml?
<intermet>It is not a module it is an executable. Why should I need the compiler
<Rutherther>intermet: if it's an app, then the other advice applies - wrap it
<intermet>Yes the name ocaml-app needs probably to be just "app"
<intermet>I see thank you. But what do you mean by "wrap it in one of the phases"?
<Rutherther>intermet: see gnu/packages/maths.scm:9819 "(define-public fama-c" for example of wrapping ocaml program
<intermet>Thank you very much, I did'nt know about such thing, I will study it
<Rutherther>it is done commonly. Python build system even has that in standard phases to wrap everything in bin. I am quite surprised other build systems don't and it's necessary to wrap in the packages instead
<intermet>Thank for your precious help!
<Guest71>What is the advantage of openssh sans x, if I have a headless machine?
<Guest71>I mean, why shouldn't I simply install just openssh
<omar_b>How can We get someone who has commit access to have a look at this issue
<omar_b> https://issues.guix.gnu.org/72925#68
<roptat>I'd like to update xdg-utils because the newer version adds support for KDE6, which I'm trying out. But it has 1615 dependents and is not part of any team. What's the process to update it?
<stochastic>Is there a way to compile a "program-file" gexp?
<stochastic>For performance gains
<ieure>Is there a Guix System service to run a caching HTTP proxy? I think I'd like to have one to reduce bandwidth when reconfiguring multiple machines. I see there's a Varnish service, but I'm not sure if that's what I want. That what folks are using for this kind of thing? Anyone have an example?
<ieure>(I already have a substitute server and cuirass on this box, maybe there's a way to use one or both of those?)
<ieure>Varnish looks like I'd need one config per upstream I want to cache, ideally I want a single endpoint I can stick in the http-proxy field of guix-configuration.
<ieure>I've used squid for this long, long ago, when my Internet connection was two channel-bonded 56k modems; looks like there's a package, but no system service for squid.
<ieure>(At least, there's nothing in the manual if there is.)
<efraim>Other than my build offload machines I setup each machine to also serve substitutes
<ieure>efraim, I'd prefer not to boot one laptop to open another when I already have an always-on machine that could do the job.
<efraim>I do have one larger machine that I do almost everything on and then offload to most of the others
<efraim>On my laptop I do guix system build ... -n and then use guix-copy to copy the drv to the main machine and then build it there, to spread it out
<efraim>then I'll either copy the result back or just have it grab it as a substitute
<kreved>have somebody experienced any issues with pam-mount after core-updates merge?
<ieure>efraim, I'd prefer not to significantly change my workflow, and just have the one machine cache stuff.
<Guest71>How do I mount a NFS share automatically on system boot? Using file-system doesn't work. The machine stops booting
<ieure>Guest71, I don't have an answer, but my guess as to why it's doing that is that it tries to mount before networking has started.
<kaelyn>Guest71 nad ieure: them trying to mount before networking had started is an issue I encountered. My solution was to add "(mount-may-fail? #t)" to each of the file-system entries for my NFS mounts, and then created a custom "late-nfs-mount" service that depends on 'networking and tries to mount all of the file-system entries with a type of "nfs".
<kaelyn>The service has a feature in its start function that tries to ping the nfs server, and retries for about 2 minutes for the nfs server to respond, then only tries mounting the shares once the server responds to a ping--to give time for e.g. dhcp to configure the interface after networking starts. The service is messy and not very extensible though, as it is configured by passing a list containing a single nfs server followed by some number
<kaelyn>of file-system entries for the mounts (since the service doesn't have access to the operating-system record).
<Guest71>Ah this makes sense. Thanks for the help
<ieure>Ugh. I really need to finish the autofs service I've been idly hacking on and contribute it.
<guix369>hello guixers !
<guix369>dummy here... by system kind of broke. How can i rollback to a previous image ( GRUB menu ? ) thanks !
<Rutherther>guix369: there is an entry that has the older versions under it, usually the last one
<guix369>yay, i entered, (and now chatting) from the previous one. HOw can instruct Guix to rollback in the one i'm currently in ?
<Rutherther>guix369: use guix system switch-generation to select a specific one
<guix369>let me try !
<guix369>ACTION loves GNU Guix and its people
<guix369>i see i have the chance to '$ guix system roll-back
<guix369>would that be more straightforward in my scenario ?
<Rutherther>guix369: if you want to switch exactly one generation, yeah, it switches one back
<guix369>yay, that's the case
<guix369>i mean, i already rolled-back kind of... via GRUB
<guix369>so... umh.
<Rutherther>guix369: that doesn't roll back, that just boots to a different generation
<Rutherther>guix369: roll back means that your new default boot will be that one generation
<guix369>That's what i'm looking forward... I see... Still, I'm wondering , just to make sure !  may '$ guix system roll-back'  in my context switch to this gneration i'm on ?
<Rutherther>yes
<guix369>Great! Last question before i roll back. May i just '$ guix pull ' afterwards ?
<guix369>ACTION guessing ^^
<Rutherther>sure, it's completely unrelated though
<guix369>great, thX a lot
<guix369>uoops, no joy :
<guix369>guix ... roll-back says "guix system: error: symlink: Permission denied: "//var/guix/gcroots/bootcfg.new.new""
<guix369>what am i missing ?
<Rutherther>you need to execute with sudo
<guix369>may i run it with sudo
<guix369>haha ok
<guix369>thX
<Rutherther>every system generation change needs root permissions
<PotentialUser-99>hello, would anyone give me a $PATH readout? I overwrote it with a bad home config and everything is borked so I can't even rollback. Just a 'which guix' even would work
<guix369>Rutherther: umh... so, then sudo guix pull, afterwards once rebooted?
<Rutherther>PotentialUser-99: which guix should be ~/.config/guix/current/bin/guix usually
<Rutherther>guix369: it's up to you, as said, it's completely unrelated. guix pull pulls your user profile guix, guix system doesn't affect it at all
<PotentialUser-99>thanks Rutherther.
<guix369>umh.. what the best practice ? '$ sudo guix pull ' ?
<guix369>or just as a regular user ... umh
<Rutherther>guix369: no, do not use sudo guix pull, that won't achieve much as you probably don't use root for executing guix
<Rutherther>guix369: if you want to update to newest channels, just guix pull as your user
<guix369>nice to know, thX again
<chloris>-v
<sneek>Welcome back chloris, you have 2 messages!
<sneek>chloris, Rutherther says: if you wanted to go the VM way I think a better way is to use Nix instead, you still have lot of benefits of Guix, and it has a mechanism for obtaining rust toolchain without downloading it as a binary, see https://ayats.org/blog/nix-rustup for details
<sneek>chloris, Rutherther says: scratch that message, I got confused
<chloris>Rutherther: I know now the problem of paps
<chloris>the paps of Guix is of 2019, an old version, that's why I can't use it in emacs to print
<chloris>what does "sneek" mean
<chloris>instead of a user I got the word in red "sneek" in weechat