OP has a pascal GPU, the Nouveau kernel drivers for pascal (which you need for nvk) aren’t really suitable because they can’t change the clock at runtime, you are stuck at boot clock (which you can configure)

The comment said that people that install systemd based systems and then fight systemd tools are stupid. In that instance a guy is dead set on using netplan and shell scripts on a system with networkd, when using just networkd would have been extremely simple. Maybe in a system without systemd it would have been easier to use netplan, but the guy decided to install systemd and then fight it, and then suggested other people do the same.

Think harder, I know you can. Maybe reread the comment. Maybe compare those lines with that one line change.

Maybe notice that I specifically said to install a system without systemd if you don’t like it, instead of fighting the tool you installed and then complaining.

Honestly, one of the worst parts of the Linux community is people trying to force 30 years old tools in systems built around systemd. If you want to use that old stuff then don’t install the modern replacement, find a different distro built around that ideal instead.

I remember a post on serverfault or askubuntu about disabling DHCP default gateway but keeping DHCP address assignment on a tap interface, and bring it up at boot, and the accepted answers was “configuring DHCP in networkd/NetworkManager systems is almost impossible, here’s how I did it” and it’s three pages of cobbling together ifconfig and netplan with startup scripts, that work by pure chance.

Wanna see the actual full networkd configuration for that?

# /etc/systemd/network/tap0.network
[Match]
Name=tap0

[Network]
DHCP=ipv4

[DHCPv4]
UseRoutes=false

That’s it, all that the post asked for is handled by six lines. “How do I discover that?” you may ask, because if it’s three lines, but I have to dig for hours before finding it than it’s not that useful. Simple, I go on the systemd documentation for .network files and search for DHCP. And this is a niche use case, the basic usage is readily available on the arch wiki as with anything else. Note, this does nothing for IPv6, and the interface will have IPv6 route configured, but this wasn’t relevant to the post, and my home’s IPv6 layout is “peculiar” so I have omitted it here.

Fucking finally

I hope everyone else follows soon. If you like it when you are trying to open a link on a new tab and your system randomly decides to spew a selection from another app into a random text box you are free to configure that yourself. Remember to configure in a cilice wrapping your thigh while you are at it, it’s unix-compliant and has been around for centuries.

It will hurt less being disabled

Hmmm, I’m pretty sure you just need to systemctl disable sshd.socket and then configure it how you like. Don’t trust every “solution” you see online, they are often full of bullshit written by people trying to fit a square peg in a round hole. I bet that kernel parameter thing is something to disable it on the first boot, before you get the chance to configure the thing properly.

Edit: @thorhop@sopuli.xyz wrote a comment about that option here, go to the thread if you are interested.

I’m in your same predicament. I think the long term path is to fuck ourselves until an Oracle comes to Faith Ekstrand (or another maintainer) in a dream and tells her how to make pascal work properly in nouveau. Or until the spirit of Christmas Past visits Jensen Huang.

A third display server would probably be either useless (as in it would not be adopted or development would not takeoff) or just Wayland but without the committee. As such it would only cause more fragmentation. Plus, not anyone has the skills and availability required to make a good display server, and as of now, most of the ones that do are working on Wayland.

P.s. if instead of Wayland like it were x11 like it would stop being “de-facto useless” and become “useless by definition”

UV is a lot more than pipx. It installs applications from pypi, without dependency hell, but it also uses hard links when possible to avoid wasting space. But it’s also a dev tool. It manages python installations, workspaces, you can use it to edit the pyproject, it can also publish to pypi, even from a GitHub action if set up from pypi. It just does a lot more.

Correction, uv isn’t in any way less an option. pypi is only the registry. If you are using pip you will end up in dependency hell, you might use something like poetry to avoid that, but uv is just better.

But… wait a minute… uv is inspired by cargo, and it’s also written in rust. That’s quite the coincidence, huh?

Also, cargo is fast, it’s rustc that’s slow, and that’s because rustc is doing advanced code analysis. Compiling rust is actually NP-hard, but in exchange for that, the compiler will catch bugs in place of the developer. Which is a good tradeoff considering that you only compile once and run many times.

“countless projects that use it that way” isn’t proof of anything. Countless projects tell you to curl a 2000 lines script into sudo bash that will fill your os with bullshit.

“the compositor takes care of things it shouldn’t” said the x11 user while using their display server that also manages printers, and also provides peripheral drivers, and also manages opengl drivers, and also provides a full graphic toolkit, and also provides remote access. And all that while treating multiple monitors like a single big monitor with a single shared refresh rate, and with no support for HDR of trackpad gestures. Yes, it really is upsetting when the screen compositor manages screen recording and double buffering. And remember x11 is so modular that the graphic server is part of the driver stack and so must be implemented for every GPU out there.

P.s.: anyone that thinks the Unix philosophy has any value should not touch xorg with a 10 meter pole. It does many things, badly. But sure, it does allow any unprivileged program to read and write to the framebuffer, so be my guest if that’s your thing.

That’s also slower than most of the stuff you could come up with, it is so slow that there is no hyperoperation fast enough to describe it. There were other approaches that were almost worse though, like “the function is a switch-case that returns false by default. As complaint tickets are opened, more cases get added to the switch-case”

Please don’t do that, I was stupid when I wrote that. But still, in very dynamic languages like python or js everything is an object, including functions, so you can just do object stuff on them.

The implementation is not very exciting, I capture a variable in python. It could have been done more cleanly.

The proof is this. But, I could have made mistakes, it was many years ago.

Note that in python you’ll never be able to run is_even(5) the stack cannot handle it

Edit: daaaamn, that variable is ugly as hell. I would never do things like that now.

For a time on Reddit (some years ago when I still used it) there was a trend of finding the worst way of implementing is_even(x: int) -> bool. My contribution to that was a function that ran Ackerman(x,x) flipping a Boolean at every iteration, and check if it was true or false at the end.

It works btw, I will find the proof later

Is that the issues your project is solving?

That’s exactly it, and also the fact that git doesn’t follow symlinks. Just a word of warning, If you are still inexperienced I suggest you run my tool manually instead of automating it with git hooks, as it is inherently less secure. In the post I linked in the description you can see some of the precautions I took to make it more secure. Still, running it manually is fine.

Feel free to give some feedback if you start using the tool 🙂

May I interest you in the program I just wrote. I think you might like the subcommand type.

I use it to automatically copy mime types “video” and extensions “.srt” from the Torrent folder to the jellyfin folder. The other two subcommands, I use them to save user dotfiles and system config files on git

Yes, that was one of the tools I considered before making this. I do not remember the precise detail on why, but much like gnu stow is only good for versioning user dotfiles and not system config. Etckeeper is good for storing either your system config files or user’s dotfiles, but not both at the same time. copicat doesn’t care what you use it for because you explicitly tell it all the locations and permissions that you want.

Yeah, it’s cool, people are mostly looking for something like your usecase. I got suggested stow or stow-like tools a lot when exploring this. And when they understood what I wanted, they just suggested ansible… Which would work when starting from scratch, but wasn’t right for me. I made copicat mostly because I am actually using it, and then decided to make it public because really I didn’t find anything like it.