The Linux Thread - The Autist's OS of Choice

[analrapist]

I'm too much of a nigger to actually learn s6.

Me too, frankly. I enjoy and use systemd-networkd, -nspawn, -resolved, and I don't have good alternatives for those. I embraced this in my current build and even my mounts are now systemd units. Now that I'm on Gentoo, I need to do more fucking around with musl because flatpaks and snaps and shit.

I think some Skarnet stuff like execline seems a bit of a solution in search of a problem, but big S has his own priorities, and I expect I'm too much of a mere mortal to properly understand. Plenty of inits are minimal C wrappers around a shell script.

 

[prollyanotherlurker]

Me too, frankly. I enjoy and use systemd-networkd, -nspawn, -resolved, and I don't have good alternatives for those. I embraced this in my current build and even my mounts are now systemd units. Now that I'm on Gentoo, I need to do more fucking around with musl because flatpaks and snaps and shit.

I think some Skarnet stuff like execline seems a bit of a solution in search of a problem, but big S has his own priorities, and I expect I'm too much of a mere mortal to properly understand. Plenty of inits are minimal C wrappers around a shell script.

I've messed with systems running musl instead of glibc. And I do like the idea of musl. But in the end, it ended up being more of a headache than it was worth.

That, and the idea of minimalism, and correctness is nice. But you end up loosing optimizations that glibc added in, and some security features. So you don't really end up with programs being faster, and more secure on musl in the end.

You don't necessarily end up with slower binaries either, or less secure ones. But you don't really see any improvements. But where you do get trouble, is trying to run proprietary programs that don't work with musl. Or ones that the devs don't want to bother supporting (or rely on something in glibc)

 

[Demonslayer1776]

What Linux really needs is "self-healing" distros. You know how if you massively fuck up Windows you can just run DISM /Online /Cleanup-image /Restorehealth and it'll magically reinstall and fix all the core elements needed to bring your system back to life, but at the same time there's nothing stopping you from fucking it up that way to begin with? That's the sort of retard proofing Linux needs. Right now it's either "if you fuck it up you need to back up /home and reinstall" or "here's a distro you can't fuck with so that you won't fuck it up".

Unless of course you're the type of "person" that believes Linux should only be for Mensa members that know what each GNU package in their system does and could set up LFS from memory if they wanted to while at the same time expecting people to choose it over Windows or Mac, in which case lol, lmfao

Nix has this rollback experience in alot of regards. Not exactly friendly for the type of users describing, but I think its a sign of trends that will continue into the future.

I also think its worth pointing out that this issue has largely been solved in cloud-native workflows, since state is completely separated from the runtime, but this will never work for desktops. Atomic Distros are the closest parallel and are probably the best solution for end users. I don't think this is a Linux specific issue though - DISM only works in some instances, more often than not I find a reinstall of Windows is needed anyways. Its nearly impossible to properly validate everything without a hard restore of the default OS config, and anything more strict either requires heavy isolation of applications, or a workflow too autistic for non-programmers.

 

[prollyanotherlurker]

Nix sisters. Today's your day.

"It's not really considered installing"

Nigger you're installing shit.

 

[Harry Kuntz]

NixOS just seemed like a giant meme and not a serious distribution last time I tried it. Striked me as something you would use when even Gentoo isn't autistic enough for you.

 

[Trans Fat 41g]

In order for me to take NixOS seriously, we need the official documentation to not suck ass and be stuck ten thousand feet below the Arch Wiki.

 

[prollyanotherlurker]

In order for me to take NixOS seriously, we need the official documentation to not suck ass and be stuck ten thousand feet below the Arch Wiki.

If you watch that interview. People like him are the reason the nixos documentation is so bad.

First right away. Autism voice. (Autism + nix users, probably gonna troon out tbh). Then he talks about everything as if people will even start to know what he is talking about as he explains. Even if you understand all the normal linux stuff. Nix is so absolutely different than any other unix like operating system, you are basically just learning something completely new.

the nix documentation is written in the exact same way. As he is talking about this stuff

 

[Betonhaus]

is there a way to get linux to email you or something if your WAN IP address changes? like if your provider changes your ip address or a dumb burgler steals your home media server and plugs it in

 

[ToroidalBoat]

What are "dmabuf", "pipe", "socket", and "sync_file"? Are those files on the HDD or SSD, or can they be in RAM, like "memfd" can also?

 

[DavidS877]

What are "socket", "pipe", and "dmabuf"? Are they files on the HDD or SSD, or can they be in RAM like "memfd"?

socket and pipe are both virtual files that just tell how things are connected.
pipe is a 1 in- 1 out file.

mkfifo bob_the_pipe
cat bob_the_pipe &
echo hello >bob_the_pipe
Allows communication between processes using a named mechanism instead of simply "process1 | process2"

sockets are similar but take a bit more legwork, similar to connecting between 2 processes over the network. (See: Unix Socket)

dmabuf is a pointer to memory that can be used by multiple processes at once(or so I understand) so also not a "real" file.

 

[ToroidalBoat]

virtual files

So even if process viewer in a Linux OS says browser have done a ridiculous amount of disk writing, much of it could really be to RAM?

 

[DavidS877]

So even if process viewer in a Linux OS says browser have done a ridiculous amount of disk writing, much of it could really be to RAM?

Depends on how the process monitor is monitoring stuff. Generally a write to a pipe/socket would not be counted as blockin/blockout. But it may be counted as IO. The best command line thing to look at just disk IO is "iostat" I usually use "iostat -xkz 10" for 10 second updates. This is ONLY things that actually touch the disk.

iotop I think covers anything that goes from the process to the kernel which may be a cache read or write. The numbers at the top for "Total" includes caches, etc. "Current" is just stuff hitting the physical disk.

 

[ToroidalBoat]

Depends on how the process monitor is monitoring stuff.

So it is possible for a system monitor to count writes to RAM towards "disk write total" then? That's odd.

And I take it that "anon_inode" could be another RAM thing?

 

[snotang]

Do you write code and commit it to repos of packages that you use when you find a bug in them? No? Then it's not my problem, someone else should make a distro like that.

That is exactly what I do. What are you, some kind of leech? Imagine bitching about download rates when you refuse to seed, ever. In many cases it is legitimately faster to identify the problem and make the fix locally than to go shopping around for alternatives, especially when the thing you're using 99% does what you want. Doing this regularly will make you a well-read programmer, skilled at debugging, comprehending, and modifying code you only found out about today.

 

[Another Char Clone]

Honestly you're probably right. But I'm too much of a nigger to actually learn s6. At least I haven't put in the time after reading about it.

But it sounded like it's probably about as fast as it gets. Since it's running its stuff as compiled binaries if I remember correctly.

Ultimately, all the hard stuff is around process and PID monitoring, restarting things as necessary if they crash. S6 is more of a supervisor, supervisors can function perfectly well as an init system. The important part about s6 is that it does not have to be PID 1. That means you can start it and use it for arbitrary processes, not just for system startup.
I even have s6 running under openrc in Gentoo for my user mode daemon (hello dbus, pipewire). But why the autism instead of just dumping everything in a script and running that instead? Because the supervisor automatically restarts anything that crashes, I let it do what it does best, automatically restarting my stuff for me instead of writing yet another solution for process monitoring.

I think some Skarnet stuff like execline seems a bit of a solution in search of a problem, but big S has his own priorities, and I expect I'm too much of a mere mortal to properly understand. Plenty of inits are minimal C wrappers around a shell script.

You can write shell scripts as run files for s6, so long as you remember to chmod +x it. It can even be perl or python or binary programs written in C, s6 doesn't care. I used a simple shell script to load some settings file to pipewire under s6. Execline is for in case you ABSOLUTELY DO NOT WANT TO FORK without relying on external executables or shell expansion uncertainties, but I have no problems with that.

The only part I don't like is needing to compile your service directories or having live directories, but I guess that's one way it monitors processes.

In many cases it is legitimately faster to identify the problem and make the fix locally than to go shopping around for alternatives, especially when the thing you're using 99% does what you want. Doing this regularly will make you a well-read programmer, skilled at debugging, comprehending, and modifying code you only found out about today.

I used gentoo enough that I have plenty of /etc/portage/patches customization. It's great I don't need to bug anyone if I want to change something for my own use, nor do I need any permission. Many of them are one-liners too. It gets auto applied when packages are updated as well, but if it doesn't, time to update the patch.

 

[DavidS877]

So it is possible for a system monitor to count writes to RAM towards "disk write total" then? That's odd.

And I take it that "anon_inode" could be another RAM thing?

It depends on a bunch of underlying shit. If it's going to a "real" filesystem then it's likely it will eventually end in a disk write. BUT if the same block is written 300 times it may show 300 writes but only get written to disk once depending on when the kernel sync process comes by and flushes it to storage. Also if it's a tmpfs like /tmp on some OSes then it probably won't be sent to a physical disk but from the process view it's a "disk" write.

anon_inodes are another weird case. Basically, you open a file, then delete it, you can continue writing to the file, and the data will even write to disk if it was opened on a normal filesystem. As soon as the file is closed by all procsses that had it open then it goes away. This is a deliberate usage of what you'll sometimes see if a process is writing a file(often logs) and you delete it to free up disk space but the space isn't freed until you restart the process.

There's an additional fun trick that you can, as I recall, go into /proc/$PID/fd/# and copy the data OUT of the file if you need to save it before the process exits because you inadvertently deleted the file.

 

[Lorekmo]

Is Wayland optimized for GPUs that support the latest Vulkan version? I'm getting tons of frame drops on it, also I heard it's not possible to disable compositor when playing video games, just like on macOS.
Until these things are fixed, LXDE on X11 is the best for me.

 

[prollyanotherlurker]

Is Wayland optimized for GPUs that support the latest Vulkan version? I'm getting tons of frame drops on it, also I heard it's not possible to disable compositor when playing video games, just like on macOS.
Until these things are fixed, LXDE on X11 is the best for me.

afaik the only place that is implemented is hyprland. talking about disabling vsync and allowing screen tearing.

because everything else is either gnome or kde (faggots), or just going off of devaults (and the other faggots) bikeshedded to hell wl-roots

 

[Harry Kuntz]

afaik the only place that is implemented is hyprland. talking about disabling vsync and allowing screen tearing.

because everything else is either gnome or kde (faggots), or just going off of devaults (and the other faggots) bikeshedded to hell wl-roots

Unfortunately, Hyprland is for troon faggots too (as a core part of being a tiling window manager). The real solution is to go back to X after all.
Also, KDE does implement this now in the latest versions, although if you aren't using a bleeding edge rolling release its probably not out just yet.

 

[prollyanotherlurker]

or quit playing games. studies show 100% of all troons, realized they were troons playing videos games with female characters. might as well cut your dick off now if you wanna keep playing those.