Does this have any proper roadmap or goal at all?

[Hedwig7s]

The feature-set outside simply running a computer is a bit barren, documentation is lacking, the 1.20 and 1.21 branches are diverged by a lot

To me the project looks a little aimless and the core feature of running computers seems to work fine so I think it might be good to have a more formal roadmap and put more focus on user facing stuff like scripting apis and world interaction

[Un1q32]

It's mostly unmaintained because all the maintainers got bored and moved on.
I've been doing some very light maintenance work, basically just keeping the Linux distro up to date, but don't expect any real new features any time soon.
The 1.21 branch was basically done but there was some issue with some dependency that prevented it from getting an actual release.

[Hedwig7s]

It's mostly unmaintained because all the maintainers got bored and moved on.

I see. Still, I think a roadmap would be useful for potential future contributors to have a basis for creating PRs adding features

[daniilfigasystems]

That's really sad. There's so many things that can be implemented: tablet, peripherals like speaker, xv6 support, X server, FreeBSD, machine optimizations

[daniilfigasystems]

Also please remove that stupid overlayfs system. Previous way where only kernel was in flash is better. Where did you see computers with whole linux distro flashed inside BIOS? That's adds more realistic, customizable way. Why linux doesn't have swap support??? We have expensive ram in this mod and no swap space??? You literally bloated kernel with useless features like 9p but disabled swap!

[Un1q32]

Before overlayfs the system was practically unusable compared to what it has now because it was so minimal, and half your hard drive still went to the distro. Removing overlayfs is not happening, you can still do things the old way if you want, just have an ext2 drive with an executable at /init or /sbin/init and it will work. The extra realism is not worth the huge loss in practicality.

[daniilfigasystems]

At least add swap support in kernel

[Un1q32]

As for optimization, there was once plans to replace the sedna emulator with either qemu or a hand written emulator in C, sedna is unfathomably slow especially when doing floating point math, and often can't maintain the advertised max clock speed even on good CPUs. Those plans ended when everyone got bored and stopped working on the project. If someone were to make some script or something to build qemu as a library with sedna-compatible Java bindings I could probably merge it. I barely know any Java, so I wouldn't be able to merge improvements to sedna itself, and even an extremely well optimized sedna would probably be dwarfed by qemu anyway.

[Un1q32]

At least add swap support in kernel

The biggest hard drive isn't big enough to make swap practically useful. Even if you somehow fit your userland into 1 byte you would only be getting like 25% more usable memory (assuming you have maxed out ram, which you probably do if you have the biggest hard drive).

[Un1q32]

Actually, the kernel does have swap support, is it just missing the userland utilities?

[daniilfigasystems]

You can just craft 4mb hard drive especially for swap. Even 1 mb would improve performance drastically. Why kernel is so big? What's the point of adding so many features to simple riscv pc in minecraft?

[daniilfigasystems]

Actually, the kernel does have swap support, is it just missing the userland utilities?

No it isn't. I checked on newest version: invalid argument

[daniilfigasystems]

As for optimization, there was once plans to replace the sedna emulator with either qemu or a hand written emulator in C, sedna is unfathomably slow especially when doing floating point math, and often can't maintain the advertised max clock speed even on good CPUs. Those plans ended when everyone got bored and stopped working on the project. If someone were to make some script or something to build qemu as a library with sedna-compatible Java bindings I could probably merge it. I barely know any Java, so I wouldn't be able to merge improvements to sedna itself, and even an extremely well optimized sedna would probably be dwarfed by qemu anyway.

Well i could help with that, but i hate java so i can write emu only in c, c++ and maybe asm. I actually wrote one

[Un1q32]

You can just craft 4mb hard drive especially for swap. Even 1 mb would improve performance drastically. Why kernel is so big? What's the point of adding so many features to simple riscv pc in minecraft?

Oh true actually, didn't think of that.

[daniilfigasystems]

As for optimization, there was once plans to replace the sedna emulator with either qemu or a hand written emulator in C, sedna is unfathomably slow especially when doing floating point math, and often can't maintain the advertised max clock speed even on good CPUs. Those plans ended when everyone got bored and stopped working on the project. If someone were to make some script or something to build qemu as a library with sedna-compatible Java bindings I could probably merge it. I barely know any Java, so I wouldn't be able to merge improvements to sedna itself, and even an extremely well optimized sedna would probably be dwarfed by qemu anyway.

Well i could help with that, but i hate java so i can write emu only in c, c++ and maybe asm. I actually wrote one

I can make it more portable so it would be much easier to adapt for oc2r that qemu. QEMU is too heavy for that.

[Un1q32]

It wouldn't be useful if it were in Java anyway, since it wouldn't be much faster than sedna, plus I don't even know any Java so I wouldn't be able to review it. I like the idea of using qemu because it will make the patch much easier to review, and because we can easily get new features for free by updating the version of qemu it uses. Is qemu that heavy? We obviously wouldn't need to build a version targetting everything qemu supports, just riscv64. If qemu is absolutely off the table I could review a custom C emulator, it would just take a lot longer. Avoid assembly and C++ because I don't know those as well and couldn't review it.