#Operating Systems

[unbfacts]

All 500 of the world's fastest supercomputers run on Linux-based operating systems.

[vxmpire-vxlle]

[monody-monody]

So I've recently been finally getting around to something I've wanted to do for a few years: switching to Linux.

To safely test out this process before applying it to my main tower machine, I decided to first do the switch over on an old Mac Mini which had a lot to gain from this as it was stuck on an outdated and unsupported version of MacOS.

But of course this all involved shuffling around a lot of old data and making sure everything was backed up. And here's where the story begins.

I backed up the Mini's MacOS install to its own partition on an external backup drive shared by my Windows tower machine. Once done, I shrank it to only the used size( which was complicated in its own right for reasons I'll explain shortly,) and then wanted to move it to the end of the drive so I could expand the Windows partition back out to fill the unused space.

Problem #1: MacOS's Disk Utility is so stripped down and lacking in functionality in the name of gloss, it literally cannot work with unallocated space on a drive in any way. You cannot see unallocated space, and you cannot create it. Want to shrink or delete a partition? It automatically creates a new one to fill the space whether you wanted it or not. Want to make a new partition in unallocated space without wiping the whole drive? Gotta use a different OS to make a dummy partition first and then replace it in Disk Utility. And the real problem… Want to move a partition? Can't.

Problem #2: Windows's Disk Management doesn't know what HFS+ is and refuses to work with it. Windows can't even read the Mac backup partition, let alone move it. But wait! I can use EaseUS Partition Manager! …Aaand most of its essential functionality is locked behind one of its multiple paywall tiers, including the ability to work with non-Windows filesystems at all.

But what about that new Linux install on the Mini?

I run 'sudo apt-get install gparted', a FOSS program, and in less than a minute I've already got it moving the partition. No hassle. No BS.

And if this whole thing doesn't just sum up the operating system trio, I don't know what does.

[fmk-polls]

[archeoweb]

Three OS-tan avatars uploaded by isc_luis_herrera to Flickr.

[techav]

On Major Milestones

I left off previously with init immediately crashing when trying to run NetBSD on Wrap030, my 68030 homebrew computer. I was completely lost and didn't know where to start looking. The error code it gave, 11, didn't tell me much.

Until now, most error codes I've gotten have been defined in kernel errno.h, which has 11 defined as:

EDEADLK 11 /* Resource deadlock avoided */

That … also isn't helpful. I'm still not entirely sure what that means, but since this is process 1 we're dealing with, I didn't think it was relevant.

Finally, I was able to find someone who had encountered the same error six years ago. Helpful soul [Martin] explained the exact cause of the error, how to fix it, and why the kernel errno didn't line up:

I'm running a NetBSD live disk on a laptop as a test host, so I mounted my disk on it and spent some time with mknod adding the essential device nodes, referencing the "majors" file for my arch. Sure enough, on next boot it skipped right past the point it had been panicking. It worked for a bit then finally printed on the console:

Enter pathname o

Enter pathname of what? The machine appeared frozen. Nothing further printed, and it responded to no input.

I was afraid this would happen. That string is 16 characters. The 16C55x UART chips I'm using have a 16-byte buffer. The system is hung up waiting for the UART to interrupt to indicate it has finished transmitting everything in its buffer.

There's just one problem — I don't have any serial interrupts wired.

I have a confession to make. Until a few weeks ago when I got my timer working, I hadn't really worked with hardware interrupts before. So between a limited understanding of how to use them effectively and limited board space, I had omitted the interrupt signals from my 8-port serial card. This was now a Problem, and I was going to have to find a solution.

I had a few options:

Force the com driver to 8250 mode so it doesn't try to use the buffers

Use my timer interrupt to check status bits on the UARTs and fake the interrupts

Deadbug an interrupt handler onto my serial card

Respin the serial card

Option 4 would've been expensive and risked passing my deadline. I wasn't sure option 1 would even help. And option 3 would have been difficult and error-prone. I decided option 2 would be the way to go so I set about researching how to accomplish it

I spent a few hours digging through the com driver. In the process I found softintr(9), a native NetBSD software interrupt process that looked like just the thing I needed. Digging in a little deeper, I realized that the com driver was already using softintr. And then I realized all it needed to do polled mode serial ports instead of interrupt-driven was to set a single variable, sc_poll_ticks, before initializing the driver. It's such a simple thing, but it's not really documented anywhere I could find, so the only way to know it was even an option was to spend hours studying the code.

With that in place, I recompiled my kernel and tried again.

It was asking for a shell. This is promising. I accepted the default shell, /bin/sh, and waited a moment. It printed a single #.

I had a shell prompt.

I typed in the first thing that came to mind, echo "hellorld" (thanks, [Usagi]). It responded:

hellorld

and printed another # prompt.

I had a working shell.

This is a major milestone. I have a modern operating system kernel loaded and running on my homebrew computer, and I have a functional root shell. I can navigate disk directories and run commands and programs.

But only as root, and only on this one console. I have seven other serial ports I want terminals on, and I certainly don't want them all running as root.

What it's running here is single-user mode. It is just the kernel and a few core services, somewhat analogous to Safe Mode in Windows. It's a fall-back for setting up or repairing a system. It's not quite the full operating system just yet.

Getting the rest of the operating system up and running is going to be a significant task, on par with getting just the kernel running. Setting up a working Unix system from scratch is not easy. It requires a lot of detailed knowledge of the various programs and libraries and config files scattered across the disk. For a sense of scale, the AT&T Unix System V manual was over 1100 pages, plus an 800 page programmer's guide and a handful of other manuals … and that was 40 years ago. That's a lot of specialized knowledge that I don't really have.

But still, this is something I've wanted to do for years and after countless hours of work, I finally have a glimpse of what it can look like. I have a lot to learn and a lot of work to do yet, but I'm certain I can figure it out.

I'm still hoping I can get this running multi-user on all those terminals in time for VCF Southwest in June. The show is just a few weeks away and I have a lot of work to do.

[smashorpassobjects2]

[serowebs]

While scrolling thru Linux posts I saw a lot with estrogen and forcefemming

What about

Testosterone

What about

Forcemascing

What about

Give me hrt please 🙏 😭

[boyslit-moving]

ok if you've been using Windows for a while which one has been your favorite. bc i just got hit with a huge nostalgia brick for the os of my youth

if you don't have a favorite you can pick which one sucked the least :) feel free to rant and bitch in the tags if you want to, we're all haters at the end of the day <3

[quasistatic-motion]

You guys might have demons in your head, I have daemons

[akarui-the-cicada-kisser-commie]

HELP 💔💔💔 I'M A STUPID DUMB DUMB 💔💔💔💔 I JUST LEARNED THAT ANDROID IS BUILT ON LINUX............... I FEEL BETRAYED!!!!!!!! FUCK YOU LINUX EAT SHIT FUCKING UGLY PENGUIN AND SUCK UP THE TARIFFS

[cerulity]

There are three projects that I consider some of the hardest in computer science:

Compiler - Requires tons of text parsing, type system configuration, and code generation. A conglomerate of data flow.

Operating system - Build everything from the ground up. You don't even start with an allocator. Implement. Everything.

Browser - So many standards need to be implemented, it's a feat of graphics programming. Just the SVG specification is 900 pages.

If I'm missing anything on that same level of difficulty, let me know.

Some things people have mentioned:

X11 Standard: Not 100% sure why, but I'd assume an entire windowing/desktop environment interface would be crazy. From light research, I found that there seems to be a lot of backwards compatibility, and that Wayland is just more straightforward.

Emulators: Simpler consoles like the NES would already be substantial to emulate, let alone something like the Switch or PS5.

Physics engines: So much math. Raycasting, collisions, islands, and much more. And if you don't get it right, everything literally just explodes.

[zombie-the-derg]

not much happening in the linux community i see

[bye2k]