OSDev.org

PeterX wrote:Only 36 KiB size with a GUI is really cool!

Thank you To save more space in the future, the system will use only SVG files for icons, wallpapers etc.

Libraries are not shared yet, so that are another Bytes to save.

For some reason I thought I had checked this thread earlier.

akasei wrote:Thank you To save more space in the future, the system will use only SVG files for icons, wallpapers etc.

Reminds me of how Windows 95 used a font for the UI icons to make them scalable:
https://en.wikipedia.org/wiki/Marlett

That predates SVG, and it's possible it may still be more compact due to TrueType being a binary format instead of XML-based (not to mention Windows had to include the TrueType renderer either way so the parser came "for free"). But unless you're really desperate to save space SVG is probably better for more complex icons nowadays.

ChaiOS currently has a port of LwIP, an I219-V ethernet driver, and an NVMe driver. That's the semi-mature bit, there's a FAT driver in progress (being tested here), and a non-functional AHCI module. NVMe has a bit of "class".

The kernel itself supports SMP (although that needs a bit of work regarding the APIC timer), and has threads and semaphores. There is also a fledgling user mode with a working syscall interface.

Currently only x86-64+UEFI, but I've written it so as to be portable to other architectures, hopefully, and the Kernel itself doesn't necessarily depend on UEFI.

Also, I have xHCI support that only works on QEMU, not my native chipset xHCI, but seems to work with an add-in card.

Machine Control System

I ported Mesa to my OS!



It's using the softpipe backend, since I haven't worked out how to cross compile LLVM for the llvmpipe backend yet.

Still working on my GPU driver and it is making me loose motivation, so I'll take a break from it.
I wanted to work on some multitasking and I looked at my code and was like WTF does that code even do... Does that ever happen to you?

Octacone wrote:... I looked at my code and was like WTF does that code even do... Does that ever happen to you?

Yes, not often, but it happens.
My try to prevent that is:
1.) Use self declaring names for variable/functions/objects...
2.) Comment a lot especially when it is an algorithm or API or something strange
3.) Write somewhere (for example in the readme) what every part does
4.) Make things understandable intuitively
Unfortunately I don't always follow my own principles... But I do follow (1) and (2).

EDIT: I forgot: modularize! That helps a lot.

Greetings
Peter

I would add: Make small commits (usually: one change = one commit) and write long commit messages. So if some time down the line you wonder what you were thinking with a particular line, you only need to run a blame on the file in question and read the log message associated with the revision of that line. git add -p is a really important feature here, as it allows you to stage only parts of a file for a commit. So if you ended up putting multiple changes in the same file before committing, you can still split them up.

@PeterX
I do most of that and my code is super readable. It is the algorithm itself that I often forget.
@nullplan
I don't use git or anything similar. I know it is a bad idea, if my SSD fail then RIP.

For some reason I was super motivated today, so I finally implemented system calls.

my first post on here

Lemon OS on real hardware

Hello, I'm segfaultdev, I'm 14, and this is flat, a small OS I'm currently working on. I've started working on it around two weeks ago:
Currently it boots from a USTAR image from a single stage bootloader into protected mode with a GDT(uses segmentation, not paging), sets up a 90x60 text mode with custom font(that includes bold characters) and palette(sweetie 16), sets up interrupt handlers(IRQs, exceptions and a system call) and loads up a program into ring 3.

It supports VGA text and serial communication, both with ANSI escape codes, and you can redirect it to other "devices", like a file.

If anyone is interested in the source code, I'll upload it on GitHub soon. I'd love to see your opinions about it and I'm open to any suggestions.

segfaultdev wrote:I'd love to see your opinions about it and I'm open to any suggestions.

* Nice colors and font, and I *really* like the use of actual bold characters instead of increasing the brightness.
* You're way ahead of me already!
* Never give your age out online unless you're an adult already. (This is old personal-security advice, but still applicable to most online forums & chat rooms, I think.)

Ok, so my OS will be a month old tomorrow. I though it might be a good time to show a screenshot.

I've oped to have my CLI embedded in a GUI (no mouse yet, so not much point). I've gone proper old school for the style. My CLI only knows one command the "dir" command so all I can do is have fun listing directories.

My cli is actually the boot task who's job it is to bring the rest of the OS up (yes, I'm a microkernel guy).

There are two characters in the screen title bar, these are little spinners which are each run by their own normal priority tasks in the background, when they stop moving I know I've had a bad crash No memory protection so one task has probably just written all over another task's code/data.

bloodline wrote:I've gone proper old school for the style.

You have! The colors remind me of my 8-bit Ataris, the title bar style my Atari ST, and the font printing in graphics mode in QBasic. (It's the VGA ROM font, isn't it?) Nostalgia fusion!