When your OS goes crazy - Screenshots

[ybilta]

When I switched parameters on my terminal scroll function, the screen scrolled when i had more than 25 characters on a line.
What you should see is a big ascii GOOS, the intro screen for my go-based kernel

Edit: the first picture was after I fixed the scroll code, I'm doing something wrong

[Bender]

Has anyone tried this? lol.

Pushing it hard
EDIT: If anyone wonders why the number of reported blocks is 8181 as the system memory is around 64MB, it's because of some reserved memory by the firmware and the kernel, the kernel reserves around 30MB for the drivers, applications and memory backups. 30MB?!?!? WUT? Who cares? It's all typedef infinte unsigned long long long long __bloat today.

[qw]

Is there a compiler with a "__bloat" modifier already?

[BASICFreak]

This is what happened the first time I implemented my PS/2 Mouse driver:



I later made a way to disable scroll, set bounds, and move in a negative direction.

[klange]

Inspired by a certain video game, I've added a new panic screen to my OS when operating in graphical mode.

The video driver desaturates the display, adds a vignette effect, and then prints the panic message in the middle using an embedded bitmap font:

[Lionel]

Thread switching did not go as well as planned

[Bender]

Inspired by a certain video game, I've added a new panic screen to my OS when operating in graphical mode.

The video driver desaturates the display, adds a vignette effect, and then prints the panic message in the middle using an embedded bitmap font:

Are your modules relocatable? Looks like that. Is that using dynamic linking btw?

[klange]

Are your modules relocatable? Looks like that. Is that using dynamic linking btw?

Indeed, they are relocatable Elf object files (the normal output with gcc -c). The module loader isn't quite as complete as a proper dynamic loader as only certain relocation types show up in static objects. Since the kernel doesn't maintain any of its Elf information (as much as I spent many hours trying to see if I could sneak the symbol table into a loadable section with a link script...), I build a separate symbol table for everything that the kernel exports. As new modules are loaded, the symbol table is expanded so modules can depend on and extend other modules (like the crash module used in these examples, which extends the kernel debug shell by adding new commands).

e: For anyone discovering this post in the future, I'm aware of the availability of ELF symbols in Multiboot, but I'm a heavy user of QEMU's -kernel option, and QEMU's Multiboot loader does not support the ELF symbol table information, so relying on it was not suitable.

[timus]

Threading make my os look artistic

[vtsman]

I forgot to check if I was putting a character within the bounds of the screen, and it had a mildly interesting result.

[vtsman]

Here's a bug I have yet to figure out:

I'm switching over my terminal to use a list of my text, but it seems to be dumping the memory of some location I have yet to find. There's more garbage that it spits out, and as far as I can tell, it loops around to the same address infinitely.

Edit: For those of you wondering what went wrong, I used a text buffer without allocating its lists, so I was essentially writing to and printing from my kernel.

[max]

started writing a terminal to test my VFS, this happens on scroll

[DavidCooper]

Threading make my os look artistic

zrzut ekranu4.png

You could maybe turn that into an app for designing new kinds of tartan for kilts.

[Primis]

Well, Sortie was telling me to fix my interrupt code to not violate the ABI calling (somthing about following a forbidden tutorial) and when I went to test the code, I got this nice screen rather than my normal output!

[Viperidae]

My kernel loads a list of symbols from the initrd image that is uses to link kernel modules. Upon rewriting my tool to create the initrd image I forget to update the kernel symbol list (So the linker was using old symbol addresses). This was the result:



Upon making some more changes to my OS (Which altered the symbol addresses)


Took me a while to figure out that the problem was just with the linking (I thought it was a bug with my filesystem)