OSDev.org

And, what happens when you don't quite get video configured properly on a Tegra2 board.

Normally my OS isn't very interesting when things go wrong, but recently I made my console rendering (rendering a text buffer to a linear framebuffer) asynchronous. Naturally, that did give a few cases suitable for this thread.

First up, the console deadlocked and if I was lucky I could get something on the screen. That meant my kprintf had no working output device, and things kind of got broken when I redirected it to the serial line.


After fixing that issue, I finally got text on the screen. But when I tried scrolling:


And when I scrolled further:


The font and console metadata was being screwed up by a stupid buffer overrun. Then later, when I tried refactor my ls and column program a bit:


And when I tried to debug the ls/column problem, things got a bit out of hand:


A couple days ago I attempted to a ppoll(2) system call, but I messed up and some long-forgotten debug code took over.


Ok, hopefully that was interesting. Normally things tend to halt and catch fire in very unremarkable manners. I'll soon start work on a display server, that'll surely give some fun craziness.

This thread went quiet? Anyways, I managed to get a weird corruption problem when working on my GUI:

sortie wrote:This thread went quiet?

You posted here just last month, as have others the months previous.

Wow !
It's crazy to think that this all is not voluntary !

You all are lucky, I don't know if it's my OS design or something, but I never get such crazy screenshots !
Instead, I have setup a special system for debugging. When a program is doing something it shouldn't (such as writing memory it shouldn't, or other things of this kind), the OS detects it and skips to a special "debug mode" and launches a (very complicated) program I made where all the symbol tables of all programs are loaded, and where I can see the (almost) whole state of the computer. Then, it starts a "recovery process", thus making the whole system operationnal again, and starts some kind of debugging console.

That said, I have not yet worked on my graphical shell nor GUI. The only strange thing I got was when writing my audio driver (at the time, it was just an emulated SoundBlaster one). It went a bit crazy... And Vivaldi's "4 Seasons" was no longer what I thought it was...

Lessons to be learned:
1) add some locking to console code
2) don't ever let all APs panic at the same time



It looked even better in a bit more realish QEMU:

Printf and pointers to member functions don't mix well it seems.



EDIT:

You know that feeling when you find an incredibly stupid mistake?

Something recent, and fun

I got something wrong (not sure yet) in my image loading code... this is what happened for a simple 16x16 "new file" icon... not sure users will make the association.

@thepowersgang

Access is spelled with two 'C's.

Hehe the one c is actually intentional. I hope =)

What's with the too few Cs? it's got an S too many

I was working on a real mode console. It seems to have broke when I tried implementing backspace.
The reason is because 1) I made an error in my backspace procedure - Might help to delete both the character and the style, it was basically picking a random one, and 2) I forgot to have the normal text procedure jump past the backspace.

anyway, see how you like it. In case you're wondering, it says, "Ghostss<backspace> are mythological creaa<backspace>tures, which seem to like the dark.

Accidentally plotted each character backwards.



(12:51:29 AM) <klange> IN SOVIET RUSSIA, BACKWARDS PRINT TEXT!

I'm not even sure what happened here.
The buffered text came out just fine...

I figured I'd improve the standards compliance of my stdio routines by adding setvbuf and implementing buffering semantics. Something went wrong.



Init wasn't pleased and the shell much less. Interestingly, the system did run - although it crashed a lot and the shell was an unreadable mess!