Device driver detection and initialization code seems awfully complicated

[sam-falvo]

Maybe it's because I'm new to the system; however, it seems like the initialization code is pretty intricate and fragile. I can't guarantee I would be able to complete this task; but, I'd like to document the current control flow and theory-of-operation, and maybe come up with simpler code that results in the same outcome.

[zuiko21]

It is complex and ugly, I agree...

Generally speaking, I try to optimize boot code for minimal size instead of speed, because it gets executed only once per session; but with the 65xx addressing limitations, I'm afraid my code is far from optimal either way :-(

We've discussed some issues about interrupts and some improvements have been designed. In order to make drivers (sooner or later) loadable on the fly, I will remove that initialization code from the kernel.s file, and get it into API functions, like DR_INST and DR_SHUT. Initially, during boot DR_INST would be called for each detected driver (currently into a static list in ROM at drivers_ad) but that could be used while running, just passing the header address to it -- when there's a way to dynamically allocate its variables, which has been spoken for.

Add to this the improvements designed for interrupt queues, and that's why I ask you a bit of time for this, as all driver stuff will be rewritten for a much more generic design.