[En-Nut-Discussion] Nut/OS Initialization

[Alain M.]

PragmaLab escreveu:
> what about getting the address of the reset vector and start debugging from
> there? Just look in the map-file. I'm not familiar with Eclipse but at least
> it must be possible to set breakpoints on (HW)addresses. 

It seams like it is easy, but I never manged to make it. If you set a 
breakpoint using the debugger, Eclipse does not enter the source level 
debugging mode. It could take an expert in Eclipse to do that, if anyone 
can find one...

> But let us stay focussed on the interesting discussion of how the HW
> initialisation can be improved in NutOS. Calling code before main: nothing
> wrong about that IMHO. Cannot think of another way to get your int i=5; done
> runtime. In NutOS quite a lot of initialisation is done before calling
> main(). Debugging has never been a problem (for me).

It has for me ... with NutOS ... on an ARM9 board from Olimex. I spent 
days to find a simple error in the startup initialization, and still 
some more to go because debugg is not working well, only program load. 
Call it a recurrent nightmare :(

That is whi, INHO, the more that I can push to after main() the simpler 
it can get :)

Alain