[avr-gcc-list] Re: maga256 patch for GCC 4.2

[Harald Kipp]

> > 1. ...why this implies to modify the stack pointer on entry. This 
> > spoils any the stack pointer modification in the .init segment.
>
> In begining of 'main'  function stack pointer should be modifed for 
> reservation of space for  'main'  local variables in stack. 
> Modification of stack pointer can be excluded only if  'main' don't 
> have local variables.
Sorry, I should have been more precise. If I remember correctly, the 
stack pointer hadn't been simply _adjusted_ to make room for local 
variables but had been _overridden_ to _fixed_ address. Fixed in the 
sense of that it is impossible to specify the stack location at runtime, 
which is fortunately offered by the .init segments. Dynamic 
determination of the stack at runtime is most useful to run the same 
firmware on different hardware layouts and sometimes for multithreading 
extensions (in my case).

> > 2. ...your statement about "In oficial GCC 4.x main() do not 
> > converted to a normal function." I'm sure that in official GCC main() 
> > is always a subroutine and always returns an int.
>
> No. Oficial GCC jump to main.
Then I was wrong here. Sorry for any confusion.

Harald