Re: [Amforth-devel] Data stack pointer problem?

[Matthias T. <mt...@we...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello Maciej,

> I have narrowed the problem to the start of COLD codeword that first
> calls SP0 to fetch the address of the stack (that should be
> stackstart) and then SP! to store the top of stack value to Y
> register.
>
> The problem is that SP0 returns its value on the stack while Y
> register is still uninitialized and contains XT_NOOP address (from
> reset routine). In the SVN this problem may be a bit hidden because
> top of stack is kept in a register pair, but the execution will start
> with savetos macro that uses Y register anyway and might corrupt the
> RAM anyway.
> 
> I think that this could be solved by adding word SP0! that just loads
> Y with stackstart value or by moving in amsforth.asm, reset routine
> the parameter stack pointer initialization just before the jump to
> Forth machine so that Y register will have the value of stackstart.

There is a much simpler bugfix, that does not need any additional
code: Move the code block for the data stack pointer in
amforth.asm to the proper postion (at the end). Just fixed in svn..

Thank you!

Bye
Matthias
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (GNU/Linux)

iD8DBQFGFK9vLo3irIddFw4RAsnJAJ42OeMDdBgrNLT75tyikwmlQ56bQgCfRKi6
CpWrXWIcHC1HkKOBMzOMz4I=
=uTW4
-----END PGP SIGNATURE-----