Re: [DotGNU] Oberon.net compat

[Rhys Weatherley]

On Monday 13 January 2003 12:41 am, Patrick Pletscher wrote:

> I attached the IL which doesnt function and also the sources of the
> program, if somebody wants to compile it itself.

I think I've fixed all the ilasm-related and engine-related bugs that caused 
this test case to fail.  There were a number of issues:

1. The "instance" flag that we already discussed.  Now, ilasm will add that 
flag automatically for "newobj", "callvirt", and "ldvirtfn" instructions.

2. There was a verifier bug related to single-dimensional arrays: the Oberon 
code was using "newobj" to create the array instead of "newarr".  Nothing 
wrong with that - it is allowed by the spec, although I originally thought 
that it wasn't.

3. The Oberon code did not have ".maxstack" declarations, so pnet's ilasm was 
assuming that the maximum stack size was zero.  Which caused other verifier 
issues.  I changed the default maximum stack size to 8, to match MS'es ilasm.

All of these changes are in CVS now.  That just leaves the "Process" bug in 
pnetlib.  Volunteers to look into that are welcome.

Cheers,

Rhys.