Re: [Pnet-developers] Porting work remaining question

[Peter Colson]

On 29/09/2004, at 11:20 PM, Gopal V wrote:

> > > > > > COP_L2I -1-
> > > > > > CVM_WORDS_PER_LONG = 1
> > > > > > sizeof(ILInt64) = 8
> > > > > > sizeof(CVMWord) = 8
> > > > > > stacktop(llx)= 80090dd2e8
> > > > > > &(stacktop[-CVM_WORDS_PER_LONG]) = 88090dd2e0
>
> This seems just about right .. unless the entire
> address of stacktop is wrong ?. My suspicions are
> more on the mload that you have there in your dump.

Yes, comparing the successful OSX CVM dump with the failing
AS/400 one, they are pretty similar up to the point of failure where
there is an mload 0, 1 occurring on the AS/400:

OSX:

0x003684FF:  set_num_args        0
0x00368501:  ckheight
0x00368506:  mk_local_1
0x00368507:  push_thread
0x00368508:  waddr               0
0x0036850A:  call_native         0x000570E0 
(DefaultEncoding.InternalCodePage "()i")
0x00368513:  return_1


AS/400:

0x09062923:  set_num_args        0
0x09062925:  ckheight
0x0906292A:  mk_local_1
0x0906292B:  push_thread
0x0906292C:  waddr               0
0x0906292E:  call_native         0x09058500 
(DefaultEncoding.InternalCodePage "()i")
0x0906293F:  mload               0, 1
0x09062942:  l2i

Be aware that on OSX the pointer size matches long (4 bytes) whereas
on the 400 the long type (4bytes) is half that of the pointer size 
(8bytes).
That seems to have an effect on padd_offset's and such which are
different in the two dumps.

Should add I'm on pnet-0.6.6 (not 6.8).

> > '&(stacktop[-CVM_WORDS_PER_LONG])' is
> > de-referenced that I get an error accessing out of
> > range memory.
>
> Very very strange ... I will run ilrun with CVM DUMP
> with current CVS and let you know if I catch on
> some difference.

Thanks, any help appreciated.

> > mscorlib.dll. Are there other supporting components
> > required at
> > this point?
>
> I18n.*.dlls would be nice :)

OK, got these now.


Regards,
Peter Colson.