Re: [DotGNU]pnet & HPUX 64

[Gopal V]

If memory serves me right, K S GUNDU RAO wrote:
> The built-in test suite for the two components treecc-0.2.0 and
> pnet-0.5.0 passed, however pnetlib-0.5.0 is failing with core dump.
> I tried to compile and run hello.cs. This is also core dumping.

You were able to compile hello.cs ? ... which is very good news, as the
compiler proves to be more portable than the engine ... I think that
proves one good thing .... :)

> Enabled the debug support by modifying the file cvm_config.h by enabling
> IL_CVM_DUMP and rebuilt the binaries.

hmm... there are actually 5 different flavours of pnet's engine which
is picked for suitability...

Switch loop -- for the ANSI C compilers
Register -- for gcc on most systems
Token threaded -- computed goto and low memory optimised portable
Direct threaded -- speed optimised portable
Unroll -- CPU specific enhancements 

I think the IA-64 will be Direct threaded ?... (could you confirm ?) ...

> address@hidden>/usr/local/bin/ildb hello.exe
> ildb 0.5.0

ildb is a debugger for the IL code and is useful only for debugging 
libraries .. It is not generally useful for debugging the runtime 
engine bugs.

Could you try it with gdb ? .. I think a gdb stacktrace would be great
for tracking down the issue.

> address@hidden>uname -a
> HP-UX toddy B.11.23 U ia64 4096752436 unlimited-user license

James McParlane has tried to compile pnet with the aCC compiler on 
HPUX on PA-RISC 2.0 .. I think he hung up on the libffi issues ...
Hopefuly the gcc team have fixed it in gcc CVS . Rhys has mentioned
that the changes will be pulled into pnet soon.

> Help to solve this problem is highly appreciated.

We'll try ... provided libffi and libgc things are sorted out , we
might be able to resolve this issue by next release (or sooner).

Thanks,
Gopal
-- 
The difference between insanity and genius is measured by success