Hi Nicolas,

Thanks for the tip again, it now worked with freshly-built neko and mysql.ndll 
from 1.8.0 distribution in both interpreted and JIT modes. Weird but ok for now.

Kind regards,
  Mike

-----Original Message-----
From: [email protected] 
[mailto:[email protected]] On Behalf Of Nicolas Cannasse
Sent: Thursday, December 11, 2008 4:25 PM
To: Neko intermediate language mailing list
Subject: Re: [Neko] Neko debug and MS VC 2008

Michael Pliskin a écrit :
> Thanks for the tip, now it fails within the interpreter, see call stack 
> below. Looks like acc is invalid. Any other pointers? :)

It's quite difficult to track such bugs.

One thing we've been doing is (under Linux) run under gdb and on 
segfault we can use the following command :

call neko_vm_dump_stack(neko_vm_current())

it will print the module callstack on stdout, which enables to know at 
least which exact operation is failing.

Another possibility would be to catch the error on Windows (using __try 
MSVC statement) and convert it into a val_throw() call to produce a 
proper Neko exception.

Once the concerned function is isolated, it's needed to know which value 
is not correct. In general it's because it's been GC'ed while it 
shouldn't have.

Which database connection are you using ? mysql5.ndll is our custom 
mysql5 driver, you might want to remove it and try again, so Neko uses 
default mysql.ndll instead.

Best,
Nicolas

-- 
Neko : One VM to run them all
(http://nekovm.org)

--
Neko : One VM to run them all
(http://nekovm.org)

Reply via email to