Re: [Pnet-developers] The stack trace bug revisited

[Rhys Weatherley]

On Monday 26 May 2003 04:18 am, Gopal V wrote:

> it is the second set_stack_trace that I don't like .... why should it
> catch something there ?.

In CVM, exception catcher blocks extend across the entire method, not just the 
parts with try blocks.  The exception filtering code at the end of the method 
sorts out which part should get the exception.  If no part wants it, then it 
rethrows it.

It was simpler this way: CVM walks up the stack to find the first exception 
catcher and passes control to it.  It doesn't has to do any type filtering, 
because the code can do it.

The engine seems to be doing the right thing execution-wise, but I'm now 
seeing the missing line number information too.  Reproducible at last!

> Also I'm removing the skipframes hack from engine/lib_diag.c now that
> we don't need it ..

Good.

Cheers,

Rhys.