[Guy Harris <gharris@xxxxxxxxx>]

Frederic Peters wrote:

> The attached patch fixes the issue but I didn't known the inner
> workings of that dissector so it may be a bit rough.
The problem is that idl2eth-generated dissectors were explicitly 
calculating the length of the top-level item for the tree for the 
specific GIOP protocol, rather than using -1, which was causing a bad 
value to be passed, causing the assertion to fail.
I've checked in a change to make the GIOP-based dissectors we have use 
-1 - and to change idl2eth (ethereal_gen.py, to be precise) to generate 
dissectors that use -1; that fixes the problem.
(The capture file in question also shows that the heuristic dissectors 
for GIOP-based dissectors aren't working correctly - they set the 
protocol column and create the top-level protocol item *before* deciding 
definitively that the GIOP protocol being used is the protocol for that 
dissector.)