[Guy Harris <gharris@xxxxxxxxx>]

Julian Cable wrote:

> I have a couple of problems with the dissectors I've written. Any ideas would be really helpful.

As Jaap noted, you're only calling dissect_one_packet() if a protocol 
tree is being built.  On the first pass through the packets, when the 
capture is being read, if there's no color filter or read filter or tap 
configured to force protocol trees to be built, the protocol tree isn't 
built, and any code inside "if (tree) { ... }" isn't executed.

You probably have different color filters, or other preferences that 
would affect whether protocol trees are built or not, configured on 
Linux and on Windows; your Linux configuration probably causes protocol 
trees to be built on the first pass, and your Windows configuration 
probably doesn't.

Code that calls subdissectors, *OR* that does reassembly (e.g., that 
calls any of the fragment_add routines), has to be called on the first 
pass through the packets, so it can't be put inside "if (tree) { ... }". 
 I'd remove the "if (tree)" stuff from dissect_msc_data().