["Craig Reese" <cfreese5@xxxxxxxxxxx>]

 

(as recent as 1.3.3)

 

 

From: Stig Bjørlykke <stig@xxxxxxxxxxxxx>
Date: Mon, 1 Jun 2009 02:06:56 +0200

I just gave it a quick try on a protocol registering to "udp.port".

The first problem occurred in an assert in splash_update, proposed
patch attached.

...

Index: gtk/about_dlg.c
===================================================================
--- gtk/about_dlg.c (revision 28551)
+++ gtk/about_dlg.c (working copy)
@@ -229,9 +229,10 @@
}
if(ul_count == 0) /* get the count of dissectors */
- ul_count = register_count() + 6; /* additional 6 for:
+ ul_count = register_count() + 8; /* additional 8 for:
dissectors, listeners,
registering plugins, handingoff plugins,
+ registering python, handingoff python,
preferences and configuration */
main_lb = g_object_get_data(G_OBJECT(win), "protocol_label");

 
The fix adds 2 (one for protocol, one for handoff) but in reality you need to
add 2 for _each_ python plugin registered.  
 
I'm not sure how to fix things, but the real problem seems to be that the counts are
"hard coded" in proto_reg_count, and handoff_reg_count in epan/dissectors/register.c 
and don't take into account that the python ones are added dynamically.
 
On a related note...  Has the Python work been orphaned?  I can't find much in
the way of others testing/fixing these basic bugs.