Some uses of context handles may require you to write a context rundown procedure in the application code for the server. If communications between the client and server are broken while the server is maintaining context for the client, RPC invokes the context rundown procedure on the server to recover the resources represented by the context handle. If you declare a context handle as a named type, you must supply a rundown procedure for that type.
When a context requires a context rundown procedure, you must define a named type that has the context_handle attribute. For each different context handle type, you must provide a context rundown procedure as part of the manager code.
The format for the rundown procedure name is as follows:
context_type_name_rundown
A rundown procedure takes one parameter, the handle of the context to be run down, and delivers no result. For example, if you declare the following:
typedef [context_handle] void * my_context;
then the rundown procedure is as follows:
void my_context_rundown (my_context this_object);
Server application code that uses a certain context handle may be executing in one or more server threads at the time that RPC detects that communications between the server and the client that is using that context have broken. The context rundown routine will not be invoked until a return of control to the server stub has happened in each of the threads that were using the context handle.
If application code in any of these threads destroys the context before returning control to the server stub from which it was called, your context rundown procedure will not be executed.