Your application's resource manager can call the Latch_Purge_by_Address_Space service to purge all granted and pending requests for a group of requestors within a group of latch sets. Latch_Purge_by_Address_Space is particularly useful when a task or address space is termi- nating, and you want to release latches that were obtained by a terminating task or programs that are running in a terminating address space.
To effectively use this service the latch_set_names and your requestor_IDs should be defined such that they have a common portion and a unique portion. Groups of latch sets can then be formed by masking off the unique portion of the latch_set_name and groups of latch requests in a latch set can then be formed by masking off the unique portion of the requestor_ID. This technique is very useful when keeping track of the unique portion of the requestor_ID might be difficult or expensive. A single purge request to can then handle multiple latch sets and multiple requests in a latch set. Recovery routines should call Latch_Purge_by_Address_Space when one or more errors prevent requestors from releasing latches.