Please report new issues athttps://github.com/DOCGroup
ooks like there is a race condition in TAO_Adapter_Registry. I happened to run into this when I was inspecting some crashes. See here -------------------------- Thread one ------- resolve_initial_references ("IORTable"); .. TAO_ORB_Core::resolve_ior_table_i () .. TAO_Adapter_Registry::insert () --------------------------------------------- --------- Thread two ---------------------------- ORB::run () .. TAO_Request_Dispatcher::dispatch () TAO_Adapter_Registry::dispatch () ------------------------------------------------- Has anyone run into this before? We have been able to work around this problem by somecode changes for now. WOuld be interested to know if someone is adding a fix without grabbing a lock to the critical path?
Looks like a refcount on dispatch pattern needs to be inplemented if we have to prevent from grabbing a lock on the critical path. Just stashing this bug away for later reference.
to pool