[
https://issues.apache.org/jira/browse/THRIFT-1184?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13055730#comment-13055730
]
Nevo Hed commented on THRIFT-1184:
----------------------------------
+1 I think the feature of running TNonBlocking on an existing event_base is an
important one, if there are any users that run this in a scenario that is not
run-forever then this would be a show-stopper for them
> event_base memory freeing by class TNonblockingServer
> -----------------------------------------------------
>
> Key: THRIFT-1184
> URL: https://issues.apache.org/jira/browse/THRIFT-1184
> Project: Thrift
> Issue Type: Bug
> Components: C++ - Library
> Reporter: Pavlin Radoslavov
> Labels: libevent
> Attachments: thrift1184_eventbase_free.patch
>
>
> The TNonblockingServer::eventBase_ pointer is assigned
> by method TNonblockingServer::registerEvents()
> so the libevent events are added to the proper libevent event base.
> Currently the TNonblockingServer destructor unconditionally
> calls event_base_free(eventBase_) to free the eventBase_ memory.
> In certain cases the libevent event base might be allocated and
> deallocated by a third-party software that uses Thrift, so it is
> not appropriate for the TNonblockingServer destructor to unconditionally
> call event_base_free(eventBase_).
> Please see attached patch for a possible solution:
> add an optional bool argument to method TNonblockingServer::registerEvents():
> the value of that flag controls whether the TNonblockingServer destructor
> should call event_base_free(eventBase_).
> The argument is optional, and defaults to true (i.e,
> event_base_free(eventBase_)
> is called) for API (and behavior) backward compatibility.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
