Using factory->create() on an endpoint that is down, causes a memroy leak

[Skmirch]

in the file cassandra_factory.cc , in the function.....

CassandraClient *CassandraFactory::createThriftClient(const string &in_host, int in_port, bool framed_transport)

In this function the open is called after the pointer to CassandraClient is created. Any exception in open causes the CassandraClient pointer to be lost resulting in a memory leak. Not being able to connect to an endpoint throws the thrift TTransport exception which therefore results in a memory leak that I am seeing.

Reversing the order like so solves the problem...

transport->open(); /* throws an exception */ CassandraClient *client= new(std::nothrow) CassandraClient(protocol);

This has been tested to work.

Thanks. -- Sandeep

[minaguib]

Thanks Sandeep. I've committed this fix in minaguib/libcassandra@cb68df9eddf50d5ad68436d1ab744a7ffd679793

I'll send Padraig a pull request including it.