sergej-klimenko

I've simplified test, replacing nested collection with some simple char fields of sufficient length. [z_plsql_types_coll_test](https://github.com/sergej-klimenko/godror/blob/nested_mem_test/z_plsql_types_coll_test.go) And this variation also failed : ``` z_plsql_types_coll_test.go:126: TestPlSqlObjColl/inOut: started with RSS 43, got 107...

> Yes there is a gradual RSS increase and it is most likely due to frequent malloc and free operations. But it not the case with same test in python-oracledb...

> [5d7d9d2](https://github.com/godror/godror/commit/5d7d9d230f519a5bbd833ffbf6dd1e76281b365c) : if I print out dpiObject.refCount in Object.Close, it seems that those PAIR objects are one instance with a hell'of a lot of references - and dpiObject_release does...

My output ``` VERBOSE=1 GODROR_WARN_DPI_OBJECT_REFCOUNT=1 go test -run=Nested/inOut 2>&1 | uniq -c 1 export GODROR_TEST_DSN="user=demo password=SECRET-*** connectString=localhost:1521/ORA19EE connectionClass=TestClassName enableEvents=1 poolIncrement=1 poolMaxSessions=16 poolMinSessions=1 poolSessionMaxLifetime=1h0m0s poolSessionTimeout=5m0s poolWaitTimeout=30s" 1 # user=demo password=SECRET-*** connectString=localhost:1521/ORA19EE...

@anthony-tuininga > I tried a similar script in python-oracledb and see a similar increase in memory. This is my variant of the test using python-oracledb, and result output - no...