Segmentation fault when work table present

[adamtwo]

Giraffez core dumps with Segmentation fault error when giraffez.BulkLoad is called but a work table is present.

Steps to reproduce:

1. Create a test database:

   CREATE DATABASE demo_db
   AS PERMANENT = 60e6,
       SPOOL = 60e6;
   

2. Execute the following python script. You may have to edit td_connection to point to your Vantage database instance.
   The script drops table demo_db.test_table_wt and then creates it. It then runs giraffez.BulkLoad against demo_db.test_table. With this naming convention, the script pretends that the work table is always there.

   import giraffez
   
   td_connection = {
       "host": "localhost",
       "username": "dbc",
       "password": "dbc"
   }
   
   try:
       with giraffez.Cmd(**td_connection) as cmd:
           cmd.execute("DROP TABLE demo_db.test_table_wt")
   except Exception:
       pass
   
   
   create_ddl = "create multiset table demo_db.test_table_wt(\
       INDEX_ID         BIGINT)"
   
   with giraffez.Cmd(**td_connection) as cmd:
       cmd.execute(create_ddl)
       show_ddl = cmd.execute("SHOW TABLE demo_db.test_table_wt")
       print(list(show_ddl))
   
   
   f = open("/tmp/temp.csv", "w")
   f.write("1")
   f.close()
   
   with giraffez.BulkLoad(**td_connection, table="demo_db.test_table") as load:
       load.from_file(
           "/tmp/temp.csv",
           delimiter=","
       )
   

3. Result:

   [Row({'request_text': 'CREATE MULTISET TABLE demo_db.test_table_wt ,FALLBACK ,\r     NO BEFORE JOURNAL,\r     NO AFTER JOURNAL,\r     CHECKSUM = DEFAULT,\r     DEFAULT MERGEBLOCKRATIO,\r     MAP = TD_MAP1\r     (\r      index_id BIGINT)\rPRIMARY INDEX ( index_id );'})]
   Segmentation fault (core dumped)
   

I've looked at the core dump. Backtrace points to giraffez/src/teradatapt.hpp:296 at if (err->Code == TD_SUCCESS):

PyObject* Exists(const char *tbl_name) {
    TeradataErr *err = NULL;
    PyObject *ret;
    ret = Py_False;
    Py_INCREF(ret);
    std::string query = "show table " + std::string(tbl_name);
    if ((err = this->ExecuteCommand(query, false)) != NULL) {
        if (err->Code != TD_ERROR_OBJECT_NOT_EXIST) {
            return NULL;
        }
        PyErr_Clear();
    }
    if (err->Code == TD_SUCCESS) { 
        // TODO: ref correct?
        ret = Py_True;
    }
    return ret;
}

When a work table exists ExecuteCommand() returns NULL. As a result, err object is also NULL. Line 296 then references memory address 0.

Many thanks to @acirtep for helping me identify the issue.