set_mapistoreURI returning MAPI_E_NOT_FOUND if mapping is alreay in database.

[qantourisc]

In the file:
mapiproxy/libmapiproxy/backends/openchangedb_mysql.c:323 MAPISTATUS set_mapistoreURI
The query can return Query OK, 0 rows affected; Rows matched: 1 Changed: 0
Meaning the row is already on the requested value.
I had to change the check if (mysql_affected_rows(conn) == 0) { to if (mysql_num_rows(conn) == 0){

Please confirm this bug and I can make a pull request.

[qantourisc]

And another possible victim: set_system_idx(

[sixstone-qq]

You're right @qantourisc . I've just added a unit test for this case with the following result:

testsuite/libmapiproxy/openchangedb.c:177:F:Openchangedb MySQL backend:test_set_mapistoreURI:0: Assertion 'retval==MAPI_E_SUCCESS' failed: retval==-2147221233, MAPI_E_SUCCESS==0

We have to see which is the right solution.

[qantourisc]

New by request by someone else: #379

[sixstone-qq]

Do you agree with that solution?

[qantourisc]

Yes and Hell No: See my patch: #378

Why yes: It looks like it will work.

Why no:
Why on earth would you do twice the work ? You read the wrong docs on mysql_***_rows, if I'm not mistaking. And you do not need to change the DB connection settings to use mysql_num_rows.

You have to look @ https://dev.mysql.com/doc/refman/5.5/en/mysql-num-rows.html not @ https://dev.mysql.com/doc/refman/5.5/en/mysql-affected-rows.html.
"mysql_num_rows() is intended for use with statements that return a result set, such as SELECT. For statements such as INSERT, UPDATE, or DELETE, the number of affected rows can be obtained with mysql_affected_rows(). " Translation (or atleast how I read it): num_rows() will select the number of rows matches by the WHERE. If you want to know howmany rows you updated: use affected_rows()

Side thought: We also ought to check for >1 cause that is also wrong, if one wishes to protect against DB-schema bugs, and code bugs. But not really required.

[sixstone-qq]

I obtained the value for mysql_num_rows after executing the UPDATE query and here you are the results:

    sql = talloc_asprintf(mem_ctx,
        "UPDATE folders f "
        "JOIN mailboxes m ON m.id = f.mailbox_id AND m.name = '%s' "
        "SET f.MAPIStoreURI = '%s' "
        "WHERE f.folder_id = %"PRIu64,
        _sql(mem_ctx, username), mapistoreURL, fid);
    OPENCHANGE_RETVAL_IF(!sql, MAPI_E_NOT_ENOUGH_MEMORY, mem_ctx);
    retval = status(execute_query(conn, sql));
        printf("Count %d affected %d\n", mysql_num_rows(conn), mysql_affected_rows(conn));

The results after running testsuite.

Count 39703056 affected 1
Count 39703056 affected 0

So for me the mysql_num_rows is not set on UPDATE statements.

[qantourisc]

Ok this is where mysql and mariadb differ it seems.So your solution is correct.