On 21/07/14 17:34, [email protected] wrote:
commit 58e9c3add8c3fbe88f930531f237f070a652249b
Author: Howard Chu <[email protected]>
Date:   Mon Jul 21 08:33:59 2014 -0700

     Fix mdb_cursor_count, reject uninit'd cursor

I think EINVAL is wrong, it need not reflect an obvious
programmer error.  MDB_NOTFOUND seems better.  Consider:

    mdb_cursor_put(cursor, &key, &data, 0);
    mdb_del(txn, dbi,  &key2, NULL));
    rc = mdb_cursor_count(cursor, &count);

where key2 happens to match key.

Makes me wonder about the EINVAL for an uninitialized xcursor.
Maybe cursor_count should initialize it in that case.

--
Hallvard

Reply via email to