Index Join executes wrongly when there's index with column has length

[winoros]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. What did you do?
   If possible, provide a recipe for reproducing the error.

create table t(a int);
create table tt(a int, b varchar(10), index idx(a, b(3)));
insert into t values(1);
insert into tt values(1, 'aaaaaaa'), (1, 'aaaabbb'), (1, 'aaaacccc');
select /*+ TIDB_HJ(tt) */ * from t, tt where t.a=tt.a and tt.b in ('aaaaaaa', 'aaaabbb', 'aaaacccc');
select /*+ TIDB_INLJ(tt) */ * from t, tt where t.a=tt.a and tt.b in ('aaaaaaa', 'aaaabbb', 'aaaacccc');

2. What did you expect to see?

Both them returns correct dataset.

3. What did you see instead?

Only hash join returned. Index join returned a empty dataset.

4. What version of TiDB are you using (tidb-server -V or run select tidb_version(); on TiDB)?

[winoros]

Another case relates to index join and index with prefix index column. Though the root cause is not the same.

select /*+ TIDB_INLJ(tt) */ * from t, tt where t.a=tt.a and (tt.b = 'aaaaaaa' or tt.b = 'aaaabbb');

Its result is also a wrong one compared with TIDB_HJ hint.