Wrong result when using index-merge

[zyguan]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

/* test */ drop table if exists t1, t2;
/* test */ create table t1  (c_int int, c_str varchar(40), c_datetime datetime, c_decimal decimal(12, 6), primary key(c_int), key(c_int), key(c_str), unique key(c_decimal), key(c_datetime));
/* test */ create table t2  like t1 ;
/* test */ insert into t1 (c_int, c_str, c_datetime, c_decimal) values (6, 'sharp payne', '2020-06-07 10:40:39', 6.117000) , (7, 'objective kare', '2020-02-05 18:47:26', 1.053000) , (8, 'thirsty pasteur', '2020-01-02 13:06:56', 2.506000) , (9, 'blissful wilbur', '2020-06-04 11:34:04', 9.144000) , (10, 'reverent mclean', '2020-02-12 07:36:26', 7.751000) ;
/* test */ insert into t2 (c_int, c_str, c_datetime, c_decimal) values (6, 'beautiful joliot', '2020-01-16 01:44:37', 5.627000) , (7, 'hopeful blackburn', '2020-05-23 21:44:20', 7.890000) , (8, 'ecstatic davinci', '2020-02-01 12:27:17', 5.648000) , (9, 'hopeful lewin', '2020-05-05 05:58:25', 7.288000) , (10, 'sharp jennings', '2020-01-28 04:35:03', 9.758000) ;
/* test */ select * from t1 where c_decimal < all (select c_decimal from t2 where t1.c_int = t2.c_int and t1.c_datetime > t2.c_datetime and t2.c_decimal = 9.060 or t2.c_str <= 'interesting shtern' and t1.c_int = t2.c_int); -- flags: QUERY

2. What did you expect to see? (Required)

/* test */ select * from t1 where c_decimal < all (select c_decimal from t2 where t1.c_int = t2.c_int and t1.c_datetime > t2.c_datetime and t2.c_decimal = 9.060 or t2.c_str <= 'interesting shtern' and t1.c_int = t2.c_int); -- flags: QUERY
-- test >> +-------+-----------------+---------------------+-----------+
-- test    | c_int |      c_str      |     c_datetime      | c_decimal |
-- test    +-------+-----------------+---------------------+-----------+
-- test    |     7 | objective kare  | 2020-02-05 18:47:26 |  1.053000 |
-- test    |     8 | thirsty pasteur | 2020-01-02 13:06:56 |  2.506000 |
-- test    |    10 | reverent mclean | 2020-02-12 07:36:26 |  7.751000 |
-- test    +-------+-----------------+---------------------+-----------+

3. What did you see instead (Required)

/* test */ select * from t1 where c_decimal < all (select c_decimal from t2 where t1.c_int = t2.c_int and t1.c_datetime > t2.c_datetime and t2.c_decimal = 9.060 or t2.c_str <= 'interesting shtern' and t1.c_int = t2.c_int); -- flags: QUERY
-- test >> +-------+-----------------+---------------------+-----------+
-- test    | c_int |      c_str      |     c_datetime      | c_decimal |
-- test    +-------+-----------------+---------------------+-----------+
-- test    |     6 | sharp payne     | 2020-06-07 10:40:39 |  6.117000 |
-- test    |     7 | objective kare  | 2020-02-05 18:47:26 |  1.053000 |
-- test    |     8 | thirsty pasteur | 2020-01-02 13:06:56 |  2.506000 |
-- test    |     9 | blissful wilbur | 2020-06-04 11:34:04 |  9.144000 |
-- test    |    10 | reverent mclean | 2020-02-12 07:36:26 |  7.751000 |
-- test    +-------+-----------------+---------------------+-----------+

4. What is your TiDB version? (Required)

master @ f1e5877

[guo-shaoge]

Looks like we forgot to rebuild range/dagPB when got correlated columns. Really thanks for your issue @zyguan .

[github-actions]

Please check whether the issue should be labeled with 'affects-x.y' or 'fixes-x.y.z', and then remove 'needs-more-info' label.