Details
-
Bug
-
Status: Stalled (View Workflow)
-
Minor
-
Resolution: Unresolved
-
1.4.2
-
None
-
2020-2
Description
see mysql/queries/working_tpch1/qa_fe_cnxFunctions/hex.sql
The query:
select cidx, CDECIMAL1, HEX(CDECIMAL1) from DataTypeTestm order by cidx; |
should return
1 -9 FFFFFFFFFFFFFFF7
|
2 -8 FFFFFFFFFFFFFFF8
|
3 -7 FFFFFFFFFFFFFFF9
|
4 -6 FFFFFFFFFFFFFFFA
|
5 -5 FFFFFFFFFFFFFFFB
|
6 5 5
|
7 6 6
|
8 7 7
|
9 8 8
|
10 9 9
|
11 0 0
|
But instead returns
1 -9 FFFF
|
2 -8 FFFF
|
3 -7 FFFF
|
4 -6 FFFF
|
5 -5 FFFF
|
6 5 5
|
7 6 6
|
8 7 7
|
9 8 8
|
10 9 9
|
11 0 0
|
This appears to only affect decimal types. I looked in the utils/funcexp/hex.cpp code and no changes there should have caused this.
Attachments
Issue Links
- is blocked by
-
-
- Closed
-
- is part of
-
MCOL-3747 Regression in 1.4 working_ssb_compareLogOnly/sub/order_limit_sub
-
- Closed
-
Activity
changed the fix version to n/a because it was messing up daniel's accounting. Change it to whatever is appropriate once it's fixed.
The ref file (mysql/queries/working_tpch1/qa_fe_cnxFunctions/hex.sql.ref.log) is temporarily changed to match (incorrect) output from ColumnStore. We need to revert it back when MDEV-20548 is fixed.
MDEV-20548 is not done, so we cannot do this for 1..5.3.
How to repeat the problem:
First, I create the table:
cd mariadb-columnstore-regression-test
|
mysql test < mysql/scripts/create_datatypetestm.sql
|
Now run this query:
select cidx, CDECIMAL1, HEX(CDECIMAL1) from datatypetestm order by cidx; |
+------+-----------+----------------+
|
| cidx | CDECIMAL1 | HEX(CDECIMAL1) |
|
+------+-----------+----------------+
|
| 1 | -9 | FFFF |
|
| 2 | -8 | FFFF |
|
| 3 | -7 | FFFF |
|
| 4 | -6 | FFFF |
|
| 5 | -5 | FFFF |
|
| 6 | 5 | 5 |
|
| 7 | 6 | 6 |
|
| 8 | 7 | 7 |
|
| 9 | 8 | 8 |
|
| 10 | 9 | 9 |
|
| 11 | 0 | 0 |
|
+------+-----------+----------------+
|
Notice, the FFFF in the last column is wrong.
Now if I change the engine from ColumnStore to InnoDB, the results become correct:
alter table datatypetestm engine=innodb; |
select cidx, CDECIMAL1, HEX(CDECIMAL1) from datatypetestm order by cidx; |
+------+-----------+------------------+
|
| cidx | CDECIMAL1 | HEX(CDECIMAL1) |
|
+------+-----------+------------------+
|
| 1 | -9 | FFFFFFFFFFFFFFF7 |
|
| 2 | -8 | FFFFFFFFFFFFFFF8 |
|
| 3 | -7 | FFFFFFFFFFFFFFF9 |
|
| 4 | -6 | FFFFFFFFFFFFFFFA |
|
| 5 | -5 | FFFFFFFFFFFFFFFB |
|
| 6 | 5 | 5 |
|
| 7 | 6 | 6 |
|
| 8 | 7 | 7 |
|
| 9 | 8 | 8 |
|
| 10 | 9 | 9 |
|
| 11 | 0 | 0 |
|
+------+-----------+------------------+
|
Investigated this issue. ColumnStore creates the correct string with length 16, however the server truncates the field length when the string is sent to it via store() call in fetchNextRow().
There is an open server issue for this: https://jira.mariadb.org/browse/MDEV-20548
Closing this issue.