Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.3(EOL)
-
None
-
bb-10.3-release 9b59f78d16fb000
Description
I couldn't find it in the current tests, so don't know if it's intentional.
create or replace table t1 (i int, s bigint unsigned as row start, e bigint unsigned as row end, period for system_time(s,e)) with system versioning; |
insert into t1 (i) values (1),(2); |
update t1 set i = 3; |
MariaDB [test]> select * from t1 for system_time all; |
+------+-----+----------------------+ |
| i | s | e |
|
+------+-----+----------------------+ |
| 3 | 229 | 18446744073709551615 |
|
| 3 | 229 | 18446744073709551615 |
|
| 1 | 222 | 229 |
|
| 2 | 222 | 229 |
|
+------+-----+----------------------+ |
4 rows in set (0.00 sec) |
|
Existing transaction, old snapshot - OK |
MariaDB [test]> select * from t1 for system_time as of transaction 222; |
+------+-----+-----+ |
| i | s | e |
|
+------+-----+-----+ |
| 1 | 222 | 229 |
|
| 2 | 222 | 229 |
|
+------+-----+-----+ |
2 rows in set (0.00 sec) |
|
Existing transaction, current snapshot - OK |
MariaDB [test]> select * from t1 for system_time as of transaction 229; |
+------+-----+----------------------+ |
| i | s | e |
|
+------+-----+----------------------+ |
| 3 | 229 | 18446744073709551615 |
|
| 3 | 229 | 18446744073709551615 |
|
+------+-----+----------------------+ |
2 rows in set (0.00 sec) |
|
Non existing transaction (greater than latest) -- questionable |
MariaDB [test]> select * from t1 for system_time as of transaction 230; |
+------+-----+----------------------+ |
| i | s | e |
|
+------+-----+----------------------+ |
| 3 | 229 | 18446744073709551615 |
|
| 3 | 229 | 18446744073709551615 |
|
+------+-----+----------------------+ |
2 rows in set (0.00 sec) |
|
Non existing transaction (less than minimal) -- not good |
MariaDB [test]> select * from t1 for system_time as of transaction 1; |
+------+-----+----------------------+ |
| i | s | e |
|
+------+-----+----------------------+ |
| 3 | 229 | 18446744073709551615 |
|
| 3 | 229 | 18446744073709551615 |
|
+------+-----+----------------------+ |
2 rows in set (0.00 sec) |
|
Non existing transaction (intermediate value) -- not good |
MariaDB [test]> select * from t1 for system_time as of transaction 227; |
+------+-----+----------------------+ |
| i | s | e |
|
+------+-----+----------------------+ |
| 3 | 229 | 18446744073709551615 |
|
| 3 | 229 | 18446744073709551615 |
|
+------+-----+----------------------+ |
2 rows in set (0.00 sec) |
Attachments
Issue Links
- is duplicated by
-
-
- Closed
-
Activity
TRX_ID is something material. If there was a transaction, there was some TRX_ID. But unknown TRX_ID means there was no such transaction. We must not use unknown TRX_ID therefore and cannot link it to some point in time.
I tend to agree with midenok. A transaction id means there was a transaction with this id. If there was no such a transaction, you cannot show a view as of nonexistent transaction.
Question is about TR_table::query_sees() algorithm.
Here is my opinion. as of transaction 1 will see nothing. as of transaction big_number will see current rows. And here is the patch:
diff --git a/sql/table.cc b/sql/table.ccindex 4f90d429ce5..d235f37f312 100644--- a/sql/table.cc+++ b/sql/table.cc@@ -8751,11 +8751,18 @@ bool TR_table::query_sees(bool &result, ulonglong trx_id1, ulonglong trx_id0,if (!commit_id1){- if (!query(trx_id1))- return true;-- commit_id1= (*this)[FLD_COMMIT_ID]->val_int();- iso_level1= iso_level();+ if (query(trx_id1))+ {+ commit_id1= (*this)[FLD_COMMIT_ID]->val_int();+ iso_level1= iso_level();+ }+ else+ {+ // User specified some random transaction id.+ // Make it instant and use default isolation level for computation.+ commit_id1= trx_id1;+ iso_level1= static_cast<enum_tx_isolation>(thd->variables.tx_isolation);+ }}Alexey(midenok) has different opinion. He says as of unknown_trx_id should be an error.
We don't know how it should work and ask you what is the correct behaviour.