Repository navigation
Conversation
DuckDB already prepared for binlog 2PC but had no recover callback, so ha_count_rw_2pc() was 0 and Gtid_log_event asserted. Advertise recover (no prepared XIDs survive restart).
| @@ -0,0 +1,20 @@ | |||
| --source ../include/have_duckdb.inc | |||
| --source include/have_log_bin.inc | |||
| --source include/not_msan.inc | |||
There was a problem hiding this comment.
Can you remove the not_msan.inc. there's currently no compatibility problem.
gkodinov
added
the
External Contribution
gkodinov
left a comment
gkodinov
left a comment
There was a problem hiding this comment.
Thank you for your contribution. This is a preliminary review.
I'd like to request changes based on the .test file comment: it seems that there're some redundant tests there.
| return 0; | ||
| } | ||
|
|
||
| #if MYSQL_VERSION_ID >= 110800 |
There was a problem hiding this comment.
This is most likely killing the messenger.
The real issue is that Gtid_log_event::Gtid_log_event() expects that ha_count_rw_2pc() would return non-zero. And it can only return non-zero if the recover hook is defined.
Why is that, I have no idea (and is a question for the final review).
Feel free to ignore this for now, but I strongly suspect that this is wrong and it will come up in the final review.
| SELECT * FROM t1; | ||
| DROP TABLE t1; | ||
|
|
||
| CREATE TABLE t1 (id INT PRIMARY KEY, v VARCHAR(16)) ENGINE=DuckDB; |
There was a problem hiding this comment.
Why do you need two separate tables here? The mdev claims that the above is enough to reproduce.
3 participants
DuckDB already prepared for binlog 2PC but had no recover callback, so ha_count_rw_2pc() was 0 and Gtid_log_event asserted.