stephane@skysql.com, MASTER_GTID_WAIT() is defined in the sense of gtid_slave_pos_pos__wait(). Could have (and should've in my taste) been taken such name, but apparently the name chosen to follow a pre-existing wait function.

We can go along with this ticket to rename it with necessary deprecation steps, I think.

HI Andrei, why not simply defining it as gtid_binlog_pos_wait() on the executed GTID binlog position it can be queued and mutexed after the binlog 2PC and would only work on a slave having log_slave_updates=1, current implementation would be rename gtid_slave_pos_wait() but i guess it also lack or the replication source or a liste of domain ids to compare with

stephane@skysql.com, bonjour! Your comment just forces me to think towards a master-slave neutral binlog-agnostic-too WAIT_FOR_GTID(list-of-gtids-to-wait-until-they-got-committed.

That is the user application would call SELECT wait_for_gtid() on any server.
But I still would like to replace the old one with it.

Please review this!

Cheers,
Andrei

Привет Andrei,

Good to hear from you and see that this task is moving.
Make sens to get an other name to not break existing application using it already and master _wait_gtid could be a alias be renamed wait_for_leader_gtid() as master was deprecated if i remember.

Cheers
/svar

Bonjour, Stephane! Would you also like the offered WAIT_FOR_GTID()?

Yes yes for sure