MDEV-37797 - main.information_schema_db fails sporadically

[svoj]

MDEV-20549 test output was affected by incompletely closed preceding connections.

Fix for similar issue was attempted by f4eb095, it introduced wait_until_disconnected.inc. For an explanation of why it didn't work see bead24b commit comment.

Another fix was attempted in bead24b. It didn't help either, because connection is removed from information_schema.processlist after COM_QUIT handling.

Make test for MDEV-20549 agnostic to concurrent connections by querying information_schema.processlist only for connections that it uses.

[Copilot]

Pull Request Overview

This PR fixes a sporadic test failure in main.information_schema_db caused by incompletely closed connections from preceding tests affecting the MDEV-20549 test output. The fix replaces broken connection cleanup patterns with a more reliable approach.

• Hardcodes $count_sessions to 1 instead of using the broken count_sessions.inc
• Removes redundant connection switching before disconnections
• Uses wait_until_count_sessions.inc to ensure connections are properly closed

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
mysql-test/main/information_schema_db.test	Implements the connection cleanup fix with hardcoded session count and simplified disconnect logic
mysql-test/main/information_schema_db.result	Updates expected test output to reflect the removed connection switching commands

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[svoj]

@vuvova thinking more on this... given that sleep(1) (suggested in MDEV) uncovers dozens of similar test failures, current approach assumes we'd have to surround each end every disconnect with wait_until_count_sessions.inc, which is ridiculous.

We could make this particular (and probably some other) test agnostic to concurrent connections by changing queries to something like select current_user(),user from information_schema.processlist where id=connection_id().

Some tests can't probably be made agnostic to concurrent connections, in this case we add wait_until_count_sessions.inc to such tests rather than to every disconnect statement.

[vuvova]

ouch! I fixed it in 090de4e, but forgot to push. Please, free free to push your fix with =connection_id() or mine with !='mysqltest_1' or leave it to me to push. As you prefer.

[svoj]

@vuvova with sleep(1) there're also testdb_1 and testdb_2, so !='mysqltest_1' is not going to be enough. I'd go for =connection_id(). Thanks for your feedback!