MDEV-39008 FederatedX: set time_zone on connection change

[PeterKoletzki]

When the underlying connection to the remote server changes (e.g. load balancer switching nodes or driver reconnect), the session variable time_zone was not set again. TIMESTAMP values could then be wrong.

Track the connection (mysql.net.vio) and re-send SET time_zone='+00:00' when the vio changes so TIMESTAMP remains consistent.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ PeterKoletzki
❌ PeterSecusmart
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[gkodinov]

Thank you for your contribution! This is a preliminary review.

I'd love if you would consider adding a regression test for this. It can be as simple as:

1. open a federatedX conneciton
2. restart the remote server
3. try a query that requires a timezone
4. try another query that requires a timezone
5. cleanup

A slightly more advanced version would be to add a DBUG_EXECUTE_IF() with a unique key name, that, when set, will disconnect the federatedX session before the next query. And then do as the above, but instead of restarting the remote just set the flag on the federatedX server.

[PeterKoletzki]

I have added the test and made the change to use a boolean flag.
You are right, it looks cleaner.

Sorry. But I have no idea what mysql.cc does.
So it is better that I do not change anything that I do not understand. or have no knowledge

[PeterKoletzki]

federated.timezone_after_reconnect 'X' w13 [ pass ] 1444

The log supposed that there is a memory leak after the test sequence related to client.cc. I have not a idea whats going on, because my test passed and my company refused me to spend more time in this testing things.

[PeterKoletzki]

Cursor AI make some comments after my request to analyse this isuue:

MDEV-39008: With timezone_after_reconnect (reconnect path, backend restart),
server shutdown reports "Warning: Memory not freed: N" (e.g. 9304). PR without
this test had no such warning. FederatedX fix itself allocates nothing; leak
likely in libmariadb reconnect path or server cleanup. Suppression added so CI
passes. Maintainers may remove this and fix the leak, or keep/refine the pm.

I have added a adjusted suite.pm by Cursor. We will see ...