Fixed #23652 -- Restored database name after destroying test database

[claudep]

Thanks Bjarkias for the report.

[timgraham]

Looks good. How do you feel about backport? Doesn't seem like it's critical functionality that is relied upon if it's been a regression since 1.4. Guess we could at least update the older docs to reflect actual behavior.

[claudep]

I don't see it as critical either. Updating older docs might be a good idea.

[dcreager]

We just ran into this with 1.6. Would it be possible to backport this patch?

(For context, we run into this via south's migrationcheck command, which tries to create a new test database for each app with migrations. Before we thought the issue was that migrationcheck wasn't creating the test databases at all, so we used a sqlite in-memory database as a workaround. We just recently tried to update this to use a proper postgres database, to match what's running in prod, which exposed this regression.)

[timgraham]

As 1.7 as been released, 1.6 only receives security updates per our supported versions policy.

[dcreager]

Yeah that's what I figured...I thought that since it was simple it might fall under the "developer's discretion" clause. 😄 Ah well, thanks anyway for your hard work!