Peter Eisentraut

Isn't this supposed to be the behavior of the existing fast-fail check? The comment at the top of `check_fast_fail()` says as much: https://github.com/pgbouncer/pgbouncer/blob/ab7957160b2ec92a183f699e3112b2d0e10bfbbb/src/objects.c#L569 Is this different from what you want?

I don't understand what this does. Your explanation could use a lot more information and details. Please squash your commits into sensible pieces.

You didn't miss anything. Someone just needs to write some code. :)

> It would be cool to be `target_session_attrs`, but is it possible without a larger rewrite ? Your analysis is not wrong.

> With the latest release and new feature allowing multiple hosts to connect to, are we any closer to implementing this feature? no ;-)

I have committed a fix for the first issue, but with the code a bit different. I'll continue to work through the other issues.

The second item has been fixed now. (In the future, I would recommend sending separate pull requests for separate issues. Otherwise it makes reviewing really hard if you don't know...

Looking at the third item, I'm not sure about the fix. Having a user with an empty password is a valid state, so we can't just take that as license...

Does the fix from #329 (currently on in master) help?

How did you build pgbouncer?