[Sqlalchemy-tickets] Issue #4252: double checkin due to handle_error called inside of _finalize_fairy -> reset (zzzeek/sqlalchemy)

[Michael B. <iss...@bi...>]

New issue 4252: double checkin due to handle_error called inside of _finalize_fairy -> reset
https://bitbucket.org/zzzeek/sqlalchemy/issues/4252/double-checkin-due-to-handle_error-called

Michael Bayer:

```
#!python

from sqlalchemy import create_engine
import mock
import psycopg2

e = create_engine("postgresql://scott:tiger@localhost/test")
# fill pool with one connection
e.connect().close()

print(e.pool._pool.queue)

# have an exception be a disconnect error
e.dialect.is_disconnect = lambda *arg, **kw: True

conn = e.connect()
trans = conn.begin()

# have a rollback raise, will cause disconnect + invalidate
with mock.patch.object(
    conn.connection, "rollback",
    side_effect=psycopg2.OperationalError("something went wrong with the rollback")
):
    conn.close()

# connection is checked in twice
print(e.pool._pool.queue)

```

at the end we get:


```
#!python

sqlalchemy.exc.OperationalError: (psycopg2.OperationalError) something went wrong with the rollback (Background on this error at: http://sqlalche.me/e/e3q8)
deque([<sqlalchemy.pool._ConnectionRecord object at 0x7f71910976a0>, <sqlalchemy.pool._ConnectionRecord object at 0x7f71910976a0>])

```

this is due to the reset agent calling _handle_error which then goes and does the invalidation sequence.