On Mon, Mar 30, 2020 at 12:48:51PM -0300, Alvaro Herrera wrote: > On 2020-Mar-30, Tom Lane wrote: >> It looks to me like there's a similar hazard a bit further down >> (line 3029): >> >> appendConnStrVal(&connstr, PQdb(o_conn)); >> >> I wonder if we should force reuse_previous to false if there's >> no o_conn, rather than fixing this stuff piecemeal. > > That was my impression too.
Good point. What do you think about the attached then? -- Michael