Hi Christoph, I am still looking into this one. Unfortunately outside of sbuild, I am not able to reproduce this issue. When I ran `./regress.sh 005.jdbc` the test is passing. But when run in sbuild, I am getting the error. I am not sure if I missed any flags or configuration. But due to this I am not able to run a debugger. By putting some debug prints inside the source code and rerunning sbuild, what I was able to observe was that inside `src/auth/pool_auth.c`, inside the function `do_SCRAM` we are getting a `AUTH_REQ_SASL` then followed by a `AUTH_REQ_SASL_CONT` for `auth_kind`. But when `auth_kind` takes the next value from `backend` using `pool_read` it is getting the value `1397113172` instead of `AUTH_REQ_SASL_FIN` (whose value is 12). At this stage we were getting the value `E` for `kind` where the expected value was `R`. I was planning to put some more debug prints inside `pg_SASL_continue` and `pg_fe_scram_exchange` and check for more data points. Please mention if you have any ideas regarding how to reproduce this in upstream source without any of the Debian helpers so that we can run a debugger on the issue (which hopefully would speed up things).
Also, this issue occurs even before the JDBC tests run, that is, when psql connects to the `pgpooltest.port` (https://salsa.debian.org/postgresql/pgpool2/-/blob/master/debian/tests/jdbc-tests?ref_type=heads#L66) Thanks, Pranav
