Test: stabilize 037.failover_session.

On some platform (in my case Rocky Linux 10 running on VitualBox)
fails to finish shutdownall after test1.  This could be caused by the
failover in the test to fail to restore the signal handler which
should accept shutdown signal from shutdownall. Adding "sleep 5"
before shutdownall seems to mitigate the problem.

Author: Tatsuo Ishii <ishii@postgresql.org>
Backpatch-through: v4.5

diff --git a/src/test/regression/tests/037.failover_session/test.sh b/src/test/regression/tests/037.failover_session/test.sh
index b75377140692355439fd26305a30318e50bbe825..7fc9845d8450c9461b3b0240862a35adfba7ff6c 100755 (executable)
--- a/src/test/regression/tests/037.failover_session/test.sh
+++ b/src/test/regression/tests/037.failover_session/test.sh
@@ -60,6 +60,8 @@ else
     echo "pgbench suceeded. test1 ok."
     r1=ok
 fi
+# give pgpool chance to complete the failover
+sleep 5
 ./shutdownall
 
 echo "=== test2: backend_weight2 = 0 and pgbench with -C option"