diff --git a/gdb/python/lib/gdb/dap/next.py b/gdb/python/lib/gdb/dap/next.py
index 1dc1d6dd74d..7e06b1b79dc 100644
--- a/gdb/python/lib/gdb/dap/next.py
+++ b/gdb/python/lib/gdb/dap/next.py
@@ -73,10 +73,10 @@ def step_in(
     exec_and_expect_stop(cmd)
 
 
-@request("stepOut", response=False)
+@request("stepOut", defer_stop_events=True)
 def step_out(*, threadId: int, singleThread: bool = False, **args):
     _handle_thread_step(threadId, singleThread, True)
-    exec_and_expect_stop("finish")
+    exec_and_expect_stop("finish &", propagate_exception=True)
 
 
 # This is a server-side request because it is funny: it wants to
diff --git a/gdb/testsuite/gdb.dap/step-out.exp b/gdb/testsuite/gdb.dap/step-out.exp
index 757f4ebdaca..31c50bb81ec 100644
--- a/gdb/testsuite/gdb.dap/step-out.exp
+++ b/gdb/testsuite/gdb.dap/step-out.exp
@@ -79,4 +79,15 @@ gdb_assert {[dict get $varlist variablesReference] > 0} \
 gdb_assert {[dict get $varlist name] == "(return)"} \
     "variable is return value"
 
+set response_and_events [dap_request_and_response stepOut {o threadId [i 1]}]
+set response [lindex $response_and_events 0]
+if {[dict get $response success] == "true"} {
+    fail "stepOut from outermost frame should not succeed"
+} else {
+    pass "stepOut from outermost frame failed like it should"
+}
+
+dap_check_request_and_response "still stopped and may request backtrace" \
+    stackTrace {o threadId [i 1]}
+
 dap_shutdown