Fix using FQCN for flush_handlers (#79057)

Fixes #79023

changelogs/fragments/79023-fix-flush_handlers-fqcn.yml

@@ -0,0 +1,2 @@
+bugfixes:
+  - "handlers - fix an issue where the ``flush_handlers`` meta task could not be used with FQCN: ``ansible.builtin.meta`` (https://github.com/ansible/ansible/issues/79023)"

lib/ansible/plugins/strategy/linear.py

@@ -118,7 +118,7 @@ class StrategyModule(StrategyBase):
         # once hosts synchronize on 'flush_handlers' lockstep enters
         # '_in_handlers' phase where handlers are run instead of tasks
         # until at least one host is in IteratingStates.HANDLERS
-        if (not self._in_handlers and cur_task.action == 'meta' and
+        if (not self._in_handlers and cur_task.action in C._ACTION_META and
                 cur_task.args.get('_raw_params') == 'flush_handlers'):
             self._in_handlers = True
 

test/integration/targets/handlers/runme.sh

@@ -166,3 +166,7 @@ ansible-playbook test_flush_handlers_rescue_always.yml -i inventory.handlers "$@
 [ "$(grep out.txt -ce 'rescue ran')" = "1" ]
 [ "$(grep out.txt -ce 'always ran')" = "2" ]
 [ "$(grep out.txt -ce 'should run for both hosts')" = "2" ]
+
+ansible-playbook test_fqcn_meta_flush_handlers.yml -i inventory.handlers "$@" 2>&1 | tee out.txt
+grep out.txt -e "handler ran"
+grep out.txt -e "after flush"

test/integration/targets/handlers/test_fqcn_meta_flush_handlers.yml

@@ -0,0 +1,14 @@
+- hosts: A
+  gather_facts: false
+  tasks:
+    - command: echo
+      notify: handler
+
+    - ansible.builtin.meta: flush_handlers
+
+    - debug:
+        msg: after flush
+  handlers:
+    - name: handler
+      debug:
+        msg: handler ran