test(inverter): cover FoxCloud in the idle-cycle button press regression

test_force_export_off_does_not_press_every_cycle named the FoxCloud/TESLA/
Enphase/Deye/Sunsynk/AlphaESS cloud types in its docstring as the affected
set but only parametrised GS_fb00 and GS, so the cloud behaviour rested on
GS_fb00 happening to share has_discharge_enable_time rather than on a test.

FoxCloud stands in for those types, and its expected press count is 0 rather
than 1: being HH:MM:SS format it never takes the H M unconditional register
rewrite, so the commit-once safety net gated on that format is skipped too
and an idle cloud inverter should be silent, not spend one press on a first
post-restart commit the way GS_fb00 does.

Also adds a positive control - entering an export window straight after the
idle cycles must still commit exactly once. Without it an inverter that had
stopped pressing altogether would satisfy the idle assertion just as well,
which is the failure mode that would actually matter to a Fox user.

inv.sleep is stubbed for the loop: TestHAInterface applies a service call
synchronously, so the write-and-poll read-back already sees the new value
and the sleep between them is pure wall clock. discharge_target_soc and the
idle time selects join saved_items, as the positive control writes them and
this test shares my_predbat/ha with the rest of the run.

Verified against two mutations of inverter.py, each reverted after: restoring
the pre-fix None comparison (times_supplied = True) makes FoxCloud press on
all 4 idle cycles, and disabling the press block trips the positive control.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: springfall2008

apps/predbat/tests/test_inverter.py

@@ -2588,6 +2588,13 @@ def test_force_export_off_does_not_press_every_cycle(test_name, ha, my_predbat):
     inverter still reports a real time - and None never compares equal, so every cycle looked like a
     change and pressed the button. That is most of the day, not just export windows.
 
+    FoxCloud stands in for the cloud types here. They reach the bug by the same route as GS_fb00 but
+    settle somewhere different: being HH:MM:SS format they never take the H M unconditional register
+    rewrite, so the commit-once safety net gated on that format is skipped too and an idle cloud
+    inverter should press zero times - not the one GS_fb00 spends on its first post-restart commit.
+    Pinning that separately keeps the cloud behaviour from resting on GS_fb00 happening to share
+    has_discharge_enable_time.
+
     Plain GS takes the midnight-override path and is not affected by this None-comparison route, so it
     is checked here too to pin the difference down. (GS was affected by a separate bug - #4711's
     unconditional H M register rewrite, which GS also uses - but that is a different code path to the
@@ -2601,13 +2608,24 @@ def test_force_export_off_does_not_press_every_cycle(test_name, ha, my_predbat):
     # later tests and making results order-dependent.
     unset = object()
     saved_args = {key: my_predbat.args.get(key, unset) for key in ("inverter_type", "discharge_start_time", "discharge_end_time", "scheduled_discharge_enable")}
-    saved_items = {key: ha.dummy_items.get(key, unset) for key in ("select.discharge_start_time", "select.discharge_end_time", "switch.scheduled_discharge_enable", "select.inverter_mode")}
+    saved_items = {
+        key: ha.dummy_items.get(key, unset)
+        for key in ("select.discharge_start_time", "select.discharge_end_time", "switch.scheduled_discharge_enable", "select.inverter_mode", "number.discharge_target_soc", "select.idle_start_time", "select.idle_end_time")
+    }
+
+    # Times for the positive control below - any window will do, it just has to differ from idle.
+    export_start = datetime.strptime("18:00:00", "%H:%M:%S")
+    export_end = datetime.strptime("19:00:00", "%H:%M:%S")
 
     try:
-        for inverter_type, expected_presses in (("GS_fb00", 1), ("GS", 1)):
+        for inverter_type, expected_presses in (("GS_fb00", 1), ("GS", 1), ("FoxCloud", 0)):
             my_predbat.args["inverter_type"] = [inverter_type]
             inv = Inverter(my_predbat, 0, quiet=True)
             inv.rest_data = None
+            # Only the press count matters here. TestHAInterface applies a service call synchronously,
+            # so the write-and-poll read-back already sees the new value and the sleep between them is
+            # pure wall clock - several seconds per register write, across three inverter types.
+            inv.sleep = lambda seconds: None
 
             ha.dummy_items["select.discharge_start_time"] = "00:00:00"
             ha.dummy_items["select.discharge_end_time"] = "00:00:00"
@@ -2628,6 +2646,15 @@ def test_force_export_off_does_not_press_every_cycle(test_name, ha, my_predbat):
             if len(presses) > expected_presses:
                 print(f"ERROR: {test_name}: {inverter_type} pressed the button {len(presses)} times over 4 idle cycles, expected at most {expected_presses}")
                 failed = True
+
+            # Positive control: the suppression has to be specific to idle cycles. An inverter that had
+            # stopped committing altogether would satisfy the check above just as well, so entering an
+            # export window straight afterwards must still commit, exactly once.
+            idle_presses = len(presses)
+            inv.adjust_force_export(True, export_start, export_end)
+            if len(presses) != idle_presses + 1:
+                print(f"ERROR: {test_name}: {inverter_type} should press once on entering an export window, pressed {len(presses) - idle_presses} times")
+                failed = True
     finally:
         for key, value in saved_args.items():
             if value is unset: