Attempt to make auto at least somewhat more reliable. Recommend that device be explicitly set if possible.

README.md

@@ -460,13 +460,15 @@ setting <action> [parameters]
 - `speech#voice=voice_name` - Voice selection (e.g., "en-us+f3")
 - `speech#module=module_name` - TTS module (e.g., "espeak-ng")
 - `speech#driver=driver_name` - Speech driver (speechdDriver/genericDriver/dectalkDriver/litetalkDriver/doubletalkDriver/tripletalkDriver)
-- `speech#hardware_device=auto` - Hardware synth serial device for dectalkDriver/litetalkDriver
+- `speech#hardware_device=/dev/ttyS0` - Hardware synth serial device for dectalkDriver/litetalkDriver
 - `speech#hardware_baud_rate=9600` - Hardware synth serial baud rate
 - `speech#history_size=50` - Number of spoken items kept in runtime speech history
 
 USB hardware synths are supported only when Linux exposes them as a serial tty
 such as `/dev/ttyACM0` or `/dev/ttyUSB0`. A USB-only TripleTalk with no tty
-device would require a separate USB protocol driver.
+device would require a separate USB protocol driver. Use an explicit
+`speech#hardware_device` path for stable hardware speech; `auto` is limited to
+devices that respond to a driver-specific probe.
 - `speech#auto_read_incoming=True/False` - Auto-read new text
 
 *Sound Settings:*

config/settings/settings.conf

@@ -79,12 +79,13 @@ volume=1.0
 # Used by dectalkDriver, litetalkDriver, doubletalkDriver, and tripletalkDriver.
 # USB serial devices are supported if Linux exposes them as /dev/ttyACM*
 # or /dev/ttyUSB*. USB-only synths with no tty device need a separate driver.
-# auto checks /dev/ttyACM* first, then /dev/ttyUSB*.
+# Set an explicit device for stable hardware speech. auto only uses devices
+# that respond to a driver-specific probe and may fail on silent synths.
 # Examples:
 # hardware_device=/dev/ttyACM0  # RPITalk USB gadget mode
 # hardware_device=/dev/ttyUSB0  # USB serial adapter
 # hardware_device=/dev/ttyS0    # built-in serial port
-hardware_device=auto
+hardware_device=/dev/ttyS0
 
 # Serial baud rate for hardware speech synthesizers.
 hardware_baud_rate=9600

docs/fenrir.adoc

@@ -1684,15 +1684,15 @@ the pico module:
 language=de-DE
 ....
 
-Hardware speech drivers use a serial device. The default `+auto+` checks
-`+/dev/ttyACM*+` first, then `+/dev/ttyUSB*+`. Set an explicit path for
-stable systems.
+Hardware speech drivers use a serial device. Set an explicit path for stable
+systems. `+auto+` is limited to devices that respond to a driver-specific
+probe and may fail on silent synths.
 
 ....
-hardware_device=auto
 hardware_device=/dev/ttyACM0
 hardware_device=/dev/ttyUSB0
 hardware_device=/dev/ttyS0
+hardware_device=auto
 ....
 
 Hardware speech drivers use 9600 baud by default.

docs/user.md

@@ -101,7 +101,7 @@ driver=speechdDriver
 rate=0.5
 pitch=0.5
 volume=1.0
-hardware_device=auto
+hardware_device=/dev/ttyS0
 hardware_baud_rate=9600
 history_size=50
 
@@ -341,10 +341,11 @@ Fenrir automatically detects and provides audio feedback for progress indicators
 - **doubletalkDriver** - Serial DoubleTalk LT-compatible hardware speech
 - **tripletalkDriver** - Serial TripleTalk-compatible hardware speech
 
-For hardware speech, set `speech#hardware_device` to `auto` or an explicit
-serial path. RPITalk gadget mode usually appears as `/dev/ttyACM0`; USB serial
-adapters usually appear as `/dev/ttyUSB0`; built-in serial ports may be
-`/dev/ttyS0`. The default baud rate is `9600`. `doubletalkDriver` targets
+For hardware speech, set `speech#hardware_device` to an explicit serial path.
+RPITalk gadget mode usually appears as `/dev/ttyACM0`; USB serial adapters
+usually appear as `/dev/ttyUSB0`; built-in serial ports may be `/dev/ttyS0`.
+The default baud rate is `9600`. `auto` is limited to devices that respond to a
+driver-specific probe and may fail on silent synths. `doubletalkDriver` targets
 DoubleTalk LT-style serial devices, not the internal DoubleTalk PC ISA card.
 USB TripleTalk devices work only if Linux exposes them as a serial tty such as
 `/dev/ttyACM0` or `/dev/ttyUSB0`; USB-only models with no tty device need a

docs/user.txt

@@ -927,11 +927,11 @@ Select the language you want Fenrir to use.
   language=english-us
 Values: Text, see your TTS synths documentation what is available.
 
-Hardware speech drivers use a serial device. The default ''auto'' checks /dev/ttyACM* first, then /dev/ttyUSB*. Set an explicit path for stable systems.
-  hardware_device=auto
+Hardware speech drivers use a serial device. Set an explicit path for stable systems. ''auto'' is limited to devices that respond to a driver-specific probe and may fail on silent synths.
   hardware_device=/dev/ttyACM0
   hardware_device=/dev/ttyUSB0
   hardware_device=/dev/ttyS0
+  hardware_device=auto
 
 Hardware speech drivers use 9600 baud by default.
   hardware_baud_rate=9600

src/fenrirscreenreader/speechDriver/hardwareSerialDriver.py

@@ -166,8 +166,6 @@ class hardware_serial_driver(speech_driver):
             )
             return
 
-        fallback_device = None
-        fallback_port = None
         for device in devices:
             port = self._open_configured_serial_port(device)
             if port is None:
@@ -178,28 +176,20 @@ class hardware_serial_driver(speech_driver):
                 )
                 return
             if self._probe_serial_port(port):
-                if fallback_port is not None:
-                    self._close_port(fallback_port)
                 self._activate_serial_port(device, port, probe_matched=True)
                 return
-            if fallback_port is None:
-                fallback_device = device
-                fallback_port = port
-            else:
-                self._close_port(port)
+            self._close_port(port)
 
-        if fallback_port is not None:
+        if auto_detect and self.hardware_probe_command:
             self._debug(
-                "Hardware speech probe did not identify a synth; "
-                f"falling back to {fallback_device}",
-                debug.DebugLevel.WARNING,
+                "Hardware speech auto did not identify a synth; set "
+                "speech#hardware_device to the known serial device",
+                debug.DebugLevel.ERROR,
                 on_any_level=True,
             )
-            self._activate_serial_port(
-                fallback_device, fallback_port, probe_matched=False
-            )
 
     def _open_configured_serial_port(self, device):
+        port = None
         try:
             port = os.open(device, os.O_RDWR | os.O_NOCTTY)
             tty.setraw(port)
@@ -214,6 +204,7 @@ class hardware_serial_driver(speech_driver):
             termios.tcsetattr(port, termios.TCSANOW, attrs)
             return port
         except (OSError, termios.error) as error:
+            self._close_port(port)
             self._debug(
                 f"Hardware speech device open failed: {device}: {error}",
                 debug.DebugLevel.ERROR,
@@ -331,7 +322,11 @@ class hardware_serial_driver(speech_driver):
             )
             return [device]
         devices = []
-        for pattern in ("/dev/ttyACM*", "/dev/ttyUSB*", "/dev/ttyS*"):
+        for pattern in (
+            "/dev/serial/by-id/*",
+            "/dev/ttyACM*",
+            "/dev/ttyUSB*",
+        ):
             matches = sorted(glob.glob(pattern))
             self._debug(
                 f"Hardware speech auto scan {pattern}: {matches}",

tests/unit/test_hardware_speech_drivers.py

@@ -105,7 +105,7 @@ def test_litetalk_driver_writes_settings_and_cancel(serial_pair):
         speech_driver.shutdown()
 
 
-def test_auto_device_detection_includes_classic_serial(
+def test_configured_device_supports_classic_serial(
     monkeypatch, serial_pair
 ):
     master_fd, slave_name = serial_pair
@@ -120,11 +120,11 @@ def test_auto_device_detection_includes_classic_serial(
         fake_glob,
     )
     speech_driver = litetalkDriver.driver()
-    speech_driver.initialize(build_environment("auto"))
+    speech_driver.initialize(build_environment(slave_name))
     try:
         assert speech_driver.device == slave_name
         speech_driver.speak("Serial")
-        assert read_available(master_fd, 9) == b"\x01ISerial\r"
+        assert read_available(master_fd, 7) == b"Serial\r"
     finally:
         speech_driver.shutdown()
 
@@ -219,7 +219,7 @@ def test_auto_device_detection_prefers_probe_response(monkeypatch):
         speech_driver.shutdown()
 
 
-def test_auto_device_detection_falls_back_without_probe_response(
+def test_auto_device_detection_fails_without_probe_response(
     monkeypatch
 ):
     opened_ports = []
@@ -265,13 +265,12 @@ def test_auto_device_detection_falls_back_without_probe_response(
     )
 
     speech_driver = litetalkDriver.driver()
-    speech_driver.initialize(build_environment("auto"))
-    try:
-        assert opened_ports == ["/dev/ttyUSB0", "/dev/ttyUSB1"]
-        assert closed_ports == [2]
-        assert speech_driver.device == "/dev/ttyUSB0"
-    finally:
-        speech_driver.shutdown()
+    with pytest.raises(RuntimeError, match="hardware speech device"):
+        speech_driver.initialize(build_environment("auto"))
+
+    assert opened_ports == ["/dev/ttyUSB0", "/dev/ttyUSB1"]
+    assert closed_ports == [1, 2]
+    assert speech_driver.device == "auto"
 
 
 def test_auto_device_detection_skips_termios_failures(monkeypatch):
@@ -280,10 +279,8 @@ def test_auto_device_detection_skips_termios_failures(monkeypatch):
 
     monkeypatch.setattr(
         "fenrirscreenreader.speechDriver.hardwareSerialDriver.glob.glob",
-        lambda pattern: ["/dev/ttyUSB0"]
+        lambda pattern: ["/dev/ttyUSB0", "/dev/ttyUSB1"]
         if pattern == "/dev/ttyUSB*"
-        else ["/dev/ttyS0"]
-        if pattern == "/dev/ttyS*"
         else [],
     )
 
@@ -293,7 +290,7 @@ def test_auto_device_detection_skips_termios_failures(monkeypatch):
         return port
 
     def fake_tcgetattr(port):
-        if port == 101:
+        if port == 100:
             raise termios.error(5, "Input/output error")
         return [0, 0, 0, 0, 0, 0, [0] * 32]
 
@@ -331,12 +328,12 @@ def test_auto_device_detection_skips_termios_failures(monkeypatch):
     )
 
     speech_driver = litetalkDriver.driver()
-    speech_driver.initialize(build_environment("auto"))
-    try:
-        assert opened_ports == [("/dev/ttyUSB0", 100), ("/dev/ttyS0", 101)]
-        assert speech_driver.device == "/dev/ttyUSB0"
-    finally:
-        speech_driver.shutdown()
+    with pytest.raises(RuntimeError, match="hardware speech device"):
+        speech_driver.initialize(build_environment("auto"))
+
+    assert opened_ports == [("/dev/ttyUSB0", 100), ("/dev/ttyUSB1", 101)]
+    assert closed_ports == [100, 101]
+    assert speech_driver.device == "auto"
 
 
 def test_auto_device_detection_fails_when_no_serial_device(monkeypatch):