Don't raise exceptions on bad unicode decode.

src/genpy/generator.py

@@ -497,7 +497,7 @@ def string_serializer_generator(package, type_, name, serialize):  # noqa: D401
                 yield '%s = str[start:end]' % (var)
             else:
                 yield 'if python3:'
-                yield INDENT+"%s = str[start:end].decode('utf-8')" % (var)  # If messages are python3-decode back to unicode
+                yield INDENT+"%s = str[start:end].decode('utf-8', 'rosmsg')" % (var)  # If messages are python3-decode back to unicode
                 yield 'else:'
                 yield INDENT+'%s = str[start:end]' % (var)
 
@@ -756,6 +756,7 @@ def deserialize_fn_generator(msg_context, spec, is_numpy=False):  # noqa: D401
     :param is_numpy: if True, generate serializer code for numpy
       datatypes instead of Python lists, ``bool``
     """
+    yield 'codecs.lookup_error("rosmsg").msg_type = self._type'
     yield 'try:'
     package = spec.package
     # Instantiate embedded type classes
@@ -816,6 +817,7 @@ def msg_generator(msg_context, spec, search_path):
 
     yield '# This Python file uses the following encoding: utf-8'
     yield '"""autogenerated by genpy from %s.msg. Do not edit."""' % spec.full_name
+    yield 'import codecs'
     yield 'import sys'
     yield 'python3 = True if sys.hexversion > 0x03000000 else False'
     yield 'import genpy\nimport struct\n'

src/genpy/message.py

@@ -37,6 +37,7 @@
 libraries for type checking and retrieving message classes by type name.
 """
 
+import codecs
 import itertools
 import math
 import struct
@@ -65,6 +66,26 @@
 
 struct_I = struct.Struct('<I')
 
+_warned_decoding_error = set()
+
+# Notify the user while not crashing in the face of errors attempting
+# to decode non-unicode data within a ROS message.
+class RosMsgUnicodeErrors:
+    def __init__(self):
+        self.msg_type = None
+
+    def __call__(self, err):
+        global _warned_decoding_error
+        if self.msg_type not in _warned_decoding_error:
+            _warned_decoding_error.add(self.msg_type)
+            # Lazy import to avoid this cost in the non-error case.
+            import logging
+            logger = logging.getLogger('rosout')
+            extra = "message %s" % self.msg_type if self.msg_type else "unknown message"
+            logger.error("Characters replaced when decoding %s (will print only once): %s", extra, err)
+        return codecs.replace_errors(err)
+codecs.register_error('rosmsg', RosMsgUnicodeErrors())
+
 
 def isstring(s):
     """Small helper version to check an object is a string in a way that works for both Python 2 and 3."""

test/files/array/string_fixed_deser.txt

@@ -6,7 +6,7 @@ for i in range(0, 2):
   start = end
   end += length
   if python3:
-    val0 = str[start:end].decode('utf-8')
+    val0 = str[start:end].decode('utf-8', 'rosmsg')
   else:
     val0 = str[start:end]
   data.append(val0)

test/files/array/string_varlen_deser.txt

@@ -9,7 +9,7 @@ for i in range(0, length):
   start = end
   end += length
   if python3:
-    val0 = str[start:end].decode('utf-8')
+    val0 = str[start:end].decode('utf-8', 'rosmsg')
   else:
     val0 = str[start:end]
   data.append(val0)

test/test_genpy_generator.py

@@ -331,7 +331,7 @@ def test_string_serializer_generator():
 start = end
 end += length
 if python3:
-  var_name = str[start:end].decode('utf-8')
+  var_name = str[start:end].decode('utf-8', 'rosmsg')
 else:
   var_name = str[start:end]"""
     # string serializer and array serializer are identical

test/test_genpy_message.py

@@ -738,3 +738,27 @@ def test_serialize_exception_msg(self):
             self.assertEqual(str(e), "<class 'struct.error'>: 'required argument is not a float' when writing '1.0'")
         except Exception:
             assert False, 'This should have raised a genpy.SerializationError instead'
+
+    @unittest.skipIf(sys.hexversion < 0x03000000, "Python 3 only test")
+    def test_deserialize_unicode_error(self):
+        from genpy.msg import TestString, TestMsgArray
+
+        m = TestString()
+        buff = b'\x00\x00\x00\x04\xF0\x9F\x92\xA9'
+        self.assertEqual(m.deserialize(buff).data, b'\xF0\x9F\x92\xA9'.decode())
+
+        m = TestString()
+        buff = b'\x00\x00\x00\x04\x41\xff\xfe\x42'
+        with self.assertLogs('rosout', level='ERROR') as cm:
+            self.assertEqual(m.deserialize(buff).data, 'A\ufffd\ufffdB')
+            self.assertEqual(m.deserialize(buff).data, 'A\ufffd\ufffdB')
+            self.assertEqual(len(cm.output), 1)
+            self.assertIn("Characters replaced when decoding message genpy/TestString (will print only once)", cm.output[0])
+
+        m = TestMsgArray()
+        buff = b'\x00\x00\x00\x00\x00\x00\x00\x04\x41\xff\xfe\x42'
+        with self.assertLogs('rosout', level='ERROR') as cm:
+            self.assertEqual(m.deserialize(buff).fixed_strings[0].data, 'A\ufffd\ufffdB')
+            self.assertEqual(m.deserialize(buff).fixed_strings[0].data, 'A\ufffd\ufffdB')
+            self.assertEqual(len(cm.output), 1)
+            self.assertIn("Characters replaced when decoding message genpy/TestMsgArray (will print only once)", cm.output[0])