diff --git a/CHANGES b/CHANGES
index c4bf31738..d6a531462 100644
--- a/CHANGES
+++ b/CHANGES
@@ -39,6 +39,7 @@ Bugs fixed
 * #4973: latex: glossary directive adds whitespace to each item
 * #4980: latex: Explicit labels on code blocks are duplicated
 * #4919: node.asdom() crashes if toctree has :numbered: option
+* #4914: autodoc: Parsing error when using dataclasses without default values
 
 Testing
 --------
diff --git a/sphinx/pycode/parser.py b/sphinx/pycode/parser.py
index deba48b1c..5c4291d3d 100644
--- a/sphinx/pycode/parser.py
+++ b/sphinx/pycode/parser.py
@@ -224,12 +224,13 @@ class AfterCommentParser(TokenProcessor):
     def parse(self):
         # type: () -> None
         """Parse the code and obtain comment after assignment."""
-        # skip lvalue (until '=' operator)
-        while self.fetch_token() != [OP, '=']:
+        # skip lvalue (or whole of AnnAssign)
+        while not self.fetch_token().match([OP, '='], NEWLINE, COMMENT):
             assert self.current
 
-        # skip rvalue
-        self.fetch_rvalue()
+        # skip rvalue (if exists)
+        if self.current == [OP, '=']:
+            self.fetch_rvalue()
 
         if self.current == COMMENT:
             self.comment = self.current.value
diff --git a/tests/test_pycode_parser.py b/tests/test_pycode_parser.py
index 09f1f41f5..29363e17e 100644
--- a/tests/test_pycode_parser.py
+++ b/tests/test_pycode_parser.py
@@ -100,11 +100,13 @@ def test_comment_picker_location():
 def test_annotated_assignment_py36():
     source = ('a: str = "Sphinx"  #: comment\n'
               'b: int = 1\n'
-              '"""string on next line"""')
+              '"""string on next line"""\n'
+              'c: int  #: comment')
     parser = Parser(source)
     parser.parse()
     assert parser.comments == {('', 'a'): 'comment',
-                               ('', 'b'): 'string on next line'}
+                               ('', 'b'): 'string on next line',
+                               ('', 'c'): 'comment'}
     assert parser.definitions == {}