diff --git a/_pytest/compat.py b/_pytest/compat.py
index 45491a722c1..7bf3bb9b89f 100644
--- a/_pytest/compat.py
+++ b/_pytest/compat.py
@@ -149,7 +149,7 @@ def _bytes_to_ascii(val):
                 # empty bytes crashes codecs.escape_encode (#1087)
                 return ''
 
-    def _ascii_escaped(val):
+    def ascii_escaped(val):
         """If val is pure ascii, returns it as a str().  Otherwise, escapes
         bytes objects into a sequence of escaped bytes:
 
@@ -177,7 +177,7 @@ def _ascii_escaped(val):
 
     from itertools import imap, izip  # NOQA
 
-    def _ascii_escaped(val):
+    def ascii_escaped(val):
         """In py2 bytes and str are the same type, so return if it's a bytes
         object, return it unchanged if it is a full ascii string,
         otherwise escape it into its binary form.
diff --git a/_pytest/python.py b/_pytest/python.py
index e3d79b2e975..0161c10e8ec 100644
--- a/_pytest/python.py
+++ b/_pytest/python.py
@@ -19,7 +19,7 @@
 from _pytest import fixtures
 from _pytest import main
 from _pytest.compat import (
-    isclass, isfunction, is_generator, _ascii_escaped,
+    isclass, isfunction, is_generator, ascii_escaped,
     REGEX_TYPE, STRING_TYPES, NoneType, NOTSET,
     get_real_func, getfslineno, safe_getattr,
     safe_str, getlocation, enum,
@@ -922,7 +922,7 @@ def _idval(val, argname, idx, idfn, config=None):
             msg += '\nUpdate your code as this will raise an error in pytest-4.0.'
             warnings.warn(msg, DeprecationWarning)
         if s:
-            return _ascii_escaped(s)
+            return ascii_escaped(s)
 
     if config:
         hook_id = config.hook.pytest_make_parametrize_id(
@@ -931,11 +931,11 @@ def _idval(val, argname, idx, idfn, config=None):
             return hook_id
 
     if isinstance(val, STRING_TYPES):
-        return _ascii_escaped(val)
+        return ascii_escaped(val)
     elif isinstance(val, (float, int, bool, NoneType)):
         return str(val)
     elif isinstance(val, REGEX_TYPE):
-        return _ascii_escaped(val.pattern)
+        return ascii_escaped(val.pattern)
     elif enum is not None and isinstance(val, enum.Enum):
         return str(val)
     elif isclass(val) and hasattr(val, '__name__'):
@@ -951,7 +951,7 @@ def _idvalset(idx, parameterset, argnames, idfn, ids, config=None):
                    for val, argname in zip(parameterset.values, argnames)]
         return "-".join(this_id)
     else:
-        return _ascii_escaped(ids[idx])
+        return ascii_escaped(ids[idx])
 
 
 def idmaker(argnames, parametersets, idfn=None, ids=None, config=None):
diff --git a/_pytest/warnings.py b/_pytest/warnings.py
index 926b1f58112..847771daa25 100644
--- a/_pytest/warnings.py
+++ b/_pytest/warnings.py
@@ -72,8 +72,8 @@ def catch_warnings_for_item(item):
             unicode_warning = False
 
             if compat._PY2 and any(isinstance(m, compat.UNICODE_TYPES) for m in warn_msg.args):
-                new_args = [compat.safe_str(m) for m in warn_msg.args]
-                unicode_warning = warn_msg.args != new_args
+                new_args = [compat.ascii_escaped(m) for m in warn_msg.args]
+                unicode_warning = list(warn_msg.args) != new_args
                 warn_msg.args = new_args
 
             msg = warnings.formatwarning(
diff --git a/changelog/2809.bugfix b/changelog/2809.bugfix
new file mode 100644
index 00000000000..6db7e8c6c57
--- /dev/null
+++ b/changelog/2809.bugfix
@@ -0,0 +1 @@
+Pytest no longer complains about warnings with unicode messages being non-ascii compatible even for ascii-compatible messages. As a result of this, warnings with unicode messages are converted first to an ascii representation for safety.
\ No newline at end of file
diff --git a/testing/test_warnings.py b/testing/test_warnings.py
index 1328cc3f2b8..4beff45480c 100644
--- a/testing/test_warnings.py
+++ b/testing/test_warnings.py
@@ -163,13 +163,33 @@ def test_func(fix):
     result.stdout.fnmatch_lines([
         '*== %s ==*' % WARNINGS_SUMMARY_HEADER,
 
-        '*test_py2_unicode.py:8: UserWarning: \u6d4b\u8bd5',
+        '*test_py2_unicode.py:8: UserWarning: \\u6d4b\\u8bd5',
         '*warnings.warn(u"\u6d4b\u8bd5")',
         '*warnings.py:*: UnicodeWarning: Warning is using unicode non*',
         '* 1 passed, 2 warnings*',
     ])
 
 
+def test_py2_unicode_ascii(testdir):
+    """Ensure that our warning about 'unicode warnings containing non-ascii messages'
+    does not trigger with ascii-convertible messages"""
+    testdir.makeini('[pytest]')
+    testdir.makepyfile('''
+        import pytest
+        import warnings
+
+        @pytest.mark.filterwarnings('always')
+        def test_func():
+            warnings.warn(u"hello")
+    ''')
+    result = testdir.runpytest()
+    result.stdout.fnmatch_lines([
+        '*== %s ==*' % WARNINGS_SUMMARY_HEADER,
+        '*warnings.warn(u"hello")',
+        '* 1 passed, 1 warnings in*'
+    ])
+
+
 def test_works_with_filterwarnings(testdir):
     """Ensure our warnings capture does not mess with pre-installed filters (#2430)."""
     testdir.makepyfile('''