From 6329607f518392ab8695d33fde7d5fba4721f4bb Mon Sep 17 00:00:00 2001
From: Zackery Spytz <zspytz@gmail.com>
Date: Fri, 14 Jun 2019 08:30:52 -0600
Subject: [PATCH 1/2] bpo-19865: ctypes.create_unicode_buffer() fails on
 non-BMP strings on Windows

---
 Lib/ctypes/__init__.py                                    | 4 ++++
 Lib/ctypes/test/test_buffers.py                           | 8 ++++++++
 .../next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst | 2 ++
 3 files changed, 14 insertions(+)
 create mode 100644 Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst

diff --git a/Lib/ctypes/__init__.py b/Lib/ctypes/__init__.py
index 4107db3e3972d79..43be113029622d4 100644
--- a/Lib/ctypes/__init__.py
+++ b/Lib/ctypes/__init__.py
@@ -275,6 +275,10 @@ def create_unicode_buffer(init, size=None):
     if isinstance(init, str):
         if size is None:
             size = len(init)+1
+            if sizeof(c_wchar) == 2:
+                for c in init:
+                    if ord(c) > 0xFFFF:
+                        size += 1
         buftype = c_wchar * size
         buf = buftype()
         buf.value = init
diff --git a/Lib/ctypes/test/test_buffers.py b/Lib/ctypes/test/test_buffers.py
index 166faaf4e4b89c9..a6f184e048481c9 100644
--- a/Lib/ctypes/test/test_buffers.py
+++ b/Lib/ctypes/test/test_buffers.py
@@ -60,5 +60,13 @@ def test_unicode_conversion(self):
         self.assertEqual(b[::2], "ac")
         self.assertEqual(b[::5], "a")
 
+    @need_symbol('c_wchar')
+    def test_create_unicode_buffer_non_bmp(self):
+        b = create_unicode_buffer('\U00010000\U00100000')
+        expected = 5 if sizeof(c_wchar) == 2 else 3
+        self.assertEqual(len(b), expected)
+        self.assertEqual(b[-1], '\0')
+
+
 if __name__ == "__main__":
     unittest.main()
diff --git a/Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst b/Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst
new file mode 100644
index 000000000000000..b7f58f0d1800b39
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst
@@ -0,0 +1,2 @@
+:func:`ctypes.create_unicode_buffer()` no longer fails on non-BMP strings on
+Windows.

From 7818fb5850b9892e679a07ce7e65d3a4fa0a18c1 Mon Sep 17 00:00:00 2001
From: Zackery Spytz <zspytz@gmail.com>
Date: Fri, 14 Jun 2019 09:07:06 -0600
Subject: [PATCH 2/2] Make the requested changes.

---
 Lib/ctypes/__init__.py                               | 12 ++++++++----
 Lib/ctypes/test/test_buffers.py                      |  7 ++++---
 .../Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst |  4 ++--
 3 files changed, 14 insertions(+), 9 deletions(-)

diff --git a/Lib/ctypes/__init__.py b/Lib/ctypes/__init__.py
index 43be113029622d4..128155dbf4f2d9e 100644
--- a/Lib/ctypes/__init__.py
+++ b/Lib/ctypes/__init__.py
@@ -274,11 +274,15 @@ def create_unicode_buffer(init, size=None):
     """
     if isinstance(init, str):
         if size is None:
-            size = len(init)+1
             if sizeof(c_wchar) == 2:
-                for c in init:
-                    if ord(c) > 0xFFFF:
-                        size += 1
+                # UTF-16 requires a surrogate pair (2 wchar_t) for non-BMP
+                # characters (outside [U+0000; U+FFFF] range). +1 for trailing
+                # NUL character.
+                size = sum(2 if ord(c) > 0xFFFF else 1 for c in init) + 1
+            else:
+                # 32-bit wchar_t (1 wchar_t per Unicode character). +1 for
+                # trailing NUL character.
+                size = len(init) + 1
         buftype = c_wchar * size
         buf = buftype()
         buf.value = init
diff --git a/Lib/ctypes/test/test_buffers.py b/Lib/ctypes/test/test_buffers.py
index a6f184e048481c9..15782be757c8535 100644
--- a/Lib/ctypes/test/test_buffers.py
+++ b/Lib/ctypes/test/test_buffers.py
@@ -62,10 +62,11 @@ def test_unicode_conversion(self):
 
     @need_symbol('c_wchar')
     def test_create_unicode_buffer_non_bmp(self):
-        b = create_unicode_buffer('\U00010000\U00100000')
         expected = 5 if sizeof(c_wchar) == 2 else 3
-        self.assertEqual(len(b), expected)
-        self.assertEqual(b[-1], '\0')
+        for s in '\U00010000\U00100000', '\U00010000\U0010ffff':
+            b = create_unicode_buffer(s)
+            self.assertEqual(len(b), expected)
+            self.assertEqual(b[-1], '\0')
 
 
 if __name__ == "__main__":
diff --git a/Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst b/Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst
index b7f58f0d1800b39..efd1f55c0135b0b 100644
--- a/Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst
+++ b/Misc/NEWS.d/next/Library/2019-06-14-08-30-16.bpo-19865.FRGH4I.rst
@@ -1,2 +1,2 @@
-:func:`ctypes.create_unicode_buffer()` no longer fails on non-BMP strings on
-Windows.
+:func:`ctypes.create_unicode_buffer()` now also supports non-BMP characters
+on platforms with 16-bit :c:type:`wchar_t` (for example, Windows and AIX).