#5057: the peepholer no longer optimizes subscription on unicode literals (e.g. u"foo"[0]) in order to produce compatible pyc files between narrow and wide builds.
diff --git a/Lib/test/test_peepholer.py b/Lib/test/test_peepholer.py
index 3e8b7ae..7e05f49 100644
--- a/Lib/test/test_peepholer.py
+++ b/Lib/test/test_peepholer.py
@@ -138,21 +138,22 @@
self.assertIn('(1000)', asm)
def test_binary_subscr_on_unicode(self):
- # valid code get optimized
+ # unicode strings don't get optimized
asm = dis_single('u"foo"[0]')
- self.assertIn("(u'f')", asm)
- self.assertNotIn('BINARY_SUBSCR', asm)
+ self.assertNotIn("(u'f')", asm)
+ self.assertIn('BINARY_SUBSCR', asm)
asm = dis_single('u"\u0061\uffff"[1]')
- self.assertIn("(u'\\uffff')", asm)
- self.assertNotIn('BINARY_SUBSCR', asm)
+ self.assertNotIn("(u'\\uffff')", asm)
+ self.assertIn('BINARY_SUBSCR', asm)
- # invalid code doesn't get optimized
# out of range
asm = dis_single('u"fuu"[10]')
self.assertIn('BINARY_SUBSCR', asm)
# non-BMP char (see #5057)
asm = dis_single('u"\U00012345"[0]')
self.assertIn('BINARY_SUBSCR', asm)
+ asm = dis_single('u"\U00012345abcdef"[3]')
+ self.assertIn('BINARY_SUBSCR', asm)
def test_folding_of_unaryops_on_constants(self):
diff --git a/Misc/NEWS b/Misc/NEWS
index 9716c6d..bd73d0a 100644
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -9,6 +9,10 @@
Core and Builtins
-----------------
+- Issue #5057: the peepholer no longer optimize subscription on unicode
+ literals (e.g. u'foo'[0]) in order to produce compatible pyc files between
+ narrow and wide builds.
+
- Issue #8401: assigning an int to a bytearray slice (e.g. b[3:4] = 5) now
raises an error.
diff --git a/Python/peephole.c b/Python/peephole.c
index ae84efa..fb6cd03 100644
--- a/Python/peephole.c
+++ b/Python/peephole.c
@@ -128,27 +128,14 @@
newconst = PyNumber_Subtract(v, w);
break;
case BINARY_SUBSCR:
- newconst = PyObject_GetItem(v, w);
/* #5057: if v is unicode, there might be differences between
- wide and narrow builds in cases like u'\U00012345'[0].
- Wide builds will return a non-BMP char, whereas narrow builds
- will return a surrogate. In both the cases skip the
- optimization in order to produce compatible pycs.
- */
-#ifdef Py_USING_UNICODE
- if (newconst != NULL &&
- PyUnicode_Check(v) && PyUnicode_Check(newconst)) {
- Py_UNICODE ch = PyUnicode_AS_UNICODE(newconst)[0];
-#ifdef Py_UNICODE_WIDE
- if (ch > 0xFFFF) {
-#else
- if (ch >= 0xD800 && ch <= 0xDFFF) {
-#endif
- Py_DECREF(newconst);
- return 0;
- }
- }
-#endif
+ wide and narrow builds in cases like '\U00012345'[0] or
+ '\U00012345abcdef'[3], so it's better to skip the optimization
+ in order to produce compatible pycs.
+ */
+ if (PyUnicode_Check(v))
+ return 0;
+ newconst = PyObject_GetItem(v, w);
break;
case BINARY_LSHIFT:
newconst = PyNumber_Lshift(v, w);