bpo-26163: Frozenset hash improvement (GH-5194) (#5198)

(cherry picked from commit b44c5169f64178d2ff2914187b315549e7ab0cb6)

diff --git a/Misc/NEWS.d/next/Core and Builtins/2018-01-14-20-32-47.bpo-26163.xv9Iuv.rst b/Misc/NEWS.d/next/Core and Builtins/2018-01-14-20-32-47.bpo-26163.xv9Iuv.rst
new file mode 100644 (file)
index 0000000..163b9e0
--- /dev/null
+++ b/Misc/NEWS.d/next/Core and Builtins/2018-01-14-20-32-47.bpo-26163.xv9Iuv.rst
@@ -0,0 +1,2 @@
+Improved frozenset() hash to create more distinct hash values when faced
+with datasets containing many similar values.

diff --git a/Objects/setobject.c b/Objects/setobject.c
index 24272b4d14a505c1dd839be2108d36b209903964..093a15f367f0290c6c7bbe312e858f44859927a7 100644 (file)
--- a/Objects/setobject.c
+++ b/Objects/setobject.c
@@ -790,6 +790,7 @@ frozenset_hash(PyObject *self)
     hash ^= ((Py_uhash_t)PySet_GET_SIZE(self) + 1) * 1927868237UL;
 
     /* Disperse patterns arising in nested frozensets */
+    hash ^= (hash >> 11) ^ (~hash >> 25);
     hash = hash * 69069U + 907133923UL;
 
     /* -1 is reserved as an error code */