Fix unit test, the code now holds more references

It cannot re-use the incoming tuple as before, because it is no longer a tuple at all.  So a new tuple must be created, which then holds references for each member.

Author: b-pass

include/pybind11/pybind11.h

@@ -1301,7 +1301,7 @@ class cpp_function : public function {
          * CPython itself implements the search this way, first comparing all pointers ... which is
          * cheap and will work if the strings are interned.  If it fails, then it falls back to a
          * second lexicographic check. This is wildly expensive for huge argument lists, but those
-         * are incredably rare so we optimize for the vastly common case of just a couple of args.
+         * are incredibly rare so we optimize for the vastly common case of just a couple of args.
          */
         auto n = PyTuple_GET_SIZE(haystack);
         auto s = reinterpret_steal<pybind11::str>(PyUnicode_InternFromString(needle));

tests/test_kwargs_and_defaults.py

@@ -458,7 +458,8 @@ def test_args_refcount():
     assert refcount(myval) == expected
 
     exp3 = refcount(myval, myval, myval)
-    assert m.args_refcount(myval, myval, myval) == (exp3, exp3, exp3)
+    # if we have to create a new tuple internally, then it will hold an extra reference for each item in it.
+    assert m.args_refcount(myval, myval, myval) == (exp3 + 3, exp3 + 3, exp3 + 3)
     assert refcount(myval) == expected
 
     # This function takes the first arg as a `py::object` and the rest as a `py::args`.  Unlike the