diff --git a/Doc/library/itertools.rst b/Doc/library/itertools.rst index 17303dd01cef06..18477f1299031c 100644 --- a/Doc/library/itertools.rst +++ b/Doc/library/itertools.rst @@ -659,6 +659,10 @@ loops that truncate the stream. used anywhere else; otherwise, the *iterable* could get advanced without the tee objects being informed. + ``tee`` iterators are not threadsafe. A :exc:`RuntimeError` may be + raised when using simultaneously iterators returned by the same :func:`tee` + call, even if the original *iterable* is threadsafe. + This itertool may require significant auxiliary storage (depending on how much temporary data needs to be stored). In general, if one iterator uses most or all of the data before another iterator starts, it is faster to use diff --git a/Lib/test/test_itertools.py b/Lib/test/test_itertools.py index 04279792065edf..2cdcbb21efcdf6 100644 --- a/Lib/test/test_itertools.py +++ b/Lib/test/test_itertools.py @@ -10,6 +10,10 @@ import copy import pickle from functools import reduce +try: + import threading +except ImportError: + threading = None maxsize = test_support.MAX_Py_ssize_t minsize = -maxsize-1 @@ -984,6 +988,43 @@ def test_tee_del_backward(self): del forward, backward raise + def test_tee_reenter(self): + class I: + first = True + def __iter__(self): + return self + def next(self): + first = self.first + self.first = False + if first: + return next(b) + + a, b = tee(I()) + with self.assertRaisesRegexp(RuntimeError, "tee"): + next(a) + + @unittest.skipUnless(threading, 'Threading required for this test.') + def test_tee_concurrent(self): + start = threading.Event() + finish = threading.Event() + class I: + def __iter__(self): + return self + def next(self): + start.set() + finish.wait() + + a, b = tee(I()) + thread = threading.Thread(target=next, args=[a]) + thread.start() + try: + start.wait() + with self.assertRaisesRegexp(RuntimeError, "tee"): + next(b) + finally: + finish.set() + thread.join() + def test_StopIteration(self): self.assertRaises(StopIteration, izip().next) diff --git a/Misc/NEWS.d/next/Library/2019-08-31-01-52-59.bpo-34410.7KbWZQ.rst b/Misc/NEWS.d/next/Library/2019-08-31-01-52-59.bpo-34410.7KbWZQ.rst new file mode 100644 index 00000000000000..64e778ee0913c2 --- /dev/null +++ b/Misc/NEWS.d/next/Library/2019-08-31-01-52-59.bpo-34410.7KbWZQ.rst @@ -0,0 +1,2 @@ +Fixed a crash in the :func:`tee` iterator when re-enter it. RuntimeError is +now raised in this case. diff --git a/Modules/itertoolsmodule.c b/Modules/itertoolsmodule.c index 04076fd087099c..edd21be337cf51 100644 --- a/Modules/itertoolsmodule.c +++ b/Modules/itertoolsmodule.c @@ -314,6 +314,7 @@ typedef struct { PyObject_HEAD PyObject *it; int numread; + int running; PyObject *nextlink; PyObject *(values[LINKCELLS]); } teedataobject; @@ -336,6 +337,7 @@ teedataobject_new(PyObject *it) if (tdo == NULL) return NULL; + tdo->running = 0; tdo->numread = 0; tdo->nextlink = NULL; Py_INCREF(it); @@ -364,7 +366,14 @@ teedataobject_getitem(teedataobject *tdo, int i) else { /* this is the lead iterator, so fetch more data */ assert(i == tdo->numread); + if (tdo->running) { + PyErr_SetString(PyExc_RuntimeError, + "cannot re-enter the tee iterator"); + return NULL; + } + tdo->running = 1; value = PyIter_Next(tdo->it); + tdo->running = 0; if (value == NULL) return NULL; tdo->numread++;