Issue #2801: fix bug in float.is_integer where ValueError
could be incorrectly raised. This is a backport of the
Py3k fix in r62939. (Should really have been fixed
in the trunk first and svnmerged into py3k.)
diff --git a/Objects/floatobject.c b/Objects/floatobject.c
index f79995c..ceb0b6d 100644
--- a/Objects/floatobject.c
+++ b/Objects/floatobject.c
@@ -1031,6 +1031,7 @@
return NULL;
if (!Py_IS_FINITE(x))
Py_RETURN_FALSE;
+ errno = 0;
PyFPE_START_PROTECT("is_integer", return NULL)
o = (floor(x) == x) ? Py_True : Py_False;
PyFPE_END_PROTECT(x)