[Python-checkins] gh-74895: getaddrinfo no longer raises OverflowError (#2435)

[gpshead]

https://github.com/python/cpython/commit/928752ce4c23f47d3175dd47ecacf08d86a99c9d
commit: 928752ce4c23f47d3175dd47ecacf08d86a99c9d
branch: main
author: Radek Smejkal <radek.smejkal at atrak.cz>
committer: gpshead <greg at krypto.org>
date: 2023-02-13T17:37:34-08:00
summary:

gh-74895: getaddrinfo no longer raises OverflowError (#2435)

`socket.getaddrinfo()` no longer raises `OverflowError` based on the **port** argument. Error reporting (or not) for its value is left up to the underlying C library `getaddrinfo()` implementation.

files:
A Misc/NEWS.d/next/Core and Builtins/2023-02-13-22-21-58.gh-issue-74895.esMNtq.rst
M Lib/test/test_socket.py
M Misc/ACKS
M Modules/getaddrinfo.c
M Modules/socketmodule.c

diff --git a/Lib/test/test_socket.py b/Lib/test/test_socket.py
index f1b4018c265e..a313da29b4a4 100644
--- a/Lib/test/test_socket.py
+++ b/Lib/test/test_socket.py
@@ -1600,6 +1600,54 @@ def testGetaddrinfo(self):
             except socket.gaierror:
                 pass
 
+    def test_getaddrinfo_int_port_overflow(self):
+        # gh-74895: Test that getaddrinfo does not raise OverflowError on port.
+        #
+        # POSIX getaddrinfo() never specify the valid range for "service"
+        # decimal port number values. For IPv4 and IPv6 they are technically
+        # unsigned 16-bit values, but the API is protocol agnostic. Which values
+        # trigger an error from the C library function varies by platform as
+        # they do not all perform validation.
+
+        # The key here is that we don't want to produce OverflowError as Python
+        # prior to 3.12 did for ints outside of a [LONG_MIN, LONG_MAX] range.
+        # Leave the error up to the underlying string based platform C API.
+
+        from _testcapi import ULONG_MAX, LONG_MAX, LONG_MIN
+        try:
+            socket.getaddrinfo(None, ULONG_MAX + 1)
+        except OverflowError:
+            # Platforms differ as to what values consitute a getaddrinfo() error
+            # return. Some fail for LONG_MAX+1, others ULONG_MAX+1, and Windows
+            # silently accepts such huge "port" aka "service" numeric values.
+            self.fail("Either no error or socket.gaierror expected.")
+        except socket.gaierror:
+            pass
+
+        try:
+            socket.getaddrinfo(None, LONG_MAX + 1)
+        except OverflowError:
+            self.fail("Either no error or socket.gaierror expected.")
+        except socket.gaierror:
+            pass
+
+        try:
+            socket.getaddrinfo(None, LONG_MAX - 0xffff + 1)
+        except OverflowError:
+            self.fail("Either no error or socket.gaierror expected.")
+        except socket.gaierror:
+            pass
+
+        try:
+            socket.getaddrinfo(None, LONG_MIN - 1)
+        except OverflowError:
+            self.fail("Either no error or socket.gaierror expected.")
+        except socket.gaierror:
+            pass
+
+        socket.getaddrinfo(None, 0)  # No error expected.
+        socket.getaddrinfo(None, 0xffff)  # No error expected.
+
     def test_getnameinfo(self):
         # only IP addresses are allowed
         self.assertRaises(OSError, socket.getnameinfo, ('mail.python.org',0), 0)
diff --git a/Misc/ACKS b/Misc/ACKS
index e12cbea0ebd6..ca92608868f2 100644
--- a/Misc/ACKS
+++ b/Misc/ACKS
@@ -1688,6 +1688,7 @@ Roman Skurikhin
 Ville Skyttä
 Michael Sloan
 Nick Sloan
+Radek Smejkal
 Václav Šmilauer
 Casper W. Smet
 Allen W. Smith
diff --git a/Misc/NEWS.d/next/Core and Builtins/2023-02-13-22-21-58.gh-issue-74895.esMNtq.rst b/Misc/NEWS.d/next/Core and Builtins/2023-02-13-22-21-58.gh-issue-74895.esMNtq.rst
new file mode 100644
index 000000000000..adbbb601634a
--- /dev/null
+++ b/Misc/NEWS.d/next/Core and Builtins/2023-02-13-22-21-58.gh-issue-74895.esMNtq.rst	
@@ -0,0 +1,5 @@
+:mod:`socket.getaddrinfo` no longer raises :class:`OverflowError` for
+:class:`int` **port** values outside of the C long range. Out of range values
+are left up to the underlying string based C library API to report. A
+:class:`socket.gaierror` ``SAI_SERVICE`` may occur instead, or no error at all
+as not all platform C libraries generate an error.
diff --git a/Modules/getaddrinfo.c b/Modules/getaddrinfo.c
index 0b4620ed683d..f1c28d7d9312 100644
--- a/Modules/getaddrinfo.c
+++ b/Modules/getaddrinfo.c
@@ -342,7 +342,11 @@ getaddrinfo(const char*hostname, const char*servname,
                 pai->ai_socktype = SOCK_DGRAM;
                 pai->ai_protocol = IPPROTO_UDP;
             }
-            port = htons((u_short)atoi(servname));
+            long maybe_port = strtol(servname, NULL, 10);
+            if (maybe_port < 0 || maybe_port > 0xffff) {
+                ERR(EAI_SERVICE);
+            }
+            port = htons((u_short)maybe_port);
         } else {
             struct servent *sp;
             const char *proto;
diff --git a/Modules/socketmodule.c b/Modules/socketmodule.c
index 0a9e46512b15..2d300f19436b 100644
--- a/Modules/socketmodule.c
+++ b/Modules/socketmodule.c
@@ -6650,7 +6650,7 @@ socket_getaddrinfo(PyObject *self, PyObject *args, PyObject* kwargs)
     struct addrinfo *res0 = NULL;
     PyObject *hobj = NULL;
     PyObject *pobj = (PyObject *)NULL;
-    char pbuf[30];
+    PyObject *pstr = NULL;
     const char *hptr, *pptr;
     int family, socktype, protocol, flags;
     int error;
@@ -6680,11 +6680,13 @@ socket_getaddrinfo(PyObject *self, PyObject *args, PyObject* kwargs)
         return NULL;
     }
     if (PyLong_CheckExact(pobj)) {
-        long value = PyLong_AsLong(pobj);
-        if (value == -1 && PyErr_Occurred())
+        pstr = PyObject_Str(pobj);
+        if (pstr == NULL)
+            goto err;
+        assert(PyUnicode_Check(pstr));
+        pptr = PyUnicode_AsUTF8(pstr);
+        if (pptr == NULL)
             goto err;
-        PyOS_snprintf(pbuf, sizeof(pbuf), "%ld", value);
-        pptr = pbuf;
     } else if (PyUnicode_Check(pobj)) {
         pptr = PyUnicode_AsUTF8(pobj);
         if (pptr == NULL)
@@ -6750,12 +6752,14 @@ socket_getaddrinfo(PyObject *self, PyObject *args, PyObject* kwargs)
         Py_DECREF(single);
     }
     Py_XDECREF(idna);
+    Py_XDECREF(pstr);
     if (res0)
         freeaddrinfo(res0);
     return all;
  err:
     Py_XDECREF(all);
     Py_XDECREF(idna);
+    Py_XDECREF(pstr);
     if (res0)
         freeaddrinfo(res0);
     return (PyObject *)NULL;