gh-161588: Ignore invalid PAX header 'size' in tarfile to extract files properly

Lib/tarfile.py

@@ -1710,23 +1710,29 @@ def _apply_pax_info(self, pax_headers, encoding, errors):
         """Replace fields with supplemental information from a previous
            pax extended or global header.
         """
+        keyword_to_remove = []
         for keyword, value in pax_headers.items():
-            if keyword == "GNU.sparse.name":
+            if keyword in ("GNU.sparse.name", "path"):
+                value = value.rstrip("/")
                 setattr(self, "path", value)
-            elif keyword == "GNU.sparse.size":
-                setattr(self, "size", int(value))
-            elif keyword == "GNU.sparse.realsize":
-                setattr(self, "size", int(value))
+            elif keyword in ("GNU.sparse.size", "GNU.sparse.realsize", "size"):
+                try:
+                    value = int(value)
+                except ValueError:
+                    keyword_to_remove.append(keyword)
+                else:
+                    setattr(self, "size", value)
             elif keyword in PAX_FIELDS:
                 if keyword in PAX_NUMBER_FIELDS:
                     try:
                         value = PAX_NUMBER_FIELDS[keyword](value)
                     except ValueError:
                         value = 0
-                if keyword == "path":
-                    value = value.rstrip("/")
                 setattr(self, keyword, value)
 
+        for k in keyword_to_remove:
+            pax_headers.pop(k)
+
         self.pax_headers = pax_headers.copy()
 
     def _decode_pax_field(self, value, encoding, fallback_encoding, fallback_errors):

Lib/test/test_tarfile.py

@@ -2759,6 +2759,32 @@ def test_pax_extended_header(self):
         finally:
             tar.close()
 
+    def test_invalid_size_pax_extended_header(self):
+        # When the value of the keyword 'size' is invalid, it should be
+        # ignored and proceeds to extract all the records instead of
+        # setting size to 0 and producing a single empty extracted file
+        t1sz = 89
+        t2sz = 64
+        with tarfile.open(tmpname, "w", format=tarfile.PAX_FORMAT,
+                          pax_headers={"size": "xyz"},
+                          encoding="iso8859-1") as tar:
+            t1 = tarfile.TarInfo()
+            t1.name = "foo"
+            t1.size = t1sz
+            tar.addfile(t1, io.BytesIO(b"a"*t1sz))
+
+            t2 = tarfile.TarInfo()
+            t2.name = "bar"
+            t2.size = t2sz
+            tar.addfile(t2, io.BytesIO(b"z"*t2sz))
+
+        t1_path = os.path.join(TEMPDIR, "foo")
+        t2_path = os.path.join(TEMPDIR, "bar")
+        with tarfile.open(tmpname, encoding="iso8859-1") as tar:
+            tar.extractall(path=TEMPDIR)
+            self.assertEqual(os.path.getsize(t1_path), t1sz)
+            self.assertEqual(os.path.getsize(t2_path), t2sz)
+
     def test_create_pax_header(self):
         # The ustar header should contain values that can be
         # represented reasonably, even if a better (e.g. higher

Misc/NEWS.d/next/Library/2026-11-07-22-52-41.gh-issue-161588.rbB5Dg.rst

@@ -0,0 +1,2 @@
+Allow complete extraction of all records in a tar archive when the PAX
+header 'size' is invalid in :mod:`tarfile`.