Issue #1531: Read fileobj from the current offset, do not seek to

the start.

(will backport to 2.5)

diff --git a/Lib/tarfile.py b/Lib/tarfile.py
index ee9922c60ab730e11c477f6c689bdc0b61d07691..ae242912e92c8b0aa8d335777736c14354bb56a2 100644 (file)
--- a/Lib/tarfile.py
+++ b/Lib/tarfile.py
@@ -1558,7 +1558,8 @@ class TarFile(object):
         self.closed = False
         self.members = []       # list of members as TarInfo objects
         self._loaded = False    # flag if all members have been read
-        self.offset = 0L        # current position in the archive file
+        self.offset = self.fileobj.tell()
+                                # current position in the archive file
         self.inodes = {}        # dictionary caching the inodes of
                                 # archive members already added
 

diff --git a/Lib/test/test_tarfile.py b/Lib/test/test_tarfile.py
index 1f0825831cd3151c28af60f8c2d9b0a7b5a4ab8d..a280bdda82e59d0f2cc2c3948c42b27b360d685c 100644 (file)
--- a/Lib/test/test_tarfile.py
+++ b/Lib/test/test_tarfile.py
@@ -160,6 +160,38 @@ class MiscReadTest(ReadTest):
         tar = tarfile.open(fileobj=fobj, mode=self.mode)
         self.assertEqual(tar.name, None)
 
+    def test_fileobj_with_offset(self):
+        # Skip the first member and store values from the second member
+        # of the testtar.
+        tar = tarfile.open(self.tarname, mode=self.mode)
+        tar.next()
+        t = tar.next()
+        name = t.name
+        offset = t.offset
+        data = tar.extractfile(t).read()
+        tar.close()
+
+        # Open the testtar and seek to the offset of the second member.
+        if self.mode.endswith(":gz"):
+            _open = gzip.GzipFile
+        elif self.mode.endswith(":bz2"):
+            _open = bz2.BZ2File
+        else:
+            _open = open
+        fobj = _open(self.tarname, "rb")
+        fobj.seek(offset)
+
+        # Test if the tarfile starts with the second member.
+        tar = tar.open(self.tarname, mode="r:", fileobj=fobj)
+        t = tar.next()
+        self.assertEqual(t.name, name)
+        # Read to the end of fileobj and test if seeking back to the
+        # beginning works.
+        tar.getmembers()
+        self.assertEqual(tar.extractfile(t).read(), data,
+                "seek back did not work")
+        tar.close()
+
     def test_fail_comp(self):
         # For Gzip and Bz2 Tests: fail with a ReadError on an uncompressed file.
         if self.mode == "r:":

diff --git a/Misc/NEWS b/Misc/NEWS
index 2023143f523ce77ba5597c6c0ee26b7501f5a08b..c32e3f93e2dd5e730bc5f82be4111e36a9195350 100644 (file)
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -304,6 +304,9 @@ Core and builtins
 Library
 -------
 
+- Issue #1531: tarfile.py: Read fileobj from the current offset, do not
+  seek to the start.
+
 - Issue #1534: Added a dictionary sys.float_info with information about the
   internal floating point type to the sys module.