Issue #26293: Fixed writing ZIP files that starts not from the start of the

file.  Offsets in ZIP file now are relative to the start of the archive in
conforming to the specification.

diff --git a/Lib/test/test_zipfile.py b/Lib/test/test_zipfile.py
index 4000b30be8dd640c8e05741ac0f83d3929f2a885..986808131e536e3fde5746be90caa604c3031b7d 100644 (file)
--- a/Lib/test/test_zipfile.py
+++ b/Lib/test/test_zipfile.py
@@ -344,6 +344,49 @@ class TestsWithSourceFile(unittest.TestCase):
             f.seek(len(data))
             with zipfile.ZipFile(f, "r") as zipfp:
                 self.assertEqual(zipfp.namelist(), [TESTFN])
+                self.assertEqual(zipfp.read(TESTFN), self.data)
+        with open(TESTFN2, 'rb') as f:
+            self.assertEqual(f.read(len(data)), data)
+            zipfiledata = f.read()
+        with io.BytesIO(zipfiledata) as bio, zipfile.ZipFile(bio) as zipfp:
+            self.assertEqual(zipfp.namelist(), [TESTFN])
+            self.assertEqual(zipfp.read(TESTFN), self.data)
+
+    def test_read_concatenated_zip_file(self):
+        with io.BytesIO() as bio:
+            with zipfile.ZipFile(bio, 'w', zipfile.ZIP_STORED) as zipfp:
+                zipfp.write(TESTFN, TESTFN)
+            zipfiledata = bio.getvalue()
+        data = b'I am not a ZipFile!'*10
+        with open(TESTFN2, 'wb') as f:
+            f.write(data)
+            f.write(zipfiledata)
+
+        with zipfile.ZipFile(TESTFN2) as zipfp:
+            self.assertEqual(zipfp.namelist(), [TESTFN])
+            self.assertEqual(zipfp.read(TESTFN), self.data)
+
+    def test_append_to_concatenated_zip_file(self):
+        with io.BytesIO() as bio:
+            with zipfile.ZipFile(bio, 'w', zipfile.ZIP_STORED) as zipfp:
+                zipfp.write(TESTFN, TESTFN)
+            zipfiledata = bio.getvalue()
+        data = b'I am not a ZipFile!'*1000000
+        with open(TESTFN2, 'wb') as f:
+            f.write(data)
+            f.write(zipfiledata)
+
+        with zipfile.ZipFile(TESTFN2, 'a') as zipfp:
+            self.assertEqual(zipfp.namelist(), [TESTFN])
+            zipfp.writestr('strfile', self.data)
+
+        with open(TESTFN2, 'rb') as f:
+            self.assertEqual(f.read(len(data)), data)
+            zipfiledata = f.read()
+        with io.BytesIO(zipfiledata) as bio, zipfile.ZipFile(bio) as zipfp:
+            self.assertEqual(zipfp.namelist(), [TESTFN, 'strfile'])
+            self.assertEqual(zipfp.read(TESTFN), self.data)
+            self.assertEqual(zipfp.read('strfile'), self.data)
 
     def test_ignores_newline_at_end(self):
         with zipfile.ZipFile(TESTFN2, "w", zipfile.ZIP_STORED) as zipfp:

diff --git a/Lib/zipfile.py b/Lib/zipfile.py
index 3ab66cea69d89fb32f8570449581a0c5b4a3bbd2..1d1065083bd51a92e77b72bac7c38cb95a2d01de 100644 (file)
--- a/Lib/zipfile.py
+++ b/Lib/zipfile.py
@@ -772,6 +772,7 @@ class ZipFile(object):
                 # set the modified flag so central directory gets written
                 # even if no files are added to the archive
                 self._didModify = True
+                self._start_disk = self.fp.tell()
             elif key == 'a':
                 try:
                     # See if file is a zip file
@@ -785,6 +786,7 @@ class ZipFile(object):
                     # set the modified flag so central directory gets written
                     # even if no files are added to the archive
                     self._didModify = True
+                    self._start_disk = self.fp.tell()
             else:
                 raise RuntimeError('Mode must be "r", "w" or "a"')
         except:
@@ -815,17 +817,18 @@ class ZipFile(object):
         offset_cd = endrec[_ECD_OFFSET]         # offset of central directory
         self._comment = endrec[_ECD_COMMENT]    # archive comment
 
-        # "concat" is zero, unless zip was concatenated to another file
-        concat = endrec[_ECD_LOCATION] - size_cd - offset_cd
+        # self._start_disk:  Position of the start of ZIP archive
+        # It is zero, unless ZIP was concatenated to another file
+        self._start_disk = endrec[_ECD_LOCATION] - size_cd - offset_cd
         if endrec[_ECD_SIGNATURE] == stringEndArchive64:
             # If Zip64 extension structures are present, account for them
-            concat -= (sizeEndCentDir64 + sizeEndCentDir64Locator)
+            self._start_disk -= (sizeEndCentDir64 + sizeEndCentDir64Locator)
 
         if self.debug > 2:
-            inferred = concat + offset_cd
-            print "given, inferred, offset", offset_cd, inferred, concat
+            inferred = self._start_disk + offset_cd
+            print "given, inferred, offset", offset_cd, inferred, self._start_disk
         # self.start_dir:  Position of start of central directory
-        self.start_dir = offset_cd + concat
+        self.start_dir = offset_cd + self._start_disk
         fp.seek(self.start_dir, 0)
         data = fp.read(size_cd)
         fp = cStringIO.StringIO(data)
@@ -855,7 +858,7 @@ class ZipFile(object):
                                      t>>11, (t>>5)&0x3F, (t&0x1F) * 2 )
 
             x._decodeExtra()
-            x.header_offset = x.header_offset + concat
+            x.header_offset = x.header_offset + self._start_disk
             x.filename = x._decodeFilename()
             self.filelist.append(x)
             self.NameToInfo[x.filename] = x
@@ -1198,7 +1201,7 @@ class ZipFile(object):
                 raise RuntimeError('Compressed size larger than uncompressed size')
         # Seek backwards and write file header (which will now include
         # correct CRC and file sizes)
-        position = self.fp.tell()       # Preserve current position in file
+        position = self.fp.tell() # Preserve current position in file
         self.fp.seek(zinfo.header_offset, 0)
         self.fp.write(zinfo.FileHeader(zip64))
         self.fp.seek(position, 0)
@@ -1284,11 +1287,10 @@ class ZipFile(object):
                         file_size = zinfo.file_size
                         compress_size = zinfo.compress_size
 
-                    if zinfo.header_offset > ZIP64_LIMIT:
-                        extra.append(zinfo.header_offset)
+                    header_offset = zinfo.header_offset - self._start_disk
+                    if header_offset > ZIP64_LIMIT:
+                        extra.append(header_offset)
                         header_offset = 0xffffffffL
-                    else:
-                        header_offset = zinfo.header_offset
 
                     extra_data = zinfo.extra
                     if extra:
@@ -1332,7 +1334,7 @@ class ZipFile(object):
                 # Write end-of-zip-archive record
                 centDirCount = len(self.filelist)
                 centDirSize = pos2 - pos1
-                centDirOffset = pos1
+                centDirOffset = pos1 - self._start_disk
                 requires_zip64 = None
                 if centDirCount > ZIP_FILECOUNT_LIMIT:
                     requires_zip64 = "Files count"

diff --git a/Misc/NEWS b/Misc/NEWS
index 6be73784fc90b4c2f2bc0b0ff7b6ff5341f8e02d..1a737b84f83220ad1215a5dfa9e0168dfdce759b 100644 (file)
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -49,6 +49,10 @@ Core and Builtins
 Library
 -------
 
+- Issue #26293: Fixed writing ZIP files that starts not from the start of the
+  file.  Offsets in ZIP file now are relative to the start of the archive in
+  conforming to the specification.
+
 - Fix possible integer overflows and crashes in the mmap module with unusual
   usage patterns.