Adjusted stream logic to make it work on all tested platforms ... .

As taken from https://github.com/gitpython-developers/gitdb/blob/master/gitdb/stream.py#L292
->
NOTE: Behavior changed in PY2.7 onward, which requires special handling to make the tests work properly.
They are thorough, and I assume it is truly working.
Why is this logic as convoluted as it is ? Please look at the table in
gitpython-developers/gitdb#19 to learn about the test-results.
Bascially, on py2.6, you want to use branch 1, whereas on all other python version, the second branch
will be the one that works.
However, the zlib VERSIONs as well as the platform check is used to further match the entries in the
table in the github issue. This is it ... it was the only way I could make this work everywhere.
IT's CERTAINLY GOING TO BITE US IN THE FUTURE ... .
<-

Fixes #19

Author: Byron

gitdb/fun.py

@@ -426,7 +426,7 @@ def pack_object_header_info(data):
     s = 4                           # starting bit-shift size
     if PY3:
         while c & 0x80:
-            c = data[i]
+            c = byte_ord(data[i])
             i += 1
             size += (c & 0x7f) << s
             s += 7

gitdb/pack.py

@@ -552,6 +552,7 @@ def _iter_objects(self, start_offset, as_stream=True):
             # the amount of compressed bytes we need to get to the next offset
 
             stream_copy(ostream.read, null.write, ostream.size, chunk_size)
+            assert ostream.stream._br == ostream.size
             cur_offset += (data_offset - ostream.pack_offset) + ostream.stream.compressed_bytes_read()
 
             # if a stream is requested, reset it beforehand

gitdb/stream.py

@@ -289,11 +289,18 @@ def read(self, size=-1):
         # if we hit the end of the stream
         # NOTE: Behavior changed in PY2.7 onward, which requires special handling to make the tests work properly.
         # They are thorough, and I assume it is truly working.
-        if PY26:
+        # Why is this logic as convoluted as it is ? Please look at the table in 
+        # https://github.com/gitpython-developers/gitdb/issues/19 to learn about the test-results.
+        # Bascially, on py2.6, you want to use branch 1, whereas on all other python version, the second branch
+        # will be the one that works. 
+        # However, the zlib VERSIONs as well as the platform check is used to further match the entries in the 
+        # table in the github issue. This is it ... it was the only way I could make this work everywhere.
+        # IT's CERTAINLY GOING TO BITE US IN THE FUTURE ... .
+        if PY26 or ((zlib.ZLIB_VERSION == '1.2.7' or zlib.ZLIB_VERSION == '1.2.5') and not sys.platform == 'darwin'):
             unused_datalen = len(self._zip.unconsumed_tail)
         else:
             unused_datalen = len(self._zip.unconsumed_tail) + len(self._zip.unused_data)
-        # end handle very special case ...
+        # # end handle very special case ...
 
         self._cbr += len(indata) - unused_datalen
         self._br += len(dcompdat)
@@ -374,7 +381,6 @@ def _set_cache_too_slow_without_c(self, attr):
         # Aggregate all deltas into one delta in reverse order. Hence we take
         # the last delta, and reverse-merge its ancestor delta, until we receive
         # the final delta data stream.
-        # print "Handling %i delta streams, sizes: %s" % (len(self._dstreams), [ds.size for ds in self._dstreams])
         dcl = connect_deltas(self._dstreams)
 
         # call len directly, as the (optional) c version doesn't implement the sequence

setup.py

@@ -118,6 +118,7 @@ def get_data_files(self):
     "Programming Language :: Python :: 2.6",
     "Programming Language :: Python :: 2.7",
     "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.2",
     "Programming Language :: Python :: 3.3",
     "Programming Language :: Python :: 3.4",
 ],)