BufferedFile.read() now returns byte strings instead of text strings

It is the right thing to do since we have no idea what encoding the file
is in, or even if the file is text data. BufferedFile.readline() is
unchanged and returns text strings assuming the file is utf-8 encoded.
This should fix the following issue:
http://comments.gmane.org/gmane.comp.sysutils.backup.obnam/252

Antoine Brenner

Conflicts:
	sites/www/changelog.rst

5 files changed, 54 insertions, 33 deletions

diff --git a/paramiko/file.py b/paramiko/file.py
index f57aa79f..725ca5f6 100644
--- a/paramiko/file.py
+++ b/paramiko/file.py
@@ -124,10 +124,15 @@ class BufferedFile (object):
         file first).  If the ``size`` argument is negative or omitted, read all
         the remaining data in the file.
 
+        `'b' mode flag is ignored (self.FLAG_BINARY in self._flags), because
+        SSH treats all files as binary, since we have no idea what encoding
+        the file is in, or even if the file is text data.
+  
+
         :param int size: maximum number of bytes to read
         :return:
-            data read from the file (as a `str`), or an empty string if EOF was
-            encountered immediately
+            data read from the file (as bytes ), or an empty string
+            if EOF was encountered immediately
         """
         if self._closed:
             raise IOError('File is closed')
@@ -148,12 +153,12 @@ class BufferedFile (object):
                 result += new_data
                 self._realpos += len(new_data)
                 self._pos += len(new_data)
-            return result if self._flags & self.FLAG_BINARY else u(result)
+            return result 
         if size <= len(self._rbuffer):
             result = self._rbuffer[:size]
             self._rbuffer = self._rbuffer[size:]
             self._pos += len(result)
-            return result if self._flags & self.FLAG_BINARY else u(result)
+            return result 
         while len(self._rbuffer) < size:
             read_size = size - len(self._rbuffer)
             if self._flags & self.FLAG_BUFFERED:
@@ -169,7 +174,7 @@ class BufferedFile (object):
         result = self._rbuffer[:size]
         self._rbuffer = self._rbuffer[size:]
         self._pos += len(result)
-        return result if self._flags & self.FLAG_BINARY else u(result)
+        return result 
 
     def readline(self, size=None):
         """
@@ -186,8 +191,12 @@ class BufferedFile (object):
 
         :param int size: maximum length of returned string.
         :return:
-            next line of the file (`str`), or an empty string if the end of the
+            next line of the file, or an empty string if the end of the
             file has been reached.
+            If the file was opened in binary 'b' mode: bytes are returned
+            Else: the encoding of the file is assumed to be utf-8 (default
+            encoding used by paramiko.py3compat.u) and character strings
+           (`str`) are returned
         """
         # it's almost silly how complex this function is.
         if self._closed:
diff --git a/sites/www/changelog.rst b/sites/www/changelog.rst
index 21ba6e12..67c4f827 100644
--- a/sites/www/changelog.rst
+++ b/sites/www/changelog.rst
@@ -2,6 +2,13 @@
 Changelog
 =========
 
+* :bug: BufferedFile.read() now returns byte strings instead of text strings
+  It is the right thing to do since we have no idea what encoding the file 
+  is in, or even if the file is text data. BufferedFile.readline() is
+  unchanged and returns text strings assuming the file is utf-8 encoded.
+  This should fix the following issue:
+  http://comments.gmane.org/gmane.comp.sysutils.backup.obnam/252  
+  Thanks Antoine Brenner
 * :bug:`-` Added self.args for exception classes. Used for unpickling. Related
   to (`Fabric #986 <https://github.com/fabric/fabric/issues/986>`_, `Fabric
   #714 <https://github.com/fabric/fabric/issues/714>`_). Thanks to Alex
diff --git a/tests/test_file.py b/tests/test_file.py
index e11d7fd5..c6edd7af 100755
--- a/tests/test_file.py
+++ b/tests/test_file.py
@@ -53,7 +53,7 @@ class BufferedFileTest (unittest.TestCase):
     def test_1_simple(self):
         f = LoopbackFile('r')
         try:
-            f.write('hi')
+            f.write(b'hi')
             self.assertTrue(False, 'no exception on write to read-only file')
         except:
             pass
@@ -69,7 +69,7 @@ class BufferedFileTest (unittest.TestCase):
 
     def test_2_readline(self):
         f = LoopbackFile('r+U')
-        f.write('First line.\nSecond line.\r\nThird line.\nFinal line non-terminated.')
+        f.write(b'First line.\nSecond line.\r\nThird line.\nFinal line non-terminated.')
         self.assertEqual(f.readline(), 'First line.\n')
         # universal newline mode should convert this linefeed:
         self.assertEqual(f.readline(), 'Second line.\n')
@@ -93,9 +93,9 @@ class BufferedFileTest (unittest.TestCase):
         try to trick the linefeed detector.
         """
         f = LoopbackFile('r+U')
-        f.write('First line.\r')
+        f.write(b'First line.\r')
         self.assertEqual(f.readline(), 'First line.\n')
-        f.write('\nSecond.\r\n')
+        f.write(b'\nSecond.\r\n')
         self.assertEqual(f.readline(), 'Second.\n')
         f.close()
         self.assertEqual(f.newlines, crlf)
@@ -105,7 +105,7 @@ class BufferedFileTest (unittest.TestCase):
         verify that write buffering is on.
         """
         f = LoopbackFile('r+', 1)
-        f.write('Complete line.\nIncomplete line.')
+        f.write(b'Complete line.\nIncomplete line.')
         self.assertEqual(f.readline(), 'Complete line.\n')
         self.assertEqual(f.readline(), '')
         f.write('..\n')
@@ -118,12 +118,12 @@ class BufferedFileTest (unittest.TestCase):
         """
         f = LoopbackFile('r+', 512)
         f.write('Not\nquite\n512 bytes.\n')
-        self.assertEqual(f.read(1), '')
+        self.assertEqual(f.read(1), b'')
         f.flush()
-        self.assertEqual(f.read(5), 'Not\nq')
-        self.assertEqual(f.read(10), 'uite\n512 b')
-        self.assertEqual(f.read(9), 'ytes.\n')
-        self.assertEqual(f.read(3), '')
+        self.assertEqual(f.read(5), b'Not\nq')
+        self.assertEqual(f.read(10), b'uite\n512 b')
+        self.assertEqual(f.read(9), b'ytes.\n')
+        self.assertEqual(f.read(3), b'')
         f.close()
 
     def test_6_buffering(self):
@@ -131,12 +131,12 @@ class BufferedFileTest (unittest.TestCase):
         verify that flushing happens automatically on buffer crossing.
         """
         f = LoopbackFile('r+', 16)
-        f.write('Too small.')
-        self.assertEqual(f.read(4), '')
-        f.write('  ')
-        self.assertEqual(f.read(4), '')
-        f.write('Enough.')
-        self.assertEqual(f.read(20), 'Too small.  Enough.')
+        f.write(b'Too small.')
+        self.assertEqual(f.read(4), b'')
+        f.write(b'  ')
+        self.assertEqual(f.read(4), b'')
+        f.write(b'Enough.')
+        self.assertEqual(f.read(20), b'Too small.  Enough.')
         f.close()
 
     def test_7_read_all(self):
@@ -144,9 +144,14 @@ class BufferedFileTest (unittest.TestCase):
         verify that read(-1) returns everything left in the file.
         """
         f = LoopbackFile('r+', 16)
-        f.write('The first thing you need to do is open your eyes. ')
-        f.write('Then, you need to close them again.\n')
+        f.write(b'The first thing you need to do is open your eyes. ')
+        f.write(b'Then, you need to close them again.\n')
         s = f.read(-1)
-        self.assertEqual(s, 'The first thing you need to do is open your eyes. Then, you ' +
-                         'need to close them again.\n')
+        self.assertEqual(s, b'The first thing you need to do is open your eyes. Then, you ' +
+                            b'need to close them again.\n')
         f.close()
+
+if __name__ == '__main__':
+    from unittest import main
+    main()
+
diff --git a/tests/test_sftp.py b/tests/test_sftp.py
index e0534eb0..720b8215 100755
--- a/tests/test_sftp.py
+++ b/tests/test_sftp.py
@@ -405,7 +405,7 @@ class SFTPTest (unittest.TestCase):
             self.assertEqual(sftp.stat(FOLDER + '/testing.txt').st_size, 13)
             with sftp.open(FOLDER + '/testing.txt', 'r') as f:
                 data = f.read(20)
-            self.assertEqual(data, 'hello kiddy.\n')
+            self.assertEqual(data, b'hello kiddy.\n')
         finally:
             sftp.remove(FOLDER + '/testing.txt')
 
@@ -466,8 +466,8 @@ class SFTPTest (unittest.TestCase):
                 f.write('?\n')
 
             with sftp.open(FOLDER + '/happy.txt', 'r') as f:
-                self.assertEqual(f.readline(), 'full line?\n')
-                self.assertEqual(f.read(7), 'partial')
+                self.assertEqual(f.readline(), u'full line?\n')
+                self.assertEqual(f.read(7), b'partial')
         finally:
             try:
                 sftp.remove(FOLDER + '/happy.txt')
@@ -662,8 +662,8 @@ class SFTPTest (unittest.TestCase):
 
         fd, localname = mkstemp()
         os.close(fd)
-        text = 'All I wanted was a plastic bunny rabbit.\n'
-        with open(localname, 'w') as f:
+        text = b'All I wanted was a plastic bunny rabbit.\n'
+        with open(localname, 'wb') as f:
             f.write(text)
         saved_progress = []
 
diff --git a/tests/test_sftp_big.py b/tests/test_sftp_big.py
index 521fbdc8..abed27b8 100644
--- a/tests/test_sftp_big.py
+++ b/tests/test_sftp_big.py
@@ -85,7 +85,7 @@ class BigSFTPTest (unittest.TestCase):
         write a 1MB file with no buffering.
         """
         sftp = get_sftp()
-        kblob = (1024 * 'x')
+        kblob = (1024 * b'x')
         start = time.time()
         try:
             with sftp.open('%s/hongry.txt' % FOLDER, 'w') as f:
@@ -231,7 +231,7 @@ class BigSFTPTest (unittest.TestCase):
         without using it, to verify that paramiko doesn't get confused.
         """
         sftp = get_sftp()
-        kblob = (1024 * 'x')
+        kblob = (1024 * b'x')
         try:
             with sftp.open('%s/hongry.txt' % FOLDER, 'w') as f:
                 f.set_pipelined(True)