Fix large SFTP reads

docs/changelog-fragments/638.bugfix.rst

@@ -0,0 +1,5 @@
+Fixed reading files over SFTP that go over the pre-defined chunk size.
+
+Prior to this change, the files could end up being corrupted, ending up with the last read chunk written to the file instead of the entire payload.
+
+-- by :user:`Jakuje`.

src/pylibsshext/sftp.pyx

@@ -89,18 +89,19 @@ cdef class SFTP:
 
         rf = sftp.sftp_open(self._libssh_sftp_session, remote_file_b, O_RDONLY, sftp.S_IRWXU)
         if rf is NULL:
-            raise LibsshSFTPException("Opening remote file [%s] for read failed with error [%s]" % (remote_file, self._get_sftp_error_str()))
-
-        while True:
-            file_data = sftp.sftp_read(rf, <void *>read_buffer, sizeof(char) * 1024)
-            if file_data == 0:
-                break
-            elif file_data < 0:
-                sftp.sftp_close(rf)
-                raise LibsshSFTPException("Reading data from remote file [%s] failed with error [%s]"
-                                          % (remote_file, self._get_sftp_error_str()))
-
-            with open(local_file, 'wb+') as f:
+            raise LibsshSFTPException("Opening remote file [%s] for read failed with error [%s]"
+                                      % (remote_file, self._get_sftp_error_str()))
+
+        with open(local_file, 'wb') as f:
+            while True:
+                file_data = sftp.sftp_read(rf, <void *>read_buffer, sizeof(char) * 1024)
+                if file_data == 0:
+                    break
+                elif file_data < 0:
+                    sftp.sftp_close(rf)
+                    raise LibsshSFTPException("Reading data from remote file [%s] failed with error [%s]"
+                                              % (remote_file, self._get_sftp_error_str()))
+
                 bytes_written = f.write(read_buffer[:file_data])
                 if bytes_written and file_data != bytes_written:
                     sftp.sftp_close(rf)

tests/unit/sftp_test.py

@@ -2,6 +2,8 @@
 
 """Tests suite for sftp."""
 
+import random
+import string
 import uuid
 
 import pytest
@@ -18,11 +20,21 @@ def sftp_session(ssh_client_session):
         del sftp_sess  # noqa: WPS420
 
 
-@pytest.fixture
-def transmit_payload():
-    """Generate a binary test payload."""
-    uuid_name = uuid.uuid4()
-    return 'Hello, {name!s}'.format(name=uuid_name).encode()
+@pytest.fixture(
+    params=(32, 1024 + 1),
+    ids=('small-payload', 'large-payload'),
+)
+def transmit_payload(request: pytest.FixtureRequest):
+    """Generate binary test payloads of assorted sizes.
+
+    The choice 32 is arbitrary small value.
+
+    The choice 1024 + 1 is meant to be 1B larger than the chunk size used in
+    sftp.pyx to make sure we excercise at least two rounds of reading/writing.
+    """
+    payload_len = request.param
+    random_bytes = [ord(random.choice(string.printable)) for _ in range(payload_len)]
+    return bytes(random_bytes)
 
 
 @pytest.fixture
@@ -48,6 +60,22 @@ def dst_path(file_paths_pair):
     return path
 
 
+@pytest.fixture
+def other_payload():
+    """Generate a binary test payload."""
+    uuid_name = uuid.uuid4()
+    return 'Original content: {name!s}'.format(name=uuid_name).encode()
+
+
+@pytest.fixture
+def dst_exists_path(file_paths_pair, other_payload):
+    """Return a data destination path."""
+    path = file_paths_pair[1]
+    path.write_bytes(other_payload)
+    assert path.exists()
+    return path
+
+
 def test_make_sftp(sftp_session):
     """Smoke-test SFTP instance creation."""
     assert sftp_session
@@ -63,3 +91,15 @@ def test_get(dst_path, src_path, sftp_session, transmit_payload):
     """Check that SFTP file download works."""
     sftp_session.get(str(src_path), str(dst_path))
     assert dst_path.read_bytes() == transmit_payload
+
+
+def test_get_existing(dst_exists_path, src_path, sftp_session, transmit_payload):
+    """Check that SFTP file download works when target file exists."""
+    sftp_session.get(str(src_path), str(dst_exists_path))
+    assert dst_exists_path.read_bytes() == transmit_payload
+
+
+def test_put_existing(dst_exists_path, src_path, sftp_session, transmit_payload):
+    """Check that SFTP file upload works when target file exists."""
+    sftp_session.put(str(src_path), str(dst_exists_path))
+    assert dst_exists_path.read_bytes() == transmit_payload