~bzr-pqm/bzr/bzr.dev

# Copyright (C) 2005, 2006, 2007, 2008, 2009 Canonical Ltd
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

"""Implementation of Transport over SFTP, using paramiko."""

# TODO: Remove the transport-based lock_read and lock_write methods.  They'll
# then raise TransportNotPossible, which will break remote access to any
# formats which rely on OS-level locks.  That should be fine as those formats
# are pretty old, but these combinations may have to be removed from the test
# suite.  Those formats all date back to 0.7; so we should be able to remove
# these methods when we officially drop support for those formats.

import bisect
import errno
import itertools
import os
import random
import select
import socket
import stat
import sys
import time
import urllib
import urlparse
import warnings

from bzrlib import (
    config,
    debug,
    errors,
    urlutils,
    )
from bzrlib.errors import (FileExists,
                           NoSuchFile, PathNotChild,
                           TransportError,
                           LockError,
                           PathError,
                           ParamikoNotPresent,
                           )
from bzrlib.osutils import pathjoin, fancy_rename, getcwd
from bzrlib.symbol_versioning import (
        deprecated_function,
        )
from bzrlib.trace import mutter, warning
from bzrlib.transport import (
    FileFileStream,
    _file_streams,
    local,
    Server,
    ssh,
    ConnectedTransport,
    )

# Disable one particular warning that comes from paramiko in Python2.5; if
# this is emitted at the wrong time it tends to cause spurious test failures
# or at least noise in the test case::
#
# [1770/7639 in 86s, 1 known failures, 50 skipped, 2 missing features]
# test_permissions.TestSftpPermissions.test_new_files
# /var/lib/python-support/python2.5/paramiko/message.py:226: DeprecationWarning: integer argument expected, got float
#  self.packet.write(struct.pack('>I', n))
warnings.filterwarnings('ignore',
        'integer argument expected, got float',
        category=DeprecationWarning,
        module='paramiko.message')

try:
    import paramiko
except ImportError, e:
    raise ParamikoNotPresent(e)
else:
    from paramiko.sftp import (SFTP_FLAG_WRITE, SFTP_FLAG_CREATE,
                               SFTP_FLAG_EXCL, SFTP_FLAG_TRUNC,
                               CMD_HANDLE, CMD_OPEN)
    from paramiko.sftp_attr import SFTPAttributes
    from paramiko.sftp_file import SFTPFile


_paramiko_version = getattr(paramiko, '__version_info__', (0, 0, 0))
# don't use prefetch unless paramiko version >= 1.5.5 (there were bugs earlier)
_default_do_prefetch = (_paramiko_version >= (1, 5, 5))


class SFTPLock(object):
    """This fakes a lock in a remote location.

    A present lock is indicated just by the existence of a file.  This
    doesn't work well on all transports and they are only used in
    deprecated storage formats.
    """

    __slots__ = ['path', 'lock_path', 'lock_file', 'transport']

    def __init__(self, path, transport):
        self.lock_file = None
        self.path = path
        self.lock_path = path + '.write-lock'
        self.transport = transport
        try:
            # RBC 20060103 FIXME should we be using private methods here ?
            abspath = transport._remote_path(self.lock_path)
            self.lock_file = transport._sftp_open_exclusive(abspath)
        except FileExists:
            raise LockError('File %r already locked' % (self.path,))

    def __del__(self):
        """Should this warn, or actually try to cleanup?"""
        if self.lock_file:
            warning("SFTPLock %r not explicitly unlocked" % (self.path,))
            self.unlock()

    def unlock(self):
        if not self.lock_file:
            return
        self.lock_file.close()
        self.lock_file = None
        try:
            self.transport.delete(self.lock_path)
        except (NoSuchFile,):
            # What specific errors should we catch here?
            pass


class _SFTPReadvHelper(object):
    """A class to help with managing the state of a readv request."""

    # See _get_requests for an explanation.
    _max_request_size = 32768

    def __init__(self, original_offsets, relpath, _report_activity):
        """Create a new readv helper.

        :param original_offsets: The original requests given by the caller of
            readv()
        :param relpath: The name of the file (if known)
        :param _report_activity: A Transport._report_activity bound method,
            to be called as data arrives.
        """
        self.original_offsets = list(original_offsets)
        self.relpath = relpath
        self._report_activity = _report_activity

    def _get_requests(self):
        """Break up the offsets into individual requests over sftp.

        The SFTP spec only requires implementers to support 32kB requests. We
        could try something larger (openssh supports 64kB), but then we have to
        handle requests that fail.
        So instead, we just break up our maximum chunks into 32kB chunks, and
        asyncronously requests them.
        Newer versions of paramiko would do the chunking for us, but we want to
        start processing results right away, so we do it ourselves.
        """
        # TODO: Because we issue async requests, we don't 'fudge' any extra
        #       data.  I'm not 100% sure that is the best choice.

        # The first thing we do, is to collapse the individual requests as much
        # as possible, so we don't issues requests <32kB
        sorted_offsets = sorted(self.original_offsets)
        coalesced = list(ConnectedTransport._coalesce_offsets(sorted_offsets,
                                                        limit=0, fudge_factor=0))
        requests = []
        for c_offset in coalesced:
            start = c_offset.start
            size = c_offset.length

            # Break this up into 32kB requests
            while size > 0:
                next_size = min(size, self._max_request_size)
                requests.append((start, next_size))
                size -= next_size
                start += next_size
        if 'sftp' in debug.debug_flags:
            mutter('SFTP.readv(%s) %s offsets => %s coalesced => %s requests',
                self.relpath, len(sorted_offsets), len(coalesced),
                len(requests))
        return requests

    def request_and_yield_offsets(self, fp):
        """Request the data from the remote machine, yielding the results.

        :param fp: A Paramiko SFTPFile object that supports readv.
        :return: Yield the data requested by the original readv caller, one by
            one.
        """
        requests = self._get_requests()
        offset_iter = iter(self.original_offsets)
        cur_offset, cur_size = offset_iter.next()
        # paramiko .readv() yields strings that are in the order of the requests
        # So we track the current request to know where the next data is
        # being returned from.
        input_start = None
        last_end = None
        buffered_data = []
        buffered_len = 0

        # This is used to buffer chunks which we couldn't process yet
        # It is (start, end, data) tuples.
        data_chunks = []
        # Create an 'unlimited' data stream, so we stop based on requests,
        # rather than just because the data stream ended. This lets us detect
        # short readv.
        data_stream = itertools.chain(fp.readv(requests),
                                      itertools.repeat(None))
        for (start, length), data in itertools.izip(requests, data_stream):
            if data is None:
                if cur_coalesced is not None:
                    raise errors.ShortReadvError(self.relpath,
                        start, length, len(data))
            if len(data) != length:
                raise errors.ShortReadvError(self.relpath,
                    start, length, len(data))
            self._report_activity(length, 'read')
            if last_end is None:
                # This is the first request, just buffer it
                buffered_data = [data]
                buffered_len = length
                input_start = start
            elif start == last_end:
                # The data we are reading fits neatly on the previous
                # buffer, so this is all part of a larger coalesced range.
                buffered_data.append(data)
                buffered_len += length
            else:
                # We have an 'interrupt' in the data stream. So we know we are
                # at a request boundary.
                if buffered_len > 0:
                    # We haven't consumed the buffer so far, so put it into
                    # data_chunks, and continue.
                    buffered = ''.join(buffered_data)
                    data_chunks.append((input_start, buffered))
                input_start = start
                buffered_data = [data]
                buffered_len = length
            last_end = start + length
            if input_start == cur_offset and cur_size <= buffered_len:
                # Simplify the next steps a bit by transforming buffered_data
                # into a single string. We also have the nice property that
                # when there is only one string ''.join([x]) == x, so there is
                # no data copying.
                buffered = ''.join(buffered_data)
                # Clean out buffered data so that we keep memory
                # consumption low
                del buffered_data[:]
                buffered_offset = 0
                # TODO: We *could* also consider the case where cur_offset is in
                #       in the buffered range, even though it doesn't *start*
                #       the buffered range. But for packs we pretty much always
                #       read in order, so you won't get any extra data in the
                #       middle.
                while (input_start == cur_offset
                       and (buffered_offset + cur_size) <= buffered_len):
                    # We've buffered enough data to process this request, spit it
                    # out
                    cur_data = buffered[buffered_offset:buffered_offset + cur_size]
                    # move the direct pointer into our buffered data
                    buffered_offset += cur_size
                    # Move the start-of-buffer pointer
                    input_start += cur_size
                    # Yield the requested data
                    yield cur_offset, cur_data
                    cur_offset, cur_size = offset_iter.next()
                # at this point, we've consumed as much of buffered as we can,
                # so break off the portion that we consumed
                if buffered_offset == len(buffered_data):
                    # No tail to leave behind
                    buffered_data = []
                    buffered_len = 0
                else:
                    buffered = buffered[buffered_offset:]
                    buffered_data = [buffered]
                    buffered_len = len(buffered)
        if buffered_len:
            buffered = ''.join(buffered_data)
            del buffered_data[:]
            data_chunks.append((input_start, buffered))
        if data_chunks:
            if 'sftp' in debug.debug_flags:
                mutter('SFTP readv left with %d out-of-order bytes',
                    sum(map(lambda x: len(x[1]), data_chunks)))
            # We've processed all the readv data, at this point, anything we
            # couldn't process is in data_chunks. This doesn't happen often, so
            # this code path isn't optimized
            # We use an interesting process for data_chunks
            # Specifically if we have "bisect_left([(start, len, entries)],
            #                                       (qstart,)])
            # If start == qstart, then we get the specific node. Otherwise we
            # get the previous node
            while True:
                idx = bisect.bisect_left(data_chunks, (cur_offset,))
                if idx < len(data_chunks) and data_chunks[idx][0] == cur_offset:
                    # The data starts here
                    data = data_chunks[idx][1][:cur_size]
                elif idx > 0:
                    # The data is in a portion of a previous page
                    idx -= 1
                    sub_offset = cur_offset - data_chunks[idx][0]
                    data = data_chunks[idx][1]
                    data = data[sub_offset:sub_offset + cur_size]
                else:
                    # We are missing the page where the data should be found,
                    # something is wrong
                    data = ''
                if len(data) != cur_size:
                    raise AssertionError('We must have miscalulated.'
                        ' We expected %d bytes, but only found %d'
                        % (cur_size, len(data)))
                yield cur_offset, data
                cur_offset, cur_size = offset_iter.next()


class SFTPTransport(ConnectedTransport):
    """Transport implementation for SFTP access."""

    _do_prefetch = _default_do_prefetch
    # TODO: jam 20060717 Conceivably these could be configurable, either
    #       by auto-tuning at run-time, or by a configuration (per host??)
    #       but the performance curve is pretty flat, so just going with
    #       reasonable defaults.
    _max_readv_combine = 200
    # Having to round trip to the server means waiting for a response,
    # so it is better to download extra bytes.
    # 8KiB had good performance for both local and remote network operations
    _bytes_to_read_before_seek = 8192

    # The sftp spec says that implementations SHOULD allow reads
    # to be at least 32K. paramiko.readv() does an async request
    # for the chunks. So we need to keep it within a single request
    # size for paramiko <= 1.6.1. paramiko 1.6.2 will probably chop
    # up the request itself, rather than us having to worry about it
    _max_request_size = 32768

    def __init__(self, base, _from_transport=None):
        super(SFTPTransport, self).__init__(base,
                                            _from_transport=_from_transport)

    def _remote_path(self, relpath):
        """Return the path to be passed along the sftp protocol for relpath.

        :param relpath: is a urlencoded string.
        """
        relative = urlutils.unescape(relpath).encode('utf-8')
        remote_path = self._combine_paths(self._path, relative)
        # the initial slash should be removed from the path, and treated as a
        # homedir relative path (the path begins with a double slash if it is
        # absolute).  see draft-ietf-secsh-scp-sftp-ssh-uri-03.txt
        # RBC 20060118 we are not using this as its too user hostile. instead
        # we are following lftp and using /~/foo to mean '~/foo'
        # vila--20070602 and leave absolute paths begin with a single slash.
        if remote_path.startswith('/~/'):
            remote_path = remote_path[3:]
        elif remote_path == '/~':
            remote_path = ''
        return remote_path

    def _create_connection(self, credentials=None):
        """Create a new connection with the provided credentials.

        :param credentials: The credentials needed to establish the connection.

        :return: The created connection and its associated credentials.

        The credentials are only the password as it may have been entered
        interactively by the user and may be different from the one provided
        in base url at transport creation time.
        """
        if credentials is None:
            password = self._password
        else:
            password = credentials

        vendor = ssh._get_ssh_vendor()
        user = self._user
        if user is None:
            auth = config.AuthenticationConfig()
            user = auth.get_user('ssh', self._host, self._port)
        connection = vendor.connect_sftp(self._user, password,
                                         self._host, self._port)
        return connection, (user, password)

    def _get_sftp(self):
        """Ensures that a connection is established"""
        connection = self._get_connection()
        if connection is None:
            # First connection ever
            connection, credentials = self._create_connection()
            self._set_connection(connection, credentials)
        return connection

    def has(self, relpath):
        """
        Does the target location exist?
        """
        try:
            self._get_sftp().stat(self._remote_path(relpath))
            # stat result is about 20 bytes, let's say
            self._report_activity(20, 'read')
            return True
        except IOError:
            return False

    def get(self, relpath):
        """Get the file at the given relative path.

        :param relpath: The relative path to the file
        """
        try:
            # FIXME: by returning the file directly, we don't pass this
            # through to report_activity.  We could try wrapping the object
            # before it's returned.  For readv and get_bytes it's handled in
            # the higher-level function.
            # -- mbp 20090126
            path = self._remote_path(relpath)
            f = self._get_sftp().file(path, mode='rb')
            if self._do_prefetch and (getattr(f, 'prefetch', None) is not None):
                f.prefetch()
            return f
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, path, ': error retrieving',
                failure_exc=errors.ReadError)

    def get_bytes(self, relpath):
        # reimplement this here so that we can report how many bytes came back
        f = self.get(relpath)
        try:
            bytes = f.read()
            self._report_activity(len(bytes), 'read')
            return bytes
        finally:
            f.close()

    def _readv(self, relpath, offsets):
        """See Transport.readv()"""
        # We overload the default readv() because we want to use a file
        # that does not have prefetch enabled.
        # Also, if we have a new paramiko, it implements an async readv()
        if not offsets:
            return

        try:
            path = self._remote_path(relpath)
            fp = self._get_sftp().file(path, mode='rb')
            readv = getattr(fp, 'readv', None)
            if readv:
                return self._sftp_readv(fp, offsets, relpath)
            if 'sftp' in debug.debug_flags:
                mutter('seek and read %s offsets', len(offsets))
            return self._seek_and_read(fp, offsets, relpath)
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, path, ': error retrieving')

    def recommended_page_size(self):
        """See Transport.recommended_page_size().

        For SFTP we suggest a large page size to reduce the overhead
        introduced by latency.
        """
        return 64 * 1024

    def _sftp_readv(self, fp, offsets, relpath):
        """Use the readv() member of fp to do async readv.

        Then read them using paramiko.readv(). paramiko.readv()
        does not support ranges > 64K, so it caps the request size, and
        just reads until it gets all the stuff it wants.
        """
        helper = _SFTPReadvHelper(offsets, relpath, self._report_activity)
        return helper.request_and_yield_offsets(fp)

    def put_file(self, relpath, f, mode=None):
        """
        Copy the file-like object into the location.

        :param relpath: Location to put the contents, relative to base.
        :param f:       File-like object.
        :param mode: The final mode for the file
        """
        final_path = self._remote_path(relpath)
        return self._put(final_path, f, mode=mode)

    def _put(self, abspath, f, mode=None):
        """Helper function so both put() and copy_abspaths can reuse the code"""
        tmp_abspath = '%s.tmp.%.9f.%d.%d' % (abspath, time.time(),
                        os.getpid(), random.randint(0,0x7FFFFFFF))
        fout = self._sftp_open_exclusive(tmp_abspath, mode=mode)
        closed = False
        try:
            try:
                fout.set_pipelined(True)
                length = self._pump(f, fout)
            except (IOError, paramiko.SSHException), e:
                self._translate_io_exception(e, tmp_abspath)
            # XXX: This doesn't truly help like we would like it to.
            #      The problem is that openssh strips sticky bits. So while we
            #      can properly set group write permission, we lose the group
            #      sticky bit. So it is probably best to stop chmodding, and
            #      just tell users that they need to set the umask correctly.
            #      The attr.st_mode = mode, in _sftp_open_exclusive
            #      will handle when the user wants the final mode to be more
            #      restrictive. And then we avoid a round trip. Unless
            #      paramiko decides to expose an async chmod()

            # This is designed to chmod() right before we close.
            # Because we set_pipelined() earlier, theoretically we might
            # avoid the round trip for fout.close()
            if mode is not None:
                self._get_sftp().chmod(tmp_abspath, mode)
            fout.close()
            closed = True
            self._rename_and_overwrite(tmp_abspath, abspath)
            return length
        except Exception, e:
            # If we fail, try to clean up the temporary file
            # before we throw the exception
            # but don't let another exception mess things up
            # Write out the traceback, because otherwise
            # the catch and throw destroys it
            import traceback
            mutter(traceback.format_exc())
            try:
                if not closed:
                    fout.close()
                self._get_sftp().remove(tmp_abspath)
            except:
                # raise the saved except
                raise e
            # raise the original with its traceback if we can.
            raise

    def _put_non_atomic_helper(self, relpath, writer, mode=None,
                               create_parent_dir=False,
                               dir_mode=None):
        abspath = self._remote_path(relpath)

        # TODO: jam 20060816 paramiko doesn't publicly expose a way to
        #       set the file mode at create time. If it does, use it.
        #       But for now, we just chmod later anyway.

        def _open_and_write_file():
            """Try to open the target file, raise error on failure"""
            fout = None
            try:
                try:
                    fout = self._get_sftp().file(abspath, mode='wb')
                    fout.set_pipelined(True)
                    writer(fout)
                except (paramiko.SSHException, IOError), e:
                    self._translate_io_exception(e, abspath,
                                                 ': unable to open')

                # This is designed to chmod() right before we close.
                # Because we set_pipelined() earlier, theoretically we might
                # avoid the round trip for fout.close()
                if mode is not None:
                    self._get_sftp().chmod(abspath, mode)
            finally:
                if fout is not None:
                    fout.close()

        if not create_parent_dir:
            _open_and_write_file()
            return

        # Try error handling to create the parent directory if we need to
        try:
            _open_and_write_file()
        except NoSuchFile:
            # Try to create the parent directory, and then go back to
            # writing the file
            parent_dir = os.path.dirname(abspath)
            self._mkdir(parent_dir, dir_mode)
            _open_and_write_file()

    def put_file_non_atomic(self, relpath, f, mode=None,
                            create_parent_dir=False,
                            dir_mode=None):
        """Copy the file-like object into the target location.

        This function is not strictly safe to use. It is only meant to
        be used when you already know that the target does not exist.
        It is not safe, because it will open and truncate the remote
        file. So there may be a time when the file has invalid contents.

        :param relpath: The remote location to put the contents.
        :param f:       File-like object.
        :param mode:    Possible access permissions for new file.
                        None means do not set remote permissions.
        :param create_parent_dir: If we cannot create the target file because
                        the parent directory does not exist, go ahead and
                        create it, and then try again.
        """
        def writer(fout):
            self._pump(f, fout)
        self._put_non_atomic_helper(relpath, writer, mode=mode,
                                    create_parent_dir=create_parent_dir,
                                    dir_mode=dir_mode)

    def put_bytes_non_atomic(self, relpath, bytes, mode=None,
                             create_parent_dir=False,
                             dir_mode=None):
        def writer(fout):
            fout.write(bytes)
        self._put_non_atomic_helper(relpath, writer, mode=mode,
                                    create_parent_dir=create_parent_dir,
                                    dir_mode=dir_mode)

    def iter_files_recursive(self):
        """Walk the relative paths of all files in this transport."""
        # progress is handled by list_dir
        queue = list(self.list_dir('.'))
        while queue:
            relpath = queue.pop(0)
            st = self.stat(relpath)
            if stat.S_ISDIR(st.st_mode):
                for i, basename in enumerate(self.list_dir(relpath)):
                    queue.insert(i, relpath+'/'+basename)
            else:
                yield relpath

    def _mkdir(self, abspath, mode=None):
        if mode is None:
            local_mode = 0777
        else:
            local_mode = mode
        try:
            self._report_activity(len(abspath), 'write')
            self._get_sftp().mkdir(abspath, local_mode)
            self._report_activity(1, 'read')
            if mode is not None:
                # chmod a dir through sftp will erase any sgid bit set
                # on the server side.  So, if the bit mode are already
                # set, avoid the chmod.  If the mode is not fine but
                # the sgid bit is set, report a warning to the user
                # with the umask fix.
                stat = self._get_sftp().lstat(abspath)
                mode = mode & 0777 # can't set special bits anyway
                if mode != stat.st_mode & 0777:
                    if stat.st_mode & 06000:
                        warning('About to chmod %s over sftp, which will result'
                                ' in its suid or sgid bits being cleared.  If'
                                ' you want to preserve those bits, change your '
                                ' environment on the server to use umask 0%03o.'
                                % (abspath, 0777 - mode))
                    self._get_sftp().chmod(abspath, mode=mode)
        except (paramiko.SSHException, IOError), e:
            self._translate_io_exception(e, abspath, ': unable to mkdir',
                failure_exc=FileExists)

    def mkdir(self, relpath, mode=None):
        """Create a directory at the given path."""
        self._mkdir(self._remote_path(relpath), mode=mode)

    def open_write_stream(self, relpath, mode=None):
        """See Transport.open_write_stream."""
        # initialise the file to zero-length
        # this is three round trips, but we don't use this
        # api more than once per write_group at the moment so
        # it is a tolerable overhead. Better would be to truncate
        # the file after opening. RBC 20070805
        self.put_bytes_non_atomic(relpath, "", mode)
        abspath = self._remote_path(relpath)
        # TODO: jam 20060816 paramiko doesn't publicly expose a way to
        #       set the file mode at create time. If it does, use it.
        #       But for now, we just chmod later anyway.
        handle = None
        try:
            handle = self._get_sftp().file(abspath, mode='wb')
            handle.set_pipelined(True)
        except (paramiko.SSHException, IOError), e:
            self._translate_io_exception(e, abspath,
                                         ': unable to open')
        _file_streams[self.abspath(relpath)] = handle
        return FileFileStream(self, relpath, handle)

    def _translate_io_exception(self, e, path, more_info='',
                                failure_exc=PathError):
        """Translate a paramiko or IOError into a friendlier exception.

        :param e: The original exception
        :param path: The path in question when the error is raised
        :param more_info: Extra information that can be included,
                          such as what was going on
        :param failure_exc: Paramiko has the super fun ability to raise completely
                           opaque errors that just set "e.args = ('Failure',)" with
                           no more information.
                           If this parameter is set, it defines the exception
                           to raise in these cases.
        """
        # paramiko seems to generate detailless errors.
        self._translate_error(e, path, raise_generic=False)
        if getattr(e, 'args', None) is not None:
            if (e.args == ('No such file or directory',) or
                e.args == ('No such file',)):
                raise NoSuchFile(path, str(e) + more_info)
            if (e.args == ('mkdir failed',) or
                e.args[0].startswith('syserr: File exists')):
                raise FileExists(path, str(e) + more_info)
            # strange but true, for the paramiko server.
            if (e.args == ('Failure',)):
                raise failure_exc(path, str(e) + more_info)
            # Can be something like args = ('Directory not empty:
            # '/srv/bazaar.launchpad.net/blah...: '
            # [Errno 39] Directory not empty',)
            if (e.args[0].startswith('Directory not empty: ')
                or getattr(e, 'errno', None) == errno.ENOTEMPTY):
                raise errors.DirectoryNotEmpty(path, str(e))
            mutter('Raising exception with args %s', e.args)
        if getattr(e, 'errno', None) is not None:
            mutter('Raising exception with errno %s', e.errno)
        raise e

    def append_file(self, relpath, f, mode=None):
        """
        Append the text in the file-like object into the final
        location.
        """
        try:
            path = self._remote_path(relpath)
            fout = self._get_sftp().file(path, 'ab')
            if mode is not None:
                self._get_sftp().chmod(path, mode)
            result = fout.tell()
            self._pump(f, fout)
            return result
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, relpath, ': unable to append')

    def rename(self, rel_from, rel_to):
        """Rename without special overwriting"""
        try:
            self._get_sftp().rename(self._remote_path(rel_from),
                              self._remote_path(rel_to))
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, rel_from,
                    ': unable to rename to %r' % (rel_to))

    def _rename_and_overwrite(self, abs_from, abs_to):
        """Do a fancy rename on the remote server.

        Using the implementation provided by osutils.
        """
        try:
            sftp = self._get_sftp()
            fancy_rename(abs_from, abs_to,
                         rename_func=sftp.rename,
                         unlink_func=sftp.remove)
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, abs_from,
                                         ': unable to rename to %r' % (abs_to))

    def move(self, rel_from, rel_to):
        """Move the item at rel_from to the location at rel_to"""
        path_from = self._remote_path(rel_from)
        path_to = self._remote_path(rel_to)
        self._rename_and_overwrite(path_from, path_to)

    def delete(self, relpath):
        """Delete the item at relpath"""
        path = self._remote_path(relpath)
        try:
            self._get_sftp().remove(path)
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, path, ': unable to delete')

    def external_url(self):
        """See bzrlib.transport.Transport.external_url."""
        # the external path for SFTP is the base
        return self.base

    def listable(self):
        """Return True if this store supports listing."""
        return True

    def list_dir(self, relpath):
        """
        Return a list of all files at the given location.
        """
        # does anything actually use this?
        # -- Unknown
        # This is at least used by copy_tree for remote upgrades.
        # -- David Allouche 2006-08-11
        path = self._remote_path(relpath)
        try:
            entries = self._get_sftp().listdir(path)
            self._report_activity(sum(map(len, entries)), 'read')
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, path, ': failed to list_dir')
        return [urlutils.escape(entry) for entry in entries]

    def rmdir(self, relpath):
        """See Transport.rmdir."""
        path = self._remote_path(relpath)
        try:
            return self._get_sftp().rmdir(path)
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, path, ': failed to rmdir')

    def stat(self, relpath):
        """Return the stat information for a file."""
        path = self._remote_path(relpath)
        try:
            return self._get_sftp().stat(path)
        except (IOError, paramiko.SSHException), e:
            self._translate_io_exception(e, path, ': unable to stat')

    def lock_read(self, relpath):
        """
        Lock the given file for shared (read) access.
        :return: A lock object, which has an unlock() member function
        """
        # FIXME: there should be something clever i can do here...
        class BogusLock(object):
            def __init__(self, path):
                self.path = path
            def unlock(self):
                pass
        return BogusLock(relpath)

    def lock_write(self, relpath):
        """
        Lock the given file for exclusive (write) access.
        WARNING: many transports do not support this, so trying avoid using it

        :return: A lock object, which has an unlock() member function
        """
        # This is a little bit bogus, but basically, we create a file
        # which should not already exist, and if it does, we assume
        # that there is a lock, and if it doesn't, the we assume
        # that we have taken the lock.
        return SFTPLock(relpath, self)

    def _sftp_open_exclusive(self, abspath, mode=None):
        """Open a remote path exclusively.

        SFTP supports O_EXCL (SFTP_FLAG_EXCL), which fails if
        the file already exists. However it does not expose this
        at the higher level of SFTPClient.open(), so we have to
        sneak away with it.

        WARNING: This breaks the SFTPClient abstraction, so it
        could easily break against an updated version of paramiko.

        :param abspath: The remote absolute path where the file should be opened
        :param mode: The mode permissions bits for the new file
        """
        # TODO: jam 20060816 Paramiko >= 1.6.2 (probably earlier) supports
        #       using the 'x' flag to indicate SFTP_FLAG_EXCL.
        #       However, there is no way to set the permission mode at open
        #       time using the sftp_client.file() functionality.
        path = self._get_sftp()._adjust_cwd(abspath)
        # mutter('sftp abspath %s => %s', abspath, path)
        attr = SFTPAttributes()
        if mode is not None:
            attr.st_mode = mode
        omode = (SFTP_FLAG_WRITE | SFTP_FLAG_CREATE
                | SFTP_FLAG_TRUNC | SFTP_FLAG_EXCL)
        try:
            t, msg = self._get_sftp()._request(CMD_OPEN, path, omode, attr)
            if t != CMD_HANDLE:
                raise TransportError('Expected an SFTP handle')
            handle = msg.get_string()
            return SFTPFile(self._get_sftp(), handle, 'wb', -1)
        except (paramiko.SSHException, IOError), e:
            self._translate_io_exception(e, abspath, ': unable to open',
                failure_exc=FileExists)

    def _can_roundtrip_unix_modebits(self):
        if sys.platform == 'win32':
            # anyone else?
            return False
        else:
            return True

# ------------- server test implementation --------------
import threading

from bzrlib.tests.stub_sftp import StubServer, StubSFTPServer

STUB_SERVER_KEY = """
-----BEGIN RSA PRIVATE KEY-----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-----END RSA PRIVATE KEY-----
"""


class SocketListener(threading.Thread):

    def __init__(self, callback):
        threading.Thread.__init__(self)
        self._callback = callback
        self._socket = socket.socket()
        self._socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
        self._socket.bind(('localhost', 0))
        self._socket.listen(1)
        self.port = self._socket.getsockname()[1]
        self._stop_event = threading.Event()

    def stop(self):
        # called from outside this thread
        self._stop_event.set()
        # use a timeout here, because if the test fails, the server thread may
        # never notice the stop_event.
        self.join(5.0)
        self._socket.close()

    def run(self):
        while True:
            readable, writable_unused, exception_unused = \
                select.select([self._socket], [], [], 0.1)
            if self._stop_event.isSet():
                return
            if len(readable) == 0:
                continue
            try:
                s, addr_unused = self._socket.accept()
                # because the loopback socket is inline, and transports are
                # never explicitly closed, best to launch a new thread.
                threading.Thread(target=self._callback, args=(s,)).start()
            except socket.error, x:
                sys.excepthook(*sys.exc_info())
                warning('Socket error during accept() within unit test server'
                        ' thread: %r' % x)
            except Exception, x:
                # probably a failed test; unit test thread will log the
                # failure/error
                sys.excepthook(*sys.exc_info())
                warning('Exception from within unit test server thread: %r' %
                        x)


class SocketDelay(object):
    """A socket decorator to make TCP appear slower.

    This changes recv, send, and sendall to add a fixed latency to each python
    call if a new roundtrip is detected. That is, when a recv is called and the
    flag new_roundtrip is set, latency is charged. Every send and send_all
    sets this flag.

    In addition every send, sendall and recv sleeps a bit per character send to
    simulate bandwidth.

    Not all methods are implemented, this is deliberate as this class is not a
    replacement for the builtin sockets layer. fileno is not implemented to
    prevent the proxy being bypassed.
    """

    simulated_time = 0
    _proxied_arguments = dict.fromkeys([
        "close", "getpeername", "getsockname", "getsockopt", "gettimeout",
        "setblocking", "setsockopt", "settimeout", "shutdown"])

    def __init__(self, sock, latency, bandwidth=1.0,
                 really_sleep=True):
        """
        :param bandwith: simulated bandwith (MegaBit)
        :param really_sleep: If set to false, the SocketDelay will just
        increase a counter, instead of calling time.sleep. This is useful for
        unittesting the SocketDelay.
        """
        self.sock = sock
        self.latency = latency
        self.really_sleep = really_sleep
        self.time_per_byte = 1 / (bandwidth / 8.0 * 1024 * 1024)
        self.new_roundtrip = False

    def sleep(self, s):
        if self.really_sleep:
            time.sleep(s)
        else:
            SocketDelay.simulated_time += s

    def __getattr__(self, attr):
        if attr in SocketDelay._proxied_arguments:
            return getattr(self.sock, attr)
        raise AttributeError("'SocketDelay' object has no attribute %r" %
                             attr)

    def dup(self):
        return SocketDelay(self.sock.dup(), self.latency, self.time_per_byte,
                           self._sleep)

    def recv(self, *args):
        data = self.sock.recv(*args)
        if data and self.new_roundtrip:
            self.new_roundtrip = False
            self.sleep(self.latency)
        self.sleep(len(data) * self.time_per_byte)
        return data

    def sendall(self, data, flags=0):
        if not self.new_roundtrip:
            self.new_roundtrip = True
            self.sleep(self.latency)
        self.sleep(len(data) * self.time_per_byte)
        return self.sock.sendall(data, flags)

    def send(self, data, flags=0):
        if not self.new_roundtrip:
            self.new_roundtrip = True
            self.sleep(self.latency)
        bytes_sent = self.sock.send(data, flags)
        self.sleep(bytes_sent * self.time_per_byte)
        return bytes_sent


class SFTPServer(Server):
    """Common code for SFTP server facilities."""

    def __init__(self, server_interface=StubServer):
        self._original_vendor = None
        self._homedir = None
        self._server_homedir = None
        self._listener = None
        self._root = None
        self._vendor = ssh.ParamikoVendor()
        self._server_interface = server_interface
        # sftp server logs
        self.logs = []
        self.add_latency = 0

    def _get_sftp_url(self, path):
        """Calculate an sftp url to this server for path."""
        return 'sftp://foo:bar@localhost:%d/%s' % (self._listener.port, path)

    def log(self, message):
        """StubServer uses this to log when a new server is created."""
        self.logs.append(message)

    def _run_server_entry(self, sock):
        """Entry point for all implementations of _run_server.

        If self.add_latency is > 0.000001 then sock is given a latency adding
        decorator.
        """
        if self.add_latency > 0.000001:
            sock = SocketDelay(sock, self.add_latency)
        return self._run_server(sock)

    def _run_server(self, s):
        ssh_server = paramiko.Transport(s)
        key_file = pathjoin(self._homedir, 'test_rsa.key')
        f = open(key_file, 'w')
        f.write(STUB_SERVER_KEY)
        f.close()
        host_key = paramiko.RSAKey.from_private_key_file(key_file)
        ssh_server.add_server_key(host_key)
        server = self._server_interface(self)
        ssh_server.set_subsystem_handler('sftp', paramiko.SFTPServer,
                                         StubSFTPServer, root=self._root,
                                         home=self._server_homedir)
        event = threading.Event()
        ssh_server.start_server(event, server)
        event.wait(5.0)

    def setUp(self, backing_server=None):
        # XXX: TODO: make sftpserver back onto backing_server rather than local
        # disk.
        if not (backing_server is None or
                isinstance(backing_server, local.LocalURLServer)):
            raise AssertionError(
                "backing_server should not be %r, because this can only serve the "
                "local current working directory." % (backing_server,))
        self._original_vendor = ssh._ssh_vendor_manager._cached_ssh_vendor
        ssh._ssh_vendor_manager._cached_ssh_vendor = self._vendor
        if sys.platform == 'win32':
            # Win32 needs to use the UNICODE api
            self._homedir = getcwd()
        else:
            # But Linux SFTP servers should just deal in bytestreams
            self._homedir = os.getcwd()
        if self._server_homedir is None:
            self._server_homedir = self._homedir
        self._root = '/'
        if sys.platform == 'win32':
            self._root = ''
        self._listener = SocketListener(self._run_server_entry)
        self._listener.setDaemon(True)
        self._listener.start()

    def tearDown(self):
        """See bzrlib.transport.Server.tearDown."""
        self._listener.stop()
        ssh._ssh_vendor_manager._cached_ssh_vendor = self._original_vendor

    def get_bogus_url(self):
        """See bzrlib.transport.Server.get_bogus_url."""
        # this is chosen to try to prevent trouble with proxies, wierd dns, etc
        # we bind a random socket, so that we get a guaranteed unused port
        # we just never listen on that port
        s = socket.socket()
        s.bind(('localhost', 0))
        return 'sftp://%s:%s/' % s.getsockname()


class SFTPFullAbsoluteServer(SFTPServer):
    """A test server for sftp transports, using absolute urls and ssh."""

    def get_url(self):
        """See bzrlib.transport.Server.get_url."""
        homedir = self._homedir
        if sys.platform != 'win32':
            # Remove the initial '/' on all platforms but win32
            homedir = homedir[1:]
        return self._get_sftp_url(urlutils.escape(homedir))


class SFTPServerWithoutSSH(SFTPServer):
    """An SFTP server that uses a simple TCP socket pair rather than SSH."""

    def __init__(self):
        super(SFTPServerWithoutSSH, self).__init__()
        self._vendor = ssh.LoopbackVendor()

    def _run_server(self, sock):
        # Re-import these as locals, so that they're still accessible during
        # interpreter shutdown (when all module globals get set to None, leading
        # to confusing errors like "'NoneType' object has no attribute 'error'".
        class FakeChannel(object):
            def get_transport(self):
                return self
            def get_log_channel(self):
                return 'paramiko'
            def get_name(self):
                return '1'
            def get_hexdump(self):
                return False
            def close(self):
                pass

        server = paramiko.SFTPServer(
            FakeChannel(), 'sftp', StubServer(self), StubSFTPServer,
            root=self._root, home=self._server_homedir)
        try:
            server.start_subsystem(
                'sftp', None, ssh.SocketAsChannelAdapter(sock))
        except socket.error, e:
            if (len(e.args) > 0) and (e.args[0] == errno.EPIPE):
                # it's okay for the client to disconnect abruptly
                # (bug in paramiko 1.6: it should absorb this exception)
                pass
            else:
                raise
        except Exception, e:
            # This typically seems to happen during interpreter shutdown, so
            # most of the useful ways to report this error are won't work.
            # Writing the exception type, and then the text of the exception,
            # seems to be the best we can do.
            import sys
            sys.stderr.write('\nEXCEPTION %r: ' % (e.__class__,))
            sys.stderr.write('%s\n\n' % (e,))
        server.finish_subsystem()


class SFTPAbsoluteServer(SFTPServerWithoutSSH):
    """A test server for sftp transports, using absolute urls."""

    def get_url(self):
        """See bzrlib.transport.Server.get_url."""
        homedir = self._homedir
        if sys.platform != 'win32':
            # Remove the initial '/' on all platforms but win32
            homedir = homedir[1:]
        return self._get_sftp_url(urlutils.escape(homedir))


class SFTPHomeDirServer(SFTPServerWithoutSSH):
    """A test server for sftp transports, using homedir relative urls."""

    def get_url(self):
        """See bzrlib.transport.Server.get_url."""
        return self._get_sftp_url("~/")


class SFTPSiblingAbsoluteServer(SFTPAbsoluteServer):
    """A test server for sftp transports where only absolute paths will work.

    It does this by serving from a deeply-nested directory that doesn't exist.
    """

    def setUp(self, backing_server=None):
        self._server_homedir = '/dev/noone/runs/tests/here'
        super(SFTPSiblingAbsoluteServer, self).setUp(backing_server)


def get_test_permutations():
    """Return the permutations to be used in testing."""
    return [(SFTPTransport, SFTPAbsoluteServer),
            (SFTPTransport, SFTPHomeDirServer),
            (SFTPTransport, SFTPSiblingAbsoluteServer),
            ]