from allmydata.util import fileutil
from allmydata.interfaces import IDirectoryNode
from allmydata.util import log
-from allmydata.util.fileutil import precondition_abspath, get_pathinfo
-from allmydata.util.assertutil import precondition
+from allmydata.util.fileutil import precondition_abspath, get_pathinfo, ConflictError
+from allmydata.util.assertutil import precondition, _assert
from allmydata.util.deferredutil import HookMixin
from allmydata.util.encodingutil import listdir_filepath, to_filepath, \
extend_filepath, unicode_from_filepath, unicode_segments_from, \
raise
+def is_new_file(pathinfo, db_entry):
+ if db_entry is None:
+ return True
+
+ if not pathinfo.exists and db_entry.size is None:
+ return False
+
+ return ((pathinfo.size, pathinfo.ctime, pathinfo.mtime) !=
+ (db_entry.size, db_entry.ctime, db_entry.mtime))
+
+
class MagicFolder(service.MultiService):
name = 'magic-folder'
- def __init__(self, client, upload_dircap, collective_dircap, local_path_u, dbfile,
- pending_delay=1.0, clock=reactor):
+ def __init__(self, client, upload_dircap, collective_dircap, local_path_u, dbfile, umask,
+ pending_delay=1.0, clock=None):
precondition_abspath(local_path_u)
service.MultiService.__init__(self)
+ immediate = clock is not None
+ clock = clock or reactor
db = magicfolderdb.get_magicfolderdb(dbfile, create_version=(magicfolderdb.SCHEMA_v1, 1))
if db is None:
return Failure(Exception('ERROR: Unable to load magic folder db.'))
self._client = client
self._db = db
- self.is_ready = False
+ upload_dirnode = self._client.create_node_from_uri(upload_dircap)
+ collective_dirnode = self._client.create_node_from_uri(collective_dircap)
- self.uploader = Uploader(client, local_path_u, db, upload_dircap, pending_delay, clock)
- self.downloader = Downloader(client, local_path_u, db, collective_dircap, clock)
+ self.uploader = Uploader(client, local_path_u, db, upload_dirnode, pending_delay, clock, immediate)
+ self.downloader = Downloader(client, local_path_u, db, collective_dirnode,
+ upload_dirnode.get_readonly_uri(), clock, self.uploader.is_pending, umask)
def startService(self):
# TODO: why is this being called more than once?
"""ready is used to signal us to start
processing the upload and download items...
"""
- self.is_ready = True
d = self.uploader.start_scanning()
d2 = self.downloader.start_scanning()
d.addCallback(lambda ign: d2)
self._deque = deque()
self._lazy_tail = defer.succeed(None)
- self._pending = set()
self._stopped = False
self._turn_delay = 0
def _get_filepath(self, relpath_u):
+ self._log("_get_filepath(%r)" % (relpath_u,))
return extend_filepath(self._local_filepath, relpath_u.split(u"/"))
def _get_relpath(self, filepath):
self._log("%s += %r" % (counter_name, delta))
self._client.stats_provider.count(ctr, delta)
+ def _logcb(self, res, msg):
+ self._log("%s: %r" % (msg, res))
+ return res
+
def _log(self, msg):
s = "Magic Folder %s %s: %s" % (quote_output(self._client.nickname), self._name, msg)
self._client.log(s)
print s
#open("events", "ab+").write(msg)
- def _append_to_deque(self, relpath_u):
- self._log("_append_to_deque(%r)" % (relpath_u,))
- if relpath_u in self._pending or magicpath.should_ignore_file(relpath_u):
- return
- self._deque.append(relpath_u)
- self._pending.add(relpath_u)
- self._count('objects_queued')
- if self.is_ready:
- self._clock.callLater(0, self._turn_deque)
-
def _turn_deque(self):
+ self._log("_turn_deque")
if self._stopped:
+ self._log("stopped")
return
try:
item = self._deque.pop()
+ self._log("popped %r" % (item,))
self._count('objects_queued', -1)
except IndexError:
self._log("deque is now empty")
class Uploader(QueueMixin):
- def __init__(self, client, local_path_u, db, upload_dircap, pending_delay, clock):
+ def __init__(self, client, local_path_u, db, upload_dirnode, pending_delay, clock,
+ immediate=False):
QueueMixin.__init__(self, client, local_path_u, db, 'uploader', clock)
self.is_ready = False
+ self._immediate = immediate
- # TODO: allow a path rather than a cap URI.
- self._upload_dirnode = self._client.create_node_from_uri(upload_dircap)
- if not IDirectoryNode.providedBy(self._upload_dirnode):
- raise AssertionError("The URI in 'private/magic_folder_dircap' does not refer to a directory.")
- if self._upload_dirnode.is_unknown() or self._upload_dirnode.is_readonly():
- raise AssertionError("The URI in 'private/magic_folder_dircap' is not a writecap to a directory.")
+ if not IDirectoryNode.providedBy(upload_dirnode):
+ raise AssertionError("The URI in '%s' does not refer to a directory."
+ % os.path.join('private', 'magic_folder_dircap'))
+ if upload_dirnode.is_unknown() or upload_dirnode.is_readonly():
+ raise AssertionError("The URI in '%s' is not a writecap to a directory."
+ % os.path.join('private', 'magic_folder_dircap'))
+ self._upload_dirnode = upload_dirnode
self._inotify = get_inotify_module()
self._notifier = self._inotify.INotify()
+ self._pending = set()
if hasattr(self._notifier, 'set_pending_delay'):
self._notifier.set_pending_delay(pending_delay)
- # We don't watch for IN_CREATE, because that would cause us to read and upload a
- # possibly-incomplete file before the application has closed it. There should always
- # be an IN_CLOSE_WRITE after an IN_CREATE (I think).
- # TODO: what about IN_MOVE_SELF, IN_MOVED_FROM, or IN_UNMOUNT?
+ # TODO: what about IN_MOVE_SELF and IN_UNMOUNT?
#
- self.mask = ( self._inotify.IN_CLOSE_WRITE
+ self.mask = ( self._inotify.IN_CREATE
+ | self._inotify.IN_CLOSE_WRITE
| self._inotify.IN_MOVED_TO
| self._inotify.IN_MOVED_FROM
| self._inotify.IN_DELETE
d = defer.succeed(None)
for child in children:
- assert isinstance(child, unicode), child
+ _assert(isinstance(child, unicode), child=child)
d.addCallback(lambda ign, child=child:
("%s/%s" % (reldir_u, child) if reldir_u else child))
def _add_pending(relpath_u):
return d
+ def is_pending(self, relpath_u):
+ return relpath_u in self._pending
+
def _notify(self, opaque, path, events_mask):
self._log("inotify event %r, %r, %r\n" % (opaque, path, ', '.join(self._inotify.humanReadableMask(events_mask))))
relpath_u = self._get_relpath(path)
- self._append_to_deque(relpath_u)
+
+ # We filter out IN_CREATE events not associated with a directory.
+ # Acting on IN_CREATE for files could cause us to read and upload
+ # a possibly-incomplete file before the application has closed it.
+ # There should always be an IN_CLOSE_WRITE after an IN_CREATE, I think.
+ # It isn't possible to avoid watching for IN_CREATE at all, because
+ # it is the only event notified for a directory creation.
+
+ if ((events_mask & self._inotify.IN_CREATE) != 0 and
+ (events_mask & self._inotify.IN_ISDIR) == 0):
+ self._log("ignoring event for %r (creation of non-directory)\n" % (relpath_u,))
+ return
+ if relpath_u in self._pending:
+ self._log("ignoring event for %r (already pending)" % (relpath_u,))
+ return
+ if magicpath.should_ignore_file(relpath_u):
+ self._log("ignoring event for %r (ignorable path)" % (relpath_u,))
+ return
+
+ self._log("appending %r to deque" % (relpath_u,))
+ self._deque.append(relpath_u)
+ self._pending.add(relpath_u)
+ self._count('objects_queued')
+ if self.is_ready:
+ if self._immediate: # for tests
+ self._turn_deque()
+ else:
+ self._clock.callLater(0, self._turn_deque)
def _when_queue_is_empty(self):
return defer.succeed(None)
def _process(self, relpath_u):
+ # Uploader
self._log("_process(%r)" % (relpath_u,))
if relpath_u is None:
return
precondition(isinstance(relpath_u, unicode), relpath_u)
+ precondition(not relpath_u.endswith(u'/'), relpath_u)
d = defer.succeed(None)
fp = self._get_filepath(relpath_u)
pathinfo = get_pathinfo(unicode_from_filepath(fp))
- self._log("pending = %r, about to remove %r" % (self._pending, relpath_u))
+ self._log("about to remove %r from pending set %r" %
+ (relpath_u, self._pending))
self._pending.remove(relpath_u)
encoded_path_u = magicpath.path2magic(relpath_u)
# FIXME merge this with the 'isfile' case.
self._log("notified object %s disappeared (this is normal)" % quote_filepath(fp))
self._count('objects_disappeared')
- if not self._db.check_file_db_exists(relpath_u):
+
+ db_entry = self._db.get_db_entry(relpath_u)
+ if db_entry is None:
return None
- last_downloaded_timestamp = now
- last_downloaded_uri = self._db.get_last_downloaded_uri(relpath_u)
+ last_downloaded_timestamp = now # is this correct?
- current_version = self._db.get_local_file_version(relpath_u)
- if current_version is None:
- new_version = 0
+ if is_new_file(pathinfo, db_entry):
+ new_version = db_entry.version + 1
else:
- new_version = current_version + 1
+ self._log("Not uploading %r" % (relpath_u,))
+ self._count('objects_not_uploaded')
+ return
metadata = { 'version': new_version,
'deleted': True,
'last_downloaded_timestamp': last_downloaded_timestamp }
- if last_downloaded_uri is not None:
- metadata['last_downloaded_uri'] = last_downloaded_uri
+ if db_entry.last_downloaded_uri is not None:
+ metadata['last_downloaded_uri'] = db_entry.last_downloaded_uri
empty_uploadable = Data("", self._client.convergence)
d2 = self._upload_dirnode.add_file(encoded_path_u, empty_uploadable,
def _add_db_entry(filenode):
filecap = filenode.get_uri()
+ last_downloaded_uri = metadata.get('last_downloaded_uri', None)
self._db.did_upload_version(relpath_u, new_version, filecap,
- last_downloaded_uri, last_downloaded_timestamp, pathinfo)
+ last_downloaded_uri, last_downloaded_timestamp,
+ pathinfo)
self._count('files_uploaded')
d2.addCallback(_add_db_entry)
return d2
self.warn("WARNING: cannot upload symlink %s" % quote_filepath(fp))
return None
elif pathinfo.isdir:
- self._notifier.watch(fp, mask=self.mask, callbacks=[self._notify], recursive=True)
+ if not getattr(self._notifier, 'recursive_includes_new_subdirectories', False):
+ self._notifier.watch(fp, mask=self.mask, callbacks=[self._notify], recursive=True)
+
uploadable = Data("", self._client.convergence)
encoded_path_u += magicpath.path2magic(u"/")
+ self._log("encoded_path_u = %r" % (encoded_path_u,))
upload_d = self._upload_dirnode.add_file(encoded_path_u, uploadable, metadata={"version":0}, overwrite=True)
def _succeeded(ign):
self._log("created subdirectory %r" % (relpath_u,))
upload_d.addCallback(lambda ign: self._scan(relpath_u))
return upload_d
elif pathinfo.isfile:
- last_downloaded_uri = self._db.get_last_downloaded_uri(relpath_u)
+ db_entry = self._db.get_db_entry(relpath_u)
+
last_downloaded_timestamp = now
- current_version = self._db.get_local_file_version(relpath_u)
- if current_version is None:
+ if db_entry is None:
new_version = 0
- elif self._db.is_new_file(pathinfo, relpath_u):
- new_version = current_version + 1
+ elif is_new_file(pathinfo, db_entry):
+ new_version = db_entry.version + 1
else:
+ self._log("Not uploading %r" % (relpath_u,))
+ self._count('objects_not_uploaded')
return None
metadata = { 'version': new_version,
'last_downloaded_timestamp': last_downloaded_timestamp }
- if last_downloaded_uri is not None:
- metadata['last_downloaded_uri'] = last_downloaded_uri
+ if db_entry is not None and db_entry.last_downloaded_uri is not None:
+ metadata['last_downloaded_uri'] = db_entry.last_downloaded_uri
uploadable = FileName(unicode_from_filepath(fp), self._client.convergence)
d2 = self._upload_dirnode.add_file(encoded_path_u, uploadable,
filecap = filenode.get_uri()
last_downloaded_uri = metadata.get('last_downloaded_uri', None)
self._db.did_upload_version(relpath_u, new_version, filecap,
- last_downloaded_uri, last_downloaded_timestamp, pathinfo)
+ last_downloaded_uri, last_downloaded_timestamp,
+ pathinfo)
self._count('files_uploaded')
d2.addCallback(_add_db_entry)
return d2
return res
def _failed(f):
self._count('objects_failed')
- self._log("%r while processing %r" % (f, relpath_u))
+ self._log("%s while processing %r" % (f, relpath_u))
return f
d.addCallbacks(_succeeded, _failed)
return d
return d
-class Downloader(QueueMixin):
- REMOTE_SCAN_INTERVAL = 3 # facilitates tests
+class WriteFileMixin(object):
+ FUDGE_SECONDS = 10.0
- def __init__(self, client, local_path_u, db, collective_dircap, clock):
- QueueMixin.__init__(self, client, local_path_u, db, 'downloader', clock)
+ def _get_conflicted_filename(self, abspath_u):
+ return abspath_u + u".conflict"
+
+ def _write_downloaded_file(self, abspath_u, file_contents, is_conflict=False, now=None):
+ self._log("_write_downloaded_file(%r, <%d bytes>, is_conflict=%r, now=%r)"
+ % (abspath_u, len(file_contents), is_conflict, now))
+
+ # 1. Write a temporary file, say .foo.tmp.
+ # 2. is_conflict determines whether this is an overwrite or a conflict.
+ # 3. Set the mtime of the replacement file to be T seconds before the
+ # current local time.
+ # 4. Perform a file replacement with backup filename foo.backup,
+ # replaced file foo, and replacement file .foo.tmp. If any step of
+ # this operation fails, reclassify as a conflict and stop.
+ #
+ # Returns the path of the destination file.
+
+ precondition_abspath(abspath_u)
+ replacement_path_u = abspath_u + u".tmp" # FIXME more unique
+ backup_path_u = abspath_u + u".backup"
+ if now is None:
+ now = time.time()
+
+ # ensure parent directory exists
+ head, tail = os.path.split(abspath_u)
+
+ old_mask = os.umask(self._umask)
+ try:
+ fileutil.make_dirs(head, (~ self._umask) & 0777)
+ fileutil.write(replacement_path_u, file_contents)
+ finally:
+ os.umask(old_mask)
+
+ os.utime(replacement_path_u, (now, now - self.FUDGE_SECONDS))
+ if is_conflict:
+ print "0x00 ------------ <><> is conflict; calling _rename_conflicted_file... %r %r" % (abspath_u, replacement_path_u)
+ return self._rename_conflicted_file(abspath_u, replacement_path_u)
+ else:
+ try:
+ fileutil.replace_file(abspath_u, replacement_path_u, backup_path_u)
+ return abspath_u
+ except fileutil.ConflictError:
+ return self._rename_conflicted_file(abspath_u, replacement_path_u)
+
+ def _rename_conflicted_file(self, abspath_u, replacement_path_u):
+ self._log("_rename_conflicted_file(%r, %r)" % (abspath_u, replacement_path_u))
+
+ conflict_path_u = self._get_conflicted_filename(abspath_u)
+ print "XXX rename %r %r" % (replacement_path_u, conflict_path_u)
+ if os.path.isfile(replacement_path_u):
+ print "%r exists" % (replacement_path_u,)
+ if os.path.isfile(conflict_path_u):
+ print "%r exists" % (conflict_path_u,)
+
+ fileutil.rename_no_overwrite(replacement_path_u, conflict_path_u)
+ return conflict_path_u
- # TODO: allow a path rather than a cap URI.
- self._collective_dirnode = self._client.create_node_from_uri(collective_dircap)
+ def _rename_deleted_file(self, abspath_u):
+ self._log('renaming deleted file to backup: %s' % (abspath_u,))
+ try:
+ fileutil.rename_no_overwrite(abspath_u, abspath_u + u'.backup')
+ except OSError:
+ self._log("Already gone: '%s'" % (abspath_u,))
+ return abspath_u
- if not IDirectoryNode.providedBy(self._collective_dirnode):
- raise AssertionError("The URI in 'private/collective_dircap' does not refer to a directory.")
- if self._collective_dirnode.is_unknown() or not self._collective_dirnode.is_readonly():
- raise AssertionError("The URI in 'private/collective_dircap' is not a readonly cap to a directory.")
+class Downloader(QueueMixin, WriteFileMixin):
+ REMOTE_SCAN_INTERVAL = 3 # facilitates tests
+
+ def __init__(self, client, local_path_u, db, collective_dirnode,
+ upload_readonly_dircap, clock, is_upload_pending, umask):
+ QueueMixin.__init__(self, client, local_path_u, db, 'downloader', clock)
+
+ if not IDirectoryNode.providedBy(collective_dirnode):
+ raise AssertionError("The URI in '%s' does not refer to a directory."
+ % os.path.join('private', 'collective_dircap'))
+ if collective_dirnode.is_unknown() or not collective_dirnode.is_readonly():
+ raise AssertionError("The URI in '%s' is not a readonly cap to a directory."
+ % os.path.join('private', 'collective_dircap'))
+
+ self._collective_dirnode = collective_dirnode
+ self._upload_readonly_dircap = upload_readonly_dircap
+ self._is_upload_pending = is_upload_pending
+ self._umask = umask
self._turn_delay = self.REMOTE_SCAN_INTERVAL
- self._download_scan_batch = {} # path -> [(filenode, metadata)]
def start_scanning(self):
self._log("start_scanning")
files = self._db.get_all_relpaths()
self._log("all files %s" % files)
- d = self._scan_remote_collective()
+ d = self._scan_remote_collective(scan_self=True)
+ d.addBoth(self._logcb, "after _scan_remote_collective 0")
self._turn_deque()
return d
We check the remote metadata version against our magic-folder db version number;
latest version wins.
"""
+ self._log("_should_download(%r, %r)" % (relpath_u, remote_version))
if magicpath.should_ignore_file(relpath_u):
+ self._log("nope")
return False
- v = self._db.get_local_file_version(relpath_u)
- return (v is None or v < remote_version)
+ self._log("yep")
+ db_entry = self._db.get_db_entry(relpath_u)
+ if db_entry is None:
+ return True
+ self._log("version %r" % (db_entry.version,))
+ return (db_entry.version < remote_version)
def _get_local_latest(self, relpath_u):
"""
"""
if not self._get_filepath(relpath_u).exists():
return None
- return self._db.get_local_file_version(relpath_u)
+ db_entry = self._db.get_db_entry(relpath_u)
+ return None if db_entry is None else db_entry.version
def _get_collective_latest_file(self, filename):
"""
collective_dirmap_d.addCallback(highest_version)
return collective_dirmap_d
- def _append_to_batch(self, name, file_node, metadata):
- if self._download_scan_batch.has_key(name):
- self._download_scan_batch[name] += [(file_node, metadata)]
- else:
- self._download_scan_batch[name] = [(file_node, metadata)]
-
- def _scan_remote(self, nickname, dirnode):
- self._log("_scan_remote nickname %r" % (nickname,))
+ def _scan_remote_dmd(self, nickname, dirnode, scan_batch):
+ self._log("_scan_remote_dmd nickname %r" % (nickname,))
d = dirnode.list()
def scan_listing(listing_map):
- for name in listing_map.keys():
- file_node, metadata = listing_map[name]
- local_version = self._get_local_latest(name)
+ for encoded_relpath_u in listing_map.keys():
+ relpath_u = magicpath.magic2path(encoded_relpath_u)
+ self._log("found %r" % (relpath_u,))
+
+ file_node, metadata = listing_map[encoded_relpath_u]
+ local_version = self._get_local_latest(relpath_u)
remote_version = metadata.get('version', None)
- self._log("%r has local version %r, remote version %r" % (name, local_version, remote_version))
+ self._log("%r has local version %r, remote version %r" % (relpath_u, local_version, remote_version))
+
if local_version is None or remote_version is None or local_version < remote_version:
- self._log("added to download queue\n")
- self._append_to_batch(name, file_node, metadata)
+ self._log("%r added to download queue" % (relpath_u,))
+ if scan_batch.has_key(relpath_u):
+ scan_batch[relpath_u] += [(file_node, metadata)]
+ else:
+ scan_batch[relpath_u] = [(file_node, metadata)]
+
d.addCallback(scan_listing)
+ d.addBoth(self._logcb, "end of _scan_remote_dmd")
return d
- def _scan_remote_collective(self):
+ def _scan_remote_collective(self, scan_self=False):
self._log("_scan_remote_collective")
- self._download_scan_batch = {} # XXX
+ scan_batch = {} # path -> [(filenode, metadata)]
+
+ d = self._collective_dirnode.list()
+ def scan_collective(dirmap):
+ d2 = defer.succeed(None)
+ for dir_name in dirmap:
+ (dirnode, metadata) = dirmap[dir_name]
+ if scan_self or dirnode.get_readonly_uri() != self._upload_readonly_dircap:
+ d2.addCallback(lambda ign, dir_name=dir_name, dirnode=dirnode:
+ self._scan_remote_dmd(dir_name, dirnode, scan_batch))
+ def _err(f, dir_name=dir_name):
+ self._log("failed to scan DMD for client %r: %s" % (dir_name, f))
+ # XXX what should we do to make this failure more visible to users?
+ d2.addErrback(_err)
- if self._collective_dirnode is None:
- return
- collective_dirmap_d = self._collective_dirnode.list()
- def do_list(result):
- others = [x for x in result.keys()]
- return result, others
- collective_dirmap_d.addCallback(do_list)
- def scan_collective(result):
- d = defer.succeed(None)
- collective_dirmap, others_list = result
- for dir_name in others_list:
- d.addCallback(lambda x, dir_name=dir_name: self._scan_remote(dir_name, collective_dirmap[dir_name][0]))
- # XXX todo add errback
- return d
- collective_dirmap_d.addCallback(scan_collective)
- collective_dirmap_d.addCallback(self._filter_scan_batch)
- collective_dirmap_d.addCallback(self._add_batch_to_download_queue)
- return collective_dirmap_d
+ return d2
+ d.addCallback(scan_collective)
+
+ def _filter_batch_to_deque(ign):
+ self._log("deque = %r, scan_batch = %r" % (self._deque, scan_batch))
+ for relpath_u in scan_batch.keys():
+ file_node, metadata = max(scan_batch[relpath_u], key=lambda x: x[1]['version'])
- def _add_batch_to_download_queue(self, result):
- self._log("result = %r" % (result,))
- self._log("deque = %r" % (self._deque,))
- self._deque.extend(result)
- self._log("deque after = %r" % (self._deque,))
- self._count('objects_queued', len(result))
- self._log("pending = %r" % (self._pending,))
- self._pending.update(map(lambda x: x[0], result))
- self._log("pending after = %r" % (self._pending,))
-
- def _filter_scan_batch(self, result):
- extension = [] # consider whether this should be a dict
- for relpath_u in self._download_scan_batch.keys():
- if relpath_u in self._pending:
- continue
- file_node, metadata = max(self._download_scan_batch[relpath_u], key=lambda x: x[1]['version'])
- if self._should_download(relpath_u, metadata['version']):
- extension += [(relpath_u, file_node, metadata)]
- return extension
+ if self._should_download(relpath_u, metadata['version']):
+ self._deque.append( (relpath_u, file_node, metadata) )
+ else:
+ self._log("Excluding %r" % (relpath_u,))
+ self._call_hook(None, 'processed')
+
+ self._log("deque after = %r" % (self._deque,))
+ d.addCallback(_filter_batch_to_deque)
+ return d
def _when_queue_is_empty(self):
d = task.deferLater(self._clock, self._turn_delay, self._scan_remote_collective)
+ d.addBoth(self._logcb, "after _scan_remote_collective 1")
d.addCallback(lambda ign: self._turn_deque())
return d
def _process(self, item, now=None):
+ # Downloader
self._log("_process(%r)" % (item,))
if now is None:
now = time.time()
(relpath_u, file_node, metadata) = item
- d = file_node.download_best_version()
- def succeeded(res):
- fp = self._get_filepath(relpath_u)
- abspath_u = unicode_from_filepath(fp)
- d2 = defer.succeed(res)
- d2.addCallback(lambda result: self._write_downloaded_file(abspath_u, result, is_conflict=False))
- def do_update_db(written_abspath_u):
- filecap = file_node.get_uri()
- last_uploaded_uri = metadata.get('last_uploaded_uri', None)
- last_downloaded_uri = filecap
- last_downloaded_timestamp = now
- written_pathinfo = get_pathinfo(written_abspath_u)
- if not written_pathinfo.exists:
- raise Exception("downloaded file %s disappeared" % quote_local_unicode_path(written_abspath_u))
-
- self._db.did_upload_version(relpath_u, metadata['version'], last_uploaded_uri,
- last_downloaded_uri, last_downloaded_timestamp, written_pathinfo)
- d2.addCallback(do_update_db)
- # XXX handle failure here with addErrback...
- self._count('objects_downloaded')
- return d2
- def failed(f):
- self._log("download failed: %s" % (str(f),))
- self._count('objects_download_failed')
- return f
- d.addCallbacks(succeeded, failed)
- def remove_from_pending(res):
- self._pending.remove(relpath_u)
- return res
- d.addBoth(remove_from_pending)
- return d
+ fp = self._get_filepath(relpath_u)
+ abspath_u = unicode_from_filepath(fp)
+ conflict_path_u = self._get_conflicted_filename(abspath_u)
- FUDGE_SECONDS = 10.0
+ d = defer.succeed(None)
- @classmethod
- def _write_downloaded_file(cls, abspath_u, file_contents, is_conflict=False, now=None):
- # 1. Write a temporary file, say .foo.tmp.
- # 2. is_conflict determines whether this is an overwrite or a conflict.
- # 3. Set the mtime of the replacement file to be T seconds before the
- # current local time.
- # 4. Perform a file replacement with backup filename foo.backup,
- # replaced file foo, and replacement file .foo.tmp. If any step of
- # this operation fails, reclassify as a conflict and stop.
- #
- # Returns the path of the destination file.
+ def do_update_db(written_abspath_u):
+ filecap = file_node.get_uri()
+ last_uploaded_uri = metadata.get('last_uploaded_uri', None)
+ last_downloaded_uri = filecap
+ last_downloaded_timestamp = now
+ written_pathinfo = get_pathinfo(written_abspath_u)
- precondition_abspath(abspath_u)
- replacement_path_u = abspath_u + u".tmp" # FIXME more unique
- backup_path_u = abspath_u + u".backup"
- if now is None:
- now = time.time()
+ if not written_pathinfo.exists and not metadata.get('deleted', False):
+ raise Exception("downloaded object %s disappeared" % quote_local_unicode_path(written_abspath_u))
- # ensure parent directory exists
- head, tail = os.path.split(abspath_u)
- mode = 0777 # XXX
- fileutil.make_dirs(head, mode)
+ self._db.did_upload_version(relpath_u, metadata['version'], last_uploaded_uri,
+ last_downloaded_uri, last_downloaded_timestamp, written_pathinfo)
+ self._count('objects_downloaded')
+ def failed(f):
+ self._log("download failed: %s" % (str(f),))
+ self._count('objects_failed')
+ return f
- fileutil.write(replacement_path_u, file_contents)
- os.utime(replacement_path_u, (now, now - cls.FUDGE_SECONDS))
- if is_conflict:
- return cls._rename_conflicted_file(abspath_u, replacement_path_u)
+ if os.path.isfile(conflict_path_u):
+ def fail(res):
+ raise ConflictError("download failed: already conflicted: %r" % (relpath_u,))
+ d.addCallback(fail)
else:
- try:
- fileutil.replace_file(abspath_u, replacement_path_u, backup_path_u)
- return abspath_u
- except fileutil.ConflictError:
- return cls._rename_conflicted_file(abspath_u, replacement_path_u)
+ is_conflict = False
+ db_entry = self._db.get_db_entry(relpath_u)
+ dmd_last_downloaded_uri = metadata.get('last_downloaded_uri', None)
+ dmd_last_uploaded_uri = metadata.get('last_uploaded_uri', None)
+ if db_entry:
+ if dmd_last_downloaded_uri is not None and db_entry.last_downloaded_uri is not None:
+ if dmd_last_downloaded_uri != db_entry.last_downloaded_uri:
+ is_conflict = True
+ self._count('objects_conflicted')
+ elif dmd_last_uploaded_uri is not None and dmd_last_uploaded_uri != db_entry.last_uploaded_uri:
+ is_conflict = True
+ self._count('objects_conflicted')
+ elif self._is_upload_pending(relpath_u):
+ is_conflict = True
+ self._count('objects_conflicted')
+
+ if relpath_u.endswith(u"/"):
+ if metadata.get('deleted', False):
+ self._log("rmdir(%r) ignored" % (abspath_u,))
+ else:
+ self._log("mkdir(%r)" % (abspath_u,))
+ d.addCallback(lambda ign: fileutil.make_dirs(abspath_u))
+ d.addCallback(lambda ign: abspath_u)
+ else:
+ if metadata.get('deleted', False):
+ d.addCallback(lambda ign: self._rename_deleted_file(abspath_u))
+ else:
+ d.addCallback(lambda ign: file_node.download_best_version())
+ d.addCallback(lambda contents: self._write_downloaded_file(abspath_u, contents,
+ is_conflict=is_conflict))
- @classmethod
- def _rename_conflicted_file(self, abspath_u, replacement_path_u):
- conflict_path_u = abspath_u + u".conflict"
- fileutil.rename_no_overwrite(replacement_path_u, conflict_path_u)
- return conflict_path_u
+ d.addCallbacks(do_update_db, failed)
+
+ def trap_conflicts(f):
+ f.trap(ConflictError)
+ return None
+ d.addErrback(trap_conflicts)
+ return d