/brz/remove-bazaar

To get this branch, use:
bzr branch http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2008-05-01 07:03:04 UTC
  • mfrom: (3388.1.3 deprecated-in)
  • Revision ID: pqm@pqm.ubuntu.com-20080501070304-s4tdgkmt0ghea1i1
(mbp) Add symbol_versioning.deprecated_in

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
38
46
    debug,
 
47
    delta,
39
48
    dirstate,
40
49
    errors,
41
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
55
    revision as _mod_revision,
44
56
    revisiontree,
 
57
    textui,
45
58
    trace,
46
59
    transform,
47
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
48
63
    )
49
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
50
66
import bzrlib.ui
51
67
""")
52
68
 
 
69
from bzrlib import symbol_versioning
53
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
56
 
from bzrlib.lock import LogicalLockResult
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
57
75
from bzrlib.mutabletree import needs_tree_write_lock
58
76
from bzrlib.osutils import (
59
77
    file_kind,
60
78
    isdir,
 
79
    normpath,
61
80
    pathjoin,
 
81
    rand_chars,
62
82
    realpath,
63
83
    safe_unicode,
 
84
    splitpath,
64
85
    )
65
 
from bzrlib.trace import mutter
 
86
from bzrlib.trace import mutter, note
66
87
from bzrlib.transport.local import LocalTransport
67
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
68
97
from bzrlib.tree import Tree
69
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
70
99
 
71
100
 
72
 
class DirStateWorkingTree(WorkingTree3):
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
73
120
    def __init__(self, basedir,
74
121
                 branch,
75
122
                 _control_files=None,
84
131
        """
85
132
        self._format = _format
86
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
87
136
        basedir = safe_unicode(basedir)
88
137
        mutter("opening working tree %r", basedir)
89
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
90
141
        self.basedir = realpath(basedir)
91
142
        # if branch is at our basedir and is a format 6 or less
92
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
93
146
        self._control_files = _control_files
94
 
        self._transport = self._control_files._transport
95
147
        self._dirty = None
96
148
        #-------------
97
149
        # during a read or write lock these objects are set, and are
101
153
        #-------------
102
154
        self._setup_directory_is_tree_reference()
103
155
        self._detect_case_handling()
104
 
        self._rules_searcher = None
105
 
        self.views = self._make_views()
106
 
        #--- allow tests to select the dirstate iter_changes implementation
107
 
        self._iter_changes = dirstate._process_entry
108
156
 
109
157
    @needs_tree_write_lock
110
158
    def _add(self, files, ids, kinds):
138
186
    @needs_tree_write_lock
139
187
    def add_reference(self, sub_tree):
140
188
        # use standard implementation, which calls back to self._add
141
 
        #
 
189
        # 
142
190
        # So we don't store the reference_revision in the working dirstate,
143
 
        # it's just recorded at the moment of commit.
 
191
        # it's just recorded at the moment of commit. 
144
192
        self._add_reference(sub_tree)
145
193
 
146
194
    def break_lock(self):
185
233
            WorkingTree3._comparison_data(self, entry, path)
186
234
        # it looks like a plain directory, but it's really a reference -- see
187
235
        # also kind()
188
 
        if (self._repo_supports_tree_reference and kind == 'directory'
189
 
            and entry is not None and entry.kind == 'tree-reference'):
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
190
239
            kind = 'tree-reference'
191
240
        return kind, executable, stat_value
192
241
 
218
267
            return self._dirstate
219
268
        local_path = self.bzrdir.get_workingtree_transport(None
220
269
            ).local_abspath('dirstate')
221
 
        self._dirstate = dirstate.DirState.on_file(local_path,
222
 
            self._sha1_provider())
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
223
271
        return self._dirstate
224
272
 
225
 
    def _sha1_provider(self):
226
 
        """A function that returns a SHA1Provider suitable for this tree.
227
 
 
228
 
        :return: None if content filtering is not supported by this tree.
229
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
230
 
          form of files, i.e. after read filters are applied.
231
 
        """
232
 
        if self.supports_content_filtering():
233
 
            return ContentFilterAwareSHA1Provider(self)
234
 
        else:
235
 
            return None
236
 
 
237
273
    def filter_unversioned_files(self, paths):
238
274
        """Filter out paths that are versioned.
239
275
 
271
307
 
272
308
    def _generate_inventory(self):
273
309
        """Create and set self.inventory from the dirstate object.
274
 
 
 
310
        
275
311
        This is relatively expensive: we have to walk the entire dirstate.
276
312
        Ideally we would not, and can deprecate this function.
277
313
        """
281
317
        state._read_dirblocks_if_needed()
282
318
        root_key, current_entry = self._get_entry(path='')
283
319
        current_id = root_key[2]
284
 
        if not (current_entry[0][0] == 'd'): # directory
285
 
            raise AssertionError(current_entry)
 
320
        assert current_entry[0][0] == 'd' # directory
286
321
        inv = Inventory(root_id=current_id)
287
322
        # Turn some things into local variables
288
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
321
356
                    # add this entry to the parent map.
322
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
323
358
                elif kind == 'tree-reference':
324
 
                    if not self._repo_supports_tree_reference:
325
 
                        raise errors.UnsupportedOperation(
326
 
                            self._generate_inventory,
327
 
                            self.branch.repository)
 
359
                    assert self._repo_supports_tree_reference, \
 
360
                        "repository of %r " \
 
361
                        "doesn't support tree references " \
 
362
                        "required by entry %r" \
 
363
                        % (self, name)
328
364
                    inv_entry.reference_revision = link_or_sha1 or None
329
365
                elif kind != 'symlink':
330
366
                    raise AssertionError("unknown kind %r" % kind)
331
367
                # These checks cost us around 40ms on a 55k entry tree
332
 
                if file_id in inv_byid:
333
 
                    raise AssertionError('file_id %s already in'
334
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
335
 
                if name_unicode in parent_ie.children:
336
 
                    raise AssertionError('name %r already in parent'
337
 
                        % (name_unicode,))
 
368
                assert file_id not in inv_byid, ('file_id %s already in'
 
369
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
370
                assert name_unicode not in parent_ie.children
338
371
                inv_byid[file_id] = inv_entry
339
372
                parent_ie.children[name_unicode] = inv_entry
340
373
        self._inventory = inv
345
378
        If either file_id or path is supplied, it is used as the key to lookup.
346
379
        If both are supplied, the fastest lookup is used, and an error is
347
380
        raised if they do not both point at the same row.
348
 
 
 
381
        
349
382
        :param file_id: An optional unicode file_id to be looked up.
350
383
        :param path: An optional unicode path to be looked up.
351
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
375
408
                    return None
376
409
                else:
377
410
                    raise
378
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
379
 
            stat_value=stat_value)
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
380
413
        if entry[1][0][0] == 'f':
381
 
            if link_or_sha1 is None:
382
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
383
 
                try:
384
 
                    sha1 = osutils.sha_file(file_obj)
385
 
                finally:
386
 
                    file_obj.close()
387
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
388
 
                return sha1
389
 
            else:
390
 
                return link_or_sha1
 
414
            return link_or_sha1
391
415
        return None
392
416
 
393
417
    def _get_inventory(self):
407
431
    @needs_read_lock
408
432
    def get_parent_ids(self):
409
433
        """See Tree.get_parent_ids.
410
 
 
 
434
        
411
435
        This implementation requests the ids list from the dirstate file.
412
436
        """
413
437
        return self.current_dirstate().get_parent_ids()
435
459
        return osutils.lexists(pathjoin(
436
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
437
461
 
438
 
    def has_or_had_id(self, file_id):
439
 
        state = self.current_dirstate()
440
 
        row, parents = self._get_entry(file_id=file_id)
441
 
        return row is not None
442
 
 
443
462
    @needs_read_lock
444
463
    def id2path(self, file_id):
445
464
        "Convert a file-id to a path."
509
528
        return iter(result)
510
529
 
511
530
    def iter_references(self):
512
 
        if not self._repo_supports_tree_reference:
513
 
            # When the repo doesn't support references, we will have nothing to
514
 
            # return
515
 
            return
516
531
        for key, tree_details in self.current_dirstate()._iter_entries():
517
532
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
518
533
                # not relevant to the working tree
520
535
            if not key[1]:
521
536
                # the root is not a reference.
522
537
                continue
523
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
538
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
524
539
            try:
525
 
                if self._kind(relpath) == 'tree-reference':
526
 
                    yield relpath, key[2]
 
540
                if self._kind(path) == 'tree-reference':
 
541
                    yield path, key[2]
527
542
            except errors.NoSuchFile:
528
543
                # path is missing on disk.
529
544
                continue
530
545
 
531
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
532
 
        """See MutableTree._observed_sha1."""
533
 
        state = self.current_dirstate()
534
 
        entry = self._get_entry(file_id=file_id, path=path)
535
 
        state._observed_sha1(entry, sha1, statvalue)
536
 
 
537
546
    def kind(self, file_id):
538
547
        """Return the kind of a file.
539
548
 
543
552
        Note: The caller is expected to take a read-lock before calling this.
544
553
        """
545
554
        relpath = self.id2path(file_id)
546
 
        if relpath is None:
547
 
            raise AssertionError(
548
 
                "path for id {%s} is None!" % file_id)
 
555
        assert relpath != None, \
 
556
            "path for id {%s} is None!" % file_id
549
557
        return self._kind(relpath)
550
558
 
551
559
    def _kind(self, relpath):
552
560
        abspath = self.abspath(relpath)
553
561
        kind = file_kind(abspath)
554
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
555
 
            entry = self._get_entry(path=relpath)
556
 
            if entry[1] is not None:
557
 
                if entry[1][0][0] == 't':
558
 
                    kind = 'tree-reference'
 
562
        if (self._repo_supports_tree_reference and
 
563
            kind == 'directory' and
 
564
            self._directory_is_tree_reference(relpath)):
 
565
            kind = 'tree-reference'
559
566
        return kind
560
567
 
561
568
    @needs_read_lock
568
575
            return _mod_revision.NULL_REVISION
569
576
 
570
577
    def lock_read(self):
571
 
        """See Branch.lock_read, and WorkingTree.unlock.
572
 
 
573
 
        :return: A bzrlib.lock.LogicalLockResult.
574
 
        """
 
578
        """See Branch.lock_read, and WorkingTree.unlock."""
575
579
        self.branch.lock_read()
576
580
        try:
577
581
            self._control_files.lock_read()
590
594
        except:
591
595
            self.branch.unlock()
592
596
            raise
593
 
        return LogicalLockResult(self.unlock)
594
597
 
595
598
    def _lock_self_write(self):
596
599
        """This should be called after the branch is locked."""
611
614
        except:
612
615
            self.branch.unlock()
613
616
            raise
614
 
        return LogicalLockResult(self.unlock)
615
617
 
616
618
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: A bzrlib.lock.LogicalLockResult.
620
 
        """
 
619
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
621
620
        self.branch.lock_read()
622
 
        return self._lock_self_write()
 
621
        self._lock_self_write()
623
622
 
624
623
    def lock_write(self):
625
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
626
 
 
627
 
        :return: A bzrlib.lock.LogicalLockResult.
628
 
        """
 
624
        """See MutableTree.lock_write, and WorkingTree.unlock."""
629
625
        self.branch.lock_write()
630
 
        return self._lock_self_write()
 
626
        self._lock_self_write()
631
627
 
632
628
    @needs_tree_write_lock
633
629
    def move(self, from_paths, to_dir, after=False):
635
631
        result = []
636
632
        if not from_paths:
637
633
            return result
 
634
 
638
635
        state = self.current_dirstate()
639
 
        if isinstance(from_paths, basestring):
640
 
            raise ValueError()
 
636
 
 
637
        assert not isinstance(from_paths, basestring)
641
638
        to_dir_utf8 = to_dir.encode('utf8')
642
639
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
643
640
        id_index = state._get_id_index()
703
700
            from_entry = self._get_entry(path=from_rel)
704
701
            if from_entry == (None, None):
705
702
                raise errors.BzrMoveFailedError(from_rel,to_dir,
706
 
                    errors.NotVersionedError(path=from_rel))
 
703
                    errors.NotVersionedError(path=str(from_rel)))
707
704
 
708
705
            from_id = from_entry[0][2]
709
706
            to_rel = pathjoin(to_dir, from_tail)
798
795
                if minikind == 'd':
799
796
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
800
797
                        """Recursively update all entries in this dirblock."""
801
 
                        if from_dir == '':
802
 
                            raise AssertionError("renaming root not supported")
 
798
                        assert from_dir != '', "renaming root not supported"
803
799
                        from_key = (from_dir, '')
804
800
                        from_block_idx, present = \
805
801
                            state._find_block_index_from_key(from_key)
818
814
 
819
815
                        # Grab a copy since move_one may update the list.
820
816
                        for entry in from_block[1][:]:
821
 
                            if not (entry[0][0] == from_dir):
822
 
                                raise AssertionError()
 
817
                            assert entry[0][0] == from_dir
823
818
                            cur_details = entry[1][0]
824
819
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
825
820
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
884
879
        for tree in trees:
885
880
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
886
881
                parents):
887
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
888
 
                    trees, require_versioned)
 
882
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
889
883
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
890
884
        # -- make all paths utf8 --
891
885
        paths_utf8 = set()
952
946
                raise errors.PathsNotVersionedError(paths)
953
947
        # -- remove redundancy in supplied paths to prevent over-scanning --
954
948
        search_paths = osutils.minimum_path_selection(paths)
955
 
        # sketch:
 
949
        # sketch: 
956
950
        # for all search_indexs in each path at or under each element of
957
951
        # search_paths, if the detail is relocated: add the id, and add the
958
952
        # relocated path as one to search if its not searched already. If the
1014
1008
 
1015
1009
    def read_working_inventory(self):
1016
1010
        """Read the working inventory.
1017
 
 
 
1011
        
1018
1012
        This is a meaningless operation for dirstate, but we obey it anyhow.
1019
1013
        """
1020
1014
        return self.inventory
1038
1032
    def set_last_revision(self, new_revision):
1039
1033
        """Change the last revision in the working tree."""
1040
1034
        parents = self.get_parent_ids()
1041
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1042
 
            if len(parents) >= 2:
1043
 
                raise AssertionError(
1044
 
                    "setting the last parent to none with a pending merge is "
1045
 
                    "unsupported.")
 
1035
        if new_revision in (NULL_REVISION, None):
 
1036
            assert len(parents) < 2, (
 
1037
                "setting the last parent to none with a pending merge is "
 
1038
                "unsupported.")
1046
1039
            self.set_parent_ids([])
1047
1040
        else:
1048
1041
            self.set_parent_ids([new_revision] + parents[1:],
1051
1044
    @needs_tree_write_lock
1052
1045
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1053
1046
        """Set the parent ids to revision_ids.
1054
 
 
 
1047
        
1055
1048
        See also set_parent_trees. This api will try to retrieve the tree data
1056
1049
        for each element of revision_ids from the trees repository. If you have
1057
1050
        tree data already available, it is more efficient to use
1089
1082
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1090
1083
        real_trees = []
1091
1084
        ghosts = []
1092
 
 
1093
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1094
 
        graph = self.branch.repository.get_graph()
1095
 
        heads = graph.heads(parent_ids)
1096
 
        accepted_revisions = set()
1097
 
 
1098
1085
        # convert absent trees to the null tree, which we convert back to
1099
1086
        # missing on access.
1100
1087
        for rev_id, tree in parents_list:
1101
 
            if len(accepted_revisions) > 0:
1102
 
                # we always accept the first tree
1103
 
                if rev_id in accepted_revisions or rev_id not in heads:
1104
 
                    # We have already included either this tree, or its
1105
 
                    # descendent, so we skip it.
1106
 
                    continue
1107
1088
            _mod_revision.check_not_reserved_id(rev_id)
1108
1089
            if tree is not None:
1109
1090
                real_trees.append((rev_id, tree))
1110
1091
            else:
1111
1092
                real_trees.append((rev_id,
1112
 
                    self.branch.repository.revision_tree(
1113
 
                        _mod_revision.NULL_REVISION)))
 
1093
                    self.branch.repository.revision_tree(None)))
1114
1094
                ghosts.append(rev_id)
1115
 
            accepted_revisions.add(rev_id)
1116
1095
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1117
1096
        self._make_dirty(reset_inventory=False)
1118
1097
 
1211
1190
                # just forget the whole block.
1212
1191
                entry_index = 0
1213
1192
                while entry_index < len(block[1]):
 
1193
                    # Mark this file id as having been removed
1214
1194
                    entry = block[1][entry_index]
1215
 
                    if entry[1][0][0] in 'ar':
1216
 
                        # don't remove absent or renamed entries
 
1195
                    ids_to_unversion.discard(entry[0][2])
 
1196
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1197
                                               # entries
 
1198
                        or not state._make_absent(entry)):
1217
1199
                        entry_index += 1
1218
 
                    else:
1219
 
                        # Mark this file id as having been removed
1220
 
                        ids_to_unversion.discard(entry[0][2])
1221
 
                        if not state._make_absent(entry):
1222
 
                            # The block has not shrunk.
1223
 
                            entry_index += 1
1224
1200
                # go to the next block. (At the moment we dont delete empty
1225
1201
                # dirblocks)
1226
1202
                block_index += 1
1264
1240
 
1265
1241
    def update_basis_by_delta(self, new_revid, delta):
1266
1242
        """See MutableTree.update_basis_by_delta."""
1267
 
        if self.last_revision() == new_revid:
1268
 
            raise AssertionError()
 
1243
        assert self.last_revision() != new_revid
1269
1244
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1270
1245
 
1271
1246
    @needs_read_lock
1275
1250
    @needs_tree_write_lock
1276
1251
    def _write_inventory(self, inv):
1277
1252
        """Write inventory as the current inventory."""
1278
 
        if self._dirty:
1279
 
            raise AssertionError("attempting to write an inventory when the "
1280
 
                "dirstate is dirty will lose pending changes")
1281
 
        had_inventory = self._inventory is not None
1282
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1283
 
        # working inventory. We do this because self.inventory may be inv, or
1284
 
        # may have been modified, and either case would prevent a clean delta
1285
 
        # being created.
1286
 
        self._inventory = None
1287
 
        # generate a delta,
1288
 
        delta = inv._make_delta(self.inventory)
1289
 
        # and apply it.
1290
 
        self.apply_inventory_delta(delta)
1291
 
        if had_inventory:
 
1253
        assert not self._dirty, ("attempting to write an inventory when the "
 
1254
            "dirstate is dirty will cause data loss")
 
1255
        self.current_dirstate().set_state_from_inventory(inv)
 
1256
        self._make_dirty(reset_inventory=False)
 
1257
        if self._inventory is not None:
1292
1258
            self._inventory = inv
1293
1259
        self.flush()
1294
1260
 
1295
1261
 
1296
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1297
 
 
1298
 
    def __init__(self, tree):
1299
 
        self.tree = tree
1300
 
 
1301
 
    def sha1(self, abspath):
1302
 
        """See dirstate.SHA1Provider.sha1()."""
1303
 
        filters = self.tree._content_filter_stack(
1304
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1305
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1306
 
 
1307
 
    def stat_and_sha1(self, abspath):
1308
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1309
 
        filters = self.tree._content_filter_stack(
1310
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1311
 
        file_obj = file(abspath, 'rb', 65000)
1312
 
        try:
1313
 
            statvalue = os.fstat(file_obj.fileno())
1314
 
            if filters:
1315
 
                file_obj = filtered_input_file(file_obj, filters)
1316
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1317
 
        finally:
1318
 
            file_obj.close()
1319
 
        return statvalue, sha1
1320
 
 
1321
 
 
1322
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1323
 
    """Dirstate working tree that supports content filtering.
1324
 
 
1325
 
    The dirstate holds the hash and size of the canonical form of the file, 
1326
 
    and most methods must return that.
1327
 
    """
1328
 
 
1329
 
    def _file_content_summary(self, path, stat_result):
1330
 
        # This is to support the somewhat obsolete path_content_summary method
1331
 
        # with content filtering: see
1332
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1333
 
        #
1334
 
        # If the dirstate cache is up to date and knows the hash and size,
1335
 
        # return that.
1336
 
        # Otherwise if there are no content filters, return the on-disk size
1337
 
        # and leave the hash blank.
1338
 
        # Otherwise, read and filter the on-disk file and use its size and
1339
 
        # hash.
1340
 
        #
1341
 
        # The dirstate doesn't store the size of the canonical form so we
1342
 
        # can't trust it for content-filtered trees.  We just return None.
1343
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1344
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1345
 
        return ('file', None, executable, dirstate_sha1)
1346
 
 
1347
 
 
1348
 
class WorkingTree4(DirStateWorkingTree):
1349
 
    """This is the Format 4 working tree.
1350
 
 
1351
 
    This differs from WorkingTree3 by:
1352
 
     - Having a consolidated internal dirstate, stored in a
1353
 
       randomly-accessible sorted file on disk.
1354
 
     - Not having a regular inventory attribute.  One can be synthesized
1355
 
       on demand but this is expensive and should be avoided.
1356
 
 
1357
 
    This is new in bzr 0.15.
1358
 
    """
1359
 
 
1360
 
 
1361
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1362
 
    """This is the Format 5 working tree.
1363
 
 
1364
 
    This differs from WorkingTree4 by:
1365
 
     - Supporting content filtering.
1366
 
 
1367
 
    This is new in bzr 1.11.
1368
 
    """
1369
 
 
1370
 
 
1371
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1372
 
    """This is the Format 6 working tree.
1373
 
 
1374
 
    This differs from WorkingTree5 by:
1375
 
     - Supporting a current view that may mask the set of files in a tree
1376
 
       impacted by most user operations.
1377
 
 
1378
 
    This is new in bzr 1.14.
1379
 
    """
1380
 
 
1381
 
    def _make_views(self):
1382
 
        return views.PathBasedViews(self)
1383
 
 
1384
 
 
1385
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
 
1262
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1263
    """The first consolidated dirstate working tree format.
 
1264
 
 
1265
    This format:
 
1266
        - exists within a metadir controlling .bzr
 
1267
        - includes an explicit version marker for the workingtree control
 
1268
          files, separate from the BzrDir format
 
1269
        - modifies the hash cache format
 
1270
        - is new in bzr 0.15
 
1271
        - uses a LockDir to guard access to it.
 
1272
    """
 
1273
 
 
1274
    upgrade_recommended = False
 
1275
 
 
1276
    def get_format_string(self):
 
1277
        """See WorkingTreeFormat.get_format_string()."""
 
1278
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1279
 
 
1280
    def get_format_description(self):
 
1281
        """See WorkingTreeFormat.get_format_description()."""
 
1282
        return "Working tree format 4"
1386
1283
 
1387
1284
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1388
1285
                   accelerator_tree=None, hardlink=False):
1406
1303
        control_files = self._open_control_files(a_bzrdir)
1407
1304
        control_files.create_lock()
1408
1305
        control_files.lock_write()
1409
 
        transport.put_bytes('format', self.get_format_string(),
1410
 
            mode=a_bzrdir._get_file_mode())
 
1306
        control_files.put_utf8('format', self.get_format_string())
1411
1307
        if from_branch is not None:
1412
1308
            branch = from_branch
1413
1309
        else:
1419
1315
        state = dirstate.DirState.initialize(local_path)
1420
1316
        state.unlock()
1421
1317
        del state
1422
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1318
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1423
1319
                         branch,
1424
1320
                         _format=self,
1425
1321
                         _bzrdir=a_bzrdir,
1427
1323
        wt._new_tree()
1428
1324
        wt.lock_tree_write()
1429
1325
        try:
1430
 
            self._init_custom_control_files(wt)
1431
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1326
            if revision_id in (None, NULL_REVISION):
1432
1327
                if branch.repository.supports_rich_root():
1433
1328
                    wt._set_root_id(generate_ids.gen_root_id())
1434
1329
                else:
1445
1340
                    pass
1446
1341
            if basis is None:
1447
1342
                basis = branch.repository.revision_tree(revision_id)
1448
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1343
            if revision_id == NULL_REVISION:
1449
1344
                parents_list = []
1450
1345
            else:
1451
1346
                parents_list = [(revision_id, basis)]
1459
1354
                if basis_root_id is not None:
1460
1355
                    wt._set_root_id(basis_root_id)
1461
1356
                    wt.flush()
1462
 
                if wt.supports_content_filtering():
1463
 
                    # The original tree may not have the same content filters
1464
 
                    # applied so we can't safely build the inventory delta from
1465
 
                    # the source tree.
1466
 
                    delta_from_tree = False
1467
 
                else:
1468
 
                    delta_from_tree = True
1469
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1470
 
                # because wt4.apply_inventory_delta does not mutate the input
1471
 
                # inventory entries.
1472
1357
                transform.build_tree(basis, wt, accelerator_tree,
1473
 
                                     hardlink=hardlink,
1474
 
                                     delta_from_tree=delta_from_tree)
 
1358
                                     hardlink=hardlink)
1475
1359
            finally:
1476
1360
                basis.unlock()
1477
1361
        finally:
1479
1363
            wt.unlock()
1480
1364
        return wt
1481
1365
 
1482
 
    def _init_custom_control_files(self, wt):
1483
 
        """Subclasses with custom control files should override this method.
1484
 
 
1485
 
        The working tree and control files are locked for writing when this
1486
 
        method is called.
1487
 
 
1488
 
        :param wt: the WorkingTree object
1489
 
        """
1490
 
 
1491
1366
    def _open(self, a_bzrdir, control_files):
1492
1367
        """Open the tree itself.
1493
1368
 
1494
1369
        :param a_bzrdir: the dir for the tree.
1495
1370
        :param control_files: the control files for the tree.
1496
1371
        """
1497
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1372
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1498
1373
                           branch=a_bzrdir.open_branch(),
1499
1374
                           _format=self,
1500
1375
                           _bzrdir=a_bzrdir,
1501
1376
                           _control_files=control_files)
1502
1377
 
1503
1378
    def __get_matchingbzrdir(self):
1504
 
        return self._get_matchingbzrdir()
1505
 
 
1506
 
    def _get_matchingbzrdir(self):
1507
 
        """Overrideable method to get a bzrdir for testing."""
1508
1379
        # please test against something that will let us do tree references
1509
1380
        return bzrdir.format_registry.make_bzrdir(
1510
1381
            'dirstate-with-subtree')
1512
1383
    _matchingbzrdir = property(__get_matchingbzrdir)
1513
1384
 
1514
1385
 
1515
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1516
 
    """The first consolidated dirstate working tree format.
1517
 
 
1518
 
    This format:
1519
 
        - exists within a metadir controlling .bzr
1520
 
        - includes an explicit version marker for the workingtree control
1521
 
          files, separate from the BzrDir format
1522
 
        - modifies the hash cache format
1523
 
        - is new in bzr 0.15
1524
 
        - uses a LockDir to guard access to it.
1525
 
    """
1526
 
 
1527
 
    upgrade_recommended = False
1528
 
 
1529
 
    _tree_class = WorkingTree4
1530
 
 
1531
 
    def get_format_string(self):
1532
 
        """See WorkingTreeFormat.get_format_string()."""
1533
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1534
 
 
1535
 
    def get_format_description(self):
1536
 
        """See WorkingTreeFormat.get_format_description()."""
1537
 
        return "Working tree format 4"
1538
 
 
1539
 
 
1540
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1541
 
    """WorkingTree format supporting content filtering.
1542
 
    """
1543
 
 
1544
 
    upgrade_recommended = False
1545
 
 
1546
 
    _tree_class = WorkingTree5
1547
 
 
1548
 
    def get_format_string(self):
1549
 
        """See WorkingTreeFormat.get_format_string()."""
1550
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1551
 
 
1552
 
    def get_format_description(self):
1553
 
        """See WorkingTreeFormat.get_format_description()."""
1554
 
        return "Working tree format 5"
1555
 
 
1556
 
    def supports_content_filtering(self):
1557
 
        return True
1558
 
 
1559
 
 
1560
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1561
 
    """WorkingTree format supporting views.
1562
 
    """
1563
 
 
1564
 
    upgrade_recommended = False
1565
 
 
1566
 
    _tree_class = WorkingTree6
1567
 
 
1568
 
    def get_format_string(self):
1569
 
        """See WorkingTreeFormat.get_format_string()."""
1570
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1571
 
 
1572
 
    def get_format_description(self):
1573
 
        """See WorkingTreeFormat.get_format_description()."""
1574
 
        return "Working tree format 6"
1575
 
 
1576
 
    def _init_custom_control_files(self, wt):
1577
 
        """Subclasses with custom control files should override this method."""
1578
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1579
 
 
1580
 
    def supports_content_filtering(self):
1581
 
        return True
1582
 
 
1583
 
    def supports_views(self):
1584
 
        return True
1585
 
 
1586
 
 
1587
1386
class DirStateRevisionTree(Tree):
1588
 
    """A revision tree pulling the inventory from a dirstate.
1589
 
    
1590
 
    Note that this is one of the historical (ie revision) trees cached in the
1591
 
    dirstate for easy access, not the workingtree.
1592
 
    """
 
1387
    """A revision tree pulling the inventory from a dirstate."""
1593
1388
 
1594
1389
    def __init__(self, dirstate, revision_id, repository):
1595
1390
        self._dirstate = dirstate
1598
1393
        self._inventory = None
1599
1394
        self._locked = 0
1600
1395
        self._dirstate_locked = False
1601
 
        self._repo_supports_tree_reference = getattr(
1602
 
            repository._format, "supports_tree_reference",
1603
 
            False)
1604
1396
 
1605
1397
    def __repr__(self):
1606
1398
        return "<%s of %s in %s>" % \
1609
1401
    def annotate_iter(self, file_id,
1610
1402
                      default_revision=_mod_revision.CURRENT_REVISION):
1611
1403
        """See Tree.annotate_iter"""
1612
 
        text_key = (file_id, self.inventory[file_id].revision)
1613
 
        annotations = self._repository.texts.annotate(text_key)
1614
 
        return [(key[-1], line) for (key, line) in annotations]
 
1404
        w = self._get_weave(file_id)
 
1405
        return w.annotate(self.inventory[file_id].revision)
1615
1406
 
1616
1407
    def _get_ancestors(self, default_revision):
1617
1408
        return set(self._repository.get_ancestry(self._revision_id,
1646
1437
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1647
1438
        return path_utf8.decode('utf8')
1648
1439
 
1649
 
    def iter_references(self):
1650
 
        if not self._repo_supports_tree_reference:
1651
 
            # When the repo doesn't support references, we will have nothing to
1652
 
            # return
1653
 
            return iter([])
1654
 
        # Otherwise, fall back to the default implementation
1655
 
        return super(DirStateRevisionTree, self).iter_references()
1656
 
 
1657
1440
    def _get_parent_index(self):
1658
1441
        """Return the index in the dirstate referenced by this tree."""
1659
1442
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1664
1447
        If either file_id or path is supplied, it is used as the key to lookup.
1665
1448
        If both are supplied, the fastest lookup is used, and an error is
1666
1449
        raised if they do not both point at the same row.
1667
 
 
 
1450
        
1668
1451
        :param file_id: An optional unicode file_id to be looked up.
1669
1452
        :param path: An optional unicode path to be looked up.
1670
1453
        :return: The dirstate row tuple for path/file_id, or (None, None)
1684
1467
 
1685
1468
        This is relatively expensive: we have to walk the entire dirstate.
1686
1469
        """
1687
 
        if not self._locked:
1688
 
            raise AssertionError(
1689
 
                'cannot generate inventory of an unlocked '
1690
 
                'dirstate revision tree')
 
1470
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1471
            'dirstate revision tree'
1691
1472
        # separate call for profiling - makes it clear where the costs are.
1692
1473
        self._dirstate._read_dirblocks_if_needed()
1693
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1694
 
            raise AssertionError(
1695
 
                'parent %s has disappeared from %s' % (
1696
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1474
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1475
            'parent %s has disappeared from %s' % (
 
1476
            self._revision_id, self._dirstate.get_parent_ids())
1697
1477
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1698
1478
        # This is identical now to the WorkingTree _generate_inventory except
1699
1479
        # for the tree index use.
1700
1480
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1701
1481
        current_id = root_key[2]
1702
 
        if current_entry[parent_index][0] != 'd':
1703
 
            raise AssertionError()
 
1482
        assert current_entry[parent_index][0] == 'd'
1704
1483
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1705
1484
        inv.root.revision = current_entry[parent_index][4]
1706
1485
        # Turn some things into local variables
1746
1525
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1747
1526
                            % entry)
1748
1527
                # These checks cost us around 40ms on a 55k entry tree
1749
 
                if file_id in inv_byid:
1750
 
                    raise AssertionError('file_id %s already in'
1751
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1752
 
                if name_unicode in parent_ie.children:
1753
 
                    raise AssertionError('name %r already in parent'
1754
 
                        % (name_unicode,))
 
1528
                assert file_id not in inv_byid
 
1529
                assert name_unicode not in parent_ie.children
1755
1530
                inv_byid[file_id] = inv_entry
1756
1531
                parent_ie.children[name_unicode] = inv_entry
1757
1532
        self._inventory = inv
1767
1542
            return None
1768
1543
        parent_index = self._get_parent_index()
1769
1544
        last_changed_revision = entry[1][parent_index][4]
1770
 
        try:
1771
 
            rev = self._repository.get_revision(last_changed_revision)
1772
 
        except errors.NoSuchRevision:
1773
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1774
 
        return rev.timestamp
 
1545
        return self._repository.get_revision(last_changed_revision).timestamp
1775
1546
 
1776
1547
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1777
1548
        entry = self._get_entry(file_id=file_id, path=path)
1781
1552
            return parent_details[1]
1782
1553
        return None
1783
1554
 
 
1555
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1556
    def get_weave(self, file_id):
 
1557
        return self._get_weave(file_id)
 
1558
 
 
1559
    def _get_weave(self, file_id):
 
1560
        return self._repository.weave_store.get_weave(file_id,
 
1561
                self._repository.get_transaction())
 
1562
 
1784
1563
    def get_file(self, file_id, path=None):
1785
1564
        return StringIO(self.get_file_text(file_id))
1786
1565
 
 
1566
    def get_file_lines(self, file_id):
 
1567
        entry = self._get_entry(file_id=file_id)[1]
 
1568
        if entry == None:
 
1569
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1570
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1571
 
1787
1572
    def get_file_size(self, file_id):
1788
1573
        """See Tree.get_file_size"""
1789
1574
        return self.inventory[file_id].text_size
1790
1575
 
1791
 
    def get_file_text(self, file_id, path=None):
1792
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1793
 
        return ''.join(content)
 
1576
    def get_file_text(self, file_id):
 
1577
        return ''.join(self.get_file_lines(file_id))
1794
1578
 
1795
1579
    def get_reference_revision(self, file_id, path=None):
1796
1580
        return self.inventory[file_id].reference_revision
1815
1599
        if entry[1][parent_index][0] != 'l':
1816
1600
            return None
1817
1601
        else:
1818
 
            target = entry[1][parent_index][1]
1819
 
            target = target.decode('utf8')
1820
 
            return target
 
1602
            # At present, none of the tree implementations supports non-ascii
 
1603
            # symlink targets. So we will just assume that the dirstate path is
 
1604
            # correct.
 
1605
            return entry[1][parent_index][1]
1821
1606
 
1822
1607
    def get_revision_id(self):
1823
1608
        """Return the revision id for this tree."""
1842
1627
 
1843
1628
    def kind(self, file_id):
1844
1629
        entry = self._get_entry(file_id=file_id)[1]
1845
 
        if entry is None:
 
1630
        if entry == None:
1846
1631
            raise errors.NoSuchId(tree=self, file_id=file_id)
1847
 
        parent_index = self._get_parent_index()
1848
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1632
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1849
1633
 
1850
1634
    def stored_kind(self, file_id):
1851
1635
        """See Tree.stored_kind"""
1871
1655
            return None
1872
1656
        return ie.executable
1873
1657
 
1874
 
    def is_locked(self):
1875
 
        return self._locked
1876
 
 
1877
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1658
    def list_files(self, include_root=False):
1878
1659
        # We use a standard implementation, because DirStateRevisionTree is
1879
1660
        # dealing with one of the parents of the current state
1880
1661
        inv = self._get_inventory()
1881
 
        if from_dir is None:
1882
 
            from_dir_id = None
1883
 
        else:
1884
 
            from_dir_id = inv.path2id(from_dir)
1885
 
            if from_dir_id is None:
1886
 
                # Directory not versioned
1887
 
                return
1888
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1889
 
        if inv.root is not None and not include_root and from_dir is None:
 
1662
        entries = inv.iter_entries()
 
1663
        if self.inventory.root is not None and not include_root:
1890
1664
            entries.next()
1891
1665
        for path, entry in entries:
1892
1666
            yield path, 'V', entry.kind, entry.file_id, entry
1893
1667
 
1894
1668
    def lock_read(self):
1895
 
        """Lock the tree for a set of operations.
1896
 
 
1897
 
        :return: A bzrlib.lock.LogicalLockResult.
1898
 
        """
 
1669
        """Lock the tree for a set of operations."""
1899
1670
        if not self._locked:
1900
1671
            self._repository.lock_read()
1901
1672
            if self._dirstate._lock_token is None:
1902
1673
                self._dirstate.lock_read()
1903
1674
                self._dirstate_locked = True
1904
1675
        self._locked += 1
1905
 
        return LogicalLockResult(self.unlock)
1906
1676
 
1907
1677
    def _must_be_locked(self):
1908
1678
        if not self._locked:
1929
1699
                self._dirstate_locked = False
1930
1700
            self._repository.unlock()
1931
1701
 
1932
 
    @needs_read_lock
1933
 
    def supports_tree_reference(self):
1934
 
        return self._repo_supports_tree_reference
1935
 
 
1936
1702
    def walkdirs(self, prefix=""):
1937
1703
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1938
 
        # implementation based on an inventory.
 
1704
        # implementation based on an inventory.  
1939
1705
        # This should be cleaned up to use the much faster Dirstate code
1940
1706
        # So for now, we just build up the parent inventory, and extract
1941
1707
        # it the same way RevisionTree does.
1970
1736
 
1971
1737
class InterDirStateTree(InterTree):
1972
1738
    """Fast path optimiser for changes_from with dirstate trees.
1973
 
 
1974
 
    This is used only when both trees are in the dirstate working file, and
1975
 
    the source is any parent within the dirstate, and the destination is
 
1739
    
 
1740
    This is used only when both trees are in the dirstate working file, and 
 
1741
    the source is any parent within the dirstate, and the destination is 
1976
1742
    the current working tree of the same dirstate.
1977
1743
    """
1978
1744
    # this could be generalized to allow comparisons between any trees in the
1991
1757
        target.set_parent_ids([revid])
1992
1758
        return target.basis_tree(), target
1993
1759
 
1994
 
    @classmethod
1995
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1996
 
        result = klass.make_source_parent_tree(source, target)
1997
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1998
 
        return result
1999
 
 
2000
 
    @classmethod
2001
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2002
 
                                                  target):
2003
 
        from bzrlib.tests.test__dirstate_helpers import \
2004
 
            compiled_dirstate_helpers_feature
2005
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2006
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2007
 
        result = klass.make_source_parent_tree(source, target)
2008
 
        result[1]._iter_changes = ProcessEntryC
2009
 
        return result
2010
 
 
2011
1760
    _matching_from_tree_format = WorkingTreeFormat4()
2012
1761
    _matching_to_tree_format = WorkingTreeFormat4()
2013
 
 
2014
 
    @classmethod
2015
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2016
 
        # This method shouldn't be called, because we have python and C
2017
 
        # specific flavours.
2018
 
        raise NotImplementedError
 
1762
    _test_mutable_trees_to_test_trees = make_source_parent_tree
2019
1763
 
2020
1764
    def iter_changes(self, include_unchanged=False,
2021
1765
                      specific_files=None, pb=None, extra_trees=[],
2039
1783
            output. An unversioned file is defined as one with (False, False)
2040
1784
            for the versioned pair.
2041
1785
        """
 
1786
        utf8_decode = cache_utf8._utf8_decode
 
1787
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1788
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1789
        # NB: show_status depends on being able to pass in non-versioned files
 
1790
        # and report them as unknown
2042
1791
        # TODO: handle extra trees in the dirstate.
2043
1792
        if (extra_trees or specific_files == []):
2044
1793
            # we can't fast-path these cases (yet)
2045
 
            return super(InterDirStateTree, self).iter_changes(
 
1794
            for f in super(InterDirStateTree, self).iter_changes(
2046
1795
                include_unchanged, specific_files, pb, extra_trees,
2047
 
                require_versioned, want_unversioned=want_unversioned)
 
1796
                require_versioned, want_unversioned=want_unversioned):
 
1797
                yield f
 
1798
            return
2048
1799
        parent_ids = self.target.get_parent_ids()
2049
 
        if not (self.source._revision_id in parent_ids
2050
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2051
 
            raise AssertionError(
2052
 
                "revision {%s} is not stored in {%s}, but %s "
2053
 
                "can only be used for trees stored in the dirstate"
2054
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1800
        assert (self.source._revision_id in parent_ids
 
1801
                or self.source._revision_id == NULL_REVISION), \
 
1802
                "revision {%s} is not stored in {%s}, but %s " \
 
1803
                "can only be used for trees stored in the dirstate" \
 
1804
                % (self.source._revision_id, self.target, self.iter_changes)
2055
1805
        target_index = 0
2056
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1806
        if self.source._revision_id == NULL_REVISION:
2057
1807
            source_index = None
2058
1808
            indices = (target_index,)
2059
1809
        else:
2060
 
            if not (self.source._revision_id in parent_ids):
2061
 
                raise AssertionError(
2062
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2063
 
                    self.source._revision_id, parent_ids))
 
1810
            assert (self.source._revision_id in parent_ids), \
 
1811
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1812
                self.source._revision_id, parent_ids)
2064
1813
            source_index = 1 + parent_ids.index(self.source._revision_id)
2065
1814
            indices = (source_index, target_index)
2066
1815
        # -- make all specific_files utf8 --
2067
1816
        if specific_files:
2068
1817
            specific_files_utf8 = set()
2069
1818
            for path in specific_files:
2070
 
                # Note, if there are many specific files, using cache_utf8
2071
 
                # would be good here.
2072
1819
                specific_files_utf8.add(path.encode('utf8'))
2073
1820
            specific_files = specific_files_utf8
2074
1821
        else:
2075
1822
            specific_files = set([''])
2076
1823
        # -- specific_files is now a utf8 path set --
2077
 
 
2078
1824
        # -- get the state object and prepare it.
2079
1825
        state = self.target.current_dirstate()
2080
1826
        state._read_dirblocks_if_needed()
 
1827
        def _entries_for_path(path):
 
1828
            """Return a list with all the entries that match path for all ids.
 
1829
            """
 
1830
            dirname, basename = os.path.split(path)
 
1831
            key = (dirname, basename, '')
 
1832
            block_index, present = state._find_block_index_from_key(key)
 
1833
            if not present:
 
1834
                # the block which should contain path is absent.
 
1835
                return []
 
1836
            result = []
 
1837
            block = state._dirblocks[block_index][1]
 
1838
            entry_index, _ = state._find_entry_index(key, block)
 
1839
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1840
            while (entry_index < len(block) and
 
1841
                block[entry_index][0][0:2] == key[0:2]):
 
1842
                result.append(block[entry_index])
 
1843
                entry_index += 1
 
1844
            return result
2081
1845
        if require_versioned:
2082
1846
            # -- check all supplied paths are versioned in a search tree. --
2083
 
            not_versioned = []
 
1847
            all_versioned = True
2084
1848
            for path in specific_files:
2085
 
                path_entries = state._entries_for_path(path)
 
1849
                path_entries = _entries_for_path(path)
2086
1850
                if not path_entries:
2087
1851
                    # this specified path is not present at all: error
2088
 
                    not_versioned.append(path)
2089
 
                    continue
 
1852
                    all_versioned = False
 
1853
                    break
2090
1854
                found_versioned = False
2091
1855
                # for each id at this path
2092
1856
                for entry in path_entries:
2099
1863
                if not found_versioned:
2100
1864
                    # none of the indexes was not 'absent' at all ids for this
2101
1865
                    # path.
2102
 
                    not_versioned.append(path)
2103
 
            if len(not_versioned) > 0:
2104
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1866
                    all_versioned = False
 
1867
                    break
 
1868
            if not all_versioned:
 
1869
                raise errors.PathsNotVersionedError(specific_files)
2105
1870
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2106
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1871
        search_specific_files = set()
 
1872
        for path in specific_files:
 
1873
            other_specific_files = specific_files.difference(set([path]))
 
1874
            if not osutils.is_inside_any(other_specific_files, path):
 
1875
                # this is a top level path, we must check it.
 
1876
                search_specific_files.add(path)
 
1877
        # sketch: 
 
1878
        # compare source_index and target_index at or under each element of search_specific_files.
 
1879
        # follow the following comparison table. Note that we only want to do diff operations when
 
1880
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1881
        # for the target.
 
1882
        # cases:
 
1883
        # 
 
1884
        # Source | Target | disk | action
 
1885
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1886
        #        |        |      | diff check on source-target
 
1887
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1888
        #        |        |      | ???
 
1889
        #   r    |  a     |      | add source to search
 
1890
        #   r    |  a     |  a   | 
 
1891
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1892
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1893
        #   a    | fdlt   |      | add new id
 
1894
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1895
        #   a    |  a     |      | not present in either tree, skip
 
1896
        #   a    |  a     |  a   | not present in any tree, skip
 
1897
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1898
        #        |        |      | may not be selected by the users list of paths.
 
1899
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1900
        #        |        |      | may not be selected by the users list of paths.
 
1901
        #  fdlt  | fdlt   |      | content in both: diff them
 
1902
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1903
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1904
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1905
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1906
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1907
        #        |        |      | this id at the other path.
 
1908
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1909
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1910
        #        |        |      | this id at the other path.
 
1911
 
 
1912
        # for all search_indexs in each path at or under each element of
 
1913
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1914
        # relocated path as one to search if its not searched already. If the
 
1915
        # detail is not relocated, add the id.
 
1916
        searched_specific_files = set()
 
1917
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1918
        # Using a list so that we can access the values and change them in
 
1919
        # nested scope. Each one is [path, file_id, entry]
 
1920
        last_source_parent = [None, None]
 
1921
        last_target_parent = [None, None]
2107
1922
 
2108
1923
        use_filesystem_for_exec = (sys.platform != 'win32')
2109
 
        iter_changes = self.target._iter_changes(include_unchanged,
2110
 
            use_filesystem_for_exec, search_specific_files, state,
2111
 
            source_index, target_index, want_unversioned, self.target)
2112
 
        return iter_changes.iter_changes()
 
1924
 
 
1925
        # Just a sentry, so that _process_entry can say that this
 
1926
        # record is handled, but isn't interesting to process (unchanged)
 
1927
        uninteresting = object()
 
1928
 
 
1929
 
 
1930
        old_dirname_to_file_id = {}
 
1931
        new_dirname_to_file_id = {}
 
1932
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1933
        #       keeping a cache of directories that we have seen.
 
1934
 
 
1935
        def _process_entry(entry, path_info):
 
1936
            """Compare an entry and real disk to generate delta information.
 
1937
 
 
1938
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1939
                the path of entry. If None, then the path is considered absent.
 
1940
                (Perhaps we should pass in a concrete entry for this ?)
 
1941
                Basename is returned as a utf8 string because we expect this
 
1942
                tuple will be ignored, and don't want to take the time to
 
1943
                decode.
 
1944
            :return: None if these don't match
 
1945
                     A tuple of information about the change, or
 
1946
                     the object 'uninteresting' if these match, but are
 
1947
                     basically identical.
 
1948
            """
 
1949
            if source_index is None:
 
1950
                source_details = NULL_PARENT_DETAILS
 
1951
            else:
 
1952
                source_details = entry[1][source_index]
 
1953
            target_details = entry[1][target_index]
 
1954
            target_minikind = target_details[0]
 
1955
            if path_info is not None and target_minikind in 'fdlt':
 
1956
                assert target_index == 0
 
1957
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1958
                                                  stat_value=path_info[3])
 
1959
                # The entry may have been modified by update_entry
 
1960
                target_details = entry[1][target_index]
 
1961
                target_minikind = target_details[0]
 
1962
            else:
 
1963
                link_or_sha1 = None
 
1964
            file_id = entry[0][2]
 
1965
            source_minikind = source_details[0]
 
1966
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1967
                # claimed content in both: diff
 
1968
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1969
                #        |        |      | diff check on source-target
 
1970
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1971
                #        |        |      | ???
 
1972
                if source_minikind in 'r':
 
1973
                    # add the source to the search path to find any children it
 
1974
                    # has.  TODO ? : only add if it is a container ?
 
1975
                    if not osutils.is_inside_any(searched_specific_files,
 
1976
                                                 source_details[1]):
 
1977
                        search_specific_files.add(source_details[1])
 
1978
                    # generate the old path; this is needed for stating later
 
1979
                    # as well.
 
1980
                    old_path = source_details[1]
 
1981
                    old_dirname, old_basename = os.path.split(old_path)
 
1982
                    path = pathjoin(entry[0][0], entry[0][1])
 
1983
                    old_entry = state._get_entry(source_index,
 
1984
                                                 path_utf8=old_path)
 
1985
                    # update the source details variable to be the real
 
1986
                    # location.
 
1987
                    if old_entry == (None, None):
 
1988
                        raise errors.CorruptDirstate(state._filename,
 
1989
                            "entry '%s/%s' is considered renamed from %r"
 
1990
                            " but source does not exist\n"
 
1991
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
1992
                    source_details = old_entry[1][source_index]
 
1993
                    source_minikind = source_details[0]
 
1994
                else:
 
1995
                    old_dirname = entry[0][0]
 
1996
                    old_basename = entry[0][1]
 
1997
                    old_path = path = None
 
1998
                if path_info is None:
 
1999
                    # the file is missing on disk, show as removed.
 
2000
                    content_change = True
 
2001
                    target_kind = None
 
2002
                    target_exec = False
 
2003
                else:
 
2004
                    # source and target are both versioned and disk file is present.
 
2005
                    target_kind = path_info[2]
 
2006
                    if target_kind == 'directory':
 
2007
                        if path is None:
 
2008
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2009
                        new_dirname_to_file_id[path] = file_id
 
2010
                        if source_minikind != 'd':
 
2011
                            content_change = True
 
2012
                        else:
 
2013
                            # directories have no fingerprint
 
2014
                            content_change = False
 
2015
                        target_exec = False
 
2016
                    elif target_kind == 'file':
 
2017
                        if source_minikind != 'f':
 
2018
                            content_change = True
 
2019
                        else:
 
2020
                            # We could check the size, but we already have the
 
2021
                            # sha1 hash.
 
2022
                            content_change = (link_or_sha1 != source_details[1])
 
2023
                        # Target details is updated at update_entry time
 
2024
                        if use_filesystem_for_exec:
 
2025
                            # We don't need S_ISREG here, because we are sure
 
2026
                            # we are dealing with a file.
 
2027
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2028
                        else:
 
2029
                            target_exec = target_details[3]
 
2030
                    elif target_kind == 'symlink':
 
2031
                        if source_minikind != 'l':
 
2032
                            content_change = True
 
2033
                        else:
 
2034
                            content_change = (link_or_sha1 != source_details[1])
 
2035
                        target_exec = False
 
2036
                    elif target_kind == 'tree-reference':
 
2037
                        if source_minikind != 't':
 
2038
                            content_change = True
 
2039
                        else:
 
2040
                            content_change = False
 
2041
                        target_exec = False
 
2042
                    else:
 
2043
                        raise Exception, "unknown kind %s" % path_info[2]
 
2044
                if source_minikind == 'd':
 
2045
                    if path is None:
 
2046
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2047
                    old_dirname_to_file_id[old_path] = file_id
 
2048
                # parent id is the entry for the path in the target tree
 
2049
                if old_dirname == last_source_parent[0]:
 
2050
                    source_parent_id = last_source_parent[1]
 
2051
                else:
 
2052
                    try:
 
2053
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2054
                    except KeyError:
 
2055
                        source_parent_entry = state._get_entry(source_index,
 
2056
                                                               path_utf8=old_dirname)
 
2057
                        source_parent_id = source_parent_entry[0][2]
 
2058
                    if source_parent_id == entry[0][2]:
 
2059
                        # This is the root, so the parent is None
 
2060
                        source_parent_id = None
 
2061
                    else:
 
2062
                        last_source_parent[0] = old_dirname
 
2063
                        last_source_parent[1] = source_parent_id
 
2064
                new_dirname = entry[0][0]
 
2065
                if new_dirname == last_target_parent[0]:
 
2066
                    target_parent_id = last_target_parent[1]
 
2067
                else:
 
2068
                    try:
 
2069
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2070
                    except KeyError:
 
2071
                        # TODO: We don't always need to do the lookup, because the
 
2072
                        #       parent entry will be the same as the source entry.
 
2073
                        target_parent_entry = state._get_entry(target_index,
 
2074
                                                               path_utf8=new_dirname)
 
2075
                        assert target_parent_entry != (None, None), (
 
2076
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2077
                            % (new_dirname, entry))
 
2078
                        target_parent_id = target_parent_entry[0][2]
 
2079
                    if target_parent_id == entry[0][2]:
 
2080
                        # This is the root, so the parent is None
 
2081
                        target_parent_id = None
 
2082
                    else:
 
2083
                        last_target_parent[0] = new_dirname
 
2084
                        last_target_parent[1] = target_parent_id
 
2085
 
 
2086
                source_exec = source_details[3]
 
2087
                if (include_unchanged
 
2088
                    or content_change
 
2089
                    or source_parent_id != target_parent_id
 
2090
                    or old_basename != entry[0][1]
 
2091
                    or source_exec != target_exec
 
2092
                    ):
 
2093
                    if old_path is None:
 
2094
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2095
                        old_path_u = utf8_decode(old_path)[0]
 
2096
                        path_u = old_path_u
 
2097
                    else:
 
2098
                        old_path_u = utf8_decode(old_path)[0]
 
2099
                        if old_path == path:
 
2100
                            path_u = old_path_u
 
2101
                        else:
 
2102
                            path_u = utf8_decode(path)[0]
 
2103
                    source_kind = _minikind_to_kind[source_minikind]
 
2104
                    return (entry[0][2],
 
2105
                           (old_path_u, path_u),
 
2106
                           content_change,
 
2107
                           (True, True),
 
2108
                           (source_parent_id, target_parent_id),
 
2109
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2110
                           (source_kind, target_kind),
 
2111
                           (source_exec, target_exec))
 
2112
                else:
 
2113
                    return uninteresting
 
2114
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2115
                # looks like a new file
 
2116
                if path_info is not None:
 
2117
                    path = pathjoin(entry[0][0], entry[0][1])
 
2118
                    # parent id is the entry for the path in the target tree
 
2119
                    # TODO: these are the same for an entire directory: cache em.
 
2120
                    parent_id = state._get_entry(target_index,
 
2121
                                                 path_utf8=entry[0][0])[0][2]
 
2122
                    if parent_id == entry[0][2]:
 
2123
                        parent_id = None
 
2124
                    if use_filesystem_for_exec:
 
2125
                        # We need S_ISREG here, because we aren't sure if this
 
2126
                        # is a file or not.
 
2127
                        target_exec = bool(
 
2128
                            stat.S_ISREG(path_info[3].st_mode)
 
2129
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2130
                    else:
 
2131
                        target_exec = target_details[3]
 
2132
                    return (entry[0][2],
 
2133
                           (None, utf8_decode(path)[0]),
 
2134
                           True,
 
2135
                           (False, True),
 
2136
                           (None, parent_id),
 
2137
                           (None, utf8_decode(entry[0][1])[0]),
 
2138
                           (None, path_info[2]),
 
2139
                           (None, target_exec))
 
2140
                else:
 
2141
                    # but its not on disk: we deliberately treat this as just
 
2142
                    # never-present. (Why ?! - RBC 20070224)
 
2143
                    pass
 
2144
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2145
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2146
                # if its still on disk, *and* theres no other entry at this
 
2147
                # path [we dont know this in this routine at the moment -
 
2148
                # perhaps we should change this - then it would be an unknown.
 
2149
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2150
                # parent id is the entry for the path in the target tree
 
2151
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2152
                if parent_id == entry[0][2]:
 
2153
                    parent_id = None
 
2154
                return (entry[0][2],
 
2155
                       (utf8_decode(old_path)[0], None),
 
2156
                       True,
 
2157
                       (True, False),
 
2158
                       (parent_id, None),
 
2159
                       (utf8_decode(entry[0][1])[0], None),
 
2160
                       (_minikind_to_kind[source_minikind], None),
 
2161
                       (source_details[3], None))
 
2162
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2163
                # a rename; could be a true rename, or a rename inherited from
 
2164
                # a renamed parent. TODO: handle this efficiently. Its not
 
2165
                # common case to rename dirs though, so a correct but slow
 
2166
                # implementation will do.
 
2167
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2168
                    search_specific_files.add(target_details[1])
 
2169
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2170
                # neither of the selected trees contain this file,
 
2171
                # so skip over it. This is not currently directly tested, but
 
2172
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2173
                pass
 
2174
            else:
 
2175
                raise AssertionError("don't know how to compare "
 
2176
                    "source_minikind=%r, target_minikind=%r"
 
2177
                    % (source_minikind, target_minikind))
 
2178
                ## import pdb;pdb.set_trace()
 
2179
            return None
 
2180
 
 
2181
        while search_specific_files:
 
2182
            # TODO: the pending list should be lexically sorted?  the
 
2183
            # interface doesn't require it.
 
2184
            current_root = search_specific_files.pop()
 
2185
            current_root_unicode = current_root.decode('utf8')
 
2186
            searched_specific_files.add(current_root)
 
2187
            # process the entries for this containing directory: the rest will be
 
2188
            # found by their parents recursively.
 
2189
            root_entries = _entries_for_path(current_root)
 
2190
            root_abspath = self.target.abspath(current_root_unicode)
 
2191
            try:
 
2192
                root_stat = os.lstat(root_abspath)
 
2193
            except OSError, e:
 
2194
                if e.errno == errno.ENOENT:
 
2195
                    # the path does not exist: let _process_entry know that.
 
2196
                    root_dir_info = None
 
2197
                else:
 
2198
                    # some other random error: hand it up.
 
2199
                    raise
 
2200
            else:
 
2201
                root_dir_info = ('', current_root,
 
2202
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2203
                    root_abspath)
 
2204
                if root_dir_info[2] == 'directory':
 
2205
                    if self.target._directory_is_tree_reference(
 
2206
                        current_root.decode('utf8')):
 
2207
                        root_dir_info = root_dir_info[:2] + \
 
2208
                            ('tree-reference',) + root_dir_info[3:]
 
2209
 
 
2210
            if not root_entries and not root_dir_info:
 
2211
                # this specified path is not present at all, skip it.
 
2212
                continue
 
2213
            path_handled = False
 
2214
            for entry in root_entries:
 
2215
                result = _process_entry(entry, root_dir_info)
 
2216
                if result is not None:
 
2217
                    path_handled = True
 
2218
                    if result is not uninteresting:
 
2219
                        yield result
 
2220
            if want_unversioned and not path_handled and root_dir_info:
 
2221
                new_executable = bool(
 
2222
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2223
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2224
                yield (None,
 
2225
                       (None, current_root_unicode),
 
2226
                       True,
 
2227
                       (False, False),
 
2228
                       (None, None),
 
2229
                       (None, splitpath(current_root_unicode)[-1]),
 
2230
                       (None, root_dir_info[2]),
 
2231
                       (None, new_executable)
 
2232
                      )
 
2233
            initial_key = (current_root, '', '')
 
2234
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2235
            if block_index == 0:
 
2236
                # we have processed the total root already, but because the
 
2237
                # initial key matched it we should skip it here.
 
2238
                block_index +=1
 
2239
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2240
                current_dir_info = None
 
2241
            else:
 
2242
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2243
                try:
 
2244
                    current_dir_info = dir_iterator.next()
 
2245
                except OSError, e:
 
2246
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2247
                    # python 2.5 has e.errno == EINVAL,
 
2248
                    #            and e.winerror == ERROR_DIRECTORY
 
2249
                    e_winerror = getattr(e, 'winerror', None)
 
2250
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2251
                    # there may be directories in the inventory even though
 
2252
                    # this path is not a file on disk: so mark it as end of
 
2253
                    # iterator
 
2254
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2255
                        current_dir_info = None
 
2256
                    elif (sys.platform == 'win32'
 
2257
                          and (e.errno in win_errors
 
2258
                               or e_winerror in win_errors)):
 
2259
                        current_dir_info = None
 
2260
                    else:
 
2261
                        raise
 
2262
                else:
 
2263
                    if current_dir_info[0][0] == '':
 
2264
                        # remove .bzr from iteration
 
2265
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2266
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2267
                        del current_dir_info[1][bzr_index]
 
2268
            # walk until both the directory listing and the versioned metadata
 
2269
            # are exhausted. 
 
2270
            if (block_index < len(state._dirblocks) and
 
2271
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2272
                current_block = state._dirblocks[block_index]
 
2273
            else:
 
2274
                current_block = None
 
2275
            while (current_dir_info is not None or
 
2276
                   current_block is not None):
 
2277
                if (current_dir_info and current_block
 
2278
                    and current_dir_info[0][0] != current_block[0]):
 
2279
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2280
                        # filesystem data refers to paths not covered by the dirblock.
 
2281
                        # this has two possibilities:
 
2282
                        # A) it is versioned but empty, so there is no block for it
 
2283
                        # B) it is not versioned.
 
2284
 
 
2285
                        # if (A) then we need to recurse into it to check for
 
2286
                        # new unknown files or directories.
 
2287
                        # if (B) then we should ignore it, because we don't
 
2288
                        # recurse into unknown directories.
 
2289
                        path_index = 0
 
2290
                        while path_index < len(current_dir_info[1]):
 
2291
                                current_path_info = current_dir_info[1][path_index]
 
2292
                                if want_unversioned:
 
2293
                                    if current_path_info[2] == 'directory':
 
2294
                                        if self.target._directory_is_tree_reference(
 
2295
                                            current_path_info[0].decode('utf8')):
 
2296
                                            current_path_info = current_path_info[:2] + \
 
2297
                                                ('tree-reference',) + current_path_info[3:]
 
2298
                                    new_executable = bool(
 
2299
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2300
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2301
                                    yield (None,
 
2302
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2303
                                        True,
 
2304
                                        (False, False),
 
2305
                                        (None, None),
 
2306
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2307
                                        (None, current_path_info[2]),
 
2308
                                        (None, new_executable))
 
2309
                                # dont descend into this unversioned path if it is
 
2310
                                # a dir
 
2311
                                if current_path_info[2] in ('directory',
 
2312
                                                            'tree-reference'):
 
2313
                                    del current_dir_info[1][path_index]
 
2314
                                    path_index -= 1
 
2315
                                path_index += 1
 
2316
 
 
2317
                        # This dir info has been handled, go to the next
 
2318
                        try:
 
2319
                            current_dir_info = dir_iterator.next()
 
2320
                        except StopIteration:
 
2321
                            current_dir_info = None
 
2322
                    else:
 
2323
                        # We have a dirblock entry for this location, but there
 
2324
                        # is no filesystem path for this. This is most likely
 
2325
                        # because a directory was removed from the disk.
 
2326
                        # We don't have to report the missing directory,
 
2327
                        # because that should have already been handled, but we
 
2328
                        # need to handle all of the files that are contained
 
2329
                        # within.
 
2330
                        for current_entry in current_block[1]:
 
2331
                            # entry referring to file not present on disk.
 
2332
                            # advance the entry only, after processing.
 
2333
                            result = _process_entry(current_entry, None)
 
2334
                            if result is not None:
 
2335
                                if result is not uninteresting:
 
2336
                                    yield result
 
2337
                        block_index +=1
 
2338
                        if (block_index < len(state._dirblocks) and
 
2339
                            osutils.is_inside(current_root,
 
2340
                                              state._dirblocks[block_index][0])):
 
2341
                            current_block = state._dirblocks[block_index]
 
2342
                        else:
 
2343
                            current_block = None
 
2344
                    continue
 
2345
                entry_index = 0
 
2346
                if current_block and entry_index < len(current_block[1]):
 
2347
                    current_entry = current_block[1][entry_index]
 
2348
                else:
 
2349
                    current_entry = None
 
2350
                advance_entry = True
 
2351
                path_index = 0
 
2352
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2353
                    current_path_info = current_dir_info[1][path_index]
 
2354
                    if current_path_info[2] == 'directory':
 
2355
                        if self.target._directory_is_tree_reference(
 
2356
                            current_path_info[0].decode('utf8')):
 
2357
                            current_path_info = current_path_info[:2] + \
 
2358
                                ('tree-reference',) + current_path_info[3:]
 
2359
                else:
 
2360
                    current_path_info = None
 
2361
                advance_path = True
 
2362
                path_handled = False
 
2363
                while (current_entry is not None or
 
2364
                    current_path_info is not None):
 
2365
                    if current_entry is None:
 
2366
                        # the check for path_handled when the path is adnvaced
 
2367
                        # will yield this path if needed.
 
2368
                        pass
 
2369
                    elif current_path_info is None:
 
2370
                        # no path is fine: the per entry code will handle it.
 
2371
                        result = _process_entry(current_entry, current_path_info)
 
2372
                        if result is not None:
 
2373
                            if result is not uninteresting:
 
2374
                                yield result
 
2375
                    elif (current_entry[0][1] != current_path_info[1]
 
2376
                          or current_entry[1][target_index][0] in 'ar'):
 
2377
                        # The current path on disk doesn't match the dirblock
 
2378
                        # record. Either the dirblock is marked as absent, or
 
2379
                        # the file on disk is not present at all in the
 
2380
                        # dirblock. Either way, report about the dirblock
 
2381
                        # entry, and let other code handle the filesystem one.
 
2382
 
 
2383
                        # Compare the basename for these files to determine
 
2384
                        # which comes first
 
2385
                        if current_path_info[1] < current_entry[0][1]:
 
2386
                            # extra file on disk: pass for now, but only
 
2387
                            # increment the path, not the entry
 
2388
                            advance_entry = False
 
2389
                        else:
 
2390
                            # entry referring to file not present on disk.
 
2391
                            # advance the entry only, after processing.
 
2392
                            result = _process_entry(current_entry, None)
 
2393
                            if result is not None:
 
2394
                                if result is not uninteresting:
 
2395
                                    yield result
 
2396
                            advance_path = False
 
2397
                    else:
 
2398
                        result = _process_entry(current_entry, current_path_info)
 
2399
                        if result is not None:
 
2400
                            path_handled = True
 
2401
                            if result is not uninteresting:
 
2402
                                yield result
 
2403
                    if advance_entry and current_entry is not None:
 
2404
                        entry_index += 1
 
2405
                        if entry_index < len(current_block[1]):
 
2406
                            current_entry = current_block[1][entry_index]
 
2407
                        else:
 
2408
                            current_entry = None
 
2409
                    else:
 
2410
                        advance_entry = True # reset the advance flaga
 
2411
                    if advance_path and current_path_info is not None:
 
2412
                        if not path_handled:
 
2413
                            # unversioned in all regards
 
2414
                            if want_unversioned:
 
2415
                                new_executable = bool(
 
2416
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2417
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2418
                                yield (None,
 
2419
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2420
                                    True,
 
2421
                                    (False, False),
 
2422
                                    (None, None),
 
2423
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2424
                                    (None, current_path_info[2]),
 
2425
                                    (None, new_executable))
 
2426
                            # dont descend into this unversioned path if it is
 
2427
                            # a dir
 
2428
                            if current_path_info[2] in ('directory'):
 
2429
                                del current_dir_info[1][path_index]
 
2430
                                path_index -= 1
 
2431
                        # dont descend the disk iterator into any tree 
 
2432
                        # paths.
 
2433
                        if current_path_info[2] == 'tree-reference':
 
2434
                            del current_dir_info[1][path_index]
 
2435
                            path_index -= 1
 
2436
                        path_index += 1
 
2437
                        if path_index < len(current_dir_info[1]):
 
2438
                            current_path_info = current_dir_info[1][path_index]
 
2439
                            if current_path_info[2] == 'directory':
 
2440
                                if self.target._directory_is_tree_reference(
 
2441
                                    current_path_info[0].decode('utf8')):
 
2442
                                    current_path_info = current_path_info[:2] + \
 
2443
                                        ('tree-reference',) + current_path_info[3:]
 
2444
                        else:
 
2445
                            current_path_info = None
 
2446
                        path_handled = False
 
2447
                    else:
 
2448
                        advance_path = True # reset the advance flagg.
 
2449
                if current_block is not None:
 
2450
                    block_index += 1
 
2451
                    if (block_index < len(state._dirblocks) and
 
2452
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2453
                        current_block = state._dirblocks[block_index]
 
2454
                    else:
 
2455
                        current_block = None
 
2456
                if current_dir_info is not None:
 
2457
                    try:
 
2458
                        current_dir_info = dir_iterator.next()
 
2459
                    except StopIteration:
 
2460
                        current_dir_info = None
 
2461
 
2113
2462
 
2114
2463
    @staticmethod
2115
2464
    def is_compatible(source, target):
2116
2465
        # the target must be a dirstate working tree
2117
 
        if not isinstance(target, DirStateWorkingTree):
 
2466
        if not isinstance(target, WorkingTree4):
2118
2467
            return False
2119
 
        # the source must be a revtree or dirstate rev tree.
 
2468
        # the source must be a revtreee or dirstate rev tree.
2120
2469
        if not isinstance(source,
2121
2470
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2122
2471
            return False
2123
2472
        # the source revid must be in the target dirstate
2124
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2473
        if not (source._revision_id == NULL_REVISION or
2125
2474
            source._revision_id in target.get_parent_ids()):
2126
 
            # TODO: what about ghosts? it may well need to
 
2475
            # TODO: what about ghosts? it may well need to 
2127
2476
            # check for them explicitly.
2128
2477
            return False
2129
2478
        return True
2139
2488
 
2140
2489
    def convert(self, tree):
2141
2490
        # lock the control files not the tree, so that we dont get tree
2142
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2491
        # on-unlock behaviours, and so that noone else diddles with the 
2143
2492
        # tree during upgrade.
2144
2493
        tree._control_files.lock_write()
2145
2494
        try:
2171
2520
 
2172
2521
    def update_format(self, tree):
2173
2522
        """Change the format marker."""
2174
 
        tree._transport.put_bytes('format',
2175
 
            self.target_format.get_format_string(),
2176
 
            mode=tree.bzrdir._get_file_mode())
2177
 
 
2178
 
 
2179
 
class Converter4to5(object):
2180
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2181
 
 
2182
 
    def __init__(self):
2183
 
        self.target_format = WorkingTreeFormat5()
2184
 
 
2185
 
    def convert(self, tree):
2186
 
        # lock the control files not the tree, so that we don't get tree
2187
 
        # on-unlock behaviours, and so that no-one else diddles with the
2188
 
        # tree during upgrade.
2189
 
        tree._control_files.lock_write()
2190
 
        try:
2191
 
            self.update_format(tree)
2192
 
        finally:
2193
 
            tree._control_files.unlock()
2194
 
 
2195
 
    def update_format(self, tree):
2196
 
        """Change the format marker."""
2197
 
        tree._transport.put_bytes('format',
2198
 
            self.target_format.get_format_string(),
2199
 
            mode=tree.bzrdir._get_file_mode())
2200
 
 
2201
 
 
2202
 
class Converter4or5to6(object):
2203
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2204
 
 
2205
 
    def __init__(self):
2206
 
        self.target_format = WorkingTreeFormat6()
2207
 
 
2208
 
    def convert(self, tree):
2209
 
        # lock the control files not the tree, so that we don't get tree
2210
 
        # on-unlock behaviours, and so that no-one else diddles with the
2211
 
        # tree during upgrade.
2212
 
        tree._control_files.lock_write()
2213
 
        try:
2214
 
            self.init_custom_control_files(tree)
2215
 
            self.update_format(tree)
2216
 
        finally:
2217
 
            tree._control_files.unlock()
2218
 
 
2219
 
    def init_custom_control_files(self, tree):
2220
 
        """Initialize custom control files."""
2221
 
        tree._transport.put_bytes('views', '',
2222
 
            mode=tree.bzrdir._get_file_mode())
2223
 
 
2224
 
    def update_format(self, tree):
2225
 
        """Change the format marker."""
2226
 
        tree._transport.put_bytes('format',
2227
 
            self.target_format.get_format_string(),
2228
 
            mode=tree.bzrdir._get_file_mode())
 
2523
        tree._control_files.put_utf8('format',
 
2524
            self.target_format.get_format_string())