~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2008-04-09 20:13:46 UTC
  • mto: This revision was merged to the branch mainline in revision 3350.
  • Revision ID: robertc@robertcollins.net-20080409201346-2m5dpim3cowzwacs
Spelling in NEWS.

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
 
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
56
75
from bzrlib.mutabletree import needs_tree_write_lock
57
76
from bzrlib.osutils import (
58
77
    file_kind,
59
78
    isdir,
 
79
    normpath,
60
80
    pathjoin,
 
81
    rand_chars,
61
82
    realpath,
62
83
    safe_unicode,
 
84
    splitpath,
63
85
    )
64
 
from bzrlib.trace import mutter
 
86
from bzrlib.trace import mutter, note
65
87
from bzrlib.transport.local import LocalTransport
66
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
        )
67
97
from bzrlib.tree import Tree
68
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
99
 
70
100
 
71
 
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
 
72
120
    def __init__(self, basedir,
73
121
                 branch,
74
122
                 _control_files=None,
83
131
        """
84
132
        self._format = _format
85
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
86
136
        basedir = safe_unicode(basedir)
87
137
        mutter("opening working tree %r", basedir)
88
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
89
141
        self.basedir = realpath(basedir)
90
142
        # if branch is at our basedir and is a format 6 or less
91
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
92
146
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
147
        self._dirty = None
95
148
        #-------------
96
149
        # during a read or write lock these objects are set, and are
100
153
        #-------------
101
154
        self._setup_directory_is_tree_reference()
102
155
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
156
 
108
157
    @needs_tree_write_lock
109
158
    def _add(self, files, ids, kinds):
137
186
    @needs_tree_write_lock
138
187
    def add_reference(self, sub_tree):
139
188
        # use standard implementation, which calls back to self._add
140
 
        #
 
189
        # 
141
190
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
191
        # it's just recorded at the moment of commit. 
143
192
        self._add_reference(sub_tree)
144
193
 
145
194
    def break_lock(self):
184
233
            WorkingTree3._comparison_data(self, entry, path)
185
234
        # it looks like a plain directory, but it's really a reference -- see
186
235
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            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)):
189
239
            kind = 'tree-reference'
190
240
        return kind, executable, stat_value
191
241
 
217
267
            return self._dirstate
218
268
        local_path = self.bzrdir.get_workingtree_transport(None
219
269
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
222
271
        return self._dirstate
223
272
 
224
 
    def _sha1_provider(self):
225
 
        """A function that returns a SHA1Provider suitable for this tree.
226
 
 
227
 
        :return: None if content filtering is not supported by this tree.
228
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
229
 
          form of files, i.e. after read filters are applied.
230
 
        """
231
 
        if self.supports_content_filtering():
232
 
            return ContentFilterAwareSHA1Provider(self)
233
 
        else:
234
 
            return None
235
 
 
236
273
    def filter_unversioned_files(self, paths):
237
274
        """Filter out paths that are versioned.
238
275
 
270
307
 
271
308
    def _generate_inventory(self):
272
309
        """Create and set self.inventory from the dirstate object.
273
 
 
 
310
        
274
311
        This is relatively expensive: we have to walk the entire dirstate.
275
312
        Ideally we would not, and can deprecate this function.
276
313
        """
280
317
        state._read_dirblocks_if_needed()
281
318
        root_key, current_entry = self._get_entry(path='')
282
319
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
320
        assert current_entry[0][0] == 'd' # directory
285
321
        inv = Inventory(root_id=current_id)
286
322
        # Turn some things into local variables
287
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
320
356
                    # add this entry to the parent map.
321
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
358
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            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)
327
364
                    inv_entry.reference_revision = link_or_sha1 or None
328
365
                elif kind != 'symlink':
329
366
                    raise AssertionError("unknown kind %r" % kind)
330
367
                # These checks cost us around 40ms on a 55k entry tree
331
 
                if file_id in inv_byid:
332
 
                    raise AssertionError('file_id %s already in'
333
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
334
 
                if name_unicode in parent_ie.children:
335
 
                    raise AssertionError('name %r already in parent'
336
 
                        % (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
337
371
                inv_byid[file_id] = inv_entry
338
372
                parent_ie.children[name_unicode] = inv_entry
339
373
        self._inventory = inv
344
378
        If either file_id or path is supplied, it is used as the key to lookup.
345
379
        If both are supplied, the fastest lookup is used, and an error is
346
380
        raised if they do not both point at the same row.
347
 
 
 
381
        
348
382
        :param file_id: An optional unicode file_id to be looked up.
349
383
        :param path: An optional unicode path to be looked up.
350
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
374
408
                    return None
375
409
                else:
376
410
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
379
413
        if entry[1][0][0] == 'f':
380
 
            if link_or_sha1 is None:
381
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
382
 
                try:
383
 
                    sha1 = osutils.sha_file(file_obj)
384
 
                finally:
385
 
                    file_obj.close()
386
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
387
 
                return sha1
388
 
            else:
389
 
                return link_or_sha1
 
414
            return link_or_sha1
390
415
        return None
391
416
 
392
417
    def _get_inventory(self):
406
431
    @needs_read_lock
407
432
    def get_parent_ids(self):
408
433
        """See Tree.get_parent_ids.
409
 
 
 
434
        
410
435
        This implementation requests the ids list from the dirstate file.
411
436
        """
412
437
        return self.current_dirstate().get_parent_ids()
434
459
        return osutils.lexists(pathjoin(
435
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
461
 
437
 
    def has_or_had_id(self, file_id):
438
 
        state = self.current_dirstate()
439
 
        row, parents = self._get_entry(file_id=file_id)
440
 
        return row is not None
441
 
 
442
462
    @needs_read_lock
443
463
    def id2path(self, file_id):
444
464
        "Convert a file-id to a path."
508
528
        return iter(result)
509
529
 
510
530
    def iter_references(self):
511
 
        if not self._repo_supports_tree_reference:
512
 
            # When the repo doesn't support references, we will have nothing to
513
 
            # return
514
 
            return
515
531
        for key, tree_details in self.current_dirstate()._iter_entries():
516
532
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
533
                # not relevant to the working tree
519
535
            if not key[1]:
520
536
                # the root is not a reference.
521
537
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
538
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
539
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
540
                if self._kind(path) == 'tree-reference':
 
541
                    yield path, key[2]
526
542
            except errors.NoSuchFile:
527
543
                # path is missing on disk.
528
544
                continue
529
545
 
530
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
531
 
        """See MutableTree._observed_sha1."""
532
 
        state = self.current_dirstate()
533
 
        entry = self._get_entry(file_id=file_id, path=path)
534
 
        state._observed_sha1(entry, sha1, statvalue)
535
 
 
536
546
    def kind(self, file_id):
537
547
        """Return the kind of a file.
538
548
 
542
552
        Note: The caller is expected to take a read-lock before calling this.
543
553
        """
544
554
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
555
        assert relpath != None, \
 
556
            "path for id {%s} is None!" % file_id
548
557
        return self._kind(relpath)
549
558
 
550
559
    def _kind(self, relpath):
551
560
        abspath = self.abspath(relpath)
552
561
        kind = file_kind(abspath)
553
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
554
 
            entry = self._get_entry(path=relpath)
555
 
            if entry[1] is not None:
556
 
                if entry[1][0][0] == 't':
557
 
                    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'
558
566
        return kind
559
567
 
560
568
    @needs_read_lock
623
631
        result = []
624
632
        if not from_paths:
625
633
            return result
 
634
 
626
635
        state = self.current_dirstate()
627
 
        if isinstance(from_paths, basestring):
628
 
            raise ValueError()
 
636
 
 
637
        assert not isinstance(from_paths, basestring)
629
638
        to_dir_utf8 = to_dir.encode('utf8')
630
639
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
631
640
        id_index = state._get_id_index()
691
700
            from_entry = self._get_entry(path=from_rel)
692
701
            if from_entry == (None, None):
693
702
                raise errors.BzrMoveFailedError(from_rel,to_dir,
694
 
                    errors.NotVersionedError(path=from_rel))
 
703
                    errors.NotVersionedError(path=str(from_rel)))
695
704
 
696
705
            from_id = from_entry[0][2]
697
706
            to_rel = pathjoin(to_dir, from_tail)
786
795
                if minikind == 'd':
787
796
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
788
797
                        """Recursively update all entries in this dirblock."""
789
 
                        if from_dir == '':
790
 
                            raise AssertionError("renaming root not supported")
 
798
                        assert from_dir != '', "renaming root not supported"
791
799
                        from_key = (from_dir, '')
792
800
                        from_block_idx, present = \
793
801
                            state._find_block_index_from_key(from_key)
806
814
 
807
815
                        # Grab a copy since move_one may update the list.
808
816
                        for entry in from_block[1][:]:
809
 
                            if not (entry[0][0] == from_dir):
810
 
                                raise AssertionError()
 
817
                            assert entry[0][0] == from_dir
811
818
                            cur_details = entry[1][0]
812
819
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
813
820
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
872
879
        for tree in trees:
873
880
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
874
881
                parents):
875
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
876
 
                    trees, require_versioned)
 
882
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
877
883
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
878
884
        # -- make all paths utf8 --
879
885
        paths_utf8 = set()
940
946
                raise errors.PathsNotVersionedError(paths)
941
947
        # -- remove redundancy in supplied paths to prevent over-scanning --
942
948
        search_paths = osutils.minimum_path_selection(paths)
943
 
        # sketch:
 
949
        # sketch: 
944
950
        # for all search_indexs in each path at or under each element of
945
951
        # search_paths, if the detail is relocated: add the id, and add the
946
952
        # relocated path as one to search if its not searched already. If the
1002
1008
 
1003
1009
    def read_working_inventory(self):
1004
1010
        """Read the working inventory.
1005
 
 
 
1011
        
1006
1012
        This is a meaningless operation for dirstate, but we obey it anyhow.
1007
1013
        """
1008
1014
        return self.inventory
1026
1032
    def set_last_revision(self, new_revision):
1027
1033
        """Change the last revision in the working tree."""
1028
1034
        parents = self.get_parent_ids()
1029
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1030
 
            if len(parents) >= 2:
1031
 
                raise AssertionError(
1032
 
                    "setting the last parent to none with a pending merge is "
1033
 
                    "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.")
1034
1039
            self.set_parent_ids([])
1035
1040
        else:
1036
1041
            self.set_parent_ids([new_revision] + parents[1:],
1039
1044
    @needs_tree_write_lock
1040
1045
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1041
1046
        """Set the parent ids to revision_ids.
1042
 
 
 
1047
        
1043
1048
        See also set_parent_trees. This api will try to retrieve the tree data
1044
1049
        for each element of revision_ids from the trees repository. If you have
1045
1050
        tree data already available, it is more efficient to use
1077
1082
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1078
1083
        real_trees = []
1079
1084
        ghosts = []
1080
 
 
1081
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1082
 
        graph = self.branch.repository.get_graph()
1083
 
        heads = graph.heads(parent_ids)
1084
 
        accepted_revisions = set()
1085
 
 
1086
1085
        # convert absent trees to the null tree, which we convert back to
1087
1086
        # missing on access.
1088
1087
        for rev_id, tree in parents_list:
1089
 
            if len(accepted_revisions) > 0:
1090
 
                # we always accept the first tree
1091
 
                if rev_id in accepted_revisions or rev_id not in heads:
1092
 
                    # We have already included either this tree, or its
1093
 
                    # descendent, so we skip it.
1094
 
                    continue
1095
1088
            _mod_revision.check_not_reserved_id(rev_id)
1096
1089
            if tree is not None:
1097
1090
                real_trees.append((rev_id, tree))
1098
1091
            else:
1099
1092
                real_trees.append((rev_id,
1100
 
                    self.branch.repository.revision_tree(
1101
 
                        _mod_revision.NULL_REVISION)))
 
1093
                    self.branch.repository.revision_tree(None)))
1102
1094
                ghosts.append(rev_id)
1103
 
            accepted_revisions.add(rev_id)
1104
1095
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1105
1096
        self._make_dirty(reset_inventory=False)
1106
1097
 
1199
1190
                # just forget the whole block.
1200
1191
                entry_index = 0
1201
1192
                while entry_index < len(block[1]):
 
1193
                    # Mark this file id as having been removed
1202
1194
                    entry = block[1][entry_index]
1203
 
                    if entry[1][0][0] in 'ar':
1204
 
                        # 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)):
1205
1199
                        entry_index += 1
1206
 
                    else:
1207
 
                        # Mark this file id as having been removed
1208
 
                        ids_to_unversion.discard(entry[0][2])
1209
 
                        if not state._make_absent(entry):
1210
 
                            # The block has not shrunk.
1211
 
                            entry_index += 1
1212
1200
                # go to the next block. (At the moment we dont delete empty
1213
1201
                # dirblocks)
1214
1202
                block_index += 1
1252
1240
 
1253
1241
    def update_basis_by_delta(self, new_revid, delta):
1254
1242
        """See MutableTree.update_basis_by_delta."""
1255
 
        if self.last_revision() == new_revid:
1256
 
            raise AssertionError()
 
1243
        assert self.last_revision() != new_revid
1257
1244
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1258
1245
 
1259
1246
    @needs_read_lock
1263
1250
    @needs_tree_write_lock
1264
1251
    def _write_inventory(self, inv):
1265
1252
        """Write inventory as the current inventory."""
1266
 
        if self._dirty:
1267
 
            raise AssertionError("attempting to write an inventory when the "
1268
 
                "dirstate is dirty will lose pending changes")
1269
 
        had_inventory = self._inventory is not None
1270
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1271
 
        # working inventory. We do this because self.inventory may be inv, or
1272
 
        # may have been modified, and either case would prevent a clean delta
1273
 
        # being created.
1274
 
        self._inventory = None
1275
 
        # generate a delta,
1276
 
        delta = inv._make_delta(self.inventory)
1277
 
        # and apply it.
1278
 
        self.apply_inventory_delta(delta)
1279
 
        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:
1280
1258
            self._inventory = inv
1281
1259
        self.flush()
1282
1260
 
1283
1261
 
1284
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1285
 
 
1286
 
    def __init__(self, tree):
1287
 
        self.tree = tree
1288
 
 
1289
 
    def sha1(self, abspath):
1290
 
        """See dirstate.SHA1Provider.sha1()."""
1291
 
        filters = self.tree._content_filter_stack(
1292
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1293
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1294
 
 
1295
 
    def stat_and_sha1(self, abspath):
1296
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1297
 
        filters = self.tree._content_filter_stack(
1298
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1299
 
        file_obj = file(abspath, 'rb', 65000)
1300
 
        try:
1301
 
            statvalue = os.fstat(file_obj.fileno())
1302
 
            if filters:
1303
 
                file_obj = filtered_input_file(file_obj, filters)
1304
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1305
 
        finally:
1306
 
            file_obj.close()
1307
 
        return statvalue, sha1
1308
 
 
1309
 
 
1310
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1311
 
    """Dirstate working tree that supports content filtering.
1312
 
 
1313
 
    The dirstate holds the hash and size of the canonical form of the file, 
1314
 
    and most methods must return that.
1315
 
    """
1316
 
 
1317
 
    def _file_content_summary(self, path, stat_result):
1318
 
        # This is to support the somewhat obsolete path_content_summary method
1319
 
        # with content filtering: see
1320
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1321
 
        #
1322
 
        # If the dirstate cache is up to date and knows the hash and size,
1323
 
        # return that.
1324
 
        # Otherwise if there are no content filters, return the on-disk size
1325
 
        # and leave the hash blank.
1326
 
        # Otherwise, read and filter the on-disk file and use its size and
1327
 
        # hash.
1328
 
        #
1329
 
        # The dirstate doesn't store the size of the canonical form so we
1330
 
        # can't trust it for content-filtered trees.  We just return None.
1331
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1332
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1333
 
        return ('file', None, executable, dirstate_sha1)
1334
 
 
1335
 
 
1336
 
class WorkingTree4(DirStateWorkingTree):
1337
 
    """This is the Format 4 working tree.
1338
 
 
1339
 
    This differs from WorkingTree3 by:
1340
 
     - Having a consolidated internal dirstate, stored in a
1341
 
       randomly-accessible sorted file on disk.
1342
 
     - Not having a regular inventory attribute.  One can be synthesized
1343
 
       on demand but this is expensive and should be avoided.
1344
 
 
1345
 
    This is new in bzr 0.15.
1346
 
    """
1347
 
 
1348
 
 
1349
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1350
 
    """This is the Format 5 working tree.
1351
 
 
1352
 
    This differs from WorkingTree4 by:
1353
 
     - Supporting content filtering.
1354
 
 
1355
 
    This is new in bzr 1.11.
1356
 
    """
1357
 
 
1358
 
 
1359
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1360
 
    """This is the Format 6 working tree.
1361
 
 
1362
 
    This differs from WorkingTree5 by:
1363
 
     - Supporting a current view that may mask the set of files in a tree
1364
 
       impacted by most user operations.
1365
 
 
1366
 
    This is new in bzr 1.14.
1367
 
    """
1368
 
 
1369
 
    def _make_views(self):
1370
 
        return views.PathBasedViews(self)
1371
 
 
1372
 
 
1373
 
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"
1374
1283
 
1375
1284
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1376
1285
                   accelerator_tree=None, hardlink=False):
1394
1303
        control_files = self._open_control_files(a_bzrdir)
1395
1304
        control_files.create_lock()
1396
1305
        control_files.lock_write()
1397
 
        transport.put_bytes('format', self.get_format_string(),
1398
 
            mode=a_bzrdir._get_file_mode())
 
1306
        control_files.put_utf8('format', self.get_format_string())
1399
1307
        if from_branch is not None:
1400
1308
            branch = from_branch
1401
1309
        else:
1407
1315
        state = dirstate.DirState.initialize(local_path)
1408
1316
        state.unlock()
1409
1317
        del state
1410
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1318
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1411
1319
                         branch,
1412
1320
                         _format=self,
1413
1321
                         _bzrdir=a_bzrdir,
1415
1323
        wt._new_tree()
1416
1324
        wt.lock_tree_write()
1417
1325
        try:
1418
 
            self._init_custom_control_files(wt)
1419
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1326
            if revision_id in (None, NULL_REVISION):
1420
1327
                if branch.repository.supports_rich_root():
1421
1328
                    wt._set_root_id(generate_ids.gen_root_id())
1422
1329
                else:
1433
1340
                    pass
1434
1341
            if basis is None:
1435
1342
                basis = branch.repository.revision_tree(revision_id)
1436
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1343
            if revision_id == NULL_REVISION:
1437
1344
                parents_list = []
1438
1345
            else:
1439
1346
                parents_list = [(revision_id, basis)]
1447
1354
                if basis_root_id is not None:
1448
1355
                    wt._set_root_id(basis_root_id)
1449
1356
                    wt.flush()
1450
 
                if wt.supports_content_filtering():
1451
 
                    # The original tree may not have the same content filters
1452
 
                    # applied so we can't safely build the inventory delta from
1453
 
                    # the source tree.
1454
 
                    delta_from_tree = False
1455
 
                else:
1456
 
                    delta_from_tree = True
1457
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1458
 
                # because wt4.apply_inventory_delta does not mutate the input
1459
 
                # inventory entries.
1460
1357
                transform.build_tree(basis, wt, accelerator_tree,
1461
 
                                     hardlink=hardlink,
1462
 
                                     delta_from_tree=delta_from_tree)
 
1358
                                     hardlink=hardlink)
1463
1359
            finally:
1464
1360
                basis.unlock()
1465
1361
        finally:
1467
1363
            wt.unlock()
1468
1364
        return wt
1469
1365
 
1470
 
    def _init_custom_control_files(self, wt):
1471
 
        """Subclasses with custom control files should override this method.
1472
 
 
1473
 
        The working tree and control files are locked for writing when this
1474
 
        method is called.
1475
 
 
1476
 
        :param wt: the WorkingTree object
1477
 
        """
1478
 
 
1479
1366
    def _open(self, a_bzrdir, control_files):
1480
1367
        """Open the tree itself.
1481
1368
 
1482
1369
        :param a_bzrdir: the dir for the tree.
1483
1370
        :param control_files: the control files for the tree.
1484
1371
        """
1485
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1372
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1486
1373
                           branch=a_bzrdir.open_branch(),
1487
1374
                           _format=self,
1488
1375
                           _bzrdir=a_bzrdir,
1489
1376
                           _control_files=control_files)
1490
1377
 
1491
1378
    def __get_matchingbzrdir(self):
1492
 
        return self._get_matchingbzrdir()
1493
 
 
1494
 
    def _get_matchingbzrdir(self):
1495
 
        """Overrideable method to get a bzrdir for testing."""
1496
1379
        # please test against something that will let us do tree references
1497
1380
        return bzrdir.format_registry.make_bzrdir(
1498
1381
            'dirstate-with-subtree')
1500
1383
    _matchingbzrdir = property(__get_matchingbzrdir)
1501
1384
 
1502
1385
 
1503
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1504
 
    """The first consolidated dirstate working tree format.
1505
 
 
1506
 
    This format:
1507
 
        - exists within a metadir controlling .bzr
1508
 
        - includes an explicit version marker for the workingtree control
1509
 
          files, separate from the BzrDir format
1510
 
        - modifies the hash cache format
1511
 
        - is new in bzr 0.15
1512
 
        - uses a LockDir to guard access to it.
1513
 
    """
1514
 
 
1515
 
    upgrade_recommended = False
1516
 
 
1517
 
    _tree_class = WorkingTree4
1518
 
 
1519
 
    def get_format_string(self):
1520
 
        """See WorkingTreeFormat.get_format_string()."""
1521
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1522
 
 
1523
 
    def get_format_description(self):
1524
 
        """See WorkingTreeFormat.get_format_description()."""
1525
 
        return "Working tree format 4"
1526
 
 
1527
 
 
1528
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1529
 
    """WorkingTree format supporting content filtering.
1530
 
    """
1531
 
 
1532
 
    upgrade_recommended = False
1533
 
 
1534
 
    _tree_class = WorkingTree5
1535
 
 
1536
 
    def get_format_string(self):
1537
 
        """See WorkingTreeFormat.get_format_string()."""
1538
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1539
 
 
1540
 
    def get_format_description(self):
1541
 
        """See WorkingTreeFormat.get_format_description()."""
1542
 
        return "Working tree format 5"
1543
 
 
1544
 
    def supports_content_filtering(self):
1545
 
        return True
1546
 
 
1547
 
 
1548
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1549
 
    """WorkingTree format supporting views.
1550
 
    """
1551
 
 
1552
 
    upgrade_recommended = False
1553
 
 
1554
 
    _tree_class = WorkingTree6
1555
 
 
1556
 
    def get_format_string(self):
1557
 
        """See WorkingTreeFormat.get_format_string()."""
1558
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1559
 
 
1560
 
    def get_format_description(self):
1561
 
        """See WorkingTreeFormat.get_format_description()."""
1562
 
        return "Working tree format 6"
1563
 
 
1564
 
    def _init_custom_control_files(self, wt):
1565
 
        """Subclasses with custom control files should override this method."""
1566
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1567
 
 
1568
 
    def supports_content_filtering(self):
1569
 
        return True
1570
 
 
1571
 
    def supports_views(self):
1572
 
        return True
1573
 
 
1574
 
 
1575
1386
class DirStateRevisionTree(Tree):
1576
 
    """A revision tree pulling the inventory from a dirstate.
1577
 
    
1578
 
    Note that this is one of the historical (ie revision) trees cached in the
1579
 
    dirstate for easy access, not the workingtree.
1580
 
    """
 
1387
    """A revision tree pulling the inventory from a dirstate."""
1581
1388
 
1582
1389
    def __init__(self, dirstate, revision_id, repository):
1583
1390
        self._dirstate = dirstate
1586
1393
        self._inventory = None
1587
1394
        self._locked = 0
1588
1395
        self._dirstate_locked = False
1589
 
        self._repo_supports_tree_reference = getattr(
1590
 
            repository._format, "supports_tree_reference",
1591
 
            False)
1592
1396
 
1593
1397
    def __repr__(self):
1594
1398
        return "<%s of %s in %s>" % \
1597
1401
    def annotate_iter(self, file_id,
1598
1402
                      default_revision=_mod_revision.CURRENT_REVISION):
1599
1403
        """See Tree.annotate_iter"""
1600
 
        text_key = (file_id, self.inventory[file_id].revision)
1601
 
        annotations = self._repository.texts.annotate(text_key)
1602
 
        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)
1603
1406
 
1604
1407
    def _get_ancestors(self, default_revision):
1605
1408
        return set(self._repository.get_ancestry(self._revision_id,
1634
1437
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1635
1438
        return path_utf8.decode('utf8')
1636
1439
 
1637
 
    def iter_references(self):
1638
 
        if not self._repo_supports_tree_reference:
1639
 
            # When the repo doesn't support references, we will have nothing to
1640
 
            # return
1641
 
            return iter([])
1642
 
        # Otherwise, fall back to the default implementation
1643
 
        return super(DirStateRevisionTree, self).iter_references()
1644
 
 
1645
1440
    def _get_parent_index(self):
1646
1441
        """Return the index in the dirstate referenced by this tree."""
1647
1442
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1652
1447
        If either file_id or path is supplied, it is used as the key to lookup.
1653
1448
        If both are supplied, the fastest lookup is used, and an error is
1654
1449
        raised if they do not both point at the same row.
1655
 
 
 
1450
        
1656
1451
        :param file_id: An optional unicode file_id to be looked up.
1657
1452
        :param path: An optional unicode path to be looked up.
1658
1453
        :return: The dirstate row tuple for path/file_id, or (None, None)
1672
1467
 
1673
1468
        This is relatively expensive: we have to walk the entire dirstate.
1674
1469
        """
1675
 
        if not self._locked:
1676
 
            raise AssertionError(
1677
 
                'cannot generate inventory of an unlocked '
1678
 
                'dirstate revision tree')
 
1470
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1471
            'dirstate revision tree'
1679
1472
        # separate call for profiling - makes it clear where the costs are.
1680
1473
        self._dirstate._read_dirblocks_if_needed()
1681
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1682
 
            raise AssertionError(
1683
 
                'parent %s has disappeared from %s' % (
1684
 
                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())
1685
1477
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1686
1478
        # This is identical now to the WorkingTree _generate_inventory except
1687
1479
        # for the tree index use.
1688
1480
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1689
1481
        current_id = root_key[2]
1690
 
        if current_entry[parent_index][0] != 'd':
1691
 
            raise AssertionError()
 
1482
        assert current_entry[parent_index][0] == 'd'
1692
1483
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1693
1484
        inv.root.revision = current_entry[parent_index][4]
1694
1485
        # Turn some things into local variables
1734
1525
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1735
1526
                            % entry)
1736
1527
                # These checks cost us around 40ms on a 55k entry tree
1737
 
                if file_id in inv_byid:
1738
 
                    raise AssertionError('file_id %s already in'
1739
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1740
 
                if name_unicode in parent_ie.children:
1741
 
                    raise AssertionError('name %r already in parent'
1742
 
                        % (name_unicode,))
 
1528
                assert file_id not in inv_byid
 
1529
                assert name_unicode not in parent_ie.children
1743
1530
                inv_byid[file_id] = inv_entry
1744
1531
                parent_ie.children[name_unicode] = inv_entry
1745
1532
        self._inventory = inv
1755
1542
            return None
1756
1543
        parent_index = self._get_parent_index()
1757
1544
        last_changed_revision = entry[1][parent_index][4]
1758
 
        try:
1759
 
            rev = self._repository.get_revision(last_changed_revision)
1760
 
        except errors.NoSuchRevision:
1761
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1762
 
        return rev.timestamp
 
1545
        return self._repository.get_revision(last_changed_revision).timestamp
1763
1546
 
1764
1547
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1765
1548
        entry = self._get_entry(file_id=file_id, path=path)
1769
1552
            return parent_details[1]
1770
1553
        return None
1771
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
 
1772
1563
    def get_file(self, file_id, path=None):
1773
1564
        return StringIO(self.get_file_text(file_id))
1774
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
 
1775
1572
    def get_file_size(self, file_id):
1776
 
        """See Tree.get_file_size"""
1777
1573
        return self.inventory[file_id].text_size
1778
1574
 
1779
 
    def get_file_text(self, file_id, path=None):
1780
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1781
 
        return ''.join(content)
 
1575
    def get_file_text(self, file_id):
 
1576
        return ''.join(self.get_file_lines(file_id))
1782
1577
 
1783
1578
    def get_reference_revision(self, file_id, path=None):
1784
1579
        return self.inventory[file_id].reference_revision
1803
1598
        if entry[1][parent_index][0] != 'l':
1804
1599
            return None
1805
1600
        else:
1806
 
            target = entry[1][parent_index][1]
1807
 
            target = target.decode('utf8')
1808
 
            return target
 
1601
            # At present, none of the tree implementations supports non-ascii
 
1602
            # symlink targets. So we will just assume that the dirstate path is
 
1603
            # correct.
 
1604
            return entry[1][parent_index][1]
1809
1605
 
1810
1606
    def get_revision_id(self):
1811
1607
        """Return the revision id for this tree."""
1830
1626
 
1831
1627
    def kind(self, file_id):
1832
1628
        entry = self._get_entry(file_id=file_id)[1]
1833
 
        if entry is None:
 
1629
        if entry == None:
1834
1630
            raise errors.NoSuchId(tree=self, file_id=file_id)
1835
 
        parent_index = self._get_parent_index()
1836
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1631
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1837
1632
 
1838
1633
    def stored_kind(self, file_id):
1839
1634
        """See Tree.stored_kind"""
1859
1654
            return None
1860
1655
        return ie.executable
1861
1656
 
1862
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1657
    def list_files(self, include_root=False):
1863
1658
        # We use a standard implementation, because DirStateRevisionTree is
1864
1659
        # dealing with one of the parents of the current state
1865
1660
        inv = self._get_inventory()
1866
 
        if from_dir is None:
1867
 
            from_dir_id = None
1868
 
        else:
1869
 
            from_dir_id = inv.path2id(from_dir)
1870
 
            if from_dir_id is None:
1871
 
                # Directory not versioned
1872
 
                return
1873
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1874
 
        if inv.root is not None and not include_root and from_dir is None:
 
1661
        entries = inv.iter_entries()
 
1662
        if self.inventory.root is not None and not include_root:
1875
1663
            entries.next()
1876
1664
        for path, entry in entries:
1877
1665
            yield path, 'V', entry.kind, entry.file_id, entry
1910
1698
                self._dirstate_locked = False
1911
1699
            self._repository.unlock()
1912
1700
 
1913
 
    @needs_read_lock
1914
 
    def supports_tree_reference(self):
1915
 
        return self._repo_supports_tree_reference
1916
 
 
1917
1701
    def walkdirs(self, prefix=""):
1918
1702
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1919
 
        # implementation based on an inventory.
 
1703
        # implementation based on an inventory.  
1920
1704
        # This should be cleaned up to use the much faster Dirstate code
1921
1705
        # So for now, we just build up the parent inventory, and extract
1922
1706
        # it the same way RevisionTree does.
1951
1735
 
1952
1736
class InterDirStateTree(InterTree):
1953
1737
    """Fast path optimiser for changes_from with dirstate trees.
1954
 
 
1955
 
    This is used only when both trees are in the dirstate working file, and
1956
 
    the source is any parent within the dirstate, and the destination is
 
1738
    
 
1739
    This is used only when both trees are in the dirstate working file, and 
 
1740
    the source is any parent within the dirstate, and the destination is 
1957
1741
    the current working tree of the same dirstate.
1958
1742
    """
1959
1743
    # this could be generalized to allow comparisons between any trees in the
1972
1756
        target.set_parent_ids([revid])
1973
1757
        return target.basis_tree(), target
1974
1758
 
1975
 
    @classmethod
1976
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1977
 
        result = klass.make_source_parent_tree(source, target)
1978
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1979
 
        return result
1980
 
 
1981
 
    @classmethod
1982
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
1983
 
                                                  target):
1984
 
        from bzrlib.tests.test__dirstate_helpers import \
1985
 
            compiled_dirstate_helpers_feature
1986
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
1987
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
1988
 
        result = klass.make_source_parent_tree(source, target)
1989
 
        result[1]._iter_changes = ProcessEntryC
1990
 
        return result
1991
 
 
1992
1759
    _matching_from_tree_format = WorkingTreeFormat4()
1993
1760
    _matching_to_tree_format = WorkingTreeFormat4()
1994
 
 
1995
 
    @classmethod
1996
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
1997
 
        # This method shouldn't be called, because we have python and C
1998
 
        # specific flavours.
1999
 
        raise NotImplementedError
 
1761
    _test_mutable_trees_to_test_trees = make_source_parent_tree
2000
1762
 
2001
1763
    def iter_changes(self, include_unchanged=False,
2002
1764
                      specific_files=None, pb=None, extra_trees=[],
2020
1782
            output. An unversioned file is defined as one with (False, False)
2021
1783
            for the versioned pair.
2022
1784
        """
 
1785
        utf8_decode = cache_utf8._utf8_decode
 
1786
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1787
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1788
        # NB: show_status depends on being able to pass in non-versioned files
 
1789
        # and report them as unknown
2023
1790
        # TODO: handle extra trees in the dirstate.
2024
1791
        if (extra_trees or specific_files == []):
2025
1792
            # we can't fast-path these cases (yet)
2026
 
            return super(InterDirStateTree, self).iter_changes(
 
1793
            for f in super(InterDirStateTree, self).iter_changes(
2027
1794
                include_unchanged, specific_files, pb, extra_trees,
2028
 
                require_versioned, want_unversioned=want_unversioned)
 
1795
                require_versioned, want_unversioned=want_unversioned):
 
1796
                yield f
 
1797
            return
2029
1798
        parent_ids = self.target.get_parent_ids()
2030
 
        if not (self.source._revision_id in parent_ids
2031
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2032
 
            raise AssertionError(
2033
 
                "revision {%s} is not stored in {%s}, but %s "
2034
 
                "can only be used for trees stored in the dirstate"
2035
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1799
        assert (self.source._revision_id in parent_ids
 
1800
                or self.source._revision_id == NULL_REVISION), \
 
1801
                "revision {%s} is not stored in {%s}, but %s " \
 
1802
                "can only be used for trees stored in the dirstate" \
 
1803
                % (self.source._revision_id, self.target, self.iter_changes)
2036
1804
        target_index = 0
2037
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1805
        if self.source._revision_id == NULL_REVISION:
2038
1806
            source_index = None
2039
1807
            indices = (target_index,)
2040
1808
        else:
2041
 
            if not (self.source._revision_id in parent_ids):
2042
 
                raise AssertionError(
2043
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2044
 
                    self.source._revision_id, parent_ids))
 
1809
            assert (self.source._revision_id in parent_ids), \
 
1810
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1811
                self.source._revision_id, parent_ids)
2045
1812
            source_index = 1 + parent_ids.index(self.source._revision_id)
2046
1813
            indices = (source_index, target_index)
2047
1814
        # -- make all specific_files utf8 --
2048
1815
        if specific_files:
2049
1816
            specific_files_utf8 = set()
2050
1817
            for path in specific_files:
2051
 
                # Note, if there are many specific files, using cache_utf8
2052
 
                # would be good here.
2053
1818
                specific_files_utf8.add(path.encode('utf8'))
2054
1819
            specific_files = specific_files_utf8
2055
1820
        else:
2056
1821
            specific_files = set([''])
2057
1822
        # -- specific_files is now a utf8 path set --
2058
 
 
2059
1823
        # -- get the state object and prepare it.
2060
1824
        state = self.target.current_dirstate()
2061
1825
        state._read_dirblocks_if_needed()
 
1826
        def _entries_for_path(path):
 
1827
            """Return a list with all the entries that match path for all ids.
 
1828
            """
 
1829
            dirname, basename = os.path.split(path)
 
1830
            key = (dirname, basename, '')
 
1831
            block_index, present = state._find_block_index_from_key(key)
 
1832
            if not present:
 
1833
                # the block which should contain path is absent.
 
1834
                return []
 
1835
            result = []
 
1836
            block = state._dirblocks[block_index][1]
 
1837
            entry_index, _ = state._find_entry_index(key, block)
 
1838
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1839
            while (entry_index < len(block) and
 
1840
                block[entry_index][0][0:2] == key[0:2]):
 
1841
                result.append(block[entry_index])
 
1842
                entry_index += 1
 
1843
            return result
2062
1844
        if require_versioned:
2063
1845
            # -- check all supplied paths are versioned in a search tree. --
2064
 
            not_versioned = []
 
1846
            all_versioned = True
2065
1847
            for path in specific_files:
2066
 
                path_entries = state._entries_for_path(path)
 
1848
                path_entries = _entries_for_path(path)
2067
1849
                if not path_entries:
2068
1850
                    # this specified path is not present at all: error
2069
 
                    not_versioned.append(path)
2070
 
                    continue
 
1851
                    all_versioned = False
 
1852
                    break
2071
1853
                found_versioned = False
2072
1854
                # for each id at this path
2073
1855
                for entry in path_entries:
2080
1862
                if not found_versioned:
2081
1863
                    # none of the indexes was not 'absent' at all ids for this
2082
1864
                    # path.
2083
 
                    not_versioned.append(path)
2084
 
            if len(not_versioned) > 0:
2085
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1865
                    all_versioned = False
 
1866
                    break
 
1867
            if not all_versioned:
 
1868
                raise errors.PathsNotVersionedError(specific_files)
2086
1869
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2087
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1870
        search_specific_files = set()
 
1871
        for path in specific_files:
 
1872
            other_specific_files = specific_files.difference(set([path]))
 
1873
            if not osutils.is_inside_any(other_specific_files, path):
 
1874
                # this is a top level path, we must check it.
 
1875
                search_specific_files.add(path)
 
1876
        # sketch: 
 
1877
        # compare source_index and target_index at or under each element of search_specific_files.
 
1878
        # follow the following comparison table. Note that we only want to do diff operations when
 
1879
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1880
        # for the target.
 
1881
        # cases:
 
1882
        # 
 
1883
        # Source | Target | disk | action
 
1884
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1885
        #        |        |      | diff check on source-target
 
1886
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1887
        #        |        |      | ???
 
1888
        #   r    |  a     |      | add source to search
 
1889
        #   r    |  a     |  a   | 
 
1890
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1891
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1892
        #   a    | fdlt   |      | add new id
 
1893
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1894
        #   a    |  a     |      | not present in either tree, skip
 
1895
        #   a    |  a     |  a   | not present in any tree, skip
 
1896
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1897
        #        |        |      | may not be selected by the users list of paths.
 
1898
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1899
        #        |        |      | may not be selected by the users list of paths.
 
1900
        #  fdlt  | fdlt   |      | content in both: diff them
 
1901
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1902
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1903
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1904
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1905
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1906
        #        |        |      | this id at the other path.
 
1907
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1908
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1909
        #        |        |      | this id at the other path.
 
1910
 
 
1911
        # for all search_indexs in each path at or under each element of
 
1912
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1913
        # relocated path as one to search if its not searched already. If the
 
1914
        # detail is not relocated, add the id.
 
1915
        searched_specific_files = set()
 
1916
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1917
        # Using a list so that we can access the values and change them in
 
1918
        # nested scope. Each one is [path, file_id, entry]
 
1919
        last_source_parent = [None, None]
 
1920
        last_target_parent = [None, None]
2088
1921
 
2089
1922
        use_filesystem_for_exec = (sys.platform != 'win32')
2090
 
        iter_changes = self.target._iter_changes(include_unchanged,
2091
 
            use_filesystem_for_exec, search_specific_files, state,
2092
 
            source_index, target_index, want_unversioned, self.target)
2093
 
        return iter_changes.iter_changes()
 
1923
 
 
1924
        # Just a sentry, so that _process_entry can say that this
 
1925
        # record is handled, but isn't interesting to process (unchanged)
 
1926
        uninteresting = object()
 
1927
 
 
1928
 
 
1929
        old_dirname_to_file_id = {}
 
1930
        new_dirname_to_file_id = {}
 
1931
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1932
        #       keeping a cache of directories that we have seen.
 
1933
 
 
1934
        def _process_entry(entry, path_info):
 
1935
            """Compare an entry and real disk to generate delta information.
 
1936
 
 
1937
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1938
                the path of entry. If None, then the path is considered absent.
 
1939
                (Perhaps we should pass in a concrete entry for this ?)
 
1940
                Basename is returned as a utf8 string because we expect this
 
1941
                tuple will be ignored, and don't want to take the time to
 
1942
                decode.
 
1943
            :return: None if these don't match
 
1944
                     A tuple of information about the change, or
 
1945
                     the object 'uninteresting' if these match, but are
 
1946
                     basically identical.
 
1947
            """
 
1948
            if source_index is None:
 
1949
                source_details = NULL_PARENT_DETAILS
 
1950
            else:
 
1951
                source_details = entry[1][source_index]
 
1952
            target_details = entry[1][target_index]
 
1953
            target_minikind = target_details[0]
 
1954
            if path_info is not None and target_minikind in 'fdlt':
 
1955
                assert target_index == 0
 
1956
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1957
                                                  stat_value=path_info[3])
 
1958
                # The entry may have been modified by update_entry
 
1959
                target_details = entry[1][target_index]
 
1960
                target_minikind = target_details[0]
 
1961
            else:
 
1962
                link_or_sha1 = None
 
1963
            file_id = entry[0][2]
 
1964
            source_minikind = source_details[0]
 
1965
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1966
                # claimed content in both: diff
 
1967
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1968
                #        |        |      | diff check on source-target
 
1969
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1970
                #        |        |      | ???
 
1971
                if source_minikind in 'r':
 
1972
                    # add the source to the search path to find any children it
 
1973
                    # has.  TODO ? : only add if it is a container ?
 
1974
                    if not osutils.is_inside_any(searched_specific_files,
 
1975
                                                 source_details[1]):
 
1976
                        search_specific_files.add(source_details[1])
 
1977
                    # generate the old path; this is needed for stating later
 
1978
                    # as well.
 
1979
                    old_path = source_details[1]
 
1980
                    old_dirname, old_basename = os.path.split(old_path)
 
1981
                    path = pathjoin(entry[0][0], entry[0][1])
 
1982
                    old_entry = state._get_entry(source_index,
 
1983
                                                 path_utf8=old_path)
 
1984
                    # update the source details variable to be the real
 
1985
                    # location.
 
1986
                    if old_entry == (None, None):
 
1987
                        raise errors.CorruptDirstate(state._filename,
 
1988
                            "entry '%s/%s' is considered renamed from %r"
 
1989
                            " but source does not exist\n"
 
1990
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
1991
                    source_details = old_entry[1][source_index]
 
1992
                    source_minikind = source_details[0]
 
1993
                else:
 
1994
                    old_dirname = entry[0][0]
 
1995
                    old_basename = entry[0][1]
 
1996
                    old_path = path = None
 
1997
                if path_info is None:
 
1998
                    # the file is missing on disk, show as removed.
 
1999
                    content_change = True
 
2000
                    target_kind = None
 
2001
                    target_exec = False
 
2002
                else:
 
2003
                    # source and target are both versioned and disk file is present.
 
2004
                    target_kind = path_info[2]
 
2005
                    if target_kind == 'directory':
 
2006
                        if path is None:
 
2007
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2008
                        new_dirname_to_file_id[path] = file_id
 
2009
                        if source_minikind != 'd':
 
2010
                            content_change = True
 
2011
                        else:
 
2012
                            # directories have no fingerprint
 
2013
                            content_change = False
 
2014
                        target_exec = False
 
2015
                    elif target_kind == 'file':
 
2016
                        if source_minikind != 'f':
 
2017
                            content_change = True
 
2018
                        else:
 
2019
                            # We could check the size, but we already have the
 
2020
                            # sha1 hash.
 
2021
                            content_change = (link_or_sha1 != source_details[1])
 
2022
                        # Target details is updated at update_entry time
 
2023
                        if use_filesystem_for_exec:
 
2024
                            # We don't need S_ISREG here, because we are sure
 
2025
                            # we are dealing with a file.
 
2026
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2027
                        else:
 
2028
                            target_exec = target_details[3]
 
2029
                    elif target_kind == 'symlink':
 
2030
                        if source_minikind != 'l':
 
2031
                            content_change = True
 
2032
                        else:
 
2033
                            content_change = (link_or_sha1 != source_details[1])
 
2034
                        target_exec = False
 
2035
                    elif target_kind == 'tree-reference':
 
2036
                        if source_minikind != 't':
 
2037
                            content_change = True
 
2038
                        else:
 
2039
                            content_change = False
 
2040
                        target_exec = False
 
2041
                    else:
 
2042
                        raise Exception, "unknown kind %s" % path_info[2]
 
2043
                if source_minikind == 'd':
 
2044
                    if path is None:
 
2045
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2046
                    old_dirname_to_file_id[old_path] = file_id
 
2047
                # parent id is the entry for the path in the target tree
 
2048
                if old_dirname == last_source_parent[0]:
 
2049
                    source_parent_id = last_source_parent[1]
 
2050
                else:
 
2051
                    try:
 
2052
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2053
                    except KeyError:
 
2054
                        source_parent_entry = state._get_entry(source_index,
 
2055
                                                               path_utf8=old_dirname)
 
2056
                        source_parent_id = source_parent_entry[0][2]
 
2057
                    if source_parent_id == entry[0][2]:
 
2058
                        # This is the root, so the parent is None
 
2059
                        source_parent_id = None
 
2060
                    else:
 
2061
                        last_source_parent[0] = old_dirname
 
2062
                        last_source_parent[1] = source_parent_id
 
2063
                new_dirname = entry[0][0]
 
2064
                if new_dirname == last_target_parent[0]:
 
2065
                    target_parent_id = last_target_parent[1]
 
2066
                else:
 
2067
                    try:
 
2068
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2069
                    except KeyError:
 
2070
                        # TODO: We don't always need to do the lookup, because the
 
2071
                        #       parent entry will be the same as the source entry.
 
2072
                        target_parent_entry = state._get_entry(target_index,
 
2073
                                                               path_utf8=new_dirname)
 
2074
                        assert target_parent_entry != (None, None), (
 
2075
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2076
                            % (new_dirname, entry))
 
2077
                        target_parent_id = target_parent_entry[0][2]
 
2078
                    if target_parent_id == entry[0][2]:
 
2079
                        # This is the root, so the parent is None
 
2080
                        target_parent_id = None
 
2081
                    else:
 
2082
                        last_target_parent[0] = new_dirname
 
2083
                        last_target_parent[1] = target_parent_id
 
2084
 
 
2085
                source_exec = source_details[3]
 
2086
                if (include_unchanged
 
2087
                    or content_change
 
2088
                    or source_parent_id != target_parent_id
 
2089
                    or old_basename != entry[0][1]
 
2090
                    or source_exec != target_exec
 
2091
                    ):
 
2092
                    if old_path is None:
 
2093
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2094
                        old_path_u = utf8_decode(old_path)[0]
 
2095
                        path_u = old_path_u
 
2096
                    else:
 
2097
                        old_path_u = utf8_decode(old_path)[0]
 
2098
                        if old_path == path:
 
2099
                            path_u = old_path_u
 
2100
                        else:
 
2101
                            path_u = utf8_decode(path)[0]
 
2102
                    source_kind = _minikind_to_kind[source_minikind]
 
2103
                    return (entry[0][2],
 
2104
                           (old_path_u, path_u),
 
2105
                           content_change,
 
2106
                           (True, True),
 
2107
                           (source_parent_id, target_parent_id),
 
2108
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2109
                           (source_kind, target_kind),
 
2110
                           (source_exec, target_exec))
 
2111
                else:
 
2112
                    return uninteresting
 
2113
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2114
                # looks like a new file
 
2115
                if path_info is not None:
 
2116
                    path = pathjoin(entry[0][0], entry[0][1])
 
2117
                    # parent id is the entry for the path in the target tree
 
2118
                    # TODO: these are the same for an entire directory: cache em.
 
2119
                    parent_id = state._get_entry(target_index,
 
2120
                                                 path_utf8=entry[0][0])[0][2]
 
2121
                    if parent_id == entry[0][2]:
 
2122
                        parent_id = None
 
2123
                    if use_filesystem_for_exec:
 
2124
                        # We need S_ISREG here, because we aren't sure if this
 
2125
                        # is a file or not.
 
2126
                        target_exec = bool(
 
2127
                            stat.S_ISREG(path_info[3].st_mode)
 
2128
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2129
                    else:
 
2130
                        target_exec = target_details[3]
 
2131
                    return (entry[0][2],
 
2132
                           (None, utf8_decode(path)[0]),
 
2133
                           True,
 
2134
                           (False, True),
 
2135
                           (None, parent_id),
 
2136
                           (None, utf8_decode(entry[0][1])[0]),
 
2137
                           (None, path_info[2]),
 
2138
                           (None, target_exec))
 
2139
                else:
 
2140
                    # but its not on disk: we deliberately treat this as just
 
2141
                    # never-present. (Why ?! - RBC 20070224)
 
2142
                    pass
 
2143
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2144
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2145
                # if its still on disk, *and* theres no other entry at this
 
2146
                # path [we dont know this in this routine at the moment -
 
2147
                # perhaps we should change this - then it would be an unknown.
 
2148
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2149
                # parent id is the entry for the path in the target tree
 
2150
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2151
                if parent_id == entry[0][2]:
 
2152
                    parent_id = None
 
2153
                return (entry[0][2],
 
2154
                       (utf8_decode(old_path)[0], None),
 
2155
                       True,
 
2156
                       (True, False),
 
2157
                       (parent_id, None),
 
2158
                       (utf8_decode(entry[0][1])[0], None),
 
2159
                       (_minikind_to_kind[source_minikind], None),
 
2160
                       (source_details[3], None))
 
2161
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2162
                # a rename; could be a true rename, or a rename inherited from
 
2163
                # a renamed parent. TODO: handle this efficiently. Its not
 
2164
                # common case to rename dirs though, so a correct but slow
 
2165
                # implementation will do.
 
2166
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2167
                    search_specific_files.add(target_details[1])
 
2168
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2169
                # neither of the selected trees contain this file,
 
2170
                # so skip over it. This is not currently directly tested, but
 
2171
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2172
                pass
 
2173
            else:
 
2174
                raise AssertionError("don't know how to compare "
 
2175
                    "source_minikind=%r, target_minikind=%r"
 
2176
                    % (source_minikind, target_minikind))
 
2177
                ## import pdb;pdb.set_trace()
 
2178
            return None
 
2179
 
 
2180
        while search_specific_files:
 
2181
            # TODO: the pending list should be lexically sorted?  the
 
2182
            # interface doesn't require it.
 
2183
            current_root = search_specific_files.pop()
 
2184
            current_root_unicode = current_root.decode('utf8')
 
2185
            searched_specific_files.add(current_root)
 
2186
            # process the entries for this containing directory: the rest will be
 
2187
            # found by their parents recursively.
 
2188
            root_entries = _entries_for_path(current_root)
 
2189
            root_abspath = self.target.abspath(current_root_unicode)
 
2190
            try:
 
2191
                root_stat = os.lstat(root_abspath)
 
2192
            except OSError, e:
 
2193
                if e.errno == errno.ENOENT:
 
2194
                    # the path does not exist: let _process_entry know that.
 
2195
                    root_dir_info = None
 
2196
                else:
 
2197
                    # some other random error: hand it up.
 
2198
                    raise
 
2199
            else:
 
2200
                root_dir_info = ('', current_root,
 
2201
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2202
                    root_abspath)
 
2203
                if root_dir_info[2] == 'directory':
 
2204
                    if self.target._directory_is_tree_reference(
 
2205
                        current_root.decode('utf8')):
 
2206
                        root_dir_info = root_dir_info[:2] + \
 
2207
                            ('tree-reference',) + root_dir_info[3:]
 
2208
 
 
2209
            if not root_entries and not root_dir_info:
 
2210
                # this specified path is not present at all, skip it.
 
2211
                continue
 
2212
            path_handled = False
 
2213
            for entry in root_entries:
 
2214
                result = _process_entry(entry, root_dir_info)
 
2215
                if result is not None:
 
2216
                    path_handled = True
 
2217
                    if result is not uninteresting:
 
2218
                        yield result
 
2219
            if want_unversioned and not path_handled and root_dir_info:
 
2220
                new_executable = bool(
 
2221
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2222
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2223
                yield (None,
 
2224
                       (None, current_root_unicode),
 
2225
                       True,
 
2226
                       (False, False),
 
2227
                       (None, None),
 
2228
                       (None, splitpath(current_root_unicode)[-1]),
 
2229
                       (None, root_dir_info[2]),
 
2230
                       (None, new_executable)
 
2231
                      )
 
2232
            initial_key = (current_root, '', '')
 
2233
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2234
            if block_index == 0:
 
2235
                # we have processed the total root already, but because the
 
2236
                # initial key matched it we should skip it here.
 
2237
                block_index +=1
 
2238
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2239
                current_dir_info = None
 
2240
            else:
 
2241
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2242
                try:
 
2243
                    current_dir_info = dir_iterator.next()
 
2244
                except OSError, e:
 
2245
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2246
                    # python 2.5 has e.errno == EINVAL,
 
2247
                    #            and e.winerror == ERROR_DIRECTORY
 
2248
                    e_winerror = getattr(e, 'winerror', None)
 
2249
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2250
                    # there may be directories in the inventory even though
 
2251
                    # this path is not a file on disk: so mark it as end of
 
2252
                    # iterator
 
2253
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2254
                        current_dir_info = None
 
2255
                    elif (sys.platform == 'win32'
 
2256
                          and (e.errno in win_errors
 
2257
                               or e_winerror in win_errors)):
 
2258
                        current_dir_info = None
 
2259
                    else:
 
2260
                        raise
 
2261
                else:
 
2262
                    if current_dir_info[0][0] == '':
 
2263
                        # remove .bzr from iteration
 
2264
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2265
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2266
                        del current_dir_info[1][bzr_index]
 
2267
            # walk until both the directory listing and the versioned metadata
 
2268
            # are exhausted. 
 
2269
            if (block_index < len(state._dirblocks) and
 
2270
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2271
                current_block = state._dirblocks[block_index]
 
2272
            else:
 
2273
                current_block = None
 
2274
            while (current_dir_info is not None or
 
2275
                   current_block is not None):
 
2276
                if (current_dir_info and current_block
 
2277
                    and current_dir_info[0][0] != current_block[0]):
 
2278
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2279
                        # filesystem data refers to paths not covered by the dirblock.
 
2280
                        # this has two possibilities:
 
2281
                        # A) it is versioned but empty, so there is no block for it
 
2282
                        # B) it is not versioned.
 
2283
 
 
2284
                        # if (A) then we need to recurse into it to check for
 
2285
                        # new unknown files or directories.
 
2286
                        # if (B) then we should ignore it, because we don't
 
2287
                        # recurse into unknown directories.
 
2288
                        path_index = 0
 
2289
                        while path_index < len(current_dir_info[1]):
 
2290
                                current_path_info = current_dir_info[1][path_index]
 
2291
                                if want_unversioned:
 
2292
                                    if current_path_info[2] == 'directory':
 
2293
                                        if self.target._directory_is_tree_reference(
 
2294
                                            current_path_info[0].decode('utf8')):
 
2295
                                            current_path_info = current_path_info[:2] + \
 
2296
                                                ('tree-reference',) + current_path_info[3:]
 
2297
                                    new_executable = bool(
 
2298
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2299
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2300
                                    yield (None,
 
2301
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2302
                                        True,
 
2303
                                        (False, False),
 
2304
                                        (None, None),
 
2305
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2306
                                        (None, current_path_info[2]),
 
2307
                                        (None, new_executable))
 
2308
                                # dont descend into this unversioned path if it is
 
2309
                                # a dir
 
2310
                                if current_path_info[2] in ('directory',
 
2311
                                                            'tree-reference'):
 
2312
                                    del current_dir_info[1][path_index]
 
2313
                                    path_index -= 1
 
2314
                                path_index += 1
 
2315
 
 
2316
                        # This dir info has been handled, go to the next
 
2317
                        try:
 
2318
                            current_dir_info = dir_iterator.next()
 
2319
                        except StopIteration:
 
2320
                            current_dir_info = None
 
2321
                    else:
 
2322
                        # We have a dirblock entry for this location, but there
 
2323
                        # is no filesystem path for this. This is most likely
 
2324
                        # because a directory was removed from the disk.
 
2325
                        # We don't have to report the missing directory,
 
2326
                        # because that should have already been handled, but we
 
2327
                        # need to handle all of the files that are contained
 
2328
                        # within.
 
2329
                        for current_entry in current_block[1]:
 
2330
                            # entry referring to file not present on disk.
 
2331
                            # advance the entry only, after processing.
 
2332
                            result = _process_entry(current_entry, None)
 
2333
                            if result is not None:
 
2334
                                if result is not uninteresting:
 
2335
                                    yield result
 
2336
                        block_index +=1
 
2337
                        if (block_index < len(state._dirblocks) and
 
2338
                            osutils.is_inside(current_root,
 
2339
                                              state._dirblocks[block_index][0])):
 
2340
                            current_block = state._dirblocks[block_index]
 
2341
                        else:
 
2342
                            current_block = None
 
2343
                    continue
 
2344
                entry_index = 0
 
2345
                if current_block and entry_index < len(current_block[1]):
 
2346
                    current_entry = current_block[1][entry_index]
 
2347
                else:
 
2348
                    current_entry = None
 
2349
                advance_entry = True
 
2350
                path_index = 0
 
2351
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2352
                    current_path_info = current_dir_info[1][path_index]
 
2353
                    if current_path_info[2] == 'directory':
 
2354
                        if self.target._directory_is_tree_reference(
 
2355
                            current_path_info[0].decode('utf8')):
 
2356
                            current_path_info = current_path_info[:2] + \
 
2357
                                ('tree-reference',) + current_path_info[3:]
 
2358
                else:
 
2359
                    current_path_info = None
 
2360
                advance_path = True
 
2361
                path_handled = False
 
2362
                while (current_entry is not None or
 
2363
                    current_path_info is not None):
 
2364
                    if current_entry is None:
 
2365
                        # the check for path_handled when the path is adnvaced
 
2366
                        # will yield this path if needed.
 
2367
                        pass
 
2368
                    elif current_path_info is None:
 
2369
                        # no path is fine: the per entry code will handle it.
 
2370
                        result = _process_entry(current_entry, current_path_info)
 
2371
                        if result is not None:
 
2372
                            if result is not uninteresting:
 
2373
                                yield result
 
2374
                    elif (current_entry[0][1] != current_path_info[1]
 
2375
                          or current_entry[1][target_index][0] in 'ar'):
 
2376
                        # The current path on disk doesn't match the dirblock
 
2377
                        # record. Either the dirblock is marked as absent, or
 
2378
                        # the file on disk is not present at all in the
 
2379
                        # dirblock. Either way, report about the dirblock
 
2380
                        # entry, and let other code handle the filesystem one.
 
2381
 
 
2382
                        # Compare the basename for these files to determine
 
2383
                        # which comes first
 
2384
                        if current_path_info[1] < current_entry[0][1]:
 
2385
                            # extra file on disk: pass for now, but only
 
2386
                            # increment the path, not the entry
 
2387
                            advance_entry = False
 
2388
                        else:
 
2389
                            # entry referring to file not present on disk.
 
2390
                            # advance the entry only, after processing.
 
2391
                            result = _process_entry(current_entry, None)
 
2392
                            if result is not None:
 
2393
                                if result is not uninteresting:
 
2394
                                    yield result
 
2395
                            advance_path = False
 
2396
                    else:
 
2397
                        result = _process_entry(current_entry, current_path_info)
 
2398
                        if result is not None:
 
2399
                            path_handled = True
 
2400
                            if result is not uninteresting:
 
2401
                                yield result
 
2402
                    if advance_entry and current_entry is not None:
 
2403
                        entry_index += 1
 
2404
                        if entry_index < len(current_block[1]):
 
2405
                            current_entry = current_block[1][entry_index]
 
2406
                        else:
 
2407
                            current_entry = None
 
2408
                    else:
 
2409
                        advance_entry = True # reset the advance flaga
 
2410
                    if advance_path and current_path_info is not None:
 
2411
                        if not path_handled:
 
2412
                            # unversioned in all regards
 
2413
                            if want_unversioned:
 
2414
                                new_executable = bool(
 
2415
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2416
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2417
                                yield (None,
 
2418
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2419
                                    True,
 
2420
                                    (False, False),
 
2421
                                    (None, None),
 
2422
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2423
                                    (None, current_path_info[2]),
 
2424
                                    (None, new_executable))
 
2425
                            # dont descend into this unversioned path if it is
 
2426
                            # a dir
 
2427
                            if current_path_info[2] in ('directory'):
 
2428
                                del current_dir_info[1][path_index]
 
2429
                                path_index -= 1
 
2430
                        # dont descend the disk iterator into any tree 
 
2431
                        # paths.
 
2432
                        if current_path_info[2] == 'tree-reference':
 
2433
                            del current_dir_info[1][path_index]
 
2434
                            path_index -= 1
 
2435
                        path_index += 1
 
2436
                        if path_index < len(current_dir_info[1]):
 
2437
                            current_path_info = current_dir_info[1][path_index]
 
2438
                            if current_path_info[2] == 'directory':
 
2439
                                if self.target._directory_is_tree_reference(
 
2440
                                    current_path_info[0].decode('utf8')):
 
2441
                                    current_path_info = current_path_info[:2] + \
 
2442
                                        ('tree-reference',) + current_path_info[3:]
 
2443
                        else:
 
2444
                            current_path_info = None
 
2445
                        path_handled = False
 
2446
                    else:
 
2447
                        advance_path = True # reset the advance flagg.
 
2448
                if current_block is not None:
 
2449
                    block_index += 1
 
2450
                    if (block_index < len(state._dirblocks) and
 
2451
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2452
                        current_block = state._dirblocks[block_index]
 
2453
                    else:
 
2454
                        current_block = None
 
2455
                if current_dir_info is not None:
 
2456
                    try:
 
2457
                        current_dir_info = dir_iterator.next()
 
2458
                    except StopIteration:
 
2459
                        current_dir_info = None
 
2460
 
2094
2461
 
2095
2462
    @staticmethod
2096
2463
    def is_compatible(source, target):
2097
2464
        # the target must be a dirstate working tree
2098
 
        if not isinstance(target, DirStateWorkingTree):
 
2465
        if not isinstance(target, WorkingTree4):
2099
2466
            return False
2100
 
        # the source must be a revtree or dirstate rev tree.
 
2467
        # the source must be a revtreee or dirstate rev tree.
2101
2468
        if not isinstance(source,
2102
2469
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2103
2470
            return False
2104
2471
        # the source revid must be in the target dirstate
2105
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2472
        if not (source._revision_id == NULL_REVISION or
2106
2473
            source._revision_id in target.get_parent_ids()):
2107
 
            # TODO: what about ghosts? it may well need to
 
2474
            # TODO: what about ghosts? it may well need to 
2108
2475
            # check for them explicitly.
2109
2476
            return False
2110
2477
        return True
2120
2487
 
2121
2488
    def convert(self, tree):
2122
2489
        # lock the control files not the tree, so that we dont get tree
2123
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2490
        # on-unlock behaviours, and so that noone else diddles with the 
2124
2491
        # tree during upgrade.
2125
2492
        tree._control_files.lock_write()
2126
2493
        try:
2152
2519
 
2153
2520
    def update_format(self, tree):
2154
2521
        """Change the format marker."""
2155
 
        tree._transport.put_bytes('format',
2156
 
            self.target_format.get_format_string(),
2157
 
            mode=tree.bzrdir._get_file_mode())
2158
 
 
2159
 
 
2160
 
class Converter4to5(object):
2161
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2162
 
 
2163
 
    def __init__(self):
2164
 
        self.target_format = WorkingTreeFormat5()
2165
 
 
2166
 
    def convert(self, tree):
2167
 
        # lock the control files not the tree, so that we don't get tree
2168
 
        # on-unlock behaviours, and so that no-one else diddles with the
2169
 
        # tree during upgrade.
2170
 
        tree._control_files.lock_write()
2171
 
        try:
2172
 
            self.update_format(tree)
2173
 
        finally:
2174
 
            tree._control_files.unlock()
2175
 
 
2176
 
    def update_format(self, tree):
2177
 
        """Change the format marker."""
2178
 
        tree._transport.put_bytes('format',
2179
 
            self.target_format.get_format_string(),
2180
 
            mode=tree.bzrdir._get_file_mode())
2181
 
 
2182
 
 
2183
 
class Converter4or5to6(object):
2184
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2185
 
 
2186
 
    def __init__(self):
2187
 
        self.target_format = WorkingTreeFormat6()
2188
 
 
2189
 
    def convert(self, tree):
2190
 
        # lock the control files not the tree, so that we don't get tree
2191
 
        # on-unlock behaviours, and so that no-one else diddles with the
2192
 
        # tree during upgrade.
2193
 
        tree._control_files.lock_write()
2194
 
        try:
2195
 
            self.init_custom_control_files(tree)
2196
 
            self.update_format(tree)
2197
 
        finally:
2198
 
            tree._control_files.unlock()
2199
 
 
2200
 
    def init_custom_control_files(self, tree):
2201
 
        """Initialize custom control files."""
2202
 
        tree._transport.put_bytes('views', '',
2203
 
            mode=tree.bzrdir._get_file_mode())
2204
 
 
2205
 
    def update_format(self, tree):
2206
 
        """Change the format marker."""
2207
 
        tree._transport.put_bytes('format',
2208
 
            self.target_format.get_format_string(),
2209
 
            mode=tree.bzrdir._get_file_mode())
 
2522
        tree._control_files.put_utf8('format',
 
2523
            self.target_format.get_format_string())