~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2008-08-16 00:09:54 UTC
  • mfrom: (3629.1.4 uncommit_ids)
  • Revision ID: pqm@pqm.ubuntu.com-20080816000954-t0401ff8s3ydnkr6
(jam) bzr uncommit now gives the revision id to restore the branch to
        the old tip.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2011 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007, 2008 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
 
 
41
import bzrlib
34
42
from bzrlib import (
35
43
    bzrdir,
36
44
    cache_utf8,
37
 
    config,
38
45
    conflicts as _mod_conflicts,
39
46
    debug,
 
47
    delta,
40
48
    dirstate,
41
49
    errors,
42
 
    filters as _mod_filters,
43
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
44
54
    osutils,
45
55
    revision as _mod_revision,
46
56
    revisiontree,
 
57
    textui,
47
58
    trace,
48
59
    transform,
49
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
50
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
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.inventory import Inventory, ROOT_ID, entry_factory
55
 
from bzrlib.lock import LogicalLockResult
56
 
from bzrlib.lockable_files import LockableFiles
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
57
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
58
75
from bzrlib.mutabletree import needs_tree_write_lock
59
76
from bzrlib.osutils import (
60
77
    file_kind,
61
78
    isdir,
 
79
    normpath,
62
80
    pathjoin,
 
81
    rand_chars,
63
82
    realpath,
64
83
    safe_unicode,
 
84
    splitpath,
65
85
    )
 
86
from bzrlib.trace import mutter, note
66
87
from bzrlib.transport.local import LocalTransport
67
 
from bzrlib.tree import (
68
 
    InterTree,
69
 
    InventoryTree,
70
 
    )
71
 
from bzrlib.workingtree import (
72
 
    InventoryWorkingTree,
73
 
    WorkingTree,
74
 
    WorkingTreeFormat,
75
 
    )
76
 
 
77
 
 
78
 
class DirStateWorkingTree(InventoryWorkingTree):
 
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
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
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
    """
79
119
 
80
120
    def __init__(self, basedir,
81
121
                 branch,
92
132
        self._format = _format
93
133
        self.bzrdir = _bzrdir
94
134
        basedir = safe_unicode(basedir)
95
 
        trace.mutter("opening working tree %r", basedir)
 
135
        mutter("opening working tree %r", basedir)
96
136
        self._branch = branch
97
137
        self.basedir = realpath(basedir)
98
138
        # if branch is at our basedir and is a format 6 or less
109
149
        self._setup_directory_is_tree_reference()
110
150
        self._detect_case_handling()
111
151
        self._rules_searcher = None
112
 
        self.views = self._make_views()
113
 
        #--- allow tests to select the dirstate iter_changes implementation
114
 
        self._iter_changes = dirstate._process_entry
115
152
 
116
153
    @needs_tree_write_lock
117
154
    def _add(self, files, ids, kinds):
132
169
            state.add(f, file_id, kind, None, '')
133
170
        self._make_dirty(reset_inventory=True)
134
171
 
135
 
    def _get_check_refs(self):
136
 
        """Return the references needed to perform a check of this tree."""
137
 
        return [('trees', self.last_revision())]
138
 
 
139
172
    def _make_dirty(self, reset_inventory):
140
173
        """Make the tree state dirty.
141
174
 
149
182
    @needs_tree_write_lock
150
183
    def add_reference(self, sub_tree):
151
184
        # use standard implementation, which calls back to self._add
152
 
        #
 
185
        # 
153
186
        # So we don't store the reference_revision in the working dirstate,
154
 
        # it's just recorded at the moment of commit.
 
187
        # it's just recorded at the moment of commit. 
155
188
        self._add_reference(sub_tree)
156
189
 
157
190
    def break_lock(self):
193
226
 
194
227
    def _comparison_data(self, entry, path):
195
228
        kind, executable, stat_value = \
196
 
            WorkingTree._comparison_data(self, entry, path)
 
229
            WorkingTree3._comparison_data(self, entry, path)
197
230
        # it looks like a plain directory, but it's really a reference -- see
198
231
        # also kind()
199
 
        if (self._repo_supports_tree_reference and kind == 'directory'
200
 
            and entry is not None and entry.kind == 'tree-reference'):
 
232
        if (self._repo_supports_tree_reference and
 
233
            kind == 'directory' and
 
234
            self._directory_is_tree_reference(path)):
201
235
            kind = 'tree-reference'
202
236
        return kind, executable, stat_value
203
237
 
205
239
    def commit(self, message=None, revprops=None, *args, **kwargs):
206
240
        # mark the tree as dirty post commit - commit
207
241
        # can change the current versioned list by doing deletes.
208
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
242
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
209
243
        self._make_dirty(reset_inventory=True)
210
244
        return result
211
245
 
229
263
            return self._dirstate
230
264
        local_path = self.bzrdir.get_workingtree_transport(None
231
265
            ).local_abspath('dirstate')
232
 
        self._dirstate = dirstate.DirState.on_file(local_path,
233
 
            self._sha1_provider(), self._worth_saving_limit())
 
266
        self._dirstate = dirstate.DirState.on_file(local_path)
234
267
        return self._dirstate
235
268
 
236
 
    def _sha1_provider(self):
237
 
        """A function that returns a SHA1Provider suitable for this tree.
238
 
 
239
 
        :return: None if content filtering is not supported by this tree.
240
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
241
 
          form of files, i.e. after read filters are applied.
242
 
        """
243
 
        if self.supports_content_filtering():
244
 
            return ContentFilterAwareSHA1Provider(self)
245
 
        else:
246
 
            return None
247
 
 
248
 
    def _worth_saving_limit(self):
249
 
        """How many hash changes are ok before we must save the dirstate.
250
 
 
251
 
        :return: an integer. -1 means never save.
252
 
        """
253
 
        # FIXME: We want a WorkingTreeStack here -- vila 20110812
254
 
        conf = config.BranchStack(self.branch)
255
 
        return conf.get('bzr.workingtree.worth_saving_limit')
256
 
 
257
269
    def filter_unversioned_files(self, paths):
258
270
        """Filter out paths that are versioned.
259
271
 
291
303
 
292
304
    def _generate_inventory(self):
293
305
        """Create and set self.inventory from the dirstate object.
294
 
 
 
306
        
295
307
        This is relatively expensive: we have to walk the entire dirstate.
296
308
        Ideally we would not, and can deprecate this function.
297
309
        """
342
354
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
343
355
                elif kind == 'tree-reference':
344
356
                    if not self._repo_supports_tree_reference:
345
 
                        raise errors.UnsupportedOperation(
346
 
                            self._generate_inventory,
347
 
                            self.branch.repository)
 
357
                        raise AssertionError(
 
358
                            "repository of %r "
 
359
                            "doesn't support tree references "
 
360
                            "required by entry %r"
 
361
                            % (self, name))
348
362
                    inv_entry.reference_revision = link_or_sha1 or None
349
363
                elif kind != 'symlink':
350
364
                    raise AssertionError("unknown kind %r" % kind)
365
379
        If either file_id or path is supplied, it is used as the key to lookup.
366
380
        If both are supplied, the fastest lookup is used, and an error is
367
381
        raised if they do not both point at the same row.
368
 
 
 
382
        
369
383
        :param file_id: An optional unicode file_id to be looked up.
370
384
        :param path: An optional unicode path to be looked up.
371
385
        :return: The dirstate row tuple for path/file_id, or (None, None)
389
403
        state = self.current_dirstate()
390
404
        if stat_value is None:
391
405
            try:
392
 
                stat_value = osutils.lstat(file_abspath)
 
406
                stat_value = os.lstat(file_abspath)
393
407
            except OSError, e:
394
408
                if e.errno == errno.ENOENT:
395
409
                    return None
396
410
                else:
397
411
                    raise
398
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
399
 
            stat_value=stat_value)
 
412
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
413
                                          stat_value=stat_value)
400
414
        if entry[1][0][0] == 'f':
401
 
            if link_or_sha1 is None:
402
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
403
 
                try:
404
 
                    sha1 = osutils.sha_file(file_obj)
405
 
                finally:
406
 
                    file_obj.close()
407
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
408
 
                return sha1
409
 
            else:
410
 
                return link_or_sha1
 
415
            return link_or_sha1
411
416
        return None
412
417
 
413
418
    def _get_inventory(self):
427
432
    @needs_read_lock
428
433
    def get_parent_ids(self):
429
434
        """See Tree.get_parent_ids.
430
 
 
 
435
        
431
436
        This implementation requests the ids list from the dirstate file.
432
437
        """
433
438
        return self.current_dirstate().get_parent_ids()
455
460
        return osutils.lexists(pathjoin(
456
461
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
457
462
 
458
 
    def has_or_had_id(self, file_id):
459
 
        state = self.current_dirstate()
460
 
        row, parents = self._get_entry(file_id=file_id)
461
 
        return row is not None
462
 
 
463
463
    @needs_read_lock
464
464
    def id2path(self, file_id):
465
465
        "Convert a file-id to a path."
498
498
            self._must_be_locked()
499
499
            if not path:
500
500
                path = self.id2path(file_id)
501
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
501
            mode = os.lstat(self.abspath(path)).st_mode
502
502
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
503
503
 
504
504
    def all_file_ids(self):
548
548
                # path is missing on disk.
549
549
                continue
550
550
 
551
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
552
 
        """See MutableTree._observed_sha1."""
553
 
        state = self.current_dirstate()
554
 
        entry = self._get_entry(file_id=file_id, path=path)
555
 
        state._observed_sha1(entry, sha1, statvalue)
556
 
 
557
551
    def kind(self, file_id):
558
552
        """Return the kind of a file.
559
553
 
571
565
    def _kind(self, relpath):
572
566
        abspath = self.abspath(relpath)
573
567
        kind = file_kind(abspath)
574
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
575
 
            entry = self._get_entry(path=relpath)
576
 
            if entry[1] is not None:
577
 
                if entry[1][0][0] == 't':
578
 
                    kind = 'tree-reference'
 
568
        if (self._repo_supports_tree_reference and
 
569
            kind == 'directory' and
 
570
            self._directory_is_tree_reference(relpath)):
 
571
            kind = 'tree-reference'
579
572
        return kind
580
573
 
581
574
    @needs_read_lock
588
581
            return _mod_revision.NULL_REVISION
589
582
 
590
583
    def lock_read(self):
591
 
        """See Branch.lock_read, and WorkingTree.unlock.
592
 
 
593
 
        :return: A bzrlib.lock.LogicalLockResult.
594
 
        """
 
584
        """See Branch.lock_read, and WorkingTree.unlock."""
595
585
        self.branch.lock_read()
596
586
        try:
597
587
            self._control_files.lock_read()
610
600
        except:
611
601
            self.branch.unlock()
612
602
            raise
613
 
        return LogicalLockResult(self.unlock)
614
603
 
615
604
    def _lock_self_write(self):
616
605
        """This should be called after the branch is locked."""
631
620
        except:
632
621
            self.branch.unlock()
633
622
            raise
634
 
        return LogicalLockResult(self.unlock)
635
623
 
636
624
    def lock_tree_write(self):
637
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
638
 
 
639
 
        :return: A bzrlib.lock.LogicalLockResult.
640
 
        """
 
625
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
641
626
        self.branch.lock_read()
642
 
        return self._lock_self_write()
 
627
        self._lock_self_write()
643
628
 
644
629
    def lock_write(self):
645
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
630
        """See MutableTree.lock_write, and WorkingTree.unlock."""
649
631
        self.branch.lock_write()
650
 
        return self._lock_self_write()
 
632
        self._lock_self_write()
651
633
 
652
634
    @needs_tree_write_lock
653
635
    def move(self, from_paths, to_dir, after=False):
723
705
            from_entry = self._get_entry(path=from_rel)
724
706
            if from_entry == (None, None):
725
707
                raise errors.BzrMoveFailedError(from_rel,to_dir,
726
 
                    errors.NotVersionedError(path=from_rel))
 
708
                    errors.NotVersionedError(path=str(from_rel)))
727
709
 
728
710
            from_id = from_entry[0][2]
729
711
            to_rel = pathjoin(to_dir, from_tail)
870
852
                rollback_rename()
871
853
                raise
872
854
            result.append((from_rel, to_rel))
873
 
            state._mark_modified()
 
855
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
874
856
            self._make_dirty(reset_inventory=False)
875
857
 
876
858
        return result
904
886
        for tree in trees:
905
887
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
906
888
                parents):
907
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
908
 
                    trees, require_versioned)
 
889
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
909
890
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
910
891
        # -- make all paths utf8 --
911
892
        paths_utf8 = set()
972
953
                raise errors.PathsNotVersionedError(paths)
973
954
        # -- remove redundancy in supplied paths to prevent over-scanning --
974
955
        search_paths = osutils.minimum_path_selection(paths)
975
 
        # sketch:
 
956
        # sketch: 
976
957
        # for all search_indexs in each path at or under each element of
977
958
        # search_paths, if the detail is relocated: add the id, and add the
978
959
        # relocated path as one to search if its not searched already. If the
1034
1015
 
1035
1016
    def read_working_inventory(self):
1036
1017
        """Read the working inventory.
1037
 
 
 
1018
        
1038
1019
        This is a meaningless operation for dirstate, but we obey it anyhow.
1039
1020
        """
1040
1021
        return self.inventory
1058
1039
    def set_last_revision(self, new_revision):
1059
1040
        """Change the last revision in the working tree."""
1060
1041
        parents = self.get_parent_ids()
1061
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
 
1042
        if new_revision in (NULL_REVISION, None):
1062
1043
            if len(parents) >= 2:
1063
1044
                raise AssertionError(
1064
1045
                    "setting the last parent to none with a pending merge is "
1071
1052
    @needs_tree_write_lock
1072
1053
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1073
1054
        """Set the parent ids to revision_ids.
1074
 
 
 
1055
        
1075
1056
        See also set_parent_trees. This api will try to retrieve the tree data
1076
1057
        for each element of revision_ids from the trees repository. If you have
1077
1058
        tree data already available, it is more efficient to use
1129
1110
                real_trees.append((rev_id, tree))
1130
1111
            else:
1131
1112
                real_trees.append((rev_id,
1132
 
                    self.branch.repository.revision_tree(
1133
 
                        _mod_revision.NULL_REVISION)))
 
1113
                    self.branch.repository.revision_tree(None)))
1134
1114
                ghosts.append(rev_id)
1135
1115
            accepted_revisions.add(rev_id)
1136
 
        updated = False
1137
 
        if (len(real_trees) == 1
1138
 
            and not ghosts
1139
 
            and self.branch.repository._format.fast_deltas
1140
 
            and isinstance(real_trees[0][1],
1141
 
                revisiontree.InventoryRevisionTree)
1142
 
            and self.get_parent_ids()):
1143
 
            rev_id, rev_tree = real_trees[0]
1144
 
            basis_id = self.get_parent_ids()[0]
1145
 
            # There are times when basis_tree won't be in
1146
 
            # self.branch.repository, (switch, for example)
1147
 
            try:
1148
 
                basis_tree = self.branch.repository.revision_tree(basis_id)
1149
 
            except errors.NoSuchRevision:
1150
 
                # Fall back to the set_parent_trees(), since we can't use
1151
 
                # _make_delta if we can't get the RevisionTree
1152
 
                pass
1153
 
            else:
1154
 
                delta = rev_tree.inventory._make_delta(basis_tree.inventory)
1155
 
                dirstate.update_basis_by_delta(delta, rev_id)
1156
 
                updated = True
1157
 
        if not updated:
1158
 
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1116
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1159
1117
        self._make_dirty(reset_inventory=False)
1160
1118
 
1161
1119
    def _set_root_id(self, file_id):
1181
1139
 
1182
1140
    def unlock(self):
1183
1141
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1142
        # do non-implementation specific cleanup
 
1143
        self._cleanup()
 
1144
 
1184
1145
        if self._control_files._lock_count == 1:
1185
 
            # do non-implementation specific cleanup
1186
 
            self._cleanup()
1187
 
 
1188
1146
            # eventually we should do signature checking during read locks for
1189
1147
            # dirstate updates.
1190
1148
            if self._control_files._lock_mode == 'w':
1253
1211
                # just forget the whole block.
1254
1212
                entry_index = 0
1255
1213
                while entry_index < len(block[1]):
 
1214
                    # Mark this file id as having been removed
1256
1215
                    entry = block[1][entry_index]
1257
 
                    if entry[1][0][0] in 'ar':
1258
 
                        # don't remove absent or renamed entries
 
1216
                    ids_to_unversion.discard(entry[0][2])
 
1217
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1218
                                               # entries
 
1219
                        or not state._make_absent(entry)):
1259
1220
                        entry_index += 1
1260
 
                    else:
1261
 
                        # Mark this file id as having been removed
1262
 
                        ids_to_unversion.discard(entry[0][2])
1263
 
                        if not state._make_absent(entry):
1264
 
                            # The block has not shrunk.
1265
 
                            entry_index += 1
1266
1221
                # go to the next block. (At the moment we dont delete empty
1267
1222
                # dirblocks)
1268
1223
                block_index += 1
1289
1244
        # have to change the legacy inventory too.
1290
1245
        if self._inventory is not None:
1291
1246
            for file_id in file_ids:
1292
 
                if self._inventory.has_id(file_id):
1293
 
                    self._inventory.remove_recursive_id(file_id)
 
1247
                self._inventory.remove_recursive_id(file_id)
1294
1248
 
1295
1249
    @needs_tree_write_lock
1296
1250
    def rename_one(self, from_rel, to_rel, after=False):
1297
1251
        """See WorkingTree.rename_one"""
1298
1252
        self.flush()
1299
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
 
1253
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1300
1254
 
1301
1255
    @needs_tree_write_lock
1302
1256
    def apply_inventory_delta(self, changes):
1321
1275
        if self._dirty:
1322
1276
            raise AssertionError("attempting to write an inventory when the "
1323
1277
                "dirstate is dirty will lose pending changes")
1324
 
        had_inventory = self._inventory is not None
1325
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1326
 
        # working inventory. We do this because self.inventory may be inv, or
1327
 
        # may have been modified, and either case would prevent a clean delta
1328
 
        # being created.
1329
 
        self._inventory = None
1330
 
        # generate a delta,
1331
 
        delta = inv._make_delta(self.inventory)
1332
 
        # and apply it.
1333
 
        self.apply_inventory_delta(delta)
1334
 
        if had_inventory:
 
1278
        self.current_dirstate().set_state_from_inventory(inv)
 
1279
        self._make_dirty(reset_inventory=False)
 
1280
        if self._inventory is not None:
1335
1281
            self._inventory = inv
1336
1282
        self.flush()
1337
1283
 
1338
 
    @needs_tree_write_lock
1339
 
    def reset_state(self, revision_ids=None):
1340
 
        """Reset the state of the working tree.
1341
 
 
1342
 
        This does a hard-reset to a last-known-good state. This is a way to
1343
 
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
1344
 
        """
1345
 
        if revision_ids is None:
1346
 
            revision_ids = self.get_parent_ids()
1347
 
        if not revision_ids:
1348
 
            base_tree = self.branch.repository.revision_tree(
1349
 
                _mod_revision.NULL_REVISION)
1350
 
            trees = []
1351
 
        else:
1352
 
            trees = zip(revision_ids,
1353
 
                        self.branch.repository.revision_trees(revision_ids))
1354
 
            base_tree = trees[0][1]
1355
 
        state = self.current_dirstate()
1356
 
        # We don't support ghosts yet
1357
 
        state.set_state_from_scratch(base_tree.inventory, trees, [])
1358
 
 
1359
 
 
1360
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1361
 
 
1362
 
    def __init__(self, tree):
1363
 
        self.tree = tree
1364
 
 
1365
 
    def sha1(self, abspath):
1366
 
        """See dirstate.SHA1Provider.sha1()."""
1367
 
        filters = self.tree._content_filter_stack(
1368
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1369
 
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
1370
 
 
1371
 
    def stat_and_sha1(self, abspath):
1372
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1373
 
        filters = self.tree._content_filter_stack(
1374
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1375
 
        file_obj = file(abspath, 'rb', 65000)
1376
 
        try:
1377
 
            statvalue = os.fstat(file_obj.fileno())
1378
 
            if filters:
1379
 
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
1380
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1381
 
        finally:
1382
 
            file_obj.close()
1383
 
        return statvalue, sha1
1384
 
 
1385
 
 
1386
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1387
 
    """Dirstate working tree that supports content filtering.
1388
 
 
1389
 
    The dirstate holds the hash and size of the canonical form of the file, 
1390
 
    and most methods must return that.
1391
 
    """
1392
 
 
1393
 
    def _file_content_summary(self, path, stat_result):
1394
 
        # This is to support the somewhat obsolete path_content_summary method
1395
 
        # with content filtering: see
1396
 
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
1397
 
        #
1398
 
        # If the dirstate cache is up to date and knows the hash and size,
1399
 
        # return that.
1400
 
        # Otherwise if there are no content filters, return the on-disk size
1401
 
        # and leave the hash blank.
1402
 
        # Otherwise, read and filter the on-disk file and use its size and
1403
 
        # hash.
1404
 
        #
1405
 
        # The dirstate doesn't store the size of the canonical form so we
1406
 
        # can't trust it for content-filtered trees.  We just return None.
1407
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1408
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1409
 
        return ('file', None, executable, dirstate_sha1)
1410
 
 
1411
 
 
1412
 
class WorkingTree4(DirStateWorkingTree):
1413
 
    """This is the Format 4 working tree.
1414
 
 
1415
 
    This differs from WorkingTree by:
1416
 
     - Having a consolidated internal dirstate, stored in a
1417
 
       randomly-accessible sorted file on disk.
1418
 
     - Not having a regular inventory attribute.  One can be synthesized
1419
 
       on demand but this is expensive and should be avoided.
1420
 
 
1421
 
    This is new in bzr 0.15.
1422
 
    """
1423
 
 
1424
 
 
1425
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1426
 
    """This is the Format 5 working tree.
1427
 
 
1428
 
    This differs from WorkingTree4 by:
1429
 
     - Supporting content filtering.
1430
 
 
1431
 
    This is new in bzr 1.11.
1432
 
    """
1433
 
 
1434
 
 
1435
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1436
 
    """This is the Format 6 working tree.
1437
 
 
1438
 
    This differs from WorkingTree5 by:
1439
 
     - Supporting a current view that may mask the set of files in a tree
1440
 
       impacted by most user operations.
1441
 
 
1442
 
    This is new in bzr 1.14.
1443
 
    """
1444
 
 
1445
 
    def _make_views(self):
1446
 
        return views.PathBasedViews(self)
1447
 
 
1448
 
 
1449
 
class DirStateWorkingTreeFormat(WorkingTreeFormat):
1450
 
 
1451
 
    missing_parent_conflicts = True
1452
 
 
1453
 
    supports_versioned_directories = True
1454
 
 
1455
 
    _lock_class = LockDir
1456
 
    _lock_file_name = 'lock'
1457
 
 
1458
 
    def _open_control_files(self, a_bzrdir):
1459
 
        transport = a_bzrdir.get_workingtree_transport(None)
1460
 
        return LockableFiles(transport, self._lock_file_name,
1461
 
                             self._lock_class)
 
1284
 
 
1285
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1286
    """The first consolidated dirstate working tree format.
 
1287
 
 
1288
    This format:
 
1289
        - exists within a metadir controlling .bzr
 
1290
        - includes an explicit version marker for the workingtree control
 
1291
          files, separate from the BzrDir format
 
1292
        - modifies the hash cache format
 
1293
        - is new in bzr 0.15
 
1294
        - uses a LockDir to guard access to it.
 
1295
    """
 
1296
 
 
1297
    upgrade_recommended = False
 
1298
 
 
1299
    _tree_class = WorkingTree4
 
1300
 
 
1301
    def get_format_string(self):
 
1302
        """See WorkingTreeFormat.get_format_string()."""
 
1303
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1304
 
 
1305
    def get_format_description(self):
 
1306
        """See WorkingTreeFormat.get_format_description()."""
 
1307
        return "Working tree format 4"
1462
1308
 
1463
1309
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1464
1310
                   accelerator_tree=None, hardlink=False):
1465
1311
        """See WorkingTreeFormat.initialize().
1466
1312
 
1467
1313
        :param revision_id: allows creating a working tree at a different
1468
 
            revision than the branch is at.
 
1314
        revision than the branch is at.
1469
1315
        :param accelerator_tree: A tree which can be used for retrieving file
1470
1316
            contents more quickly than the revision tree, i.e. a workingtree.
1471
1317
            The revision tree will be used for cases where accelerator_tree's
1504
1350
        wt.lock_tree_write()
1505
1351
        try:
1506
1352
            self._init_custom_control_files(wt)
1507
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1353
            if revision_id in (None, NULL_REVISION):
1508
1354
                if branch.repository.supports_rich_root():
1509
1355
                    wt._set_root_id(generate_ids.gen_root_id())
1510
1356
                else:
1521
1367
                    pass
1522
1368
            if basis is None:
1523
1369
                basis = branch.repository.revision_tree(revision_id)
1524
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1370
            if revision_id == NULL_REVISION:
1525
1371
                parents_list = []
1526
1372
            else:
1527
1373
                parents_list = [(revision_id, basis)]
1535
1381
                if basis_root_id is not None:
1536
1382
                    wt._set_root_id(basis_root_id)
1537
1383
                    wt.flush()
1538
 
                if wt.supports_content_filtering():
1539
 
                    # The original tree may not have the same content filters
1540
 
                    # applied so we can't safely build the inventory delta from
1541
 
                    # the source tree.
1542
 
                    delta_from_tree = False
1543
 
                else:
1544
 
                    delta_from_tree = True
1545
1384
                # delta_from_tree is safe even for DirStateRevisionTrees,
1546
1385
                # because wt4.apply_inventory_delta does not mutate the input
1547
1386
                # inventory entries.
1548
1387
                transform.build_tree(basis, wt, accelerator_tree,
1549
 
                                     hardlink=hardlink,
1550
 
                                     delta_from_tree=delta_from_tree)
 
1388
                                     hardlink=hardlink, delta_from_tree=True)
1551
1389
            finally:
1552
1390
                basis.unlock()
1553
1391
        finally:
1557
1395
 
1558
1396
    def _init_custom_control_files(self, wt):
1559
1397
        """Subclasses with custom control files should override this method.
1560
 
 
 
1398
        
1561
1399
        The working tree and control files are locked for writing when this
1562
1400
        method is called.
1563
 
 
 
1401
        
1564
1402
        :param wt: the WorkingTree object
1565
1403
        """
1566
1404
 
1567
 
    def open(self, a_bzrdir, _found=False):
1568
 
        """Return the WorkingTree object for a_bzrdir
1569
 
 
1570
 
        _found is a private parameter, do not use it. It is used to indicate
1571
 
               if format probing has already been done.
1572
 
        """
1573
 
        if not _found:
1574
 
            # we are being called directly and must probe.
1575
 
            raise NotImplementedError
1576
 
        if not isinstance(a_bzrdir.transport, LocalTransport):
1577
 
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1578
 
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
1579
 
        return wt
1580
 
 
1581
1405
    def _open(self, a_bzrdir, control_files):
1582
1406
        """Open the tree itself.
1583
1407
 
1591
1415
                           _control_files=control_files)
1592
1416
 
1593
1417
    def __get_matchingbzrdir(self):
1594
 
        return self._get_matchingbzrdir()
1595
 
 
1596
 
    def _get_matchingbzrdir(self):
1597
 
        """Overrideable method to get a bzrdir for testing."""
1598
1418
        # please test against something that will let us do tree references
1599
1419
        return bzrdir.format_registry.make_bzrdir(
1600
1420
            'dirstate-with-subtree')
1602
1422
    _matchingbzrdir = property(__get_matchingbzrdir)
1603
1423
 
1604
1424
 
1605
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1606
 
    """The first consolidated dirstate working tree format.
1607
 
 
1608
 
    This format:
1609
 
        - exists within a metadir controlling .bzr
1610
 
        - includes an explicit version marker for the workingtree control
1611
 
          files, separate from the BzrDir format
1612
 
        - modifies the hash cache format
1613
 
        - is new in bzr 0.15
1614
 
        - uses a LockDir to guard access to it.
1615
 
    """
1616
 
 
1617
 
    upgrade_recommended = False
1618
 
 
1619
 
    _tree_class = WorkingTree4
1620
 
 
1621
 
    def get_format_string(self):
1622
 
        """See WorkingTreeFormat.get_format_string()."""
1623
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1624
 
 
1625
 
    def get_format_description(self):
1626
 
        """See WorkingTreeFormat.get_format_description()."""
1627
 
        return "Working tree format 4"
1628
 
 
1629
 
 
1630
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1631
 
    """WorkingTree format supporting content filtering.
1632
 
    """
1633
 
 
1634
 
    upgrade_recommended = False
1635
 
 
1636
 
    _tree_class = WorkingTree5
1637
 
 
1638
 
    def get_format_string(self):
1639
 
        """See WorkingTreeFormat.get_format_string()."""
1640
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1641
 
 
1642
 
    def get_format_description(self):
1643
 
        """See WorkingTreeFormat.get_format_description()."""
1644
 
        return "Working tree format 5"
1645
 
 
1646
 
    def supports_content_filtering(self):
1647
 
        return True
1648
 
 
1649
 
 
1650
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1651
 
    """WorkingTree format supporting views.
1652
 
    """
1653
 
 
1654
 
    upgrade_recommended = False
1655
 
 
1656
 
    _tree_class = WorkingTree6
1657
 
 
1658
 
    def get_format_string(self):
1659
 
        """See WorkingTreeFormat.get_format_string()."""
1660
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1661
 
 
1662
 
    def get_format_description(self):
1663
 
        """See WorkingTreeFormat.get_format_description()."""
1664
 
        return "Working tree format 6"
1665
 
 
1666
 
    def _init_custom_control_files(self, wt):
1667
 
        """Subclasses with custom control files should override this method."""
1668
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1669
 
 
1670
 
    def supports_content_filtering(self):
1671
 
        return True
1672
 
 
1673
 
    def supports_views(self):
1674
 
        return True
1675
 
 
1676
 
 
1677
 
class DirStateRevisionTree(InventoryTree):
1678
 
    """A revision tree pulling the inventory from a dirstate.
1679
 
    
1680
 
    Note that this is one of the historical (ie revision) trees cached in the
1681
 
    dirstate for easy access, not the workingtree.
1682
 
    """
 
1425
class DirStateRevisionTree(Tree):
 
1426
    """A revision tree pulling the inventory from a dirstate."""
1683
1427
 
1684
1428
    def __init__(self, dirstate, revision_id, repository):
1685
1429
        self._dirstate = dirstate
1699
1443
    def annotate_iter(self, file_id,
1700
1444
                      default_revision=_mod_revision.CURRENT_REVISION):
1701
1445
        """See Tree.annotate_iter"""
1702
 
        text_key = (file_id, self.get_file_revision(file_id))
 
1446
        text_key = (file_id, self.inventory[file_id].revision)
1703
1447
        annotations = self._repository.texts.annotate(text_key)
1704
1448
        return [(key[-1], line) for (key, line) in annotations]
1705
1449
 
 
1450
    def _get_ancestors(self, default_revision):
 
1451
        return set(self._repository.get_ancestry(self._revision_id,
 
1452
                                                 topo_sorted=False))
1706
1453
    def _comparison_data(self, entry, path):
1707
1454
        """See Tree._comparison_data."""
1708
1455
        if entry is None:
1751
1498
        If either file_id or path is supplied, it is used as the key to lookup.
1752
1499
        If both are supplied, the fastest lookup is used, and an error is
1753
1500
        raised if they do not both point at the same row.
1754
 
 
 
1501
        
1755
1502
        :param file_id: An optional unicode file_id to be looked up.
1756
1503
        :param path: An optional unicode path to be looked up.
1757
1504
        :return: The dirstate row tuple for path/file_id, or (None, None)
1824
1571
                elif kind == 'directory':
1825
1572
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1826
1573
                elif kind == 'symlink':
 
1574
                    inv_entry.executable = False
 
1575
                    inv_entry.text_size = None
1827
1576
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1828
1577
                elif kind == 'tree-reference':
1829
1578
                    inv_entry.reference_revision = fingerprint or None
1852
1601
            return None
1853
1602
        parent_index = self._get_parent_index()
1854
1603
        last_changed_revision = entry[1][parent_index][4]
1855
 
        try:
1856
 
            rev = self._repository.get_revision(last_changed_revision)
1857
 
        except errors.NoSuchRevision:
1858
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1859
 
        return rev.timestamp
 
1604
        return self._repository.get_revision(last_changed_revision).timestamp
1860
1605
 
1861
1606
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1862
1607
        entry = self._get_entry(file_id=file_id, path=path)
1866
1611
            return parent_details[1]
1867
1612
        return None
1868
1613
 
1869
 
    @needs_read_lock
1870
 
    def get_file_revision(self, file_id):
1871
 
        return self.inventory[file_id].revision
1872
 
 
1873
1614
    def get_file(self, file_id, path=None):
1874
1615
        return StringIO(self.get_file_text(file_id))
1875
1616
 
 
1617
    def get_file_lines(self, file_id):
 
1618
        return osutils.split_lines(self.get_file_text(file_id))
 
1619
 
1876
1620
    def get_file_size(self, file_id):
1877
1621
        """See Tree.get_file_size"""
1878
1622
        return self.inventory[file_id].text_size
1879
1623
 
1880
 
    def get_file_text(self, file_id, path=None):
1881
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1882
 
        return ''.join(content)
 
1624
    def get_file_text(self, file_id):
 
1625
        return list(self.iter_files_bytes([(file_id, None)]))[0][1]
1883
1626
 
1884
1627
    def get_reference_revision(self, file_id, path=None):
1885
1628
        return self.inventory[file_id].reference_revision
1898
1641
                                       identifier))
1899
1642
        return self._repository.iter_files_bytes(repo_desired_files)
1900
1643
 
1901
 
    def get_symlink_target(self, file_id, path=None):
 
1644
    def get_symlink_target(self, file_id):
1902
1645
        entry = self._get_entry(file_id=file_id)
1903
1646
        parent_index = self._get_parent_index()
1904
1647
        if entry[1][parent_index][0] != 'l':
1905
1648
            return None
1906
1649
        else:
1907
 
            target = entry[1][parent_index][1]
1908
 
            target = target.decode('utf8')
1909
 
            return target
 
1650
            # At present, none of the tree implementations supports non-ascii
 
1651
            # symlink targets. So we will just assume that the dirstate path is
 
1652
            # correct.
 
1653
            return entry[1][parent_index][1]
1910
1654
 
1911
1655
    def get_revision_id(self):
1912
1656
        """Return the revision id for this tree."""
1933
1677
        entry = self._get_entry(file_id=file_id)[1]
1934
1678
        if entry is None:
1935
1679
            raise errors.NoSuchId(tree=self, file_id=file_id)
1936
 
        parent_index = self._get_parent_index()
1937
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1680
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1938
1681
 
1939
1682
    def stored_kind(self, file_id):
1940
1683
        """See Tree.stored_kind"""
1957
1700
    def is_executable(self, file_id, path=None):
1958
1701
        ie = self.inventory[file_id]
1959
1702
        if ie.kind != "file":
1960
 
            return False
 
1703
            return None
1961
1704
        return ie.executable
1962
1705
 
1963
 
    def is_locked(self):
1964
 
        return self._locked
1965
 
 
1966
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1706
    def list_files(self, include_root=False):
1967
1707
        # We use a standard implementation, because DirStateRevisionTree is
1968
1708
        # dealing with one of the parents of the current state
1969
1709
        inv = self._get_inventory()
1970
 
        if from_dir is None:
1971
 
            from_dir_id = None
1972
 
        else:
1973
 
            from_dir_id = inv.path2id(from_dir)
1974
 
            if from_dir_id is None:
1975
 
                # Directory not versioned
1976
 
                return
1977
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1978
 
        if inv.root is not None and not include_root and from_dir is None:
 
1710
        entries = inv.iter_entries()
 
1711
        if self.inventory.root is not None and not include_root:
1979
1712
            entries.next()
1980
1713
        for path, entry in entries:
1981
1714
            yield path, 'V', entry.kind, entry.file_id, entry
1982
1715
 
1983
1716
    def lock_read(self):
1984
 
        """Lock the tree for a set of operations.
1985
 
 
1986
 
        :return: A bzrlib.lock.LogicalLockResult.
1987
 
        """
 
1717
        """Lock the tree for a set of operations."""
1988
1718
        if not self._locked:
1989
1719
            self._repository.lock_read()
1990
1720
            if self._dirstate._lock_token is None:
1991
1721
                self._dirstate.lock_read()
1992
1722
                self._dirstate_locked = True
1993
1723
        self._locked += 1
1994
 
        return LogicalLockResult(self.unlock)
1995
1724
 
1996
1725
    def _must_be_locked(self):
1997
1726
        if not self._locked:
2024
1753
 
2025
1754
    def walkdirs(self, prefix=""):
2026
1755
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2027
 
        # implementation based on an inventory.
 
1756
        # implementation based on an inventory.  
2028
1757
        # This should be cleaned up to use the much faster Dirstate code
2029
1758
        # So for now, we just build up the parent inventory, and extract
2030
1759
        # it the same way RevisionTree does.
2059
1788
 
2060
1789
class InterDirStateTree(InterTree):
2061
1790
    """Fast path optimiser for changes_from with dirstate trees.
2062
 
 
2063
 
    This is used only when both trees are in the dirstate working file, and
2064
 
    the source is any parent within the dirstate, and the destination is
 
1791
    
 
1792
    This is used only when both trees are in the dirstate working file, and 
 
1793
    the source is any parent within the dirstate, and the destination is 
2065
1794
    the current working tree of the same dirstate.
2066
1795
    """
2067
1796
    # this could be generalized to allow comparisons between any trees in the
2076
1805
    def make_source_parent_tree(source, target):
2077
1806
        """Change the source tree into a parent of the target."""
2078
1807
        revid = source.commit('record tree')
2079
 
        target.branch.fetch(source.branch, revid)
 
1808
        target.branch.repository.fetch(source.branch.repository, revid)
2080
1809
        target.set_parent_ids([revid])
2081
1810
        return target.basis_tree(), target
2082
1811
 
2083
 
    @classmethod
2084
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
2085
 
        result = klass.make_source_parent_tree(source, target)
2086
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2087
 
        return result
2088
 
 
2089
 
    @classmethod
2090
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2091
 
                                                  target):
2092
 
        from bzrlib.tests.test__dirstate_helpers import \
2093
 
            compiled_dirstate_helpers_feature
2094
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2095
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2096
 
        result = klass.make_source_parent_tree(source, target)
2097
 
        result[1]._iter_changes = ProcessEntryC
2098
 
        return result
2099
 
 
2100
1812
    _matching_from_tree_format = WorkingTreeFormat4()
2101
1813
    _matching_to_tree_format = WorkingTreeFormat4()
2102
 
 
2103
 
    @classmethod
2104
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2105
 
        # This method shouldn't be called, because we have python and C
2106
 
        # specific flavours.
2107
 
        raise NotImplementedError
 
1814
    _test_mutable_trees_to_test_trees = make_source_parent_tree
2108
1815
 
2109
1816
    def iter_changes(self, include_unchanged=False,
2110
1817
                      specific_files=None, pb=None, extra_trees=[],
2128
1835
            output. An unversioned file is defined as one with (False, False)
2129
1836
            for the versioned pair.
2130
1837
        """
 
1838
        utf8_decode = cache_utf8._utf8_decode
 
1839
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1840
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1841
        # NB: show_status depends on being able to pass in non-versioned files
 
1842
        # and report them as unknown
2131
1843
        # TODO: handle extra trees in the dirstate.
2132
1844
        if (extra_trees or specific_files == []):
2133
1845
            # we can't fast-path these cases (yet)
2134
 
            return super(InterDirStateTree, self).iter_changes(
 
1846
            for f in super(InterDirStateTree, self).iter_changes(
2135
1847
                include_unchanged, specific_files, pb, extra_trees,
2136
 
                require_versioned, want_unversioned=want_unversioned)
 
1848
                require_versioned, want_unversioned=want_unversioned):
 
1849
                yield f
 
1850
            return
2137
1851
        parent_ids = self.target.get_parent_ids()
2138
1852
        if not (self.source._revision_id in parent_ids
2139
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
 
1853
                or self.source._revision_id == NULL_REVISION):
2140
1854
            raise AssertionError(
2141
1855
                "revision {%s} is not stored in {%s}, but %s "
2142
1856
                "can only be used for trees stored in the dirstate"
2143
1857
                % (self.source._revision_id, self.target, self.iter_changes))
2144
1858
        target_index = 0
2145
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1859
        if self.source._revision_id == NULL_REVISION:
2146
1860
            source_index = None
2147
1861
            indices = (target_index,)
2148
1862
        else:
2156
1870
        if specific_files:
2157
1871
            specific_files_utf8 = set()
2158
1872
            for path in specific_files:
2159
 
                # Note, if there are many specific files, using cache_utf8
2160
 
                # would be good here.
2161
1873
                specific_files_utf8.add(path.encode('utf8'))
2162
1874
            specific_files = specific_files_utf8
2163
1875
        else:
2164
1876
            specific_files = set([''])
2165
1877
        # -- specific_files is now a utf8 path set --
2166
 
 
2167
1878
        # -- get the state object and prepare it.
2168
1879
        state = self.target.current_dirstate()
2169
1880
        state._read_dirblocks_if_needed()
 
1881
        def _entries_for_path(path):
 
1882
            """Return a list with all the entries that match path for all ids.
 
1883
            """
 
1884
            dirname, basename = os.path.split(path)
 
1885
            key = (dirname, basename, '')
 
1886
            block_index, present = state._find_block_index_from_key(key)
 
1887
            if not present:
 
1888
                # the block which should contain path is absent.
 
1889
                return []
 
1890
            result = []
 
1891
            block = state._dirblocks[block_index][1]
 
1892
            entry_index, _ = state._find_entry_index(key, block)
 
1893
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1894
            while (entry_index < len(block) and
 
1895
                block[entry_index][0][0:2] == key[0:2]):
 
1896
                result.append(block[entry_index])
 
1897
                entry_index += 1
 
1898
            return result
2170
1899
        if require_versioned:
2171
1900
            # -- check all supplied paths are versioned in a search tree. --
2172
 
            not_versioned = []
 
1901
            all_versioned = True
2173
1902
            for path in specific_files:
2174
 
                path_entries = state._entries_for_path(path)
 
1903
                path_entries = _entries_for_path(path)
2175
1904
                if not path_entries:
2176
1905
                    # this specified path is not present at all: error
2177
 
                    not_versioned.append(path)
2178
 
                    continue
 
1906
                    all_versioned = False
 
1907
                    break
2179
1908
                found_versioned = False
2180
1909
                # for each id at this path
2181
1910
                for entry in path_entries:
2188
1917
                if not found_versioned:
2189
1918
                    # none of the indexes was not 'absent' at all ids for this
2190
1919
                    # path.
2191
 
                    not_versioned.append(path)
2192
 
            if len(not_versioned) > 0:
2193
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1920
                    all_versioned = False
 
1921
                    break
 
1922
            if not all_versioned:
 
1923
                raise errors.PathsNotVersionedError(specific_files)
2194
1924
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2195
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1925
        search_specific_files = set()
 
1926
        for path in specific_files:
 
1927
            other_specific_files = specific_files.difference(set([path]))
 
1928
            if not osutils.is_inside_any(other_specific_files, path):
 
1929
                # this is a top level path, we must check it.
 
1930
                search_specific_files.add(path)
 
1931
        # sketch: 
 
1932
        # compare source_index and target_index at or under each element of search_specific_files.
 
1933
        # follow the following comparison table. Note that we only want to do diff operations when
 
1934
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1935
        # for the target.
 
1936
        # cases:
 
1937
        # 
 
1938
        # Source | Target | disk | action
 
1939
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1940
        #        |        |      | diff check on source-target
 
1941
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1942
        #        |        |      | ???
 
1943
        #   r    |  a     |      | add source to search
 
1944
        #   r    |  a     |  a   | 
 
1945
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1946
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1947
        #   a    | fdlt   |      | add new id
 
1948
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1949
        #   a    |  a     |      | not present in either tree, skip
 
1950
        #   a    |  a     |  a   | not present in any tree, skip
 
1951
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1952
        #        |        |      | may not be selected by the users list of paths.
 
1953
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1954
        #        |        |      | may not be selected by the users list of paths.
 
1955
        #  fdlt  | fdlt   |      | content in both: diff them
 
1956
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1957
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1958
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1959
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1960
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1961
        #        |        |      | this id at the other path.
 
1962
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1963
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1964
        #        |        |      | this id at the other path.
 
1965
 
 
1966
        # for all search_indexs in each path at or under each element of
 
1967
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1968
        # relocated path as one to search if its not searched already. If the
 
1969
        # detail is not relocated, add the id.
 
1970
        searched_specific_files = set()
 
1971
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1972
        # Using a list so that we can access the values and change them in
 
1973
        # nested scope. Each one is [path, file_id, entry]
 
1974
        last_source_parent = [None, None]
 
1975
        last_target_parent = [None, None]
2196
1976
 
2197
1977
        use_filesystem_for_exec = (sys.platform != 'win32')
2198
 
        iter_changes = self.target._iter_changes(include_unchanged,
2199
 
            use_filesystem_for_exec, search_specific_files, state,
2200
 
            source_index, target_index, want_unversioned, self.target)
2201
 
        return iter_changes.iter_changes()
 
1978
 
 
1979
        # Just a sentry, so that _process_entry can say that this
 
1980
        # record is handled, but isn't interesting to process (unchanged)
 
1981
        uninteresting = object()
 
1982
 
 
1983
        old_dirname_to_file_id = {}
 
1984
        new_dirname_to_file_id = {}
 
1985
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1986
        #       keeping a cache of directories that we have seen.
 
1987
 
 
1988
        def _process_entry(entry, path_info):
 
1989
            """Compare an entry and real disk to generate delta information.
 
1990
 
 
1991
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1992
                the path of entry. If None, then the path is considered absent.
 
1993
                (Perhaps we should pass in a concrete entry for this ?)
 
1994
                Basename is returned as a utf8 string because we expect this
 
1995
                tuple will be ignored, and don't want to take the time to
 
1996
                decode.
 
1997
            :return: None if these don't match
 
1998
                     A tuple of information about the change, or
 
1999
                     the object 'uninteresting' if these match, but are
 
2000
                     basically identical.
 
2001
            """
 
2002
            if source_index is None:
 
2003
                source_details = NULL_PARENT_DETAILS
 
2004
            else:
 
2005
                source_details = entry[1][source_index]
 
2006
            target_details = entry[1][target_index]
 
2007
            target_minikind = target_details[0]
 
2008
            if path_info is not None and target_minikind in 'fdlt':
 
2009
                if not (target_index == 0):
 
2010
                    raise AssertionError()
 
2011
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
2012
                                                  stat_value=path_info[3])
 
2013
                # The entry may have been modified by update_entry
 
2014
                target_details = entry[1][target_index]
 
2015
                target_minikind = target_details[0]
 
2016
            else:
 
2017
                link_or_sha1 = None
 
2018
            file_id = entry[0][2]
 
2019
            source_minikind = source_details[0]
 
2020
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
2021
                # claimed content in both: diff
 
2022
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
2023
                #        |        |      | diff check on source-target
 
2024
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
2025
                #        |        |      | ???
 
2026
                if source_minikind in 'r':
 
2027
                    # add the source to the search path to find any children it
 
2028
                    # has.  TODO ? : only add if it is a container ?
 
2029
                    if not osutils.is_inside_any(searched_specific_files,
 
2030
                                                 source_details[1]):
 
2031
                        search_specific_files.add(source_details[1])
 
2032
                    # generate the old path; this is needed for stating later
 
2033
                    # as well.
 
2034
                    old_path = source_details[1]
 
2035
                    old_dirname, old_basename = os.path.split(old_path)
 
2036
                    path = pathjoin(entry[0][0], entry[0][1])
 
2037
                    old_entry = state._get_entry(source_index,
 
2038
                                                 path_utf8=old_path)
 
2039
                    # update the source details variable to be the real
 
2040
                    # location.
 
2041
                    if old_entry == (None, None):
 
2042
                        raise errors.CorruptDirstate(state._filename,
 
2043
                            "entry '%s/%s' is considered renamed from %r"
 
2044
                            " but source does not exist\n"
 
2045
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
2046
                    source_details = old_entry[1][source_index]
 
2047
                    source_minikind = source_details[0]
 
2048
                else:
 
2049
                    old_dirname = entry[0][0]
 
2050
                    old_basename = entry[0][1]
 
2051
                    old_path = path = None
 
2052
                if path_info is None:
 
2053
                    # the file is missing on disk, show as removed.
 
2054
                    content_change = True
 
2055
                    target_kind = None
 
2056
                    target_exec = False
 
2057
                else:
 
2058
                    # source and target are both versioned and disk file is present.
 
2059
                    target_kind = path_info[2]
 
2060
                    if target_kind == 'directory':
 
2061
                        if path is None:
 
2062
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2063
                        new_dirname_to_file_id[path] = file_id
 
2064
                        if source_minikind != 'd':
 
2065
                            content_change = True
 
2066
                        else:
 
2067
                            # directories have no fingerprint
 
2068
                            content_change = False
 
2069
                        target_exec = False
 
2070
                    elif target_kind == 'file':
 
2071
                        if source_minikind != 'f':
 
2072
                            content_change = True
 
2073
                        else:
 
2074
                            # We could check the size, but we already have the
 
2075
                            # sha1 hash.
 
2076
                            content_change = (link_or_sha1 != source_details[1])
 
2077
                        # Target details is updated at update_entry time
 
2078
                        if use_filesystem_for_exec:
 
2079
                            # We don't need S_ISREG here, because we are sure
 
2080
                            # we are dealing with a file.
 
2081
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2082
                        else:
 
2083
                            target_exec = target_details[3]
 
2084
                    elif target_kind == 'symlink':
 
2085
                        if source_minikind != 'l':
 
2086
                            content_change = True
 
2087
                        else:
 
2088
                            content_change = (link_or_sha1 != source_details[1])
 
2089
                        target_exec = False
 
2090
                    elif target_kind == 'tree-reference':
 
2091
                        if source_minikind != 't':
 
2092
                            content_change = True
 
2093
                        else:
 
2094
                            content_change = False
 
2095
                        target_exec = False
 
2096
                    else:
 
2097
                        raise Exception, "unknown kind %s" % path_info[2]
 
2098
                if source_minikind == 'd':
 
2099
                    if path is None:
 
2100
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2101
                    old_dirname_to_file_id[old_path] = file_id
 
2102
                # parent id is the entry for the path in the target tree
 
2103
                if old_dirname == last_source_parent[0]:
 
2104
                    source_parent_id = last_source_parent[1]
 
2105
                else:
 
2106
                    try:
 
2107
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2108
                    except KeyError:
 
2109
                        source_parent_entry = state._get_entry(source_index,
 
2110
                                                               path_utf8=old_dirname)
 
2111
                        source_parent_id = source_parent_entry[0][2]
 
2112
                    if source_parent_id == entry[0][2]:
 
2113
                        # This is the root, so the parent is None
 
2114
                        source_parent_id = None
 
2115
                    else:
 
2116
                        last_source_parent[0] = old_dirname
 
2117
                        last_source_parent[1] = source_parent_id
 
2118
                new_dirname = entry[0][0]
 
2119
                if new_dirname == last_target_parent[0]:
 
2120
                    target_parent_id = last_target_parent[1]
 
2121
                else:
 
2122
                    try:
 
2123
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2124
                    except KeyError:
 
2125
                        # TODO: We don't always need to do the lookup, because the
 
2126
                        #       parent entry will be the same as the source entry.
 
2127
                        target_parent_entry = state._get_entry(target_index,
 
2128
                                                               path_utf8=new_dirname)
 
2129
                        if target_parent_entry == (None, None):
 
2130
                            raise AssertionError(
 
2131
                                "Could not find target parent in wt: %s\nparent of: %s"
 
2132
                                % (new_dirname, entry))
 
2133
                        target_parent_id = target_parent_entry[0][2]
 
2134
                    if target_parent_id == entry[0][2]:
 
2135
                        # This is the root, so the parent is None
 
2136
                        target_parent_id = None
 
2137
                    else:
 
2138
                        last_target_parent[0] = new_dirname
 
2139
                        last_target_parent[1] = target_parent_id
 
2140
 
 
2141
                source_exec = source_details[3]
 
2142
                if (include_unchanged
 
2143
                    or content_change
 
2144
                    or source_parent_id != target_parent_id
 
2145
                    or old_basename != entry[0][1]
 
2146
                    or source_exec != target_exec
 
2147
                    ):
 
2148
                    if old_path is None:
 
2149
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2150
                        old_path_u = utf8_decode(old_path)[0]
 
2151
                        path_u = old_path_u
 
2152
                    else:
 
2153
                        old_path_u = utf8_decode(old_path)[0]
 
2154
                        if old_path == path:
 
2155
                            path_u = old_path_u
 
2156
                        else:
 
2157
                            path_u = utf8_decode(path)[0]
 
2158
                    source_kind = _minikind_to_kind[source_minikind]
 
2159
                    return (entry[0][2],
 
2160
                           (old_path_u, path_u),
 
2161
                           content_change,
 
2162
                           (True, True),
 
2163
                           (source_parent_id, target_parent_id),
 
2164
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2165
                           (source_kind, target_kind),
 
2166
                           (source_exec, target_exec))
 
2167
                else:
 
2168
                    return uninteresting
 
2169
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2170
                # looks like a new file
 
2171
                path = pathjoin(entry[0][0], entry[0][1])
 
2172
                # parent id is the entry for the path in the target tree
 
2173
                # TODO: these are the same for an entire directory: cache em.
 
2174
                parent_id = state._get_entry(target_index,
 
2175
                                             path_utf8=entry[0][0])[0][2]
 
2176
                if parent_id == entry[0][2]:
 
2177
                    parent_id = None
 
2178
                if path_info is not None:
 
2179
                    # Present on disk:
 
2180
                    if use_filesystem_for_exec:
 
2181
                        # We need S_ISREG here, because we aren't sure if this
 
2182
                        # is a file or not.
 
2183
                        target_exec = bool(
 
2184
                            stat.S_ISREG(path_info[3].st_mode)
 
2185
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2186
                    else:
 
2187
                        target_exec = target_details[3]
 
2188
                    return (entry[0][2],
 
2189
                           (None, utf8_decode(path)[0]),
 
2190
                           True,
 
2191
                           (False, True),
 
2192
                           (None, parent_id),
 
2193
                           (None, utf8_decode(entry[0][1])[0]),
 
2194
                           (None, path_info[2]),
 
2195
                           (None, target_exec))
 
2196
                else:
 
2197
                    # Its a missing file, report it as such.
 
2198
                    return (entry[0][2],
 
2199
                           (None, utf8_decode(path)[0]),
 
2200
                           False,
 
2201
                           (False, True),
 
2202
                           (None, parent_id),
 
2203
                           (None, utf8_decode(entry[0][1])[0]),
 
2204
                           (None, None),
 
2205
                           (None, False))
 
2206
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2207
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2208
                # if its still on disk, *and* theres no other entry at this
 
2209
                # path [we dont know this in this routine at the moment -
 
2210
                # perhaps we should change this - then it would be an unknown.
 
2211
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2212
                # parent id is the entry for the path in the target tree
 
2213
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2214
                if parent_id == entry[0][2]:
 
2215
                    parent_id = None
 
2216
                return (entry[0][2],
 
2217
                       (utf8_decode(old_path)[0], None),
 
2218
                       True,
 
2219
                       (True, False),
 
2220
                       (parent_id, None),
 
2221
                       (utf8_decode(entry[0][1])[0], None),
 
2222
                       (_minikind_to_kind[source_minikind], None),
 
2223
                       (source_details[3], None))
 
2224
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2225
                # a rename; could be a true rename, or a rename inherited from
 
2226
                # a renamed parent. TODO: handle this efficiently. Its not
 
2227
                # common case to rename dirs though, so a correct but slow
 
2228
                # implementation will do.
 
2229
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2230
                    search_specific_files.add(target_details[1])
 
2231
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2232
                # neither of the selected trees contain this file,
 
2233
                # so skip over it. This is not currently directly tested, but
 
2234
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2235
                pass
 
2236
            else:
 
2237
                raise AssertionError("don't know how to compare "
 
2238
                    "source_minikind=%r, target_minikind=%r"
 
2239
                    % (source_minikind, target_minikind))
 
2240
                ## import pdb;pdb.set_trace()
 
2241
            return None
 
2242
 
 
2243
        while search_specific_files:
 
2244
            # TODO: the pending list should be lexically sorted?  the
 
2245
            # interface doesn't require it.
 
2246
            current_root = search_specific_files.pop()
 
2247
            current_root_unicode = current_root.decode('utf8')
 
2248
            searched_specific_files.add(current_root)
 
2249
            # process the entries for this containing directory: the rest will be
 
2250
            # found by their parents recursively.
 
2251
            root_entries = _entries_for_path(current_root)
 
2252
            root_abspath = self.target.abspath(current_root_unicode)
 
2253
            try:
 
2254
                root_stat = os.lstat(root_abspath)
 
2255
            except OSError, e:
 
2256
                if e.errno == errno.ENOENT:
 
2257
                    # the path does not exist: let _process_entry know that.
 
2258
                    root_dir_info = None
 
2259
                else:
 
2260
                    # some other random error: hand it up.
 
2261
                    raise
 
2262
            else:
 
2263
                root_dir_info = ('', current_root,
 
2264
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2265
                    root_abspath)
 
2266
                if root_dir_info[2] == 'directory':
 
2267
                    if self.target._directory_is_tree_reference(
 
2268
                        current_root.decode('utf8')):
 
2269
                        root_dir_info = root_dir_info[:2] + \
 
2270
                            ('tree-reference',) + root_dir_info[3:]
 
2271
 
 
2272
            if not root_entries and not root_dir_info:
 
2273
                # this specified path is not present at all, skip it.
 
2274
                continue
 
2275
            path_handled = False
 
2276
            for entry in root_entries:
 
2277
                result = _process_entry(entry, root_dir_info)
 
2278
                if result is not None:
 
2279
                    path_handled = True
 
2280
                    if result is not uninteresting:
 
2281
                        yield result
 
2282
            if want_unversioned and not path_handled and root_dir_info:
 
2283
                new_executable = bool(
 
2284
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2285
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2286
                yield (None,
 
2287
                       (None, current_root_unicode),
 
2288
                       True,
 
2289
                       (False, False),
 
2290
                       (None, None),
 
2291
                       (None, splitpath(current_root_unicode)[-1]),
 
2292
                       (None, root_dir_info[2]),
 
2293
                       (None, new_executable)
 
2294
                      )
 
2295
            initial_key = (current_root, '', '')
 
2296
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2297
            if block_index == 0:
 
2298
                # we have processed the total root already, but because the
 
2299
                # initial key matched it we should skip it here.
 
2300
                block_index +=1
 
2301
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2302
                current_dir_info = None
 
2303
            else:
 
2304
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2305
                try:
 
2306
                    current_dir_info = dir_iterator.next()
 
2307
                except OSError, e:
 
2308
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2309
                    # python 2.5 has e.errno == EINVAL,
 
2310
                    #            and e.winerror == ERROR_DIRECTORY
 
2311
                    e_winerror = getattr(e, 'winerror', None)
 
2312
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2313
                    # there may be directories in the inventory even though
 
2314
                    # this path is not a file on disk: so mark it as end of
 
2315
                    # iterator
 
2316
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2317
                        current_dir_info = None
 
2318
                    elif (sys.platform == 'win32'
 
2319
                          and (e.errno in win_errors
 
2320
                               or e_winerror in win_errors)):
 
2321
                        current_dir_info = None
 
2322
                    else:
 
2323
                        raise
 
2324
                else:
 
2325
                    if current_dir_info[0][0] == '':
 
2326
                        # remove .bzr from iteration
 
2327
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2328
                        if current_dir_info[1][bzr_index][0] != '.bzr':
 
2329
                            raise AssertionError()
 
2330
                        del current_dir_info[1][bzr_index]
 
2331
            # walk until both the directory listing and the versioned metadata
 
2332
            # are exhausted. 
 
2333
            if (block_index < len(state._dirblocks) and
 
2334
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2335
                current_block = state._dirblocks[block_index]
 
2336
            else:
 
2337
                current_block = None
 
2338
            while (current_dir_info is not None or
 
2339
                   current_block is not None):
 
2340
                if (current_dir_info and current_block
 
2341
                    and current_dir_info[0][0] != current_block[0]):
 
2342
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2343
                        # filesystem data refers to paths not covered by the dirblock.
 
2344
                        # this has two possibilities:
 
2345
                        # A) it is versioned but empty, so there is no block for it
 
2346
                        # B) it is not versioned.
 
2347
 
 
2348
                        # if (A) then we need to recurse into it to check for
 
2349
                        # new unknown files or directories.
 
2350
                        # if (B) then we should ignore it, because we don't
 
2351
                        # recurse into unknown directories.
 
2352
                        path_index = 0
 
2353
                        while path_index < len(current_dir_info[1]):
 
2354
                                current_path_info = current_dir_info[1][path_index]
 
2355
                                if want_unversioned:
 
2356
                                    if current_path_info[2] == 'directory':
 
2357
                                        if self.target._directory_is_tree_reference(
 
2358
                                            current_path_info[0].decode('utf8')):
 
2359
                                            current_path_info = current_path_info[:2] + \
 
2360
                                                ('tree-reference',) + current_path_info[3:]
 
2361
                                    new_executable = bool(
 
2362
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2363
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2364
                                    yield (None,
 
2365
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2366
                                        True,
 
2367
                                        (False, False),
 
2368
                                        (None, None),
 
2369
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2370
                                        (None, current_path_info[2]),
 
2371
                                        (None, new_executable))
 
2372
                                # dont descend into this unversioned path if it is
 
2373
                                # a dir
 
2374
                                if current_path_info[2] in ('directory',
 
2375
                                                            'tree-reference'):
 
2376
                                    del current_dir_info[1][path_index]
 
2377
                                    path_index -= 1
 
2378
                                path_index += 1
 
2379
 
 
2380
                        # This dir info has been handled, go to the next
 
2381
                        try:
 
2382
                            current_dir_info = dir_iterator.next()
 
2383
                        except StopIteration:
 
2384
                            current_dir_info = None
 
2385
                    else:
 
2386
                        # We have a dirblock entry for this location, but there
 
2387
                        # is no filesystem path for this. This is most likely
 
2388
                        # because a directory was removed from the disk.
 
2389
                        # We don't have to report the missing directory,
 
2390
                        # because that should have already been handled, but we
 
2391
                        # need to handle all of the files that are contained
 
2392
                        # within.
 
2393
                        for current_entry in current_block[1]:
 
2394
                            # entry referring to file not present on disk.
 
2395
                            # advance the entry only, after processing.
 
2396
                            result = _process_entry(current_entry, None)
 
2397
                            if result is not None:
 
2398
                                if result is not uninteresting:
 
2399
                                    yield result
 
2400
                        block_index +=1
 
2401
                        if (block_index < len(state._dirblocks) and
 
2402
                            osutils.is_inside(current_root,
 
2403
                                              state._dirblocks[block_index][0])):
 
2404
                            current_block = state._dirblocks[block_index]
 
2405
                        else:
 
2406
                            current_block = None
 
2407
                    continue
 
2408
                entry_index = 0
 
2409
                if current_block and entry_index < len(current_block[1]):
 
2410
                    current_entry = current_block[1][entry_index]
 
2411
                else:
 
2412
                    current_entry = None
 
2413
                advance_entry = True
 
2414
                path_index = 0
 
2415
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2416
                    current_path_info = current_dir_info[1][path_index]
 
2417
                    if current_path_info[2] == 'directory':
 
2418
                        if self.target._directory_is_tree_reference(
 
2419
                            current_path_info[0].decode('utf8')):
 
2420
                            current_path_info = current_path_info[:2] + \
 
2421
                                ('tree-reference',) + current_path_info[3:]
 
2422
                else:
 
2423
                    current_path_info = None
 
2424
                advance_path = True
 
2425
                path_handled = False
 
2426
                while (current_entry is not None or
 
2427
                    current_path_info is not None):
 
2428
                    if current_entry is None:
 
2429
                        # the check for path_handled when the path is adnvaced
 
2430
                        # will yield this path if needed.
 
2431
                        pass
 
2432
                    elif current_path_info is None:
 
2433
                        # no path is fine: the per entry code will handle it.
 
2434
                        result = _process_entry(current_entry, current_path_info)
 
2435
                        if result is not None:
 
2436
                            if result is not uninteresting:
 
2437
                                yield result
 
2438
                    elif (current_entry[0][1] != current_path_info[1]
 
2439
                          or current_entry[1][target_index][0] in 'ar'):
 
2440
                        # The current path on disk doesn't match the dirblock
 
2441
                        # record. Either the dirblock is marked as absent, or
 
2442
                        # the file on disk is not present at all in the
 
2443
                        # dirblock. Either way, report about the dirblock
 
2444
                        # entry, and let other code handle the filesystem one.
 
2445
 
 
2446
                        # Compare the basename for these files to determine
 
2447
                        # which comes first
 
2448
                        if current_path_info[1] < current_entry[0][1]:
 
2449
                            # extra file on disk: pass for now, but only
 
2450
                            # increment the path, not the entry
 
2451
                            advance_entry = False
 
2452
                        else:
 
2453
                            # entry referring to file not present on disk.
 
2454
                            # advance the entry only, after processing.
 
2455
                            result = _process_entry(current_entry, None)
 
2456
                            if result is not None:
 
2457
                                if result is not uninteresting:
 
2458
                                    yield result
 
2459
                            advance_path = False
 
2460
                    else:
 
2461
                        result = _process_entry(current_entry, current_path_info)
 
2462
                        if result is not None:
 
2463
                            path_handled = True
 
2464
                            if result is not uninteresting:
 
2465
                                yield result
 
2466
                    if advance_entry and current_entry is not None:
 
2467
                        entry_index += 1
 
2468
                        if entry_index < len(current_block[1]):
 
2469
                            current_entry = current_block[1][entry_index]
 
2470
                        else:
 
2471
                            current_entry = None
 
2472
                    else:
 
2473
                        advance_entry = True # reset the advance flaga
 
2474
                    if advance_path and current_path_info is not None:
 
2475
                        if not path_handled:
 
2476
                            # unversioned in all regards
 
2477
                            if want_unversioned:
 
2478
                                new_executable = bool(
 
2479
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2480
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2481
                                try:
 
2482
                                    relpath_unicode = utf8_decode(current_path_info[0])[0]
 
2483
                                except UnicodeDecodeError:
 
2484
                                    raise errors.BadFilenameEncoding(
 
2485
                                        current_path_info[0], osutils._fs_enc)
 
2486
                                yield (None,
 
2487
                                    (None, relpath_unicode),
 
2488
                                    True,
 
2489
                                    (False, False),
 
2490
                                    (None, None),
 
2491
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2492
                                    (None, current_path_info[2]),
 
2493
                                    (None, new_executable))
 
2494
                            # dont descend into this unversioned path if it is
 
2495
                            # a dir
 
2496
                            if current_path_info[2] in ('directory'):
 
2497
                                del current_dir_info[1][path_index]
 
2498
                                path_index -= 1
 
2499
                        # dont descend the disk iterator into any tree 
 
2500
                        # paths.
 
2501
                        if current_path_info[2] == 'tree-reference':
 
2502
                            del current_dir_info[1][path_index]
 
2503
                            path_index -= 1
 
2504
                        path_index += 1
 
2505
                        if path_index < len(current_dir_info[1]):
 
2506
                            current_path_info = current_dir_info[1][path_index]
 
2507
                            if current_path_info[2] == 'directory':
 
2508
                                if self.target._directory_is_tree_reference(
 
2509
                                    current_path_info[0].decode('utf8')):
 
2510
                                    current_path_info = current_path_info[:2] + \
 
2511
                                        ('tree-reference',) + current_path_info[3:]
 
2512
                        else:
 
2513
                            current_path_info = None
 
2514
                        path_handled = False
 
2515
                    else:
 
2516
                        advance_path = True # reset the advance flagg.
 
2517
                if current_block is not None:
 
2518
                    block_index += 1
 
2519
                    if (block_index < len(state._dirblocks) and
 
2520
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2521
                        current_block = state._dirblocks[block_index]
 
2522
                    else:
 
2523
                        current_block = None
 
2524
                if current_dir_info is not None:
 
2525
                    try:
 
2526
                        current_dir_info = dir_iterator.next()
 
2527
                    except StopIteration:
 
2528
                        current_dir_info = None
2202
2529
 
2203
2530
    @staticmethod
2204
2531
    def is_compatible(source, target):
2205
2532
        # the target must be a dirstate working tree
2206
 
        if not isinstance(target, DirStateWorkingTree):
 
2533
        if not isinstance(target, WorkingTree4):
2207
2534
            return False
2208
 
        # the source must be a revtree or dirstate rev tree.
 
2535
        # the source must be a revtreee or dirstate rev tree.
2209
2536
        if not isinstance(source,
2210
2537
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2211
2538
            return False
2212
2539
        # the source revid must be in the target dirstate
2213
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2540
        if not (source._revision_id == NULL_REVISION or
2214
2541
            source._revision_id in target.get_parent_ids()):
2215
 
            # TODO: what about ghosts? it may well need to
 
2542
            # TODO: what about ghosts? it may well need to 
2216
2543
            # check for them explicitly.
2217
2544
            return False
2218
2545
        return True
2228
2555
 
2229
2556
    def convert(self, tree):
2230
2557
        # lock the control files not the tree, so that we dont get tree
2231
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2558
        # on-unlock behaviours, and so that noone else diddles with the 
2232
2559
        # tree during upgrade.
2233
2560
        tree._control_files.lock_write()
2234
2561
        try:
2263
2590
        tree._transport.put_bytes('format',
2264
2591
            self.target_format.get_format_string(),
2265
2592
            mode=tree.bzrdir._get_file_mode())
2266
 
 
2267
 
 
2268
 
class Converter4to5(object):
2269
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2270
 
 
2271
 
    def __init__(self):
2272
 
        self.target_format = WorkingTreeFormat5()
2273
 
 
2274
 
    def convert(self, tree):
2275
 
        # lock the control files not the tree, so that we don't get tree
2276
 
        # on-unlock behaviours, and so that no-one else diddles with the
2277
 
        # tree during upgrade.
2278
 
        tree._control_files.lock_write()
2279
 
        try:
2280
 
            self.update_format(tree)
2281
 
        finally:
2282
 
            tree._control_files.unlock()
2283
 
 
2284
 
    def update_format(self, tree):
2285
 
        """Change the format marker."""
2286
 
        tree._transport.put_bytes('format',
2287
 
            self.target_format.get_format_string(),
2288
 
            mode=tree.bzrdir._get_file_mode())
2289
 
 
2290
 
 
2291
 
class Converter4or5to6(object):
2292
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2293
 
 
2294
 
    def __init__(self):
2295
 
        self.target_format = WorkingTreeFormat6()
2296
 
 
2297
 
    def convert(self, tree):
2298
 
        # lock the control files not the tree, so that we don't get tree
2299
 
        # on-unlock behaviours, and so that no-one else diddles with the
2300
 
        # tree during upgrade.
2301
 
        tree._control_files.lock_write()
2302
 
        try:
2303
 
            self.init_custom_control_files(tree)
2304
 
            self.update_format(tree)
2305
 
        finally:
2306
 
            tree._control_files.unlock()
2307
 
 
2308
 
    def init_custom_control_files(self, tree):
2309
 
        """Initialize custom control files."""
2310
 
        tree._transport.put_bytes('views', '',
2311
 
            mode=tree.bzrdir._get_file_mode())
2312
 
 
2313
 
    def update_format(self, tree):
2314
 
        """Change the format marker."""
2315
 
        tree._transport.put_bytes('format',
2316
 
            self.target_format.get_format_string(),
2317
 
            mode=tree.bzrdir._get_file_mode())