~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Jelmer Vernooij
  • Date: 2012-01-27 19:05:43 UTC
  • mto: This revision was merged to the branch mainline in revision 6450.
  • Revision ID: jelmer@samba.org-20120127190543-vk350mv4a0c7aug2
Fix weave test.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
1
# Copyright (C) 2007-2011 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
22
22
WorkingTree.open(dir).
23
23
"""
24
24
 
 
25
from __future__ import absolute_import
 
26
 
25
27
from cStringIO import StringIO
26
28
import os
27
29
import sys
28
30
 
29
31
from bzrlib.lazy_import import lazy_import
30
32
lazy_import(globals(), """
31
 
from bisect import bisect_left
32
 
import collections
33
 
from copy import deepcopy
34
33
import errno
35
 
import itertools
36
 
import operator
37
34
import stat
38
 
from time import time
39
 
import warnings
40
35
 
41
 
import bzrlib
42
36
from bzrlib import (
43
37
    bzrdir,
44
38
    cache_utf8,
 
39
    config,
45
40
    conflicts as _mod_conflicts,
46
41
    debug,
47
 
    delta,
48
42
    dirstate,
49
43
    errors,
 
44
    filters as _mod_filters,
50
45
    generate_ids,
51
 
    globbing,
52
 
    ignores,
53
 
    merge,
54
46
    osutils,
55
47
    revision as _mod_revision,
56
48
    revisiontree,
57
 
    textui,
58
49
    trace,
59
50
    transform,
60
 
    urlutils,
61
 
    xml5,
62
 
    xml6,
 
51
    views,
63
52
    )
64
 
import bzrlib.branch
65
 
from bzrlib.transport import get_transport
66
 
import bzrlib.ui
67
53
""")
68
54
 
69
 
from bzrlib import symbol_versioning
70
55
from bzrlib.decorators import needs_read_lock, needs_write_lock
71
 
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
72
 
from bzrlib.lockable_files import LockableFiles, TransportLock
 
56
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
57
from bzrlib.lock import LogicalLockResult
 
58
from bzrlib.lockable_files import LockableFiles
73
59
from bzrlib.lockdir import LockDir
74
 
import bzrlib.mutabletree
75
 
from bzrlib.mutabletree import needs_tree_write_lock
 
60
from bzrlib.mutabletree import (
 
61
    MutableTree,
 
62
    needs_tree_write_lock,
 
63
    )
76
64
from bzrlib.osutils import (
77
65
    file_kind,
78
66
    isdir,
79
 
    normpath,
80
67
    pathjoin,
81
 
    rand_chars,
82
68
    realpath,
83
69
    safe_unicode,
84
 
    splitpath,
85
70
    )
86
 
from bzrlib.trace import mutter, note
87
71
from bzrlib.transport.local import LocalTransport
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
 
    """
 
72
from bzrlib.tree import (
 
73
    InterTree,
 
74
    InventoryTree,
 
75
    )
 
76
from bzrlib.workingtree import (
 
77
    InventoryWorkingTree,
 
78
    WorkingTree,
 
79
    WorkingTreeFormatMetaDir,
 
80
    )
 
81
 
 
82
 
 
83
class DirStateWorkingTree(InventoryWorkingTree):
119
84
 
120
85
    def __init__(self, basedir,
121
86
                 branch,
132
97
        self._format = _format
133
98
        self.bzrdir = _bzrdir
134
99
        basedir = safe_unicode(basedir)
135
 
        mutter("opening working tree %r", basedir)
 
100
        trace.mutter("opening working tree %r", basedir)
136
101
        self._branch = branch
137
102
        self.basedir = realpath(basedir)
138
103
        # if branch is at our basedir and is a format 6 or less
139
104
        # assume all other formats have their own control files.
140
105
        self._control_files = _control_files
 
106
        self._transport = self._control_files._transport
141
107
        self._dirty = None
142
108
        #-------------
143
109
        # during a read or write lock these objects are set, and are
147
113
        #-------------
148
114
        self._setup_directory_is_tree_reference()
149
115
        self._detect_case_handling()
 
116
        self._rules_searcher = None
 
117
        self.views = self._make_views()
 
118
        #--- allow tests to select the dirstate iter_changes implementation
 
119
        self._iter_changes = dirstate._process_entry
150
120
 
151
121
    @needs_tree_write_lock
152
122
    def _add(self, files, ids, kinds):
167
137
            state.add(f, file_id, kind, None, '')
168
138
        self._make_dirty(reset_inventory=True)
169
139
 
 
140
    def _get_check_refs(self):
 
141
        """Return the references needed to perform a check of this tree."""
 
142
        return [('trees', self.last_revision())]
 
143
 
170
144
    def _make_dirty(self, reset_inventory):
171
145
        """Make the tree state dirty.
172
146
 
180
154
    @needs_tree_write_lock
181
155
    def add_reference(self, sub_tree):
182
156
        # use standard implementation, which calls back to self._add
183
 
        # 
 
157
        #
184
158
        # So we don't store the reference_revision in the working dirstate,
185
 
        # it's just recorded at the moment of commit. 
 
159
        # it's just recorded at the moment of commit.
186
160
        self._add_reference(sub_tree)
187
161
 
188
162
    def break_lock(self):
224
198
 
225
199
    def _comparison_data(self, entry, path):
226
200
        kind, executable, stat_value = \
227
 
            WorkingTree3._comparison_data(self, entry, path)
 
201
            WorkingTree._comparison_data(self, entry, path)
228
202
        # it looks like a plain directory, but it's really a reference -- see
229
203
        # also kind()
230
 
        if (self._repo_supports_tree_reference and
231
 
            kind == 'directory' and
232
 
            self._directory_is_tree_reference(path)):
 
204
        if (self._repo_supports_tree_reference and kind == 'directory'
 
205
            and entry is not None and entry.kind == 'tree-reference'):
233
206
            kind = 'tree-reference'
234
207
        return kind, executable, stat_value
235
208
 
237
210
    def commit(self, message=None, revprops=None, *args, **kwargs):
238
211
        # mark the tree as dirty post commit - commit
239
212
        # can change the current versioned list by doing deletes.
240
 
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
213
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
241
214
        self._make_dirty(reset_inventory=True)
242
215
        return result
243
216
 
261
234
            return self._dirstate
262
235
        local_path = self.bzrdir.get_workingtree_transport(None
263
236
            ).local_abspath('dirstate')
264
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
237
        self._dirstate = dirstate.DirState.on_file(local_path,
 
238
            self._sha1_provider(), self._worth_saving_limit())
265
239
        return self._dirstate
266
240
 
 
241
    def _sha1_provider(self):
 
242
        """A function that returns a SHA1Provider suitable for this tree.
 
243
 
 
244
        :return: None if content filtering is not supported by this tree.
 
245
          Otherwise, a SHA1Provider is returned that sha's the canonical
 
246
          form of files, i.e. after read filters are applied.
 
247
        """
 
248
        if self.supports_content_filtering():
 
249
            return ContentFilterAwareSHA1Provider(self)
 
250
        else:
 
251
            return None
 
252
 
 
253
    def _worth_saving_limit(self):
 
254
        """How many hash changes are ok before we must save the dirstate.
 
255
 
 
256
        :return: an integer. -1 means never save.
 
257
        """
 
258
        # FIXME: We want a WorkingTreeStack here -- vila 20110812
 
259
        conf = config.BranchStack(self.branch)
 
260
        return conf.get('bzr.workingtree.worth_saving_limit')
 
261
 
267
262
    def filter_unversioned_files(self, paths):
268
263
        """Filter out paths that are versioned.
269
264
 
301
296
 
302
297
    def _generate_inventory(self):
303
298
        """Create and set self.inventory from the dirstate object.
304
 
        
 
299
 
305
300
        This is relatively expensive: we have to walk the entire dirstate.
306
301
        Ideally we would not, and can deprecate this function.
307
302
        """
352
347
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
353
348
                elif kind == 'tree-reference':
354
349
                    if not self._repo_supports_tree_reference:
355
 
                        raise AssertionError(
356
 
                            "repository of %r "
357
 
                            "doesn't support tree references "
358
 
                            "required by entry %r"
359
 
                            % (self, name))
 
350
                        raise errors.UnsupportedOperation(
 
351
                            self._generate_inventory,
 
352
                            self.branch.repository)
360
353
                    inv_entry.reference_revision = link_or_sha1 or None
361
354
                elif kind != 'symlink':
362
355
                    raise AssertionError("unknown kind %r" % kind)
377
370
        If either file_id or path is supplied, it is used as the key to lookup.
378
371
        If both are supplied, the fastest lookup is used, and an error is
379
372
        raised if they do not both point at the same row.
380
 
        
 
373
 
381
374
        :param file_id: An optional unicode file_id to be looked up.
382
375
        :param path: An optional unicode path to be looked up.
383
376
        :return: The dirstate row tuple for path/file_id, or (None, None)
401
394
        state = self.current_dirstate()
402
395
        if stat_value is None:
403
396
            try:
404
 
                stat_value = os.lstat(file_abspath)
 
397
                stat_value = osutils.lstat(file_abspath)
405
398
            except OSError, e:
406
399
                if e.errno == errno.ENOENT:
407
400
                    return None
408
401
                else:
409
402
                    raise
410
 
        link_or_sha1 = state.update_entry(entry, file_abspath,
411
 
                                          stat_value=stat_value)
 
403
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
404
            stat_value=stat_value)
412
405
        if entry[1][0][0] == 'f':
413
 
            return link_or_sha1
 
406
            if link_or_sha1 is None:
 
407
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
 
408
                try:
 
409
                    sha1 = osutils.sha_file(file_obj)
 
410
                finally:
 
411
                    file_obj.close()
 
412
                self._observed_sha1(file_id, path, (sha1, statvalue))
 
413
                return sha1
 
414
            else:
 
415
                return link_or_sha1
414
416
        return None
415
417
 
416
418
    def _get_inventory(self):
430
432
    @needs_read_lock
431
433
    def get_parent_ids(self):
432
434
        """See Tree.get_parent_ids.
433
 
        
 
435
 
434
436
        This implementation requests the ids list from the dirstate file.
435
437
        """
436
438
        return self.current_dirstate().get_parent_ids()
458
460
        return osutils.lexists(pathjoin(
459
461
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
460
462
 
 
463
    def has_or_had_id(self, file_id):
 
464
        state = self.current_dirstate()
 
465
        row, parents = self._get_entry(file_id=file_id)
 
466
        return row is not None
 
467
 
461
468
    @needs_read_lock
462
469
    def id2path(self, file_id):
463
470
        "Convert a file-id to a path."
474
481
            return False # Missing entries are not executable
475
482
        return entry[1][0][3] # Executable?
476
483
 
477
 
    if not osutils.supports_executable():
478
 
        def is_executable(self, file_id, path=None):
479
 
            """Test if a file is executable or not.
 
484
    def is_executable(self, file_id, path=None):
 
485
        """Test if a file is executable or not.
480
486
 
481
 
            Note: The caller is expected to take a read-lock before calling this.
482
 
            """
 
487
        Note: The caller is expected to take a read-lock before calling this.
 
488
        """
 
489
        if not self._supports_executable():
483
490
            entry = self._get_entry(file_id=file_id, path=path)
484
491
            if entry == (None, None):
485
492
                return False
486
493
            return entry[1][0][3]
487
 
 
488
 
        _is_executable_from_path_and_stat = \
489
 
            _is_executable_from_path_and_stat_from_basis
490
 
    else:
491
 
        def is_executable(self, file_id, path=None):
492
 
            """Test if a file is executable or not.
493
 
 
494
 
            Note: The caller is expected to take a read-lock before calling this.
495
 
            """
 
494
        else:
496
495
            self._must_be_locked()
497
496
            if not path:
498
497
                path = self.id2path(file_id)
499
 
            mode = os.lstat(self.abspath(path)).st_mode
 
498
            mode = osutils.lstat(self.abspath(path)).st_mode
500
499
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
501
500
 
502
501
    def all_file_ids(self):
527
526
        return iter(result)
528
527
 
529
528
    def iter_references(self):
 
529
        if not self._repo_supports_tree_reference:
 
530
            # When the repo doesn't support references, we will have nothing to
 
531
            # return
 
532
            return
530
533
        for key, tree_details in self.current_dirstate()._iter_entries():
531
534
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
532
535
                # not relevant to the working tree
534
537
            if not key[1]:
535
538
                # the root is not a reference.
536
539
                continue
537
 
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
540
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
538
541
            try:
539
 
                if self._kind(path) == 'tree-reference':
540
 
                    yield path, key[2]
 
542
                if self._kind(relpath) == 'tree-reference':
 
543
                    yield relpath, key[2]
541
544
            except errors.NoSuchFile:
542
545
                # path is missing on disk.
543
546
                continue
544
547
 
 
548
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
 
549
        """See MutableTree._observed_sha1."""
 
550
        state = self.current_dirstate()
 
551
        entry = self._get_entry(file_id=file_id, path=path)
 
552
        state._observed_sha1(entry, sha1, statvalue)
 
553
 
545
554
    def kind(self, file_id):
546
555
        """Return the kind of a file.
547
556
 
559
568
    def _kind(self, relpath):
560
569
        abspath = self.abspath(relpath)
561
570
        kind = file_kind(abspath)
562
 
        if (self._repo_supports_tree_reference and
563
 
            kind == 'directory' and
564
 
            self._directory_is_tree_reference(relpath)):
565
 
            kind = 'tree-reference'
 
571
        if (self._repo_supports_tree_reference and kind == 'directory'):
 
572
            entry = self._get_entry(path=relpath)
 
573
            if entry[1] is not None:
 
574
                if entry[1][0][0] == 't':
 
575
                    kind = 'tree-reference'
566
576
        return kind
567
577
 
568
578
    @needs_read_lock
575
585
            return _mod_revision.NULL_REVISION
576
586
 
577
587
    def lock_read(self):
578
 
        """See Branch.lock_read, and WorkingTree.unlock."""
 
588
        """See Branch.lock_read, and WorkingTree.unlock.
 
589
 
 
590
        :return: A bzrlib.lock.LogicalLockResult.
 
591
        """
579
592
        self.branch.lock_read()
580
593
        try:
581
594
            self._control_files.lock_read()
594
607
        except:
595
608
            self.branch.unlock()
596
609
            raise
 
610
        return LogicalLockResult(self.unlock)
597
611
 
598
612
    def _lock_self_write(self):
599
613
        """This should be called after the branch is locked."""
614
628
        except:
615
629
            self.branch.unlock()
616
630
            raise
 
631
        return LogicalLockResult(self.unlock)
617
632
 
618
633
    def lock_tree_write(self):
619
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
634
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
 
635
 
 
636
        :return: A bzrlib.lock.LogicalLockResult.
 
637
        """
620
638
        self.branch.lock_read()
621
 
        self._lock_self_write()
 
639
        return self._lock_self_write()
622
640
 
623
641
    def lock_write(self):
624
 
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
642
        """See MutableTree.lock_write, and WorkingTree.unlock.
 
643
 
 
644
        :return: A bzrlib.lock.LogicalLockResult.
 
645
        """
625
646
        self.branch.lock_write()
626
 
        self._lock_self_write()
 
647
        return self._lock_self_write()
627
648
 
628
649
    @needs_tree_write_lock
629
650
    def move(self, from_paths, to_dir, after=False):
699
720
            from_entry = self._get_entry(path=from_rel)
700
721
            if from_entry == (None, None):
701
722
                raise errors.BzrMoveFailedError(from_rel,to_dir,
702
 
                    errors.NotVersionedError(path=str(from_rel)))
 
723
                    errors.NotVersionedError(path=from_rel))
703
724
 
704
725
            from_id = from_entry[0][2]
705
726
            to_rel = pathjoin(to_dir, from_tail)
846
867
                rollback_rename()
847
868
                raise
848
869
            result.append((from_rel, to_rel))
849
 
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
870
            state._mark_modified()
850
871
            self._make_dirty(reset_inventory=False)
851
872
 
852
873
        return result
880
901
        for tree in trees:
881
902
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
882
903
                parents):
883
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
904
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
905
                    trees, require_versioned)
884
906
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
885
907
        # -- make all paths utf8 --
886
908
        paths_utf8 = set()
944
966
                    all_versioned = False
945
967
                    break
946
968
            if not all_versioned:
947
 
                raise errors.PathsNotVersionedError(paths)
 
969
                raise errors.PathsNotVersionedError(
 
970
                    [p.decode('utf-8') for p in paths])
948
971
        # -- remove redundancy in supplied paths to prevent over-scanning --
949
972
        search_paths = osutils.minimum_path_selection(paths)
950
 
        # sketch: 
 
973
        # sketch:
951
974
        # for all search_indexs in each path at or under each element of
952
975
        # search_paths, if the detail is relocated: add the id, and add the
953
976
        # relocated path as one to search if its not searched already. If the
999
1022
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1000
1023
            for dir_name in split_paths:
1001
1024
                if dir_name not in found_dir_names:
1002
 
                    raise errors.PathsNotVersionedError(paths)
 
1025
                    raise errors.PathsNotVersionedError(
 
1026
                        [p.decode('utf-8') for p in paths])
1003
1027
 
1004
1028
        for dir_name_id, trees_info in found.iteritems():
1005
1029
            for index in search_indexes:
1009
1033
 
1010
1034
    def read_working_inventory(self):
1011
1035
        """Read the working inventory.
1012
 
        
 
1036
 
1013
1037
        This is a meaningless operation for dirstate, but we obey it anyhow.
1014
1038
        """
1015
1039
        return self.inventory
1033
1057
    def set_last_revision(self, new_revision):
1034
1058
        """Change the last revision in the working tree."""
1035
1059
        parents = self.get_parent_ids()
1036
 
        if new_revision in (NULL_REVISION, None):
 
1060
        if new_revision in (_mod_revision.NULL_REVISION, None):
1037
1061
            if len(parents) >= 2:
1038
1062
                raise AssertionError(
1039
1063
                    "setting the last parent to none with a pending merge is "
1046
1070
    @needs_tree_write_lock
1047
1071
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1048
1072
        """Set the parent ids to revision_ids.
1049
 
        
 
1073
 
1050
1074
        See also set_parent_trees. This api will try to retrieve the tree data
1051
1075
        for each element of revision_ids from the trees repository. If you have
1052
1076
        tree data already available, it is more efficient to use
1084
1108
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1085
1109
        real_trees = []
1086
1110
        ghosts = []
 
1111
 
 
1112
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1113
        graph = self.branch.repository.get_graph()
 
1114
        heads = graph.heads(parent_ids)
 
1115
        accepted_revisions = set()
 
1116
 
1087
1117
        # convert absent trees to the null tree, which we convert back to
1088
1118
        # missing on access.
1089
1119
        for rev_id, tree in parents_list:
 
1120
            if len(accepted_revisions) > 0:
 
1121
                # we always accept the first tree
 
1122
                if rev_id in accepted_revisions or rev_id not in heads:
 
1123
                    # We have already included either this tree, or its
 
1124
                    # descendent, so we skip it.
 
1125
                    continue
1090
1126
            _mod_revision.check_not_reserved_id(rev_id)
1091
1127
            if tree is not None:
1092
1128
                real_trees.append((rev_id, tree))
1093
1129
            else:
1094
1130
                real_trees.append((rev_id,
1095
 
                    self.branch.repository.revision_tree(None)))
 
1131
                    self.branch.repository.revision_tree(
 
1132
                        _mod_revision.NULL_REVISION)))
1096
1133
                ghosts.append(rev_id)
1097
 
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1134
            accepted_revisions.add(rev_id)
 
1135
        updated = False
 
1136
        if (len(real_trees) == 1
 
1137
            and not ghosts
 
1138
            and self.branch.repository._format.fast_deltas
 
1139
            and isinstance(real_trees[0][1],
 
1140
                revisiontree.InventoryRevisionTree)
 
1141
            and self.get_parent_ids()):
 
1142
            rev_id, rev_tree = real_trees[0]
 
1143
            basis_id = self.get_parent_ids()[0]
 
1144
            # There are times when basis_tree won't be in
 
1145
            # self.branch.repository, (switch, for example)
 
1146
            try:
 
1147
                basis_tree = self.branch.repository.revision_tree(basis_id)
 
1148
            except errors.NoSuchRevision:
 
1149
                # Fall back to the set_parent_trees(), since we can't use
 
1150
                # _make_delta if we can't get the RevisionTree
 
1151
                pass
 
1152
            else:
 
1153
                delta = rev_tree.inventory._make_delta(basis_tree.inventory)
 
1154
                dirstate.update_basis_by_delta(delta, rev_id)
 
1155
                updated = True
 
1156
        if not updated:
 
1157
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1098
1158
        self._make_dirty(reset_inventory=False)
1099
1159
 
1100
1160
    def _set_root_id(self, file_id):
1120
1180
 
1121
1181
    def unlock(self):
1122
1182
        """Unlock in format 4 trees needs to write the entire dirstate."""
1123
 
        # do non-implementation specific cleanup
1124
 
        self._cleanup()
1125
 
 
1126
1183
        if self._control_files._lock_count == 1:
 
1184
            # do non-implementation specific cleanup
 
1185
            self._cleanup()
 
1186
 
1127
1187
            # eventually we should do signature checking during read locks for
1128
1188
            # dirstate updates.
1129
1189
            if self._control_files._lock_mode == 'w':
1192
1252
                # just forget the whole block.
1193
1253
                entry_index = 0
1194
1254
                while entry_index < len(block[1]):
1195
 
                    # Mark this file id as having been removed
1196
1255
                    entry = block[1][entry_index]
1197
 
                    ids_to_unversion.discard(entry[0][2])
1198
 
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
1199
 
                                               # entries
1200
 
                        or not state._make_absent(entry)):
 
1256
                    if entry[1][0][0] in 'ar':
 
1257
                        # don't remove absent or renamed entries
1201
1258
                        entry_index += 1
 
1259
                    else:
 
1260
                        # Mark this file id as having been removed
 
1261
                        ids_to_unversion.discard(entry[0][2])
 
1262
                        if not state._make_absent(entry):
 
1263
                            # The block has not shrunk.
 
1264
                            entry_index += 1
1202
1265
                # go to the next block. (At the moment we dont delete empty
1203
1266
                # dirblocks)
1204
1267
                block_index += 1
1225
1288
        # have to change the legacy inventory too.
1226
1289
        if self._inventory is not None:
1227
1290
            for file_id in file_ids:
1228
 
                self._inventory.remove_recursive_id(file_id)
 
1291
                if self._inventory.has_id(file_id):
 
1292
                    self._inventory.remove_recursive_id(file_id)
1229
1293
 
1230
1294
    @needs_tree_write_lock
1231
1295
    def rename_one(self, from_rel, to_rel, after=False):
1232
1296
        """See WorkingTree.rename_one"""
1233
1297
        self.flush()
1234
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1298
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
1235
1299
 
1236
1300
    @needs_tree_write_lock
1237
1301
    def apply_inventory_delta(self, changes):
1256
1320
        if self._dirty:
1257
1321
            raise AssertionError("attempting to write an inventory when the "
1258
1322
                "dirstate is dirty will lose pending changes")
1259
 
        self.current_dirstate().set_state_from_inventory(inv)
1260
 
        self._make_dirty(reset_inventory=False)
1261
 
        if self._inventory is not None:
 
1323
        had_inventory = self._inventory is not None
 
1324
        # Setting self._inventory = None forces the dirstate to regenerate the
 
1325
        # working inventory. We do this because self.inventory may be inv, or
 
1326
        # may have been modified, and either case would prevent a clean delta
 
1327
        # being created.
 
1328
        self._inventory = None
 
1329
        # generate a delta,
 
1330
        delta = inv._make_delta(self.inventory)
 
1331
        # and apply it.
 
1332
        self.apply_inventory_delta(delta)
 
1333
        if had_inventory:
1262
1334
            self._inventory = inv
1263
1335
        self.flush()
1264
1336
 
1265
 
 
1266
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1267
 
    """The first consolidated dirstate working tree format.
1268
 
 
1269
 
    This format:
1270
 
        - exists within a metadir controlling .bzr
1271
 
        - includes an explicit version marker for the workingtree control
1272
 
          files, separate from the BzrDir format
1273
 
        - modifies the hash cache format
1274
 
        - is new in bzr 0.15
1275
 
        - uses a LockDir to guard access to it.
1276
 
    """
1277
 
 
1278
 
    upgrade_recommended = False
1279
 
 
1280
 
    def get_format_string(self):
1281
 
        """See WorkingTreeFormat.get_format_string()."""
1282
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1283
 
 
1284
 
    def get_format_description(self):
1285
 
        """See WorkingTreeFormat.get_format_description()."""
1286
 
        return "Working tree format 4"
 
1337
    @needs_tree_write_lock
 
1338
    def reset_state(self, revision_ids=None):
 
1339
        """Reset the state of the working tree.
 
1340
 
 
1341
        This does a hard-reset to a last-known-good state. This is a way to
 
1342
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
 
1343
        """
 
1344
        if revision_ids is None:
 
1345
            revision_ids = self.get_parent_ids()
 
1346
        if not revision_ids:
 
1347
            base_tree = self.branch.repository.revision_tree(
 
1348
                _mod_revision.NULL_REVISION)
 
1349
            trees = []
 
1350
        else:
 
1351
            trees = zip(revision_ids,
 
1352
                        self.branch.repository.revision_trees(revision_ids))
 
1353
            base_tree = trees[0][1]
 
1354
        state = self.current_dirstate()
 
1355
        # We don't support ghosts yet
 
1356
        state.set_state_from_scratch(base_tree.inventory, trees, [])
 
1357
 
 
1358
 
 
1359
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
 
1360
 
 
1361
    def __init__(self, tree):
 
1362
        self.tree = tree
 
1363
 
 
1364
    def sha1(self, abspath):
 
1365
        """See dirstate.SHA1Provider.sha1()."""
 
1366
        filters = self.tree._content_filter_stack(
 
1367
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1368
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
 
1369
 
 
1370
    def stat_and_sha1(self, abspath):
 
1371
        """See dirstate.SHA1Provider.stat_and_sha1()."""
 
1372
        filters = self.tree._content_filter_stack(
 
1373
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1374
        file_obj = file(abspath, 'rb', 65000)
 
1375
        try:
 
1376
            statvalue = os.fstat(file_obj.fileno())
 
1377
            if filters:
 
1378
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
 
1379
            sha1 = osutils.size_sha_file(file_obj)[1]
 
1380
        finally:
 
1381
            file_obj.close()
 
1382
        return statvalue, sha1
 
1383
 
 
1384
 
 
1385
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
 
1386
    """Dirstate working tree that supports content filtering.
 
1387
 
 
1388
    The dirstate holds the hash and size of the canonical form of the file, 
 
1389
    and most methods must return that.
 
1390
    """
 
1391
 
 
1392
    def _file_content_summary(self, path, stat_result):
 
1393
        # This is to support the somewhat obsolete path_content_summary method
 
1394
        # with content filtering: see
 
1395
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
 
1396
        #
 
1397
        # If the dirstate cache is up to date and knows the hash and size,
 
1398
        # return that.
 
1399
        # Otherwise if there are no content filters, return the on-disk size
 
1400
        # and leave the hash blank.
 
1401
        # Otherwise, read and filter the on-disk file and use its size and
 
1402
        # hash.
 
1403
        #
 
1404
        # The dirstate doesn't store the size of the canonical form so we
 
1405
        # can't trust it for content-filtered trees.  We just return None.
 
1406
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
 
1407
        executable = self._is_executable_from_path_and_stat(path, stat_result)
 
1408
        return ('file', None, executable, dirstate_sha1)
 
1409
 
 
1410
 
 
1411
class WorkingTree4(DirStateWorkingTree):
 
1412
    """This is the Format 4 working tree.
 
1413
 
 
1414
    This differs from WorkingTree by:
 
1415
     - Having a consolidated internal dirstate, stored in a
 
1416
       randomly-accessible sorted file on disk.
 
1417
     - Not having a regular inventory attribute.  One can be synthesized
 
1418
       on demand but this is expensive and should be avoided.
 
1419
 
 
1420
    This is new in bzr 0.15.
 
1421
    """
 
1422
 
 
1423
 
 
1424
class WorkingTree5(ContentFilteringDirStateWorkingTree):
 
1425
    """This is the Format 5 working tree.
 
1426
 
 
1427
    This differs from WorkingTree4 by:
 
1428
     - Supporting content filtering.
 
1429
 
 
1430
    This is new in bzr 1.11.
 
1431
    """
 
1432
 
 
1433
 
 
1434
class WorkingTree6(ContentFilteringDirStateWorkingTree):
 
1435
    """This is the Format 6 working tree.
 
1436
 
 
1437
    This differs from WorkingTree5 by:
 
1438
     - Supporting a current view that may mask the set of files in a tree
 
1439
       impacted by most user operations.
 
1440
 
 
1441
    This is new in bzr 1.14.
 
1442
    """
 
1443
 
 
1444
    def _make_views(self):
 
1445
        return views.PathBasedViews(self)
 
1446
 
 
1447
 
 
1448
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
 
1449
 
 
1450
    missing_parent_conflicts = True
 
1451
 
 
1452
    supports_versioned_directories = True
 
1453
 
 
1454
    _lock_class = LockDir
 
1455
    _lock_file_name = 'lock'
 
1456
 
 
1457
    def _open_control_files(self, a_bzrdir):
 
1458
        transport = a_bzrdir.get_workingtree_transport(None)
 
1459
        return LockableFiles(transport, self._lock_file_name,
 
1460
                             self._lock_class)
1287
1461
 
1288
1462
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1289
1463
                   accelerator_tree=None, hardlink=False):
1290
1464
        """See WorkingTreeFormat.initialize().
1291
1465
 
1292
1466
        :param revision_id: allows creating a working tree at a different
1293
 
        revision than the branch is at.
 
1467
            revision than the branch is at.
1294
1468
        :param accelerator_tree: A tree which can be used for retrieving file
1295
1469
            contents more quickly than the revision tree, i.e. a workingtree.
1296
1470
            The revision tree will be used for cases where accelerator_tree's
1307
1481
        control_files = self._open_control_files(a_bzrdir)
1308
1482
        control_files.create_lock()
1309
1483
        control_files.lock_write()
1310
 
        control_files.put_utf8('format', self.get_format_string())
 
1484
        transport.put_bytes('format', self.as_string(),
 
1485
            mode=a_bzrdir._get_file_mode())
1311
1486
        if from_branch is not None:
1312
1487
            branch = from_branch
1313
1488
        else:
1319
1494
        state = dirstate.DirState.initialize(local_path)
1320
1495
        state.unlock()
1321
1496
        del state
1322
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1497
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1323
1498
                         branch,
1324
1499
                         _format=self,
1325
1500
                         _bzrdir=a_bzrdir,
1327
1502
        wt._new_tree()
1328
1503
        wt.lock_tree_write()
1329
1504
        try:
1330
 
            if revision_id in (None, NULL_REVISION):
 
1505
            self._init_custom_control_files(wt)
 
1506
            if revision_id in (None, _mod_revision.NULL_REVISION):
1331
1507
                if branch.repository.supports_rich_root():
1332
1508
                    wt._set_root_id(generate_ids.gen_root_id())
1333
1509
                else:
1344
1520
                    pass
1345
1521
            if basis is None:
1346
1522
                basis = branch.repository.revision_tree(revision_id)
1347
 
            if revision_id == NULL_REVISION:
 
1523
            if revision_id == _mod_revision.NULL_REVISION:
1348
1524
                parents_list = []
1349
1525
            else:
1350
1526
                parents_list = [(revision_id, basis)]
1358
1534
                if basis_root_id is not None:
1359
1535
                    wt._set_root_id(basis_root_id)
1360
1536
                    wt.flush()
 
1537
                if wt.supports_content_filtering():
 
1538
                    # The original tree may not have the same content filters
 
1539
                    # applied so we can't safely build the inventory delta from
 
1540
                    # the source tree.
 
1541
                    delta_from_tree = False
 
1542
                else:
 
1543
                    delta_from_tree = True
 
1544
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1545
                # because wt4.apply_inventory_delta does not mutate the input
 
1546
                # inventory entries.
1361
1547
                transform.build_tree(basis, wt, accelerator_tree,
1362
 
                                     hardlink=hardlink)
 
1548
                                     hardlink=hardlink,
 
1549
                                     delta_from_tree=delta_from_tree)
 
1550
                for hook in MutableTree.hooks['post_build_tree']:
 
1551
                    hook(wt)
1363
1552
            finally:
1364
1553
                basis.unlock()
1365
1554
        finally:
1367
1556
            wt.unlock()
1368
1557
        return wt
1369
1558
 
 
1559
    def _init_custom_control_files(self, wt):
 
1560
        """Subclasses with custom control files should override this method.
 
1561
 
 
1562
        The working tree and control files are locked for writing when this
 
1563
        method is called.
 
1564
 
 
1565
        :param wt: the WorkingTree object
 
1566
        """
 
1567
 
 
1568
    def open(self, a_bzrdir, _found=False):
 
1569
        """Return the WorkingTree object for a_bzrdir
 
1570
 
 
1571
        _found is a private parameter, do not use it. It is used to indicate
 
1572
               if format probing has already been done.
 
1573
        """
 
1574
        if not _found:
 
1575
            # we are being called directly and must probe.
 
1576
            raise NotImplementedError
 
1577
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1578
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1579
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
 
1580
        return wt
 
1581
 
1370
1582
    def _open(self, a_bzrdir, control_files):
1371
1583
        """Open the tree itself.
1372
1584
 
1373
1585
        :param a_bzrdir: the dir for the tree.
1374
1586
        :param control_files: the control files for the tree.
1375
1587
        """
1376
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1588
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1377
1589
                           branch=a_bzrdir.open_branch(),
1378
1590
                           _format=self,
1379
1591
                           _bzrdir=a_bzrdir,
1380
1592
                           _control_files=control_files)
1381
1593
 
1382
1594
    def __get_matchingbzrdir(self):
 
1595
        return self._get_matchingbzrdir()
 
1596
 
 
1597
    def _get_matchingbzrdir(self):
 
1598
        """Overrideable method to get a bzrdir for testing."""
1383
1599
        # please test against something that will let us do tree references
1384
1600
        return bzrdir.format_registry.make_bzrdir(
1385
1601
            'dirstate-with-subtree')
1387
1603
    _matchingbzrdir = property(__get_matchingbzrdir)
1388
1604
 
1389
1605
 
1390
 
class DirStateRevisionTree(Tree):
1391
 
    """A revision tree pulling the inventory from a dirstate."""
 
1606
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
 
1607
    """The first consolidated dirstate working tree format.
 
1608
 
 
1609
    This format:
 
1610
        - exists within a metadir controlling .bzr
 
1611
        - includes an explicit version marker for the workingtree control
 
1612
          files, separate from the ControlDir format
 
1613
        - modifies the hash cache format
 
1614
        - is new in bzr 0.15
 
1615
        - uses a LockDir to guard access to it.
 
1616
    """
 
1617
 
 
1618
    upgrade_recommended = False
 
1619
 
 
1620
    _tree_class = WorkingTree4
 
1621
 
 
1622
    @classmethod
 
1623
    def get_format_string(cls):
 
1624
        """See WorkingTreeFormat.get_format_string()."""
 
1625
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1626
 
 
1627
    def get_format_description(self):
 
1628
        """See WorkingTreeFormat.get_format_description()."""
 
1629
        return "Working tree format 4"
 
1630
 
 
1631
 
 
1632
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
 
1633
    """WorkingTree format supporting content filtering.
 
1634
    """
 
1635
 
 
1636
    upgrade_recommended = False
 
1637
 
 
1638
    _tree_class = WorkingTree5
 
1639
 
 
1640
    @classmethod
 
1641
    def get_format_string(cls):
 
1642
        """See WorkingTreeFormat.get_format_string()."""
 
1643
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
 
1644
 
 
1645
    def get_format_description(self):
 
1646
        """See WorkingTreeFormat.get_format_description()."""
 
1647
        return "Working tree format 5"
 
1648
 
 
1649
    def supports_content_filtering(self):
 
1650
        return True
 
1651
 
 
1652
 
 
1653
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
 
1654
    """WorkingTree format supporting views.
 
1655
    """
 
1656
 
 
1657
    upgrade_recommended = False
 
1658
 
 
1659
    _tree_class = WorkingTree6
 
1660
 
 
1661
    @classmethod
 
1662
    def get_format_string(cls):
 
1663
        """See WorkingTreeFormat.get_format_string()."""
 
1664
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
 
1665
 
 
1666
    def get_format_description(self):
 
1667
        """See WorkingTreeFormat.get_format_description()."""
 
1668
        return "Working tree format 6"
 
1669
 
 
1670
    def _init_custom_control_files(self, wt):
 
1671
        """Subclasses with custom control files should override this method."""
 
1672
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
 
1673
 
 
1674
    def supports_content_filtering(self):
 
1675
        return True
 
1676
 
 
1677
    def supports_views(self):
 
1678
        return True
 
1679
 
 
1680
 
 
1681
class DirStateRevisionTree(InventoryTree):
 
1682
    """A revision tree pulling the inventory from a dirstate.
 
1683
    
 
1684
    Note that this is one of the historical (ie revision) trees cached in the
 
1685
    dirstate for easy access, not the workingtree.
 
1686
    """
1392
1687
 
1393
1688
    def __init__(self, dirstate, revision_id, repository):
1394
1689
        self._dirstate = dirstate
1397
1692
        self._inventory = None
1398
1693
        self._locked = 0
1399
1694
        self._dirstate_locked = False
 
1695
        self._repo_supports_tree_reference = getattr(
 
1696
            repository._format, "supports_tree_reference",
 
1697
            False)
1400
1698
 
1401
1699
    def __repr__(self):
1402
1700
        return "<%s of %s in %s>" % \
1405
1703
    def annotate_iter(self, file_id,
1406
1704
                      default_revision=_mod_revision.CURRENT_REVISION):
1407
1705
        """See Tree.annotate_iter"""
1408
 
        w = self._get_weave(file_id)
1409
 
        return w.annotate(self.inventory[file_id].revision)
 
1706
        text_key = (file_id, self.get_file_revision(file_id))
 
1707
        annotations = self._repository.texts.annotate(text_key)
 
1708
        return [(key[-1], line) for (key, line) in annotations]
1410
1709
 
1411
 
    def _get_ancestors(self, default_revision):
1412
 
        return set(self._repository.get_ancestry(self._revision_id,
1413
 
                                                 topo_sorted=False))
1414
1710
    def _comparison_data(self, entry, path):
1415
1711
        """See Tree._comparison_data."""
1416
1712
        if entry is None:
1441
1737
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1442
1738
        return path_utf8.decode('utf8')
1443
1739
 
 
1740
    def iter_references(self):
 
1741
        if not self._repo_supports_tree_reference:
 
1742
            # When the repo doesn't support references, we will have nothing to
 
1743
            # return
 
1744
            return iter([])
 
1745
        # Otherwise, fall back to the default implementation
 
1746
        return super(DirStateRevisionTree, self).iter_references()
 
1747
 
1444
1748
    def _get_parent_index(self):
1445
1749
        """Return the index in the dirstate referenced by this tree."""
1446
1750
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1451
1755
        If either file_id or path is supplied, it is used as the key to lookup.
1452
1756
        If both are supplied, the fastest lookup is used, and an error is
1453
1757
        raised if they do not both point at the same row.
1454
 
        
 
1758
 
1455
1759
        :param file_id: An optional unicode file_id to be looked up.
1456
1760
        :param path: An optional unicode path to be looked up.
1457
1761
        :return: The dirstate row tuple for path/file_id, or (None, None)
1524
1828
                elif kind == 'directory':
1525
1829
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1526
1830
                elif kind == 'symlink':
1527
 
                    inv_entry.executable = False
1528
 
                    inv_entry.text_size = None
1529
1831
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1530
1832
                elif kind == 'tree-reference':
1531
1833
                    inv_entry.reference_revision = fingerprint or None
1551
1853
        # Make sure the file exists
1552
1854
        entry = self._get_entry(file_id, path=path)
1553
1855
        if entry == (None, None): # do we raise?
1554
 
            return None
 
1856
            raise errors.NoSuchId(self, file_id)
1555
1857
        parent_index = self._get_parent_index()
1556
1858
        last_changed_revision = entry[1][parent_index][4]
1557
 
        return self._repository.get_revision(last_changed_revision).timestamp
 
1859
        try:
 
1860
            rev = self._repository.get_revision(last_changed_revision)
 
1861
        except errors.NoSuchRevision:
 
1862
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
 
1863
        return rev.timestamp
1558
1864
 
1559
1865
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1560
1866
        entry = self._get_entry(file_id=file_id, path=path)
1564
1870
            return parent_details[1]
1565
1871
        return None
1566
1872
 
1567
 
    def _get_weave(self, file_id):
1568
 
        return self._repository.weave_store.get_weave(file_id,
1569
 
                self._repository.get_transaction())
 
1873
    @needs_read_lock
 
1874
    def get_file_revision(self, file_id):
 
1875
        return self.inventory[file_id].revision
1570
1876
 
1571
1877
    def get_file(self, file_id, path=None):
1572
1878
        return StringIO(self.get_file_text(file_id))
1573
1879
 
1574
 
    def get_file_lines(self, file_id):
1575
 
        entry = self._get_entry(file_id=file_id)[1]
1576
 
        if entry is None:
1577
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1578
 
        return self._get_weave(file_id).get_lines(entry[1][4])
1579
 
 
1580
1880
    def get_file_size(self, file_id):
1581
1881
        """See Tree.get_file_size"""
1582
1882
        return self.inventory[file_id].text_size
1583
1883
 
1584
 
    def get_file_text(self, file_id):
1585
 
        return ''.join(self.get_file_lines(file_id))
 
1884
    def get_file_text(self, file_id, path=None):
 
1885
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
 
1886
        return ''.join(content)
1586
1887
 
1587
1888
    def get_reference_revision(self, file_id, path=None):
1588
1889
        return self.inventory[file_id].reference_revision
1601
1902
                                       identifier))
1602
1903
        return self._repository.iter_files_bytes(repo_desired_files)
1603
1904
 
1604
 
    def get_symlink_target(self, file_id):
 
1905
    def get_symlink_target(self, file_id, path=None):
1605
1906
        entry = self._get_entry(file_id=file_id)
1606
1907
        parent_index = self._get_parent_index()
1607
1908
        if entry[1][parent_index][0] != 'l':
1608
1909
            return None
1609
1910
        else:
1610
 
            # At present, none of the tree implementations supports non-ascii
1611
 
            # symlink targets. So we will just assume that the dirstate path is
1612
 
            # correct.
1613
 
            return entry[1][parent_index][1]
 
1911
            target = entry[1][parent_index][1]
 
1912
            target = target.decode('utf8')
 
1913
            return target
1614
1914
 
1615
1915
    def get_revision_id(self):
1616
1916
        """Return the revision id for this tree."""
1637
1937
        entry = self._get_entry(file_id=file_id)[1]
1638
1938
        if entry is None:
1639
1939
            raise errors.NoSuchId(tree=self, file_id=file_id)
1640
 
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1940
        parent_index = self._get_parent_index()
 
1941
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1641
1942
 
1642
1943
    def stored_kind(self, file_id):
1643
1944
        """See Tree.stored_kind"""
1660
1961
    def is_executable(self, file_id, path=None):
1661
1962
        ie = self.inventory[file_id]
1662
1963
        if ie.kind != "file":
1663
 
            return None
 
1964
            return False
1664
1965
        return ie.executable
1665
1966
 
1666
 
    def list_files(self, include_root=False):
 
1967
    def is_locked(self):
 
1968
        return self._locked
 
1969
 
 
1970
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1667
1971
        # We use a standard implementation, because DirStateRevisionTree is
1668
1972
        # dealing with one of the parents of the current state
1669
1973
        inv = self._get_inventory()
1670
 
        entries = inv.iter_entries()
1671
 
        if self.inventory.root is not None and not include_root:
 
1974
        if from_dir is None:
 
1975
            from_dir_id = None
 
1976
        else:
 
1977
            from_dir_id = inv.path2id(from_dir)
 
1978
            if from_dir_id is None:
 
1979
                # Directory not versioned
 
1980
                return
 
1981
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
 
1982
        if inv.root is not None and not include_root and from_dir is None:
1672
1983
            entries.next()
1673
1984
        for path, entry in entries:
1674
1985
            yield path, 'V', entry.kind, entry.file_id, entry
1675
1986
 
1676
1987
    def lock_read(self):
1677
 
        """Lock the tree for a set of operations."""
 
1988
        """Lock the tree for a set of operations.
 
1989
 
 
1990
        :return: A bzrlib.lock.LogicalLockResult.
 
1991
        """
1678
1992
        if not self._locked:
1679
1993
            self._repository.lock_read()
1680
1994
            if self._dirstate._lock_token is None:
1681
1995
                self._dirstate.lock_read()
1682
1996
                self._dirstate_locked = True
1683
1997
        self._locked += 1
 
1998
        return LogicalLockResult(self.unlock)
1684
1999
 
1685
2000
    def _must_be_locked(self):
1686
2001
        if not self._locked:
1707
2022
                self._dirstate_locked = False
1708
2023
            self._repository.unlock()
1709
2024
 
 
2025
    @needs_read_lock
 
2026
    def supports_tree_reference(self):
 
2027
        return self._repo_supports_tree_reference
 
2028
 
1710
2029
    def walkdirs(self, prefix=""):
1711
2030
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1712
 
        # implementation based on an inventory.  
 
2031
        # implementation based on an inventory.
1713
2032
        # This should be cleaned up to use the much faster Dirstate code
1714
2033
        # So for now, we just build up the parent inventory, and extract
1715
2034
        # it the same way RevisionTree does.
1744
2063
 
1745
2064
class InterDirStateTree(InterTree):
1746
2065
    """Fast path optimiser for changes_from with dirstate trees.
1747
 
    
1748
 
    This is used only when both trees are in the dirstate working file, and 
1749
 
    the source is any parent within the dirstate, and the destination is 
 
2066
 
 
2067
    This is used only when both trees are in the dirstate working file, and
 
2068
    the source is any parent within the dirstate, and the destination is
1750
2069
    the current working tree of the same dirstate.
1751
2070
    """
1752
2071
    # this could be generalized to allow comparisons between any trees in the
1761
2080
    def make_source_parent_tree(source, target):
1762
2081
        """Change the source tree into a parent of the target."""
1763
2082
        revid = source.commit('record tree')
1764
 
        target.branch.repository.fetch(source.branch.repository, revid)
 
2083
        target.branch.fetch(source.branch, revid)
1765
2084
        target.set_parent_ids([revid])
1766
2085
        return target.basis_tree(), target
1767
2086
 
 
2087
    @classmethod
 
2088
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
 
2089
        result = klass.make_source_parent_tree(source, target)
 
2090
        result[1]._iter_changes = dirstate.ProcessEntryPython
 
2091
        return result
 
2092
 
 
2093
    @classmethod
 
2094
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
 
2095
                                                  target):
 
2096
        from bzrlib.tests.test__dirstate_helpers import \
 
2097
            compiled_dirstate_helpers_feature
 
2098
        test_case.requireFeature(compiled_dirstate_helpers_feature)
 
2099
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
 
2100
        result = klass.make_source_parent_tree(source, target)
 
2101
        result[1]._iter_changes = ProcessEntryC
 
2102
        return result
 
2103
 
1768
2104
    _matching_from_tree_format = WorkingTreeFormat4()
1769
2105
    _matching_to_tree_format = WorkingTreeFormat4()
1770
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
2106
 
 
2107
    @classmethod
 
2108
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
 
2109
        # This method shouldn't be called, because we have python and C
 
2110
        # specific flavours.
 
2111
        raise NotImplementedError
1771
2112
 
1772
2113
    def iter_changes(self, include_unchanged=False,
1773
2114
                      specific_files=None, pb=None, extra_trees=[],
1791
2132
            output. An unversioned file is defined as one with (False, False)
1792
2133
            for the versioned pair.
1793
2134
        """
1794
 
        utf8_decode = cache_utf8._utf8_decode
1795
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1796
 
        cmp_by_dirs = dirstate.cmp_by_dirs
1797
 
        # NB: show_status depends on being able to pass in non-versioned files
1798
 
        # and report them as unknown
1799
2135
        # TODO: handle extra trees in the dirstate.
1800
2136
        if (extra_trees or specific_files == []):
1801
2137
            # we can't fast-path these cases (yet)
1802
 
            for f in super(InterDirStateTree, self).iter_changes(
 
2138
            return super(InterDirStateTree, self).iter_changes(
1803
2139
                include_unchanged, specific_files, pb, extra_trees,
1804
 
                require_versioned, want_unversioned=want_unversioned):
1805
 
                yield f
1806
 
            return
 
2140
                require_versioned, want_unversioned=want_unversioned)
1807
2141
        parent_ids = self.target.get_parent_ids()
1808
2142
        if not (self.source._revision_id in parent_ids
1809
 
                or self.source._revision_id == NULL_REVISION):
 
2143
                or self.source._revision_id == _mod_revision.NULL_REVISION):
1810
2144
            raise AssertionError(
1811
2145
                "revision {%s} is not stored in {%s}, but %s "
1812
2146
                "can only be used for trees stored in the dirstate"
1813
2147
                % (self.source._revision_id, self.target, self.iter_changes))
1814
2148
        target_index = 0
1815
 
        if self.source._revision_id == NULL_REVISION:
 
2149
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1816
2150
            source_index = None
1817
2151
            indices = (target_index,)
1818
2152
        else:
1826
2160
        if specific_files:
1827
2161
            specific_files_utf8 = set()
1828
2162
            for path in specific_files:
 
2163
                # Note, if there are many specific files, using cache_utf8
 
2164
                # would be good here.
1829
2165
                specific_files_utf8.add(path.encode('utf8'))
1830
2166
            specific_files = specific_files_utf8
1831
2167
        else:
1832
2168
            specific_files = set([''])
1833
2169
        # -- specific_files is now a utf8 path set --
 
2170
 
1834
2171
        # -- get the state object and prepare it.
1835
2172
        state = self.target.current_dirstate()
1836
2173
        state._read_dirblocks_if_needed()
1837
 
        def _entries_for_path(path):
1838
 
            """Return a list with all the entries that match path for all ids.
1839
 
            """
1840
 
            dirname, basename = os.path.split(path)
1841
 
            key = (dirname, basename, '')
1842
 
            block_index, present = state._find_block_index_from_key(key)
1843
 
            if not present:
1844
 
                # the block which should contain path is absent.
1845
 
                return []
1846
 
            result = []
1847
 
            block = state._dirblocks[block_index][1]
1848
 
            entry_index, _ = state._find_entry_index(key, block)
1849
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1850
 
            while (entry_index < len(block) and
1851
 
                block[entry_index][0][0:2] == key[0:2]):
1852
 
                result.append(block[entry_index])
1853
 
                entry_index += 1
1854
 
            return result
1855
2174
        if require_versioned:
1856
2175
            # -- check all supplied paths are versioned in a search tree. --
1857
 
            all_versioned = True
 
2176
            not_versioned = []
1858
2177
            for path in specific_files:
1859
 
                path_entries = _entries_for_path(path)
 
2178
                path_entries = state._entries_for_path(path)
1860
2179
                if not path_entries:
1861
2180
                    # this specified path is not present at all: error
1862
 
                    all_versioned = False
1863
 
                    break
 
2181
                    not_versioned.append(path.decode('utf-8'))
 
2182
                    continue
1864
2183
                found_versioned = False
1865
2184
                # for each id at this path
1866
2185
                for entry in path_entries:
1873
2192
                if not found_versioned:
1874
2193
                    # none of the indexes was not 'absent' at all ids for this
1875
2194
                    # path.
1876
 
                    all_versioned = False
1877
 
                    break
1878
 
            if not all_versioned:
1879
 
                raise errors.PathsNotVersionedError(specific_files)
 
2195
                    not_versioned.append(path.decode('utf-8'))
 
2196
            if len(not_versioned) > 0:
 
2197
                raise errors.PathsNotVersionedError(not_versioned)
1880
2198
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1881
 
        search_specific_files = set()
1882
 
        for path in specific_files:
1883
 
            other_specific_files = specific_files.difference(set([path]))
1884
 
            if not osutils.is_inside_any(other_specific_files, path):
1885
 
                # this is a top level path, we must check it.
1886
 
                search_specific_files.add(path)
1887
 
        # sketch: 
1888
 
        # compare source_index and target_index at or under each element of search_specific_files.
1889
 
        # follow the following comparison table. Note that we only want to do diff operations when
1890
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1891
 
        # for the target.
1892
 
        # cases:
1893
 
        # 
1894
 
        # Source | Target | disk | action
1895
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1896
 
        #        |        |      | diff check on source-target
1897
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1898
 
        #        |        |      | ???
1899
 
        #   r    |  a     |      | add source to search
1900
 
        #   r    |  a     |  a   | 
1901
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1902
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1903
 
        #   a    | fdlt   |      | add new id
1904
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1905
 
        #   a    |  a     |      | not present in either tree, skip
1906
 
        #   a    |  a     |  a   | not present in any tree, skip
1907
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1908
 
        #        |        |      | may not be selected by the users list of paths.
1909
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1910
 
        #        |        |      | may not be selected by the users list of paths.
1911
 
        #  fdlt  | fdlt   |      | content in both: diff them
1912
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1913
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1914
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1915
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1916
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1917
 
        #        |        |      | this id at the other path.
1918
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1919
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1920
 
        #        |        |      | this id at the other path.
1921
 
 
1922
 
        # for all search_indexs in each path at or under each element of
1923
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1924
 
        # relocated path as one to search if its not searched already. If the
1925
 
        # detail is not relocated, add the id.
1926
 
        searched_specific_files = set()
1927
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1928
 
        # Using a list so that we can access the values and change them in
1929
 
        # nested scope. Each one is [path, file_id, entry]
1930
 
        last_source_parent = [None, None]
1931
 
        last_target_parent = [None, None]
 
2199
        search_specific_files = osutils.minimum_path_selection(specific_files)
1932
2200
 
1933
2201
        use_filesystem_for_exec = (sys.platform != 'win32')
1934
 
 
1935
 
        # Just a sentry, so that _process_entry can say that this
1936
 
        # record is handled, but isn't interesting to process (unchanged)
1937
 
        uninteresting = object()
1938
 
 
1939
 
 
1940
 
        old_dirname_to_file_id = {}
1941
 
        new_dirname_to_file_id = {}
1942
 
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1943
 
        #       keeping a cache of directories that we have seen.
1944
 
 
1945
 
        def _process_entry(entry, path_info):
1946
 
            """Compare an entry and real disk to generate delta information.
1947
 
 
1948
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1949
 
                the path of entry. If None, then the path is considered absent.
1950
 
                (Perhaps we should pass in a concrete entry for this ?)
1951
 
                Basename is returned as a utf8 string because we expect this
1952
 
                tuple will be ignored, and don't want to take the time to
1953
 
                decode.
1954
 
            :return: None if these don't match
1955
 
                     A tuple of information about the change, or
1956
 
                     the object 'uninteresting' if these match, but are
1957
 
                     basically identical.
1958
 
            """
1959
 
            if source_index is None:
1960
 
                source_details = NULL_PARENT_DETAILS
1961
 
            else:
1962
 
                source_details = entry[1][source_index]
1963
 
            target_details = entry[1][target_index]
1964
 
            target_minikind = target_details[0]
1965
 
            if path_info is not None and target_minikind in 'fdlt':
1966
 
                if not (target_index == 0):
1967
 
                    raise AssertionError()
1968
 
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
1969
 
                                                  stat_value=path_info[3])
1970
 
                # The entry may have been modified by update_entry
1971
 
                target_details = entry[1][target_index]
1972
 
                target_minikind = target_details[0]
1973
 
            else:
1974
 
                link_or_sha1 = None
1975
 
            file_id = entry[0][2]
1976
 
            source_minikind = source_details[0]
1977
 
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
1978
 
                # claimed content in both: diff
1979
 
                #   r    | fdlt   |      | add source to search, add id path move and perform
1980
 
                #        |        |      | diff check on source-target
1981
 
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
1982
 
                #        |        |      | ???
1983
 
                if source_minikind in 'r':
1984
 
                    # add the source to the search path to find any children it
1985
 
                    # has.  TODO ? : only add if it is a container ?
1986
 
                    if not osutils.is_inside_any(searched_specific_files,
1987
 
                                                 source_details[1]):
1988
 
                        search_specific_files.add(source_details[1])
1989
 
                    # generate the old path; this is needed for stating later
1990
 
                    # as well.
1991
 
                    old_path = source_details[1]
1992
 
                    old_dirname, old_basename = os.path.split(old_path)
1993
 
                    path = pathjoin(entry[0][0], entry[0][1])
1994
 
                    old_entry = state._get_entry(source_index,
1995
 
                                                 path_utf8=old_path)
1996
 
                    # update the source details variable to be the real
1997
 
                    # location.
1998
 
                    if old_entry == (None, None):
1999
 
                        raise errors.CorruptDirstate(state._filename,
2000
 
                            "entry '%s/%s' is considered renamed from %r"
2001
 
                            " but source does not exist\n"
2002
 
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
2003
 
                    source_details = old_entry[1][source_index]
2004
 
                    source_minikind = source_details[0]
2005
 
                else:
2006
 
                    old_dirname = entry[0][0]
2007
 
                    old_basename = entry[0][1]
2008
 
                    old_path = path = None
2009
 
                if path_info is None:
2010
 
                    # the file is missing on disk, show as removed.
2011
 
                    content_change = True
2012
 
                    target_kind = None
2013
 
                    target_exec = False
2014
 
                else:
2015
 
                    # source and target are both versioned and disk file is present.
2016
 
                    target_kind = path_info[2]
2017
 
                    if target_kind == 'directory':
2018
 
                        if path is None:
2019
 
                            old_path = path = pathjoin(old_dirname, old_basename)
2020
 
                        new_dirname_to_file_id[path] = file_id
2021
 
                        if source_minikind != 'd':
2022
 
                            content_change = True
2023
 
                        else:
2024
 
                            # directories have no fingerprint
2025
 
                            content_change = False
2026
 
                        target_exec = False
2027
 
                    elif target_kind == 'file':
2028
 
                        if source_minikind != 'f':
2029
 
                            content_change = True
2030
 
                        else:
2031
 
                            # We could check the size, but we already have the
2032
 
                            # sha1 hash.
2033
 
                            content_change = (link_or_sha1 != source_details[1])
2034
 
                        # Target details is updated at update_entry time
2035
 
                        if use_filesystem_for_exec:
2036
 
                            # We don't need S_ISREG here, because we are sure
2037
 
                            # we are dealing with a file.
2038
 
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
2039
 
                        else:
2040
 
                            target_exec = target_details[3]
2041
 
                    elif target_kind == 'symlink':
2042
 
                        if source_minikind != 'l':
2043
 
                            content_change = True
2044
 
                        else:
2045
 
                            content_change = (link_or_sha1 != source_details[1])
2046
 
                        target_exec = False
2047
 
                    elif target_kind == 'tree-reference':
2048
 
                        if source_minikind != 't':
2049
 
                            content_change = True
2050
 
                        else:
2051
 
                            content_change = False
2052
 
                        target_exec = False
2053
 
                    else:
2054
 
                        raise Exception, "unknown kind %s" % path_info[2]
2055
 
                if source_minikind == 'd':
2056
 
                    if path is None:
2057
 
                        old_path = path = pathjoin(old_dirname, old_basename)
2058
 
                    old_dirname_to_file_id[old_path] = file_id
2059
 
                # parent id is the entry for the path in the target tree
2060
 
                if old_dirname == last_source_parent[0]:
2061
 
                    source_parent_id = last_source_parent[1]
2062
 
                else:
2063
 
                    try:
2064
 
                        source_parent_id = old_dirname_to_file_id[old_dirname]
2065
 
                    except KeyError:
2066
 
                        source_parent_entry = state._get_entry(source_index,
2067
 
                                                               path_utf8=old_dirname)
2068
 
                        source_parent_id = source_parent_entry[0][2]
2069
 
                    if source_parent_id == entry[0][2]:
2070
 
                        # This is the root, so the parent is None
2071
 
                        source_parent_id = None
2072
 
                    else:
2073
 
                        last_source_parent[0] = old_dirname
2074
 
                        last_source_parent[1] = source_parent_id
2075
 
                new_dirname = entry[0][0]
2076
 
                if new_dirname == last_target_parent[0]:
2077
 
                    target_parent_id = last_target_parent[1]
2078
 
                else:
2079
 
                    try:
2080
 
                        target_parent_id = new_dirname_to_file_id[new_dirname]
2081
 
                    except KeyError:
2082
 
                        # TODO: We don't always need to do the lookup, because the
2083
 
                        #       parent entry will be the same as the source entry.
2084
 
                        target_parent_entry = state._get_entry(target_index,
2085
 
                                                               path_utf8=new_dirname)
2086
 
                        if target_parent_entry == (None, None):
2087
 
                            raise AssertionError(
2088
 
                                "Could not find target parent in wt: %s\nparent of: %s"
2089
 
                                % (new_dirname, entry))
2090
 
                        target_parent_id = target_parent_entry[0][2]
2091
 
                    if target_parent_id == entry[0][2]:
2092
 
                        # This is the root, so the parent is None
2093
 
                        target_parent_id = None
2094
 
                    else:
2095
 
                        last_target_parent[0] = new_dirname
2096
 
                        last_target_parent[1] = target_parent_id
2097
 
 
2098
 
                source_exec = source_details[3]
2099
 
                if (include_unchanged
2100
 
                    or content_change
2101
 
                    or source_parent_id != target_parent_id
2102
 
                    or old_basename != entry[0][1]
2103
 
                    or source_exec != target_exec
2104
 
                    ):
2105
 
                    if old_path is None:
2106
 
                        old_path = path = pathjoin(old_dirname, old_basename)
2107
 
                        old_path_u = utf8_decode(old_path)[0]
2108
 
                        path_u = old_path_u
2109
 
                    else:
2110
 
                        old_path_u = utf8_decode(old_path)[0]
2111
 
                        if old_path == path:
2112
 
                            path_u = old_path_u
2113
 
                        else:
2114
 
                            path_u = utf8_decode(path)[0]
2115
 
                    source_kind = _minikind_to_kind[source_minikind]
2116
 
                    return (entry[0][2],
2117
 
                           (old_path_u, path_u),
2118
 
                           content_change,
2119
 
                           (True, True),
2120
 
                           (source_parent_id, target_parent_id),
2121
 
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
2122
 
                           (source_kind, target_kind),
2123
 
                           (source_exec, target_exec))
2124
 
                else:
2125
 
                    return uninteresting
2126
 
            elif source_minikind in 'a' and target_minikind in 'fdlt':
2127
 
                # looks like a new file
2128
 
                if path_info is not None:
2129
 
                    path = pathjoin(entry[0][0], entry[0][1])
2130
 
                    # parent id is the entry for the path in the target tree
2131
 
                    # TODO: these are the same for an entire directory: cache em.
2132
 
                    parent_id = state._get_entry(target_index,
2133
 
                                                 path_utf8=entry[0][0])[0][2]
2134
 
                    if parent_id == entry[0][2]:
2135
 
                        parent_id = None
2136
 
                    if use_filesystem_for_exec:
2137
 
                        # We need S_ISREG here, because we aren't sure if this
2138
 
                        # is a file or not.
2139
 
                        target_exec = bool(
2140
 
                            stat.S_ISREG(path_info[3].st_mode)
2141
 
                            and stat.S_IEXEC & path_info[3].st_mode)
2142
 
                    else:
2143
 
                        target_exec = target_details[3]
2144
 
                    return (entry[0][2],
2145
 
                           (None, utf8_decode(path)[0]),
2146
 
                           True,
2147
 
                           (False, True),
2148
 
                           (None, parent_id),
2149
 
                           (None, utf8_decode(entry[0][1])[0]),
2150
 
                           (None, path_info[2]),
2151
 
                           (None, target_exec))
2152
 
                else:
2153
 
                    # but its not on disk: we deliberately treat this as just
2154
 
                    # never-present. (Why ?! - RBC 20070224)
2155
 
                    pass
2156
 
            elif source_minikind in 'fdlt' and target_minikind in 'a':
2157
 
                # unversioned, possibly, or possibly not deleted: we dont care.
2158
 
                # if its still on disk, *and* theres no other entry at this
2159
 
                # path [we dont know this in this routine at the moment -
2160
 
                # perhaps we should change this - then it would be an unknown.
2161
 
                old_path = pathjoin(entry[0][0], entry[0][1])
2162
 
                # parent id is the entry for the path in the target tree
2163
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
2164
 
                if parent_id == entry[0][2]:
2165
 
                    parent_id = None
2166
 
                return (entry[0][2],
2167
 
                       (utf8_decode(old_path)[0], None),
2168
 
                       True,
2169
 
                       (True, False),
2170
 
                       (parent_id, None),
2171
 
                       (utf8_decode(entry[0][1])[0], None),
2172
 
                       (_minikind_to_kind[source_minikind], None),
2173
 
                       (source_details[3], None))
2174
 
            elif source_minikind in 'fdlt' and target_minikind in 'r':
2175
 
                # a rename; could be a true rename, or a rename inherited from
2176
 
                # a renamed parent. TODO: handle this efficiently. Its not
2177
 
                # common case to rename dirs though, so a correct but slow
2178
 
                # implementation will do.
2179
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
2180
 
                    search_specific_files.add(target_details[1])
2181
 
            elif source_minikind in 'ra' and target_minikind in 'ra':
2182
 
                # neither of the selected trees contain this file,
2183
 
                # so skip over it. This is not currently directly tested, but
2184
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
2185
 
                pass
2186
 
            else:
2187
 
                raise AssertionError("don't know how to compare "
2188
 
                    "source_minikind=%r, target_minikind=%r"
2189
 
                    % (source_minikind, target_minikind))
2190
 
                ## import pdb;pdb.set_trace()
2191
 
            return None
2192
 
 
2193
 
        while search_specific_files:
2194
 
            # TODO: the pending list should be lexically sorted?  the
2195
 
            # interface doesn't require it.
2196
 
            current_root = search_specific_files.pop()
2197
 
            current_root_unicode = current_root.decode('utf8')
2198
 
            searched_specific_files.add(current_root)
2199
 
            # process the entries for this containing directory: the rest will be
2200
 
            # found by their parents recursively.
2201
 
            root_entries = _entries_for_path(current_root)
2202
 
            root_abspath = self.target.abspath(current_root_unicode)
2203
 
            try:
2204
 
                root_stat = os.lstat(root_abspath)
2205
 
            except OSError, e:
2206
 
                if e.errno == errno.ENOENT:
2207
 
                    # the path does not exist: let _process_entry know that.
2208
 
                    root_dir_info = None
2209
 
                else:
2210
 
                    # some other random error: hand it up.
2211
 
                    raise
2212
 
            else:
2213
 
                root_dir_info = ('', current_root,
2214
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
2215
 
                    root_abspath)
2216
 
                if root_dir_info[2] == 'directory':
2217
 
                    if self.target._directory_is_tree_reference(
2218
 
                        current_root.decode('utf8')):
2219
 
                        root_dir_info = root_dir_info[:2] + \
2220
 
                            ('tree-reference',) + root_dir_info[3:]
2221
 
 
2222
 
            if not root_entries and not root_dir_info:
2223
 
                # this specified path is not present at all, skip it.
2224
 
                continue
2225
 
            path_handled = False
2226
 
            for entry in root_entries:
2227
 
                result = _process_entry(entry, root_dir_info)
2228
 
                if result is not None:
2229
 
                    path_handled = True
2230
 
                    if result is not uninteresting:
2231
 
                        yield result
2232
 
            if want_unversioned and not path_handled and root_dir_info:
2233
 
                new_executable = bool(
2234
 
                    stat.S_ISREG(root_dir_info[3].st_mode)
2235
 
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
2236
 
                yield (None,
2237
 
                       (None, current_root_unicode),
2238
 
                       True,
2239
 
                       (False, False),
2240
 
                       (None, None),
2241
 
                       (None, splitpath(current_root_unicode)[-1]),
2242
 
                       (None, root_dir_info[2]),
2243
 
                       (None, new_executable)
2244
 
                      )
2245
 
            initial_key = (current_root, '', '')
2246
 
            block_index, _ = state._find_block_index_from_key(initial_key)
2247
 
            if block_index == 0:
2248
 
                # we have processed the total root already, but because the
2249
 
                # initial key matched it we should skip it here.
2250
 
                block_index +=1
2251
 
            if root_dir_info and root_dir_info[2] == 'tree-reference':
2252
 
                current_dir_info = None
2253
 
            else:
2254
 
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
2255
 
                try:
2256
 
                    current_dir_info = dir_iterator.next()
2257
 
                except OSError, e:
2258
 
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
2259
 
                    # python 2.5 has e.errno == EINVAL,
2260
 
                    #            and e.winerror == ERROR_DIRECTORY
2261
 
                    e_winerror = getattr(e, 'winerror', None)
2262
 
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
2263
 
                    # there may be directories in the inventory even though
2264
 
                    # this path is not a file on disk: so mark it as end of
2265
 
                    # iterator
2266
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
2267
 
                        current_dir_info = None
2268
 
                    elif (sys.platform == 'win32'
2269
 
                          and (e.errno in win_errors
2270
 
                               or e_winerror in win_errors)):
2271
 
                        current_dir_info = None
2272
 
                    else:
2273
 
                        raise
2274
 
                else:
2275
 
                    if current_dir_info[0][0] == '':
2276
 
                        # remove .bzr from iteration
2277
 
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
2278
 
                        if current_dir_info[1][bzr_index][0] != '.bzr':
2279
 
                            raise AssertionError()
2280
 
                        del current_dir_info[1][bzr_index]
2281
 
            # walk until both the directory listing and the versioned metadata
2282
 
            # are exhausted. 
2283
 
            if (block_index < len(state._dirblocks) and
2284
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2285
 
                current_block = state._dirblocks[block_index]
2286
 
            else:
2287
 
                current_block = None
2288
 
            while (current_dir_info is not None or
2289
 
                   current_block is not None):
2290
 
                if (current_dir_info and current_block
2291
 
                    and current_dir_info[0][0] != current_block[0]):
2292
 
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
2293
 
                        # filesystem data refers to paths not covered by the dirblock.
2294
 
                        # this has two possibilities:
2295
 
                        # A) it is versioned but empty, so there is no block for it
2296
 
                        # B) it is not versioned.
2297
 
 
2298
 
                        # if (A) then we need to recurse into it to check for
2299
 
                        # new unknown files or directories.
2300
 
                        # if (B) then we should ignore it, because we don't
2301
 
                        # recurse into unknown directories.
2302
 
                        path_index = 0
2303
 
                        while path_index < len(current_dir_info[1]):
2304
 
                                current_path_info = current_dir_info[1][path_index]
2305
 
                                if want_unversioned:
2306
 
                                    if current_path_info[2] == 'directory':
2307
 
                                        if self.target._directory_is_tree_reference(
2308
 
                                            current_path_info[0].decode('utf8')):
2309
 
                                            current_path_info = current_path_info[:2] + \
2310
 
                                                ('tree-reference',) + current_path_info[3:]
2311
 
                                    new_executable = bool(
2312
 
                                        stat.S_ISREG(current_path_info[3].st_mode)
2313
 
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
2314
 
                                    yield (None,
2315
 
                                        (None, utf8_decode(current_path_info[0])[0]),
2316
 
                                        True,
2317
 
                                        (False, False),
2318
 
                                        (None, None),
2319
 
                                        (None, utf8_decode(current_path_info[1])[0]),
2320
 
                                        (None, current_path_info[2]),
2321
 
                                        (None, new_executable))
2322
 
                                # dont descend into this unversioned path if it is
2323
 
                                # a dir
2324
 
                                if current_path_info[2] in ('directory',
2325
 
                                                            'tree-reference'):
2326
 
                                    del current_dir_info[1][path_index]
2327
 
                                    path_index -= 1
2328
 
                                path_index += 1
2329
 
 
2330
 
                        # This dir info has been handled, go to the next
2331
 
                        try:
2332
 
                            current_dir_info = dir_iterator.next()
2333
 
                        except StopIteration:
2334
 
                            current_dir_info = None
2335
 
                    else:
2336
 
                        # We have a dirblock entry for this location, but there
2337
 
                        # is no filesystem path for this. This is most likely
2338
 
                        # because a directory was removed from the disk.
2339
 
                        # We don't have to report the missing directory,
2340
 
                        # because that should have already been handled, but we
2341
 
                        # need to handle all of the files that are contained
2342
 
                        # within.
2343
 
                        for current_entry in current_block[1]:
2344
 
                            # entry referring to file not present on disk.
2345
 
                            # advance the entry only, after processing.
2346
 
                            result = _process_entry(current_entry, None)
2347
 
                            if result is not None:
2348
 
                                if result is not uninteresting:
2349
 
                                    yield result
2350
 
                        block_index +=1
2351
 
                        if (block_index < len(state._dirblocks) and
2352
 
                            osutils.is_inside(current_root,
2353
 
                                              state._dirblocks[block_index][0])):
2354
 
                            current_block = state._dirblocks[block_index]
2355
 
                        else:
2356
 
                            current_block = None
2357
 
                    continue
2358
 
                entry_index = 0
2359
 
                if current_block and entry_index < len(current_block[1]):
2360
 
                    current_entry = current_block[1][entry_index]
2361
 
                else:
2362
 
                    current_entry = None
2363
 
                advance_entry = True
2364
 
                path_index = 0
2365
 
                if current_dir_info and path_index < len(current_dir_info[1]):
2366
 
                    current_path_info = current_dir_info[1][path_index]
2367
 
                    if current_path_info[2] == 'directory':
2368
 
                        if self.target._directory_is_tree_reference(
2369
 
                            current_path_info[0].decode('utf8')):
2370
 
                            current_path_info = current_path_info[:2] + \
2371
 
                                ('tree-reference',) + current_path_info[3:]
2372
 
                else:
2373
 
                    current_path_info = None
2374
 
                advance_path = True
2375
 
                path_handled = False
2376
 
                while (current_entry is not None or
2377
 
                    current_path_info is not None):
2378
 
                    if current_entry is None:
2379
 
                        # the check for path_handled when the path is adnvaced
2380
 
                        # will yield this path if needed.
2381
 
                        pass
2382
 
                    elif current_path_info is None:
2383
 
                        # no path is fine: the per entry code will handle it.
2384
 
                        result = _process_entry(current_entry, current_path_info)
2385
 
                        if result is not None:
2386
 
                            if result is not uninteresting:
2387
 
                                yield result
2388
 
                    elif (current_entry[0][1] != current_path_info[1]
2389
 
                          or current_entry[1][target_index][0] in 'ar'):
2390
 
                        # The current path on disk doesn't match the dirblock
2391
 
                        # record. Either the dirblock is marked as absent, or
2392
 
                        # the file on disk is not present at all in the
2393
 
                        # dirblock. Either way, report about the dirblock
2394
 
                        # entry, and let other code handle the filesystem one.
2395
 
 
2396
 
                        # Compare the basename for these files to determine
2397
 
                        # which comes first
2398
 
                        if current_path_info[1] < current_entry[0][1]:
2399
 
                            # extra file on disk: pass for now, but only
2400
 
                            # increment the path, not the entry
2401
 
                            advance_entry = False
2402
 
                        else:
2403
 
                            # entry referring to file not present on disk.
2404
 
                            # advance the entry only, after processing.
2405
 
                            result = _process_entry(current_entry, None)
2406
 
                            if result is not None:
2407
 
                                if result is not uninteresting:
2408
 
                                    yield result
2409
 
                            advance_path = False
2410
 
                    else:
2411
 
                        result = _process_entry(current_entry, current_path_info)
2412
 
                        if result is not None:
2413
 
                            path_handled = True
2414
 
                            if result is not uninteresting:
2415
 
                                yield result
2416
 
                    if advance_entry and current_entry is not None:
2417
 
                        entry_index += 1
2418
 
                        if entry_index < len(current_block[1]):
2419
 
                            current_entry = current_block[1][entry_index]
2420
 
                        else:
2421
 
                            current_entry = None
2422
 
                    else:
2423
 
                        advance_entry = True # reset the advance flaga
2424
 
                    if advance_path and current_path_info is not None:
2425
 
                        if not path_handled:
2426
 
                            # unversioned in all regards
2427
 
                            if want_unversioned:
2428
 
                                new_executable = bool(
2429
 
                                    stat.S_ISREG(current_path_info[3].st_mode)
2430
 
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
2431
 
                                yield (None,
2432
 
                                    (None, utf8_decode(current_path_info[0])[0]),
2433
 
                                    True,
2434
 
                                    (False, False),
2435
 
                                    (None, None),
2436
 
                                    (None, utf8_decode(current_path_info[1])[0]),
2437
 
                                    (None, current_path_info[2]),
2438
 
                                    (None, new_executable))
2439
 
                            # dont descend into this unversioned path if it is
2440
 
                            # a dir
2441
 
                            if current_path_info[2] in ('directory'):
2442
 
                                del current_dir_info[1][path_index]
2443
 
                                path_index -= 1
2444
 
                        # dont descend the disk iterator into any tree 
2445
 
                        # paths.
2446
 
                        if current_path_info[2] == 'tree-reference':
2447
 
                            del current_dir_info[1][path_index]
2448
 
                            path_index -= 1
2449
 
                        path_index += 1
2450
 
                        if path_index < len(current_dir_info[1]):
2451
 
                            current_path_info = current_dir_info[1][path_index]
2452
 
                            if current_path_info[2] == 'directory':
2453
 
                                if self.target._directory_is_tree_reference(
2454
 
                                    current_path_info[0].decode('utf8')):
2455
 
                                    current_path_info = current_path_info[:2] + \
2456
 
                                        ('tree-reference',) + current_path_info[3:]
2457
 
                        else:
2458
 
                            current_path_info = None
2459
 
                        path_handled = False
2460
 
                    else:
2461
 
                        advance_path = True # reset the advance flagg.
2462
 
                if current_block is not None:
2463
 
                    block_index += 1
2464
 
                    if (block_index < len(state._dirblocks) and
2465
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2466
 
                        current_block = state._dirblocks[block_index]
2467
 
                    else:
2468
 
                        current_block = None
2469
 
                if current_dir_info is not None:
2470
 
                    try:
2471
 
                        current_dir_info = dir_iterator.next()
2472
 
                    except StopIteration:
2473
 
                        current_dir_info = None
2474
 
 
 
2202
        iter_changes = self.target._iter_changes(include_unchanged,
 
2203
            use_filesystem_for_exec, search_specific_files, state,
 
2204
            source_index, target_index, want_unversioned, self.target)
 
2205
        return iter_changes.iter_changes()
2475
2206
 
2476
2207
    @staticmethod
2477
2208
    def is_compatible(source, target):
2478
2209
        # the target must be a dirstate working tree
2479
 
        if not isinstance(target, WorkingTree4):
 
2210
        if not isinstance(target, DirStateWorkingTree):
2480
2211
            return False
2481
 
        # the source must be a revtreee or dirstate rev tree.
 
2212
        # the source must be a revtree or dirstate rev tree.
2482
2213
        if not isinstance(source,
2483
2214
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2484
2215
            return False
2485
2216
        # the source revid must be in the target dirstate
2486
 
        if not (source._revision_id == NULL_REVISION or
 
2217
        if not (source._revision_id == _mod_revision.NULL_REVISION or
2487
2218
            source._revision_id in target.get_parent_ids()):
2488
 
            # TODO: what about ghosts? it may well need to 
 
2219
            # TODO: what about ghosts? it may well need to
2489
2220
            # check for them explicitly.
2490
2221
            return False
2491
2222
        return True
2501
2232
 
2502
2233
    def convert(self, tree):
2503
2234
        # lock the control files not the tree, so that we dont get tree
2504
 
        # on-unlock behaviours, and so that noone else diddles with the 
 
2235
        # on-unlock behaviours, and so that noone else diddles with the
2505
2236
        # tree during upgrade.
2506
2237
        tree._control_files.lock_write()
2507
2238
        try:
2533
2264
 
2534
2265
    def update_format(self, tree):
2535
2266
        """Change the format marker."""
2536
 
        tree._control_files.put_utf8('format',
2537
 
            self.target_format.get_format_string())
 
2267
        tree._transport.put_bytes('format',
 
2268
            self.target_format.as_string(),
 
2269
            mode=tree.bzrdir._get_file_mode())
 
2270
 
 
2271
 
 
2272
class Converter4to5(object):
 
2273
    """Perform an in-place upgrade of format 4 to format 5 trees."""
 
2274
 
 
2275
    def __init__(self):
 
2276
        self.target_format = WorkingTreeFormat5()
 
2277
 
 
2278
    def convert(self, tree):
 
2279
        # lock the control files not the tree, so that we don't get tree
 
2280
        # on-unlock behaviours, and so that no-one else diddles with the
 
2281
        # tree during upgrade.
 
2282
        tree._control_files.lock_write()
 
2283
        try:
 
2284
            self.update_format(tree)
 
2285
        finally:
 
2286
            tree._control_files.unlock()
 
2287
 
 
2288
    def update_format(self, tree):
 
2289
        """Change the format marker."""
 
2290
        tree._transport.put_bytes('format',
 
2291
            self.target_format.as_string(),
 
2292
            mode=tree.bzrdir._get_file_mode())
 
2293
 
 
2294
 
 
2295
class Converter4or5to6(object):
 
2296
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
 
2297
 
 
2298
    def __init__(self):
 
2299
        self.target_format = WorkingTreeFormat6()
 
2300
 
 
2301
    def convert(self, tree):
 
2302
        # lock the control files not the tree, so that we don't get tree
 
2303
        # on-unlock behaviours, and so that no-one else diddles with the
 
2304
        # tree during upgrade.
 
2305
        tree._control_files.lock_write()
 
2306
        try:
 
2307
            self.init_custom_control_files(tree)
 
2308
            self.update_format(tree)
 
2309
        finally:
 
2310
            tree._control_files.unlock()
 
2311
 
 
2312
    def init_custom_control_files(self, tree):
 
2313
        """Initialize custom control files."""
 
2314
        tree._transport.put_bytes('views', '',
 
2315
            mode=tree.bzrdir._get_file_mode())
 
2316
 
 
2317
    def update_format(self, tree):
 
2318
        """Change the format marker."""
 
2319
        tree._transport.put_bytes('format',
 
2320
            self.target_format.as_string(),
 
2321
            mode=tree.bzrdir._get_file_mode())