~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2010-04-08 04:34:03 UTC
  • mfrom: (5138 +trunk)
  • mto: This revision was merged to the branch mainline in revision 5139.
  • Revision ID: robertc@robertcollins.net-20100408043403-56z0d07vdqrx7f3t
Update bugfix for 528114 to trunk.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
1
# Copyright (C) 2007-2010 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
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
31
 
from bisect import bisect_left
32
 
import collections
33
 
from copy import deepcopy
34
31
import errno
35
 
import itertools
36
 
import operator
37
32
import stat
38
 
from time import time
39
 
import warnings
40
33
 
41
34
import bzrlib
42
35
from bzrlib import (
43
36
    bzrdir,
44
37
    cache_utf8,
45
 
    conflicts as _mod_conflicts,
46
38
    debug,
47
 
    delta,
48
39
    dirstate,
49
40
    errors,
50
41
    generate_ids,
51
 
    globbing,
52
 
    ignores,
53
 
    merge,
54
42
    osutils,
55
43
    revision as _mod_revision,
56
44
    revisiontree,
57
 
    textui,
58
45
    trace,
59
46
    transform,
60
 
    urlutils,
61
 
    xml5,
62
 
    xml6,
 
47
    views,
63
48
    )
64
49
import bzrlib.branch
65
 
from bzrlib.transport import get_transport
66
50
import bzrlib.ui
67
51
""")
68
52
 
69
 
from bzrlib import symbol_versioning
70
53
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
73
 
from bzrlib.lockdir import LockDir
 
54
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
 
55
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
74
56
import bzrlib.mutabletree
75
57
from bzrlib.mutabletree import needs_tree_write_lock
76
58
from bzrlib.osutils import (
77
59
    file_kind,
78
60
    isdir,
79
 
    normpath,
80
61
    pathjoin,
81
 
    rand_chars,
82
62
    realpath,
83
63
    safe_unicode,
84
 
    splitpath,
85
64
    )
86
 
from bzrlib.trace import mutter, note
 
65
from bzrlib.trace import mutter
87
66
from bzrlib.transport.local import LocalTransport
88
67
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
68
from bzrlib.tree import Tree
98
69
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
99
70
 
100
71
 
101
 
# This is the Windows equivalent of ENOTDIR
102
 
# It is defined in pywin32.winerror, but we don't want a strong dependency for
103
 
# just an error code.
104
 
ERROR_PATH_NOT_FOUND = 3
105
 
ERROR_DIRECTORY = 267
106
 
 
107
 
 
108
 
class WorkingTree4(WorkingTree3):
109
 
    """This is the Format 4 working tree.
110
 
 
111
 
    This differs from WorkingTree3 by:
112
 
     - Having a consolidated internal dirstate, stored in a
113
 
       randomly-accessible sorted file on disk.
114
 
     - Not having a regular inventory attribute.  One can be synthesized 
115
 
       on demand but this is expensive and should be avoided.
116
 
 
117
 
    This is new in bzr 0.15.
118
 
    """
119
 
 
 
72
class DirStateWorkingTree(WorkingTree3):
120
73
    def __init__(self, basedir,
121
74
                 branch,
122
75
                 _control_files=None,
131
84
        """
132
85
        self._format = _format
133
86
        self.bzrdir = _bzrdir
134
 
        assert isinstance(basedir, basestring), \
135
 
            "base directory %r is not a string" % basedir
136
87
        basedir = safe_unicode(basedir)
137
88
        mutter("opening working tree %r", basedir)
138
89
        self._branch = branch
139
 
        assert isinstance(self.branch, bzrlib.branch.Branch), \
140
 
            "branch %r is not a Branch" % self.branch
141
90
        self.basedir = realpath(basedir)
142
91
        # if branch is at our basedir and is a format 6 or less
143
92
        # assume all other formats have their own control files.
144
 
        assert isinstance(_control_files, LockableFiles), \
145
 
            "_control_files must be a LockableFiles, not %r" % _control_files
146
93
        self._control_files = _control_files
 
94
        self._transport = self._control_files._transport
147
95
        self._dirty = None
148
96
        #-------------
149
97
        # during a read or write lock these objects are set, and are
153
101
        #-------------
154
102
        self._setup_directory_is_tree_reference()
155
103
        self._detect_case_handling()
 
104
        self._rules_searcher = None
 
105
        self.views = self._make_views()
 
106
        #--- allow tests to select the dirstate iter_changes implementation
 
107
        self._iter_changes = dirstate._process_entry
156
108
 
157
109
    @needs_tree_write_lock
158
110
    def _add(self, files, ids, kinds):
186
138
    @needs_tree_write_lock
187
139
    def add_reference(self, sub_tree):
188
140
        # use standard implementation, which calls back to self._add
189
 
        # 
 
141
        #
190
142
        # So we don't store the reference_revision in the working dirstate,
191
 
        # it's just recorded at the moment of commit. 
 
143
        # it's just recorded at the moment of commit.
192
144
        self._add_reference(sub_tree)
193
145
 
194
146
    def break_lock(self):
233
185
            WorkingTree3._comparison_data(self, entry, path)
234
186
        # it looks like a plain directory, but it's really a reference -- see
235
187
        # also kind()
236
 
        if (self._repo_supports_tree_reference and
237
 
            kind == 'directory' and
238
 
            self._directory_is_tree_reference(path)):
 
188
        if (self._repo_supports_tree_reference and kind == 'directory'
 
189
            and entry is not None and entry.kind == 'tree-reference'):
239
190
            kind = 'tree-reference'
240
191
        return kind, executable, stat_value
241
192
 
267
218
            return self._dirstate
268
219
        local_path = self.bzrdir.get_workingtree_transport(None
269
220
            ).local_abspath('dirstate')
270
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
221
        self._dirstate = dirstate.DirState.on_file(local_path,
 
222
            self._sha1_provider())
271
223
        return self._dirstate
272
224
 
 
225
    def _sha1_provider(self):
 
226
        """A function that returns a SHA1Provider suitable for this tree.
 
227
 
 
228
        :return: None if content filtering is not supported by this tree.
 
229
          Otherwise, a SHA1Provider is returned that sha's the canonical
 
230
          form of files, i.e. after read filters are applied.
 
231
        """
 
232
        if self.supports_content_filtering():
 
233
            return ContentFilterAwareSHA1Provider(self)
 
234
        else:
 
235
            return None
 
236
 
273
237
    def filter_unversioned_files(self, paths):
274
238
        """Filter out paths that are versioned.
275
239
 
307
271
 
308
272
    def _generate_inventory(self):
309
273
        """Create and set self.inventory from the dirstate object.
310
 
        
 
274
 
311
275
        This is relatively expensive: we have to walk the entire dirstate.
312
276
        Ideally we would not, and can deprecate this function.
313
277
        """
317
281
        state._read_dirblocks_if_needed()
318
282
        root_key, current_entry = self._get_entry(path='')
319
283
        current_id = root_key[2]
320
 
        assert current_entry[0][0] == 'd' # directory
 
284
        if not (current_entry[0][0] == 'd'): # directory
 
285
            raise AssertionError(current_entry)
321
286
        inv = Inventory(root_id=current_id)
322
287
        # Turn some things into local variables
323
288
        minikind_to_kind = dirstate.DirState._minikind_to_kind
356
321
                    # add this entry to the parent map.
357
322
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
358
323
                elif kind == 'tree-reference':
359
 
                    assert self._repo_supports_tree_reference, \
360
 
                        "repository of %r " \
361
 
                        "doesn't support tree references " \
362
 
                        "required by entry %r" \
363
 
                        % (self, name)
 
324
                    if not self._repo_supports_tree_reference:
 
325
                        raise errors.UnsupportedOperation(
 
326
                            self._generate_inventory,
 
327
                            self.branch.repository)
364
328
                    inv_entry.reference_revision = link_or_sha1 or None
365
329
                elif kind != 'symlink':
366
330
                    raise AssertionError("unknown kind %r" % kind)
367
331
                # These checks cost us around 40ms on a 55k entry tree
368
 
                assert file_id not in inv_byid, ('file_id %s already in'
369
 
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
370
 
                assert name_unicode not in parent_ie.children
 
332
                if file_id in inv_byid:
 
333
                    raise AssertionError('file_id %s already in'
 
334
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
335
                if name_unicode in parent_ie.children:
 
336
                    raise AssertionError('name %r already in parent'
 
337
                        % (name_unicode,))
371
338
                inv_byid[file_id] = inv_entry
372
339
                parent_ie.children[name_unicode] = inv_entry
373
340
        self._inventory = inv
378
345
        If either file_id or path is supplied, it is used as the key to lookup.
379
346
        If both are supplied, the fastest lookup is used, and an error is
380
347
        raised if they do not both point at the same row.
381
 
        
 
348
 
382
349
        :param file_id: An optional unicode file_id to be looked up.
383
350
        :param path: An optional unicode path to be looked up.
384
351
        :return: The dirstate row tuple for path/file_id, or (None, None)
408
375
                    return None
409
376
                else:
410
377
                    raise
411
 
        link_or_sha1 = state.update_entry(entry, file_abspath,
412
 
                                          stat_value=stat_value)
 
378
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
379
            stat_value=stat_value)
413
380
        if entry[1][0][0] == 'f':
414
 
            return link_or_sha1
 
381
            if link_or_sha1 is None:
 
382
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
 
383
                try:
 
384
                    sha1 = osutils.sha_file(file_obj)
 
385
                finally:
 
386
                    file_obj.close()
 
387
                self._observed_sha1(file_id, path, (sha1, statvalue))
 
388
                return sha1
 
389
            else:
 
390
                return link_or_sha1
415
391
        return None
416
392
 
417
393
    def _get_inventory(self):
431
407
    @needs_read_lock
432
408
    def get_parent_ids(self):
433
409
        """See Tree.get_parent_ids.
434
 
        
 
410
 
435
411
        This implementation requests the ids list from the dirstate file.
436
412
        """
437
413
        return self.current_dirstate().get_parent_ids()
459
435
        return osutils.lexists(pathjoin(
460
436
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
461
437
 
 
438
    def has_or_had_id(self, file_id):
 
439
        state = self.current_dirstate()
 
440
        row, parents = self._get_entry(file_id=file_id)
 
441
        return row is not None
 
442
 
462
443
    @needs_read_lock
463
444
    def id2path(self, file_id):
464
445
        "Convert a file-id to a path."
494
475
 
495
476
            Note: The caller is expected to take a read-lock before calling this.
496
477
            """
 
478
            self._must_be_locked()
497
479
            if not path:
498
480
                path = self.id2path(file_id)
499
481
            mode = os.lstat(self.abspath(path)).st_mode
500
482
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
501
483
 
 
484
    def all_file_ids(self):
 
485
        """See Tree.iter_all_file_ids"""
 
486
        self._must_be_locked()
 
487
        result = set()
 
488
        for key, tree_details in self.current_dirstate()._iter_entries():
 
489
            if tree_details[0][0] in ('a', 'r'): # relocated
 
490
                continue
 
491
            result.add(key[2])
 
492
        return result
 
493
 
502
494
    @needs_read_lock
503
495
    def __iter__(self):
504
496
        """Iterate through file_ids for this tree.
517
509
        return iter(result)
518
510
 
519
511
    def iter_references(self):
 
512
        if not self._repo_supports_tree_reference:
 
513
            # When the repo doesn't support references, we will have nothing to
 
514
            # return
 
515
            return
520
516
        for key, tree_details in self.current_dirstate()._iter_entries():
521
517
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
522
518
                # not relevant to the working tree
524
520
            if not key[1]:
525
521
                # the root is not a reference.
526
522
                continue
527
 
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
523
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
528
524
            try:
529
 
                if self._kind(path) == 'tree-reference':
530
 
                    yield path, key[2]
 
525
                if self._kind(relpath) == 'tree-reference':
 
526
                    yield relpath, key[2]
531
527
            except errors.NoSuchFile:
532
528
                # path is missing on disk.
533
529
                continue
534
530
 
 
531
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
 
532
        """See MutableTree._observed_sha1."""
 
533
        state = self.current_dirstate()
 
534
        entry = self._get_entry(file_id=file_id, path=path)
 
535
        state._observed_sha1(entry, sha1, statvalue)
 
536
 
535
537
    def kind(self, file_id):
536
538
        """Return the kind of a file.
537
539
 
541
543
        Note: The caller is expected to take a read-lock before calling this.
542
544
        """
543
545
        relpath = self.id2path(file_id)
544
 
        assert relpath != None, \
545
 
            "path for id {%s} is None!" % file_id
 
546
        if relpath is None:
 
547
            raise AssertionError(
 
548
                "path for id {%s} is None!" % file_id)
546
549
        return self._kind(relpath)
547
550
 
548
551
    def _kind(self, relpath):
549
552
        abspath = self.abspath(relpath)
550
553
        kind = file_kind(abspath)
551
 
        if (self._repo_supports_tree_reference and
552
 
            kind == 'directory' and
553
 
            self._directory_is_tree_reference(relpath)):
554
 
            kind = 'tree-reference'
 
554
        if (self._repo_supports_tree_reference and kind == 'directory'):
 
555
            entry = self._get_entry(path=relpath)
 
556
            if entry[1] is not None:
 
557
                if entry[1][0][0] == 't':
 
558
                    kind = 'tree-reference'
555
559
        return kind
556
560
 
557
561
    @needs_read_lock
620
624
        result = []
621
625
        if not from_paths:
622
626
            return result
623
 
 
624
627
        state = self.current_dirstate()
625
 
 
626
 
        assert not isinstance(from_paths, basestring)
 
628
        if isinstance(from_paths, basestring):
 
629
            raise ValueError()
627
630
        to_dir_utf8 = to_dir.encode('utf8')
628
631
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
629
632
        id_index = state._get_id_index()
651
654
        if self._inventory is not None:
652
655
            update_inventory = True
653
656
            inv = self.inventory
 
657
            to_dir_id = to_entry[0][2]
654
658
            to_dir_ie = inv[to_dir_id]
655
 
            to_dir_id = to_entry[0][2]
656
659
        else:
657
660
            update_inventory = False
658
661
 
689
692
            from_entry = self._get_entry(path=from_rel)
690
693
            if from_entry == (None, None):
691
694
                raise errors.BzrMoveFailedError(from_rel,to_dir,
692
 
                    errors.NotVersionedError(path=str(from_rel)))
 
695
                    errors.NotVersionedError(path=from_rel))
693
696
 
694
697
            from_id = from_entry[0][2]
695
698
            to_rel = pathjoin(to_dir, from_tail)
784
787
                if minikind == 'd':
785
788
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
786
789
                        """Recursively update all entries in this dirblock."""
787
 
                        assert from_dir != '', "renaming root not supported"
 
790
                        if from_dir == '':
 
791
                            raise AssertionError("renaming root not supported")
788
792
                        from_key = (from_dir, '')
789
793
                        from_block_idx, present = \
790
794
                            state._find_block_index_from_key(from_key)
803
807
 
804
808
                        # Grab a copy since move_one may update the list.
805
809
                        for entry in from_block[1][:]:
806
 
                            assert entry[0][0] == from_dir
 
810
                            if not (entry[0][0] == from_dir):
 
811
                                raise AssertionError()
807
812
                            cur_details = entry[1][0]
808
813
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
809
814
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
868
873
        for tree in trees:
869
874
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
870
875
                parents):
871
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
876
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
877
                    trees, require_versioned)
872
878
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
873
879
        # -- make all paths utf8 --
874
880
        paths_utf8 = set()
935
941
                raise errors.PathsNotVersionedError(paths)
936
942
        # -- remove redundancy in supplied paths to prevent over-scanning --
937
943
        search_paths = osutils.minimum_path_selection(paths)
938
 
        # sketch: 
 
944
        # sketch:
939
945
        # for all search_indexs in each path at or under each element of
940
946
        # search_paths, if the detail is relocated: add the id, and add the
941
947
        # relocated path as one to search if its not searched already. If the
997
1003
 
998
1004
    def read_working_inventory(self):
999
1005
        """Read the working inventory.
1000
 
        
 
1006
 
1001
1007
        This is a meaningless operation for dirstate, but we obey it anyhow.
1002
1008
        """
1003
1009
        return self.inventory
1021
1027
    def set_last_revision(self, new_revision):
1022
1028
        """Change the last revision in the working tree."""
1023
1029
        parents = self.get_parent_ids()
1024
 
        if new_revision in (NULL_REVISION, None):
1025
 
            assert len(parents) < 2, (
1026
 
                "setting the last parent to none with a pending merge is "
1027
 
                "unsupported.")
 
1030
        if new_revision in (_mod_revision.NULL_REVISION, None):
 
1031
            if len(parents) >= 2:
 
1032
                raise AssertionError(
 
1033
                    "setting the last parent to none with a pending merge is "
 
1034
                    "unsupported.")
1028
1035
            self.set_parent_ids([])
1029
1036
        else:
1030
1037
            self.set_parent_ids([new_revision] + parents[1:],
1033
1040
    @needs_tree_write_lock
1034
1041
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1035
1042
        """Set the parent ids to revision_ids.
1036
 
        
 
1043
 
1037
1044
        See also set_parent_trees. This api will try to retrieve the tree data
1038
1045
        for each element of revision_ids from the trees repository. If you have
1039
1046
        tree data already available, it is more efficient to use
1071
1078
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1072
1079
        real_trees = []
1073
1080
        ghosts = []
 
1081
 
 
1082
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1083
        graph = self.branch.repository.get_graph()
 
1084
        heads = graph.heads(parent_ids)
 
1085
        accepted_revisions = set()
 
1086
 
1074
1087
        # convert absent trees to the null tree, which we convert back to
1075
1088
        # missing on access.
1076
1089
        for rev_id, tree in parents_list:
 
1090
            if len(accepted_revisions) > 0:
 
1091
                # we always accept the first tree
 
1092
                if rev_id in accepted_revisions or rev_id not in heads:
 
1093
                    # We have already included either this tree, or its
 
1094
                    # descendent, so we skip it.
 
1095
                    continue
1077
1096
            _mod_revision.check_not_reserved_id(rev_id)
1078
1097
            if tree is not None:
1079
1098
                real_trees.append((rev_id, tree))
1080
1099
            else:
1081
1100
                real_trees.append((rev_id,
1082
 
                    self.branch.repository.revision_tree(None)))
 
1101
                    self.branch.repository.revision_tree(
 
1102
                        _mod_revision.NULL_REVISION)))
1083
1103
                ghosts.append(rev_id)
 
1104
            accepted_revisions.add(rev_id)
1084
1105
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1085
1106
        self._make_dirty(reset_inventory=False)
1086
1107
 
1179
1200
                # just forget the whole block.
1180
1201
                entry_index = 0
1181
1202
                while entry_index < len(block[1]):
1182
 
                    # Mark this file id as having been removed
1183
1203
                    entry = block[1][entry_index]
1184
 
                    ids_to_unversion.discard(entry[0][2])
1185
 
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
1186
 
                                               # entries
1187
 
                        or not state._make_absent(entry)):
 
1204
                    if entry[1][0][0] in 'ar':
 
1205
                        # don't remove absent or renamed entries
1188
1206
                        entry_index += 1
 
1207
                    else:
 
1208
                        # Mark this file id as having been removed
 
1209
                        ids_to_unversion.discard(entry[0][2])
 
1210
                        if not state._make_absent(entry):
 
1211
                            # The block has not shrunk.
 
1212
                            entry_index += 1
1189
1213
                # go to the next block. (At the moment we dont delete empty
1190
1214
                # dirblocks)
1191
1215
                block_index += 1
1214
1238
            for file_id in file_ids:
1215
1239
                self._inventory.remove_recursive_id(file_id)
1216
1240
 
 
1241
    @needs_tree_write_lock
 
1242
    def rename_one(self, from_rel, to_rel, after=False):
 
1243
        """See WorkingTree.rename_one"""
 
1244
        self.flush()
 
1245
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1246
 
 
1247
    @needs_tree_write_lock
 
1248
    def apply_inventory_delta(self, changes):
 
1249
        """See MutableTree.apply_inventory_delta"""
 
1250
        state = self.current_dirstate()
 
1251
        state.update_by_delta(changes)
 
1252
        self._make_dirty(reset_inventory=True)
 
1253
 
1217
1254
    def update_basis_by_delta(self, new_revid, delta):
1218
1255
        """See MutableTree.update_basis_by_delta."""
1219
 
        assert self.last_revision() != new_revid
 
1256
        if self.last_revision() == new_revid:
 
1257
            raise AssertionError()
1220
1258
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1221
1259
 
1222
1260
    @needs_read_lock
1226
1264
    @needs_tree_write_lock
1227
1265
    def _write_inventory(self, inv):
1228
1266
        """Write inventory as the current inventory."""
1229
 
        assert not self._dirty, ("attempting to write an inventory when the "
1230
 
            "dirstate is dirty will cause data loss")
1231
 
        self.current_dirstate().set_state_from_inventory(inv)
1232
 
        self._make_dirty(reset_inventory=False)
1233
 
        if self._inventory is not None:
 
1267
        if self._dirty:
 
1268
            raise AssertionError("attempting to write an inventory when the "
 
1269
                "dirstate is dirty will lose pending changes")
 
1270
        had_inventory = self._inventory is not None
 
1271
        # Setting self._inventory = None forces the dirstate to regenerate the
 
1272
        # working inventory. We do this because self.inventory may be inv, or
 
1273
        # may have been modified, and either case would prevent a clean delta
 
1274
        # being created.
 
1275
        self._inventory = None
 
1276
        # generate a delta,
 
1277
        delta = inv._make_delta(self.inventory)
 
1278
        # and apply it.
 
1279
        self.apply_inventory_delta(delta)
 
1280
        if had_inventory:
1234
1281
            self._inventory = inv
1235
1282
        self.flush()
1236
1283
 
1237
1284
 
1238
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1239
 
    """The first consolidated dirstate working tree format.
1240
 
 
1241
 
    This format:
1242
 
        - exists within a metadir controlling .bzr
1243
 
        - includes an explicit version marker for the workingtree control
1244
 
          files, separate from the BzrDir format
1245
 
        - modifies the hash cache format
1246
 
        - is new in bzr 0.15
1247
 
        - uses a LockDir to guard access to it.
1248
 
    """
1249
 
 
1250
 
    upgrade_recommended = False
1251
 
 
1252
 
    def get_format_string(self):
1253
 
        """See WorkingTreeFormat.get_format_string()."""
1254
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1255
 
 
1256
 
    def get_format_description(self):
1257
 
        """See WorkingTreeFormat.get_format_description()."""
1258
 
        return "Working tree format 4"
 
1285
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
 
1286
 
 
1287
    def __init__(self, tree):
 
1288
        self.tree = tree
 
1289
 
 
1290
    def sha1(self, abspath):
 
1291
        """See dirstate.SHA1Provider.sha1()."""
 
1292
        filters = self.tree._content_filter_stack(
 
1293
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1294
        return internal_size_sha_file_byname(abspath, filters)[1]
 
1295
 
 
1296
    def stat_and_sha1(self, abspath):
 
1297
        """See dirstate.SHA1Provider.stat_and_sha1()."""
 
1298
        filters = self.tree._content_filter_stack(
 
1299
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1300
        file_obj = file(abspath, 'rb', 65000)
 
1301
        try:
 
1302
            statvalue = os.fstat(file_obj.fileno())
 
1303
            if filters:
 
1304
                file_obj = filtered_input_file(file_obj, filters)
 
1305
            sha1 = osutils.size_sha_file(file_obj)[1]
 
1306
        finally:
 
1307
            file_obj.close()
 
1308
        return statvalue, sha1
 
1309
 
 
1310
 
 
1311
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
 
1312
    """Dirstate working tree that supports content filtering.
 
1313
 
 
1314
    The dirstate holds the hash and size of the canonical form of the file, 
 
1315
    and most methods must return that.
 
1316
    """
 
1317
 
 
1318
    def _file_content_summary(self, path, stat_result):
 
1319
        # This is to support the somewhat obsolete path_content_summary method
 
1320
        # with content filtering: see
 
1321
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
 
1322
        #
 
1323
        # If the dirstate cache is up to date and knows the hash and size,
 
1324
        # return that.
 
1325
        # Otherwise if there are no content filters, return the on-disk size
 
1326
        # and leave the hash blank.
 
1327
        # Otherwise, read and filter the on-disk file and use its size and
 
1328
        # hash.
 
1329
        #
 
1330
        # The dirstate doesn't store the size of the canonical form so we
 
1331
        # can't trust it for content-filtered trees.  We just return None.
 
1332
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
 
1333
        executable = self._is_executable_from_path_and_stat(path, stat_result)
 
1334
        return ('file', None, executable, dirstate_sha1)
 
1335
 
 
1336
 
 
1337
class WorkingTree4(DirStateWorkingTree):
 
1338
    """This is the Format 4 working tree.
 
1339
 
 
1340
    This differs from WorkingTree3 by:
 
1341
     - Having a consolidated internal dirstate, stored in a
 
1342
       randomly-accessible sorted file on disk.
 
1343
     - Not having a regular inventory attribute.  One can be synthesized
 
1344
       on demand but this is expensive and should be avoided.
 
1345
 
 
1346
    This is new in bzr 0.15.
 
1347
    """
 
1348
 
 
1349
 
 
1350
class WorkingTree5(ContentFilteringDirStateWorkingTree):
 
1351
    """This is the Format 5 working tree.
 
1352
 
 
1353
    This differs from WorkingTree4 by:
 
1354
     - Supporting content filtering.
 
1355
 
 
1356
    This is new in bzr 1.11.
 
1357
    """
 
1358
 
 
1359
 
 
1360
class WorkingTree6(ContentFilteringDirStateWorkingTree):
 
1361
    """This is the Format 6 working tree.
 
1362
 
 
1363
    This differs from WorkingTree5 by:
 
1364
     - Supporting a current view that may mask the set of files in a tree
 
1365
       impacted by most user operations.
 
1366
 
 
1367
    This is new in bzr 1.14.
 
1368
    """
 
1369
 
 
1370
    def _make_views(self):
 
1371
        return views.PathBasedViews(self)
 
1372
 
 
1373
 
 
1374
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1259
1375
 
1260
1376
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1261
 
                   accelerator_tree=None):
 
1377
                   accelerator_tree=None, hardlink=False):
1262
1378
        """See WorkingTreeFormat.initialize().
1263
1379
 
1264
1380
        :param revision_id: allows creating a working tree at a different
1267
1383
            contents more quickly than the revision tree, i.e. a workingtree.
1268
1384
            The revision tree will be used for cases where accelerator_tree's
1269
1385
            content is different.
 
1386
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1387
            where possible.
1270
1388
 
1271
1389
        These trees get an initial random root id, if their repository supports
1272
1390
        rich root data, TREE_ROOT otherwise.
1277
1395
        control_files = self._open_control_files(a_bzrdir)
1278
1396
        control_files.create_lock()
1279
1397
        control_files.lock_write()
1280
 
        control_files.put_utf8('format', self.get_format_string())
 
1398
        transport.put_bytes('format', self.get_format_string(),
 
1399
            mode=a_bzrdir._get_file_mode())
1281
1400
        if from_branch is not None:
1282
1401
            branch = from_branch
1283
1402
        else:
1289
1408
        state = dirstate.DirState.initialize(local_path)
1290
1409
        state.unlock()
1291
1410
        del state
1292
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1411
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1293
1412
                         branch,
1294
1413
                         _format=self,
1295
1414
                         _bzrdir=a_bzrdir,
1297
1416
        wt._new_tree()
1298
1417
        wt.lock_tree_write()
1299
1418
        try:
1300
 
            if revision_id in (None, NULL_REVISION):
 
1419
            self._init_custom_control_files(wt)
 
1420
            if revision_id in (None, _mod_revision.NULL_REVISION):
1301
1421
                if branch.repository.supports_rich_root():
1302
1422
                    wt._set_root_id(generate_ids.gen_root_id())
1303
1423
                else:
1314
1434
                    pass
1315
1435
            if basis is None:
1316
1436
                basis = branch.repository.revision_tree(revision_id)
1317
 
            if revision_id == NULL_REVISION:
 
1437
            if revision_id == _mod_revision.NULL_REVISION:
1318
1438
                parents_list = []
1319
1439
            else:
1320
1440
                parents_list = [(revision_id, basis)]
1321
1441
            basis.lock_read()
1322
 
            wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1323
 
            wt.flush()
1324
 
            # if the basis has a root id we have to use that; otherwise we use
1325
 
            # a new random one
1326
 
            basis_root_id = basis.get_root_id()
1327
 
            if basis_root_id is not None:
1328
 
                wt._set_root_id(basis_root_id)
 
1442
            try:
 
1443
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1329
1444
                wt.flush()
1330
 
            transform.build_tree(basis, wt, accelerator_tree)
1331
 
            basis.unlock()
 
1445
                # if the basis has a root id we have to use that; otherwise we
 
1446
                # use a new random one
 
1447
                basis_root_id = basis.get_root_id()
 
1448
                if basis_root_id is not None:
 
1449
                    wt._set_root_id(basis_root_id)
 
1450
                    wt.flush()
 
1451
                if wt.supports_content_filtering():
 
1452
                    # The original tree may not have the same content filters
 
1453
                    # applied so we can't safely build the inventory delta from
 
1454
                    # the source tree.
 
1455
                    delta_from_tree = False
 
1456
                else:
 
1457
                    delta_from_tree = True
 
1458
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1459
                # because wt4.apply_inventory_delta does not mutate the input
 
1460
                # inventory entries.
 
1461
                transform.build_tree(basis, wt, accelerator_tree,
 
1462
                                     hardlink=hardlink,
 
1463
                                     delta_from_tree=delta_from_tree)
 
1464
            finally:
 
1465
                basis.unlock()
1332
1466
        finally:
1333
1467
            control_files.unlock()
1334
1468
            wt.unlock()
1335
1469
        return wt
1336
1470
 
 
1471
    def _init_custom_control_files(self, wt):
 
1472
        """Subclasses with custom control files should override this method.
 
1473
 
 
1474
        The working tree and control files are locked for writing when this
 
1475
        method is called.
 
1476
 
 
1477
        :param wt: the WorkingTree object
 
1478
        """
 
1479
 
1337
1480
    def _open(self, a_bzrdir, control_files):
1338
1481
        """Open the tree itself.
1339
1482
 
1340
1483
        :param a_bzrdir: the dir for the tree.
1341
1484
        :param control_files: the control files for the tree.
1342
1485
        """
1343
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1486
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1344
1487
                           branch=a_bzrdir.open_branch(),
1345
1488
                           _format=self,
1346
1489
                           _bzrdir=a_bzrdir,
1347
1490
                           _control_files=control_files)
1348
1491
 
1349
1492
    def __get_matchingbzrdir(self):
 
1493
        return self._get_matchingbzrdir()
 
1494
 
 
1495
    def _get_matchingbzrdir(self):
 
1496
        """Overrideable method to get a bzrdir for testing."""
1350
1497
        # please test against something that will let us do tree references
1351
1498
        return bzrdir.format_registry.make_bzrdir(
1352
1499
            'dirstate-with-subtree')
1354
1501
    _matchingbzrdir = property(__get_matchingbzrdir)
1355
1502
 
1356
1503
 
 
1504
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
 
1505
    """The first consolidated dirstate working tree format.
 
1506
 
 
1507
    This format:
 
1508
        - exists within a metadir controlling .bzr
 
1509
        - includes an explicit version marker for the workingtree control
 
1510
          files, separate from the BzrDir format
 
1511
        - modifies the hash cache format
 
1512
        - is new in bzr 0.15
 
1513
        - uses a LockDir to guard access to it.
 
1514
    """
 
1515
 
 
1516
    upgrade_recommended = False
 
1517
 
 
1518
    _tree_class = WorkingTree4
 
1519
 
 
1520
    def get_format_string(self):
 
1521
        """See WorkingTreeFormat.get_format_string()."""
 
1522
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1523
 
 
1524
    def get_format_description(self):
 
1525
        """See WorkingTreeFormat.get_format_description()."""
 
1526
        return "Working tree format 4"
 
1527
 
 
1528
 
 
1529
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
 
1530
    """WorkingTree format supporting content filtering.
 
1531
    """
 
1532
 
 
1533
    upgrade_recommended = False
 
1534
 
 
1535
    _tree_class = WorkingTree5
 
1536
 
 
1537
    def get_format_string(self):
 
1538
        """See WorkingTreeFormat.get_format_string()."""
 
1539
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
 
1540
 
 
1541
    def get_format_description(self):
 
1542
        """See WorkingTreeFormat.get_format_description()."""
 
1543
        return "Working tree format 5"
 
1544
 
 
1545
    def supports_content_filtering(self):
 
1546
        return True
 
1547
 
 
1548
 
 
1549
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
 
1550
    """WorkingTree format supporting views.
 
1551
    """
 
1552
 
 
1553
    upgrade_recommended = False
 
1554
 
 
1555
    _tree_class = WorkingTree6
 
1556
 
 
1557
    def get_format_string(self):
 
1558
        """See WorkingTreeFormat.get_format_string()."""
 
1559
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
 
1560
 
 
1561
    def get_format_description(self):
 
1562
        """See WorkingTreeFormat.get_format_description()."""
 
1563
        return "Working tree format 6"
 
1564
 
 
1565
    def _init_custom_control_files(self, wt):
 
1566
        """Subclasses with custom control files should override this method."""
 
1567
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
 
1568
 
 
1569
    def supports_content_filtering(self):
 
1570
        return True
 
1571
 
 
1572
    def supports_views(self):
 
1573
        return True
 
1574
 
 
1575
 
1357
1576
class DirStateRevisionTree(Tree):
1358
 
    """A revision tree pulling the inventory from a dirstate."""
 
1577
    """A revision tree pulling the inventory from a dirstate.
 
1578
    
 
1579
    Note that this is one of the historical (ie revision) trees cached in the
 
1580
    dirstate for easy access, not the workingtree.
 
1581
    """
1359
1582
 
1360
1583
    def __init__(self, dirstate, revision_id, repository):
1361
1584
        self._dirstate = dirstate
1364
1587
        self._inventory = None
1365
1588
        self._locked = 0
1366
1589
        self._dirstate_locked = False
 
1590
        self._repo_supports_tree_reference = getattr(
 
1591
            repository._format, "supports_tree_reference",
 
1592
            False)
1367
1593
 
1368
1594
    def __repr__(self):
1369
1595
        return "<%s of %s in %s>" % \
1372
1598
    def annotate_iter(self, file_id,
1373
1599
                      default_revision=_mod_revision.CURRENT_REVISION):
1374
1600
        """See Tree.annotate_iter"""
1375
 
        w = self._get_weave(file_id)
1376
 
        return w.annotate_iter(self.inventory[file_id].revision)
 
1601
        text_key = (file_id, self.inventory[file_id].revision)
 
1602
        annotations = self._repository.texts.annotate(text_key)
 
1603
        return [(key[-1], line) for (key, line) in annotations]
1377
1604
 
1378
1605
    def _get_ancestors(self, default_revision):
1379
1606
        return set(self._repository.get_ancestry(self._revision_id,
1408
1635
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1409
1636
        return path_utf8.decode('utf8')
1410
1637
 
 
1638
    def iter_references(self):
 
1639
        if not self._repo_supports_tree_reference:
 
1640
            # When the repo doesn't support references, we will have nothing to
 
1641
            # return
 
1642
            return iter([])
 
1643
        # Otherwise, fall back to the default implementation
 
1644
        return super(DirStateRevisionTree, self).iter_references()
 
1645
 
1411
1646
    def _get_parent_index(self):
1412
1647
        """Return the index in the dirstate referenced by this tree."""
1413
1648
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1418
1653
        If either file_id or path is supplied, it is used as the key to lookup.
1419
1654
        If both are supplied, the fastest lookup is used, and an error is
1420
1655
        raised if they do not both point at the same row.
1421
 
        
 
1656
 
1422
1657
        :param file_id: An optional unicode file_id to be looked up.
1423
1658
        :param path: An optional unicode path to be looked up.
1424
1659
        :return: The dirstate row tuple for path/file_id, or (None, None)
1438
1673
 
1439
1674
        This is relatively expensive: we have to walk the entire dirstate.
1440
1675
        """
1441
 
        assert self._locked, 'cannot generate inventory of an unlocked '\
1442
 
            'dirstate revision tree'
 
1676
        if not self._locked:
 
1677
            raise AssertionError(
 
1678
                'cannot generate inventory of an unlocked '
 
1679
                'dirstate revision tree')
1443
1680
        # separate call for profiling - makes it clear where the costs are.
1444
1681
        self._dirstate._read_dirblocks_if_needed()
1445
 
        assert self._revision_id in self._dirstate.get_parent_ids(), \
1446
 
            'parent %s has disappeared from %s' % (
1447
 
            self._revision_id, self._dirstate.get_parent_ids())
 
1682
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1683
            raise AssertionError(
 
1684
                'parent %s has disappeared from %s' % (
 
1685
                self._revision_id, self._dirstate.get_parent_ids()))
1448
1686
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1449
1687
        # This is identical now to the WorkingTree _generate_inventory except
1450
1688
        # for the tree index use.
1451
1689
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1452
1690
        current_id = root_key[2]
1453
 
        assert current_entry[parent_index][0] == 'd'
 
1691
        if current_entry[parent_index][0] != 'd':
 
1692
            raise AssertionError()
1454
1693
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1455
1694
        inv.root.revision = current_entry[parent_index][4]
1456
1695
        # Turn some things into local variables
1496
1735
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1497
1736
                            % entry)
1498
1737
                # These checks cost us around 40ms on a 55k entry tree
1499
 
                assert file_id not in inv_byid
1500
 
                assert name_unicode not in parent_ie.children
 
1738
                if file_id in inv_byid:
 
1739
                    raise AssertionError('file_id %s already in'
 
1740
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1741
                if name_unicode in parent_ie.children:
 
1742
                    raise AssertionError('name %r already in parent'
 
1743
                        % (name_unicode,))
1501
1744
                inv_byid[file_id] = inv_entry
1502
1745
                parent_ie.children[name_unicode] = inv_entry
1503
1746
        self._inventory = inv
1513
1756
            return None
1514
1757
        parent_index = self._get_parent_index()
1515
1758
        last_changed_revision = entry[1][parent_index][4]
1516
 
        return self._repository.get_revision(last_changed_revision).timestamp
 
1759
        try:
 
1760
            rev = self._repository.get_revision(last_changed_revision)
 
1761
        except errors.NoSuchRevision:
 
1762
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
 
1763
        return rev.timestamp
1517
1764
 
1518
1765
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1519
1766
        entry = self._get_entry(file_id=file_id, path=path)
1523
1770
            return parent_details[1]
1524
1771
        return None
1525
1772
 
1526
 
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
1527
 
    def get_weave(self, file_id):
1528
 
        return self._get_weave(file_id)
1529
 
 
1530
 
    def _get_weave(self, file_id):
1531
 
        return self._repository.weave_store.get_weave(file_id,
1532
 
                self._repository.get_transaction())
1533
 
 
1534
1773
    def get_file(self, file_id, path=None):
1535
1774
        return StringIO(self.get_file_text(file_id))
1536
1775
 
1537
 
    def get_file_lines(self, file_id):
1538
 
        entry = self._get_entry(file_id=file_id)[1]
1539
 
        if entry == None:
1540
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1541
 
        return self._get_weave(file_id).get_lines(entry[1][4])
1542
 
 
1543
1776
    def get_file_size(self, file_id):
 
1777
        """See Tree.get_file_size"""
1544
1778
        return self.inventory[file_id].text_size
1545
1779
 
1546
 
    def get_file_text(self, file_id):
1547
 
        return ''.join(self.get_file_lines(file_id))
 
1780
    def get_file_text(self, file_id, path=None):
 
1781
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
 
1782
        return ''.join(content)
1548
1783
 
1549
1784
    def get_reference_revision(self, file_id, path=None):
1550
1785
        return self.inventory[file_id].reference_revision
1569
1804
        if entry[1][parent_index][0] != 'l':
1570
1805
            return None
1571
1806
        else:
1572
 
            # At present, none of the tree implementations supports non-ascii
1573
 
            # symlink targets. So we will just assume that the dirstate path is
1574
 
            # correct.
1575
 
            return entry[1][parent_index][1]
 
1807
            target = entry[1][parent_index][1]
 
1808
            target = target.decode('utf8')
 
1809
            return target
1576
1810
 
1577
1811
    def get_revision_id(self):
1578
1812
        """Return the revision id for this tree."""
1597
1831
 
1598
1832
    def kind(self, file_id):
1599
1833
        entry = self._get_entry(file_id=file_id)[1]
1600
 
        if entry == None:
 
1834
        if entry is None:
1601
1835
            raise errors.NoSuchId(tree=self, file_id=file_id)
1602
 
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1836
        parent_index = self._get_parent_index()
 
1837
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1838
 
 
1839
    def stored_kind(self, file_id):
 
1840
        """See Tree.stored_kind"""
 
1841
        return self.kind(file_id)
1603
1842
 
1604
1843
    def path_content_summary(self, path):
1605
1844
        """See Tree.path_content_summary."""
1621
1860
            return None
1622
1861
        return ie.executable
1623
1862
 
1624
 
    def list_files(self, include_root=False):
 
1863
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1625
1864
        # We use a standard implementation, because DirStateRevisionTree is
1626
1865
        # dealing with one of the parents of the current state
1627
1866
        inv = self._get_inventory()
1628
 
        entries = inv.iter_entries()
1629
 
        if self.inventory.root is not None and not include_root:
 
1867
        if from_dir is None:
 
1868
            from_dir_id = None
 
1869
        else:
 
1870
            from_dir_id = inv.path2id(from_dir)
 
1871
            if from_dir_id is None:
 
1872
                # Directory not versioned
 
1873
                return
 
1874
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
 
1875
        if inv.root is not None and not include_root and from_dir is None:
1630
1876
            entries.next()
1631
1877
        for path, entry in entries:
1632
1878
            yield path, 'V', entry.kind, entry.file_id, entry
1665
1911
                self._dirstate_locked = False
1666
1912
            self._repository.unlock()
1667
1913
 
 
1914
    @needs_read_lock
 
1915
    def supports_tree_reference(self):
 
1916
        return self._repo_supports_tree_reference
 
1917
 
1668
1918
    def walkdirs(self, prefix=""):
1669
1919
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1670
 
        # implementation based on an inventory.  
 
1920
        # implementation based on an inventory.
1671
1921
        # This should be cleaned up to use the much faster Dirstate code
1672
1922
        # So for now, we just build up the parent inventory, and extract
1673
1923
        # it the same way RevisionTree does.
1702
1952
 
1703
1953
class InterDirStateTree(InterTree):
1704
1954
    """Fast path optimiser for changes_from with dirstate trees.
1705
 
    
1706
 
    This is used only when both trees are in the dirstate working file, and 
1707
 
    the source is any parent within the dirstate, and the destination is 
 
1955
 
 
1956
    This is used only when both trees are in the dirstate working file, and
 
1957
    the source is any parent within the dirstate, and the destination is
1708
1958
    the current working tree of the same dirstate.
1709
1959
    """
1710
1960
    # this could be generalized to allow comparisons between any trees in the
1723
1973
        target.set_parent_ids([revid])
1724
1974
        return target.basis_tree(), target
1725
1975
 
 
1976
    @classmethod
 
1977
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
 
1978
        result = klass.make_source_parent_tree(source, target)
 
1979
        result[1]._iter_changes = dirstate.ProcessEntryPython
 
1980
        return result
 
1981
 
 
1982
    @classmethod
 
1983
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
 
1984
                                                  target):
 
1985
        from bzrlib.tests.test__dirstate_helpers import \
 
1986
            compiled_dirstate_helpers_feature
 
1987
        test_case.requireFeature(compiled_dirstate_helpers_feature)
 
1988
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
 
1989
        result = klass.make_source_parent_tree(source, target)
 
1990
        result[1]._iter_changes = ProcessEntryC
 
1991
        return result
 
1992
 
1726
1993
    _matching_from_tree_format = WorkingTreeFormat4()
1727
1994
    _matching_to_tree_format = WorkingTreeFormat4()
1728
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
1729
 
 
1730
 
    def _iter_changes(self, include_unchanged=False,
 
1995
 
 
1996
    @classmethod
 
1997
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
 
1998
        # This method shouldn't be called, because we have python and C
 
1999
        # specific flavours.
 
2000
        raise NotImplementedError
 
2001
 
 
2002
    def iter_changes(self, include_unchanged=False,
1731
2003
                      specific_files=None, pb=None, extra_trees=[],
1732
2004
                      require_versioned=True, want_unversioned=False):
1733
2005
        """Return the changes from source to target.
1734
2006
 
1735
 
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
2007
        :return: An iterator that yields tuples. See InterTree.iter_changes
1736
2008
            for details.
1737
2009
        :param specific_files: An optional list of file paths to restrict the
1738
2010
            comparison to. When mapping filenames to ids, all matches in all
1749
2021
            output. An unversioned file is defined as one with (False, False)
1750
2022
            for the versioned pair.
1751
2023
        """
1752
 
        utf8_decode = cache_utf8._utf8_decode
1753
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1754
 
        cmp_by_dirs = dirstate.cmp_by_dirs
1755
 
        # NB: show_status depends on being able to pass in non-versioned files
1756
 
        # and report them as unknown
1757
2024
        # TODO: handle extra trees in the dirstate.
1758
2025
        if (extra_trees or specific_files == []):
1759
2026
            # we can't fast-path these cases (yet)
1760
 
            for f in super(InterDirStateTree, self)._iter_changes(
 
2027
            return super(InterDirStateTree, self).iter_changes(
1761
2028
                include_unchanged, specific_files, pb, extra_trees,
1762
 
                require_versioned, want_unversioned=want_unversioned):
1763
 
                yield f
1764
 
            return
 
2029
                require_versioned, want_unversioned=want_unversioned)
1765
2030
        parent_ids = self.target.get_parent_ids()
1766
 
        assert (self.source._revision_id in parent_ids
1767
 
                or self.source._revision_id == NULL_REVISION), \
1768
 
                "revision {%s} is not stored in {%s}, but %s " \
1769
 
                "can only be used for trees stored in the dirstate" \
1770
 
                % (self.source._revision_id, self.target, self._iter_changes)
 
2031
        if not (self.source._revision_id in parent_ids
 
2032
                or self.source._revision_id == _mod_revision.NULL_REVISION):
 
2033
            raise AssertionError(
 
2034
                "revision {%s} is not stored in {%s}, but %s "
 
2035
                "can only be used for trees stored in the dirstate"
 
2036
                % (self.source._revision_id, self.target, self.iter_changes))
1771
2037
        target_index = 0
1772
 
        if self.source._revision_id == NULL_REVISION:
 
2038
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1773
2039
            source_index = None
1774
2040
            indices = (target_index,)
1775
2041
        else:
1776
 
            assert (self.source._revision_id in parent_ids), \
1777
 
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
1778
 
                self.source._revision_id, parent_ids)
 
2042
            if not (self.source._revision_id in parent_ids):
 
2043
                raise AssertionError(
 
2044
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
2045
                    self.source._revision_id, parent_ids))
1779
2046
            source_index = 1 + parent_ids.index(self.source._revision_id)
1780
2047
            indices = (source_index, target_index)
1781
2048
        # -- make all specific_files utf8 --
1782
2049
        if specific_files:
1783
2050
            specific_files_utf8 = set()
1784
2051
            for path in specific_files:
 
2052
                # Note, if there are many specific files, using cache_utf8
 
2053
                # would be good here.
1785
2054
                specific_files_utf8.add(path.encode('utf8'))
1786
2055
            specific_files = specific_files_utf8
1787
2056
        else:
1788
2057
            specific_files = set([''])
1789
2058
        # -- specific_files is now a utf8 path set --
 
2059
 
1790
2060
        # -- get the state object and prepare it.
1791
2061
        state = self.target.current_dirstate()
1792
2062
        state._read_dirblocks_if_needed()
1793
 
        def _entries_for_path(path):
1794
 
            """Return a list with all the entries that match path for all ids.
1795
 
            """
1796
 
            dirname, basename = os.path.split(path)
1797
 
            key = (dirname, basename, '')
1798
 
            block_index, present = state._find_block_index_from_key(key)
1799
 
            if not present:
1800
 
                # the block which should contain path is absent.
1801
 
                return []
1802
 
            result = []
1803
 
            block = state._dirblocks[block_index][1]
1804
 
            entry_index, _ = state._find_entry_index(key, block)
1805
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1806
 
            while (entry_index < len(block) and
1807
 
                block[entry_index][0][0:2] == key[0:2]):
1808
 
                result.append(block[entry_index])
1809
 
                entry_index += 1
1810
 
            return result
1811
2063
        if require_versioned:
1812
2064
            # -- check all supplied paths are versioned in a search tree. --
1813
 
            all_versioned = True
 
2065
            not_versioned = []
1814
2066
            for path in specific_files:
1815
 
                path_entries = _entries_for_path(path)
 
2067
                path_entries = state._entries_for_path(path)
1816
2068
                if not path_entries:
1817
2069
                    # this specified path is not present at all: error
1818
 
                    all_versioned = False
1819
 
                    break
 
2070
                    not_versioned.append(path)
 
2071
                    continue
1820
2072
                found_versioned = False
1821
2073
                # for each id at this path
1822
2074
                for entry in path_entries:
1829
2081
                if not found_versioned:
1830
2082
                    # none of the indexes was not 'absent' at all ids for this
1831
2083
                    # path.
1832
 
                    all_versioned = False
1833
 
                    break
1834
 
            if not all_versioned:
1835
 
                raise errors.PathsNotVersionedError(specific_files)
 
2084
                    not_versioned.append(path)
 
2085
            if len(not_versioned) > 0:
 
2086
                raise errors.PathsNotVersionedError(not_versioned)
1836
2087
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1837
 
        search_specific_files = set()
1838
 
        for path in specific_files:
1839
 
            other_specific_files = specific_files.difference(set([path]))
1840
 
            if not osutils.is_inside_any(other_specific_files, path):
1841
 
                # this is a top level path, we must check it.
1842
 
                search_specific_files.add(path)
1843
 
        # sketch: 
1844
 
        # compare source_index and target_index at or under each element of search_specific_files.
1845
 
        # follow the following comparison table. Note that we only want to do diff operations when
1846
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1847
 
        # for the target.
1848
 
        # cases:
1849
 
        # 
1850
 
        # Source | Target | disk | action
1851
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1852
 
        #        |        |      | diff check on source-target
1853
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1854
 
        #        |        |      | ???
1855
 
        #   r    |  a     |      | add source to search
1856
 
        #   r    |  a     |  a   | 
1857
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1858
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1859
 
        #   a    | fdlt   |      | add new id
1860
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1861
 
        #   a    |  a     |      | not present in either tree, skip
1862
 
        #   a    |  a     |  a   | not present in any tree, skip
1863
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1864
 
        #        |        |      | may not be selected by the users list of paths.
1865
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1866
 
        #        |        |      | may not be selected by the users list of paths.
1867
 
        #  fdlt  | fdlt   |      | content in both: diff them
1868
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1869
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1870
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1871
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1872
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1873
 
        #        |        |      | this id at the other path.
1874
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1875
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1876
 
        #        |        |      | this id at the other path.
1877
 
 
1878
 
        # for all search_indexs in each path at or under each element of
1879
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1880
 
        # relocated path as one to search if its not searched already. If the
1881
 
        # detail is not relocated, add the id.
1882
 
        searched_specific_files = set()
1883
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1884
 
        # Using a list so that we can access the values and change them in
1885
 
        # nested scope. Each one is [path, file_id, entry]
1886
 
        last_source_parent = [None, None]
1887
 
        last_target_parent = [None, None]
 
2088
        search_specific_files = osutils.minimum_path_selection(specific_files)
1888
2089
 
1889
2090
        use_filesystem_for_exec = (sys.platform != 'win32')
1890
 
 
1891
 
        # Just a sentry, so that _process_entry can say that this
1892
 
        # record is handled, but isn't interesting to process (unchanged)
1893
 
        uninteresting = object()
1894
 
 
1895
 
 
1896
 
        old_dirname_to_file_id = {}
1897
 
        new_dirname_to_file_id = {}
1898
 
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1899
 
        #       keeping a cache of directories that we have seen.
1900
 
 
1901
 
        def _process_entry(entry, path_info):
1902
 
            """Compare an entry and real disk to generate delta information.
1903
 
 
1904
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1905
 
                the path of entry. If None, then the path is considered absent.
1906
 
                (Perhaps we should pass in a concrete entry for this ?)
1907
 
                Basename is returned as a utf8 string because we expect this
1908
 
                tuple will be ignored, and don't want to take the time to
1909
 
                decode.
1910
 
            :return: None if these don't match
1911
 
                     A tuple of information about the change, or
1912
 
                     the object 'uninteresting' if these match, but are
1913
 
                     basically identical.
1914
 
            """
1915
 
            if source_index is None:
1916
 
                source_details = NULL_PARENT_DETAILS
1917
 
            else:
1918
 
                source_details = entry[1][source_index]
1919
 
            target_details = entry[1][target_index]
1920
 
            target_minikind = target_details[0]
1921
 
            if path_info is not None and target_minikind in 'fdlt':
1922
 
                assert target_index == 0
1923
 
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
1924
 
                                                  stat_value=path_info[3])
1925
 
                # The entry may have been modified by update_entry
1926
 
                target_details = entry[1][target_index]
1927
 
                target_minikind = target_details[0]
1928
 
            else:
1929
 
                link_or_sha1 = None
1930
 
            file_id = entry[0][2]
1931
 
            source_minikind = source_details[0]
1932
 
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
1933
 
                # claimed content in both: diff
1934
 
                #   r    | fdlt   |      | add source to search, add id path move and perform
1935
 
                #        |        |      | diff check on source-target
1936
 
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
1937
 
                #        |        |      | ???
1938
 
                if source_minikind in 'r':
1939
 
                    # add the source to the search path to find any children it
1940
 
                    # has.  TODO ? : only add if it is a container ?
1941
 
                    if not osutils.is_inside_any(searched_specific_files,
1942
 
                                                 source_details[1]):
1943
 
                        search_specific_files.add(source_details[1])
1944
 
                    # generate the old path; this is needed for stating later
1945
 
                    # as well.
1946
 
                    old_path = source_details[1]
1947
 
                    old_dirname, old_basename = os.path.split(old_path)
1948
 
                    path = pathjoin(entry[0][0], entry[0][1])
1949
 
                    old_entry = state._get_entry(source_index,
1950
 
                                                 path_utf8=old_path)
1951
 
                    # update the source details variable to be the real
1952
 
                    # location.
1953
 
                    source_details = old_entry[1][source_index]
1954
 
                    source_minikind = source_details[0]
1955
 
                else:
1956
 
                    old_dirname = entry[0][0]
1957
 
                    old_basename = entry[0][1]
1958
 
                    old_path = path = None
1959
 
                if path_info is None:
1960
 
                    # the file is missing on disk, show as removed.
1961
 
                    content_change = True
1962
 
                    target_kind = None
1963
 
                    target_exec = False
1964
 
                else:
1965
 
                    # source and target are both versioned and disk file is present.
1966
 
                    target_kind = path_info[2]
1967
 
                    if target_kind == 'directory':
1968
 
                        if path is None:
1969
 
                            old_path = path = pathjoin(old_dirname, old_basename)
1970
 
                        new_dirname_to_file_id[path] = file_id
1971
 
                        if source_minikind != 'd':
1972
 
                            content_change = True
1973
 
                        else:
1974
 
                            # directories have no fingerprint
1975
 
                            content_change = False
1976
 
                        target_exec = False
1977
 
                    elif target_kind == 'file':
1978
 
                        if source_minikind != 'f':
1979
 
                            content_change = True
1980
 
                        else:
1981
 
                            # We could check the size, but we already have the
1982
 
                            # sha1 hash.
1983
 
                            content_change = (link_or_sha1 != source_details[1])
1984
 
                        # Target details is updated at update_entry time
1985
 
                        if use_filesystem_for_exec:
1986
 
                            # We don't need S_ISREG here, because we are sure
1987
 
                            # we are dealing with a file.
1988
 
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
1989
 
                        else:
1990
 
                            target_exec = target_details[3]
1991
 
                    elif target_kind == 'symlink':
1992
 
                        if source_minikind != 'l':
1993
 
                            content_change = True
1994
 
                        else:
1995
 
                            content_change = (link_or_sha1 != source_details[1])
1996
 
                        target_exec = False
1997
 
                    elif target_kind == 'tree-reference':
1998
 
                        if source_minikind != 't':
1999
 
                            content_change = True
2000
 
                        else:
2001
 
                            content_change = False
2002
 
                        target_exec = False
2003
 
                    else:
2004
 
                        raise Exception, "unknown kind %s" % path_info[2]
2005
 
                if source_minikind == 'd':
2006
 
                    if path is None:
2007
 
                        old_path = path = pathjoin(old_dirname, old_basename)
2008
 
                    old_dirname_to_file_id[old_path] = file_id
2009
 
                # parent id is the entry for the path in the target tree
2010
 
                if old_dirname == last_source_parent[0]:
2011
 
                    source_parent_id = last_source_parent[1]
2012
 
                else:
2013
 
                    try:
2014
 
                        source_parent_id = old_dirname_to_file_id[old_dirname]
2015
 
                    except KeyError:
2016
 
                        source_parent_entry = state._get_entry(source_index,
2017
 
                                                               path_utf8=old_dirname)
2018
 
                        source_parent_id = source_parent_entry[0][2]
2019
 
                    if source_parent_id == entry[0][2]:
2020
 
                        # This is the root, so the parent is None
2021
 
                        source_parent_id = None
2022
 
                    else:
2023
 
                        last_source_parent[0] = old_dirname
2024
 
                        last_source_parent[1] = source_parent_id
2025
 
                new_dirname = entry[0][0]
2026
 
                if new_dirname == last_target_parent[0]:
2027
 
                    target_parent_id = last_target_parent[1]
2028
 
                else:
2029
 
                    try:
2030
 
                        target_parent_id = new_dirname_to_file_id[new_dirname]
2031
 
                    except KeyError:
2032
 
                        # TODO: We don't always need to do the lookup, because the
2033
 
                        #       parent entry will be the same as the source entry.
2034
 
                        target_parent_entry = state._get_entry(target_index,
2035
 
                                                               path_utf8=new_dirname)
2036
 
                        assert target_parent_entry != (None, None), (
2037
 
                            "Could not find target parent in wt: %s\nparent of: %s"
2038
 
                            % (new_dirname, entry))
2039
 
                        target_parent_id = target_parent_entry[0][2]
2040
 
                    if target_parent_id == entry[0][2]:
2041
 
                        # This is the root, so the parent is None
2042
 
                        target_parent_id = None
2043
 
                    else:
2044
 
                        last_target_parent[0] = new_dirname
2045
 
                        last_target_parent[1] = target_parent_id
2046
 
 
2047
 
                source_exec = source_details[3]
2048
 
                if (include_unchanged
2049
 
                    or content_change
2050
 
                    or source_parent_id != target_parent_id
2051
 
                    or old_basename != entry[0][1]
2052
 
                    or source_exec != target_exec
2053
 
                    ):
2054
 
                    if old_path is None:
2055
 
                        old_path = path = pathjoin(old_dirname, old_basename)
2056
 
                        old_path_u = utf8_decode(old_path)[0]
2057
 
                        path_u = old_path_u
2058
 
                    else:
2059
 
                        old_path_u = utf8_decode(old_path)[0]
2060
 
                        if old_path == path:
2061
 
                            path_u = old_path_u
2062
 
                        else:
2063
 
                            path_u = utf8_decode(path)[0]
2064
 
                    source_kind = _minikind_to_kind[source_minikind]
2065
 
                    return (entry[0][2],
2066
 
                           (old_path_u, path_u),
2067
 
                           content_change,
2068
 
                           (True, True),
2069
 
                           (source_parent_id, target_parent_id),
2070
 
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
2071
 
                           (source_kind, target_kind),
2072
 
                           (source_exec, target_exec))
2073
 
                else:
2074
 
                    return uninteresting
2075
 
            elif source_minikind in 'a' and target_minikind in 'fdlt':
2076
 
                # looks like a new file
2077
 
                if path_info is not None:
2078
 
                    path = pathjoin(entry[0][0], entry[0][1])
2079
 
                    # parent id is the entry for the path in the target tree
2080
 
                    # TODO: these are the same for an entire directory: cache em.
2081
 
                    parent_id = state._get_entry(target_index,
2082
 
                                                 path_utf8=entry[0][0])[0][2]
2083
 
                    if parent_id == entry[0][2]:
2084
 
                        parent_id = None
2085
 
                    if use_filesystem_for_exec:
2086
 
                        # We need S_ISREG here, because we aren't sure if this
2087
 
                        # is a file or not.
2088
 
                        target_exec = bool(
2089
 
                            stat.S_ISREG(path_info[3].st_mode)
2090
 
                            and stat.S_IEXEC & path_info[3].st_mode)
2091
 
                    else:
2092
 
                        target_exec = target_details[3]
2093
 
                    return (entry[0][2],
2094
 
                           (None, utf8_decode(path)[0]),
2095
 
                           True,
2096
 
                           (False, True),
2097
 
                           (None, parent_id),
2098
 
                           (None, utf8_decode(entry[0][1])[0]),
2099
 
                           (None, path_info[2]),
2100
 
                           (None, target_exec))
2101
 
                else:
2102
 
                    # but its not on disk: we deliberately treat this as just
2103
 
                    # never-present. (Why ?! - RBC 20070224)
2104
 
                    pass
2105
 
            elif source_minikind in 'fdlt' and target_minikind in 'a':
2106
 
                # unversioned, possibly, or possibly not deleted: we dont care.
2107
 
                # if its still on disk, *and* theres no other entry at this
2108
 
                # path [we dont know this in this routine at the moment -
2109
 
                # perhaps we should change this - then it would be an unknown.
2110
 
                old_path = pathjoin(entry[0][0], entry[0][1])
2111
 
                # parent id is the entry for the path in the target tree
2112
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
2113
 
                if parent_id == entry[0][2]:
2114
 
                    parent_id = None
2115
 
                return (entry[0][2],
2116
 
                       (utf8_decode(old_path)[0], None),
2117
 
                       True,
2118
 
                       (True, False),
2119
 
                       (parent_id, None),
2120
 
                       (utf8_decode(entry[0][1])[0], None),
2121
 
                       (_minikind_to_kind[source_minikind], None),
2122
 
                       (source_details[3], None))
2123
 
            elif source_minikind in 'fdlt' and target_minikind in 'r':
2124
 
                # a rename; could be a true rename, or a rename inherited from
2125
 
                # a renamed parent. TODO: handle this efficiently. Its not
2126
 
                # common case to rename dirs though, so a correct but slow
2127
 
                # implementation will do.
2128
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
2129
 
                    search_specific_files.add(target_details[1])
2130
 
            elif source_minikind in 'ra' and target_minikind in 'ra':
2131
 
                # neither of the selected trees contain this file,
2132
 
                # so skip over it. This is not currently directly tested, but
2133
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
2134
 
                pass
2135
 
            else:
2136
 
                raise AssertionError("don't know how to compare "
2137
 
                    "source_minikind=%r, target_minikind=%r"
2138
 
                    % (source_minikind, target_minikind))
2139
 
                ## import pdb;pdb.set_trace()
2140
 
            return None
2141
 
 
2142
 
        while search_specific_files:
2143
 
            # TODO: the pending list should be lexically sorted?  the
2144
 
            # interface doesn't require it.
2145
 
            current_root = search_specific_files.pop()
2146
 
            current_root_unicode = current_root.decode('utf8')
2147
 
            searched_specific_files.add(current_root)
2148
 
            # process the entries for this containing directory: the rest will be
2149
 
            # found by their parents recursively.
2150
 
            root_entries = _entries_for_path(current_root)
2151
 
            root_abspath = self.target.abspath(current_root_unicode)
2152
 
            try:
2153
 
                root_stat = os.lstat(root_abspath)
2154
 
            except OSError, e:
2155
 
                if e.errno == errno.ENOENT:
2156
 
                    # the path does not exist: let _process_entry know that.
2157
 
                    root_dir_info = None
2158
 
                else:
2159
 
                    # some other random error: hand it up.
2160
 
                    raise
2161
 
            else:
2162
 
                root_dir_info = ('', current_root,
2163
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
2164
 
                    root_abspath)
2165
 
                if root_dir_info[2] == 'directory':
2166
 
                    if self.target._directory_is_tree_reference(
2167
 
                        current_root.decode('utf8')):
2168
 
                        root_dir_info = root_dir_info[:2] + \
2169
 
                            ('tree-reference',) + root_dir_info[3:]
2170
 
 
2171
 
            if not root_entries and not root_dir_info:
2172
 
                # this specified path is not present at all, skip it.
2173
 
                continue
2174
 
            path_handled = False
2175
 
            for entry in root_entries:
2176
 
                result = _process_entry(entry, root_dir_info)
2177
 
                if result is not None:
2178
 
                    path_handled = True
2179
 
                    if result is not uninteresting:
2180
 
                        yield result
2181
 
            if want_unversioned and not path_handled and root_dir_info:
2182
 
                new_executable = bool(
2183
 
                    stat.S_ISREG(root_dir_info[3].st_mode)
2184
 
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
2185
 
                yield (None,
2186
 
                       (None, current_root_unicode),
2187
 
                       True,
2188
 
                       (False, False),
2189
 
                       (None, None),
2190
 
                       (None, splitpath(current_root_unicode)[-1]),
2191
 
                       (None, root_dir_info[2]),
2192
 
                       (None, new_executable)
2193
 
                      )
2194
 
            initial_key = (current_root, '', '')
2195
 
            block_index, _ = state._find_block_index_from_key(initial_key)
2196
 
            if block_index == 0:
2197
 
                # we have processed the total root already, but because the
2198
 
                # initial key matched it we should skip it here.
2199
 
                block_index +=1
2200
 
            if root_dir_info and root_dir_info[2] == 'tree-reference':
2201
 
                current_dir_info = None
2202
 
            else:
2203
 
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
2204
 
                try:
2205
 
                    current_dir_info = dir_iterator.next()
2206
 
                except OSError, e:
2207
 
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
2208
 
                    # python 2.5 has e.errno == EINVAL,
2209
 
                    #            and e.winerror == ERROR_DIRECTORY
2210
 
                    e_winerror = getattr(e, 'winerror', None)
2211
 
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
2212
 
                    # there may be directories in the inventory even though
2213
 
                    # this path is not a file on disk: so mark it as end of
2214
 
                    # iterator
2215
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
2216
 
                        current_dir_info = None
2217
 
                    elif (sys.platform == 'win32'
2218
 
                          and (e.errno in win_errors
2219
 
                               or e_winerror in win_errors)):
2220
 
                        current_dir_info = None
2221
 
                    else:
2222
 
                        raise
2223
 
                else:
2224
 
                    if current_dir_info[0][0] == '':
2225
 
                        # remove .bzr from iteration
2226
 
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
2227
 
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
2228
 
                        del current_dir_info[1][bzr_index]
2229
 
            # walk until both the directory listing and the versioned metadata
2230
 
            # are exhausted. 
2231
 
            if (block_index < len(state._dirblocks) and
2232
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2233
 
                current_block = state._dirblocks[block_index]
2234
 
            else:
2235
 
                current_block = None
2236
 
            while (current_dir_info is not None or
2237
 
                   current_block is not None):
2238
 
                if (current_dir_info and current_block
2239
 
                    and current_dir_info[0][0] != current_block[0]):
2240
 
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
2241
 
                        # filesystem data refers to paths not covered by the dirblock.
2242
 
                        # this has two possibilities:
2243
 
                        # A) it is versioned but empty, so there is no block for it
2244
 
                        # B) it is not versioned.
2245
 
 
2246
 
                        # if (A) then we need to recurse into it to check for
2247
 
                        # new unknown files or directories.
2248
 
                        # if (B) then we should ignore it, because we don't
2249
 
                        # recurse into unknown directories.
2250
 
                        path_index = 0
2251
 
                        while path_index < len(current_dir_info[1]):
2252
 
                                current_path_info = current_dir_info[1][path_index]
2253
 
                                if want_unversioned:
2254
 
                                    if current_path_info[2] == 'directory':
2255
 
                                        if self.target._directory_is_tree_reference(
2256
 
                                            current_path_info[0].decode('utf8')):
2257
 
                                            current_path_info = current_path_info[:2] + \
2258
 
                                                ('tree-reference',) + current_path_info[3:]
2259
 
                                    new_executable = bool(
2260
 
                                        stat.S_ISREG(current_path_info[3].st_mode)
2261
 
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
2262
 
                                    yield (None,
2263
 
                                        (None, utf8_decode(current_path_info[0])[0]),
2264
 
                                        True,
2265
 
                                        (False, False),
2266
 
                                        (None, None),
2267
 
                                        (None, utf8_decode(current_path_info[1])[0]),
2268
 
                                        (None, current_path_info[2]),
2269
 
                                        (None, new_executable))
2270
 
                                # dont descend into this unversioned path if it is
2271
 
                                # a dir
2272
 
                                if current_path_info[2] in ('directory',
2273
 
                                                            'tree-reference'):
2274
 
                                    del current_dir_info[1][path_index]
2275
 
                                    path_index -= 1
2276
 
                                path_index += 1
2277
 
 
2278
 
                        # This dir info has been handled, go to the next
2279
 
                        try:
2280
 
                            current_dir_info = dir_iterator.next()
2281
 
                        except StopIteration:
2282
 
                            current_dir_info = None
2283
 
                    else:
2284
 
                        # We have a dirblock entry for this location, but there
2285
 
                        # is no filesystem path for this. This is most likely
2286
 
                        # because a directory was removed from the disk.
2287
 
                        # We don't have to report the missing directory,
2288
 
                        # because that should have already been handled, but we
2289
 
                        # need to handle all of the files that are contained
2290
 
                        # within.
2291
 
                        for current_entry in current_block[1]:
2292
 
                            # entry referring to file not present on disk.
2293
 
                            # advance the entry only, after processing.
2294
 
                            result = _process_entry(current_entry, None)
2295
 
                            if result is not None:
2296
 
                                if result is not uninteresting:
2297
 
                                    yield result
2298
 
                        block_index +=1
2299
 
                        if (block_index < len(state._dirblocks) and
2300
 
                            osutils.is_inside(current_root,
2301
 
                                              state._dirblocks[block_index][0])):
2302
 
                            current_block = state._dirblocks[block_index]
2303
 
                        else:
2304
 
                            current_block = None
2305
 
                    continue
2306
 
                entry_index = 0
2307
 
                if current_block and entry_index < len(current_block[1]):
2308
 
                    current_entry = current_block[1][entry_index]
2309
 
                else:
2310
 
                    current_entry = None
2311
 
                advance_entry = True
2312
 
                path_index = 0
2313
 
                if current_dir_info and path_index < len(current_dir_info[1]):
2314
 
                    current_path_info = current_dir_info[1][path_index]
2315
 
                    if current_path_info[2] == 'directory':
2316
 
                        if self.target._directory_is_tree_reference(
2317
 
                            current_path_info[0].decode('utf8')):
2318
 
                            current_path_info = current_path_info[:2] + \
2319
 
                                ('tree-reference',) + current_path_info[3:]
2320
 
                else:
2321
 
                    current_path_info = None
2322
 
                advance_path = True
2323
 
                path_handled = False
2324
 
                while (current_entry is not None or
2325
 
                    current_path_info is not None):
2326
 
                    if current_entry is None:
2327
 
                        # the check for path_handled when the path is adnvaced
2328
 
                        # will yield this path if needed.
2329
 
                        pass
2330
 
                    elif current_path_info is None:
2331
 
                        # no path is fine: the per entry code will handle it.
2332
 
                        result = _process_entry(current_entry, current_path_info)
2333
 
                        if result is not None:
2334
 
                            if result is not uninteresting:
2335
 
                                yield result
2336
 
                    elif (current_entry[0][1] != current_path_info[1]
2337
 
                          or current_entry[1][target_index][0] in 'ar'):
2338
 
                        # The current path on disk doesn't match the dirblock
2339
 
                        # record. Either the dirblock is marked as absent, or
2340
 
                        # the file on disk is not present at all in the
2341
 
                        # dirblock. Either way, report about the dirblock
2342
 
                        # entry, and let other code handle the filesystem one.
2343
 
 
2344
 
                        # Compare the basename for these files to determine
2345
 
                        # which comes first
2346
 
                        if current_path_info[1] < current_entry[0][1]:
2347
 
                            # extra file on disk: pass for now, but only
2348
 
                            # increment the path, not the entry
2349
 
                            advance_entry = False
2350
 
                        else:
2351
 
                            # entry referring to file not present on disk.
2352
 
                            # advance the entry only, after processing.
2353
 
                            result = _process_entry(current_entry, None)
2354
 
                            if result is not None:
2355
 
                                if result is not uninteresting:
2356
 
                                    yield result
2357
 
                            advance_path = False
2358
 
                    else:
2359
 
                        result = _process_entry(current_entry, current_path_info)
2360
 
                        if result is not None:
2361
 
                            path_handled = True
2362
 
                            if result is not uninteresting:
2363
 
                                yield result
2364
 
                    if advance_entry and current_entry is not None:
2365
 
                        entry_index += 1
2366
 
                        if entry_index < len(current_block[1]):
2367
 
                            current_entry = current_block[1][entry_index]
2368
 
                        else:
2369
 
                            current_entry = None
2370
 
                    else:
2371
 
                        advance_entry = True # reset the advance flaga
2372
 
                    if advance_path and current_path_info is not None:
2373
 
                        if not path_handled:
2374
 
                            # unversioned in all regards
2375
 
                            if want_unversioned:
2376
 
                                new_executable = bool(
2377
 
                                    stat.S_ISREG(current_path_info[3].st_mode)
2378
 
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
2379
 
                                yield (None,
2380
 
                                    (None, utf8_decode(current_path_info[0])[0]),
2381
 
                                    True,
2382
 
                                    (False, False),
2383
 
                                    (None, None),
2384
 
                                    (None, utf8_decode(current_path_info[1])[0]),
2385
 
                                    (None, current_path_info[2]),
2386
 
                                    (None, new_executable))
2387
 
                            # dont descend into this unversioned path if it is
2388
 
                            # a dir
2389
 
                            if current_path_info[2] in ('directory'):
2390
 
                                del current_dir_info[1][path_index]
2391
 
                                path_index -= 1
2392
 
                        # dont descend the disk iterator into any tree 
2393
 
                        # paths.
2394
 
                        if current_path_info[2] == 'tree-reference':
2395
 
                            del current_dir_info[1][path_index]
2396
 
                            path_index -= 1
2397
 
                        path_index += 1
2398
 
                        if path_index < len(current_dir_info[1]):
2399
 
                            current_path_info = current_dir_info[1][path_index]
2400
 
                            if current_path_info[2] == 'directory':
2401
 
                                if self.target._directory_is_tree_reference(
2402
 
                                    current_path_info[0].decode('utf8')):
2403
 
                                    current_path_info = current_path_info[:2] + \
2404
 
                                        ('tree-reference',) + current_path_info[3:]
2405
 
                        else:
2406
 
                            current_path_info = None
2407
 
                        path_handled = False
2408
 
                    else:
2409
 
                        advance_path = True # reset the advance flagg.
2410
 
                if current_block is not None:
2411
 
                    block_index += 1
2412
 
                    if (block_index < len(state._dirblocks) and
2413
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2414
 
                        current_block = state._dirblocks[block_index]
2415
 
                    else:
2416
 
                        current_block = None
2417
 
                if current_dir_info is not None:
2418
 
                    try:
2419
 
                        current_dir_info = dir_iterator.next()
2420
 
                    except StopIteration:
2421
 
                        current_dir_info = None
2422
 
 
 
2091
        iter_changes = self.target._iter_changes(include_unchanged,
 
2092
            use_filesystem_for_exec, search_specific_files, state,
 
2093
            source_index, target_index, want_unversioned, self.target)
 
2094
        return iter_changes.iter_changes()
2423
2095
 
2424
2096
    @staticmethod
2425
2097
    def is_compatible(source, target):
2426
2098
        # the target must be a dirstate working tree
2427
 
        if not isinstance(target, WorkingTree4):
 
2099
        if not isinstance(target, DirStateWorkingTree):
2428
2100
            return False
2429
 
        # the source must be a revtreee or dirstate rev tree.
 
2101
        # the source must be a revtree or dirstate rev tree.
2430
2102
        if not isinstance(source,
2431
2103
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2432
2104
            return False
2433
2105
        # the source revid must be in the target dirstate
2434
 
        if not (source._revision_id == NULL_REVISION or
 
2106
        if not (source._revision_id == _mod_revision.NULL_REVISION or
2435
2107
            source._revision_id in target.get_parent_ids()):
2436
 
            # TODO: what about ghosts? it may well need to 
 
2108
            # TODO: what about ghosts? it may well need to
2437
2109
            # check for them explicitly.
2438
2110
            return False
2439
2111
        return True
2449
2121
 
2450
2122
    def convert(self, tree):
2451
2123
        # lock the control files not the tree, so that we dont get tree
2452
 
        # on-unlock behaviours, and so that noone else diddles with the 
 
2124
        # on-unlock behaviours, and so that noone else diddles with the
2453
2125
        # tree during upgrade.
2454
2126
        tree._control_files.lock_write()
2455
2127
        try:
2481
2153
 
2482
2154
    def update_format(self, tree):
2483
2155
        """Change the format marker."""
2484
 
        tree._control_files.put_utf8('format',
2485
 
            self.target_format.get_format_string())
 
2156
        tree._transport.put_bytes('format',
 
2157
            self.target_format.get_format_string(),
 
2158
            mode=tree.bzrdir._get_file_mode())
 
2159
 
 
2160
 
 
2161
class Converter4to5(object):
 
2162
    """Perform an in-place upgrade of format 4 to format 5 trees."""
 
2163
 
 
2164
    def __init__(self):
 
2165
        self.target_format = WorkingTreeFormat5()
 
2166
 
 
2167
    def convert(self, tree):
 
2168
        # lock the control files not the tree, so that we don't get tree
 
2169
        # on-unlock behaviours, and so that no-one else diddles with the
 
2170
        # tree during upgrade.
 
2171
        tree._control_files.lock_write()
 
2172
        try:
 
2173
            self.update_format(tree)
 
2174
        finally:
 
2175
            tree._control_files.unlock()
 
2176
 
 
2177
    def update_format(self, tree):
 
2178
        """Change the format marker."""
 
2179
        tree._transport.put_bytes('format',
 
2180
            self.target_format.get_format_string(),
 
2181
            mode=tree.bzrdir._get_file_mode())
 
2182
 
 
2183
 
 
2184
class Converter4or5to6(object):
 
2185
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
 
2186
 
 
2187
    def __init__(self):
 
2188
        self.target_format = WorkingTreeFormat6()
 
2189
 
 
2190
    def convert(self, tree):
 
2191
        # lock the control files not the tree, so that we don't get tree
 
2192
        # on-unlock behaviours, and so that no-one else diddles with the
 
2193
        # tree during upgrade.
 
2194
        tree._control_files.lock_write()
 
2195
        try:
 
2196
            self.init_custom_control_files(tree)
 
2197
            self.update_format(tree)
 
2198
        finally:
 
2199
            tree._control_files.unlock()
 
2200
 
 
2201
    def init_custom_control_files(self, tree):
 
2202
        """Initialize custom control files."""
 
2203
        tree._transport.put_bytes('views', '',
 
2204
            mode=tree.bzrdir._get_file_mode())
 
2205
 
 
2206
    def update_format(self, tree):
 
2207
        """Change the format marker."""
 
2208
        tree._transport.put_bytes('format',
 
2209
            self.target_format.get_format_string(),
 
2210
            mode=tree.bzrdir._get_file_mode())