/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Vincent Ladeuil
  • Date: 2008-01-05 22:09:47 UTC
  • mto: (3928.1.1 bzr.integration)
  • mto: This revision was merged to the branch mainline in revision 3929.
  • Revision ID: v.ladeuil+lp@free.fr-20080105220947-t2kymulzeqf1g5n5
Fix the server name in script and ssl files.

* bzrlib/tests/ssl_certs/create_ssls.py:
(ssl_params): We use localhost, not 127.0.0.1.

* bzrlib/tests/ssl_certs/server_without_pass.key,
bzrlib/tests/ssl_certs/server_with_pass.key,
bzrlib/tests/ssl_certs/server.csr,
bzrlib/tests/ssl_certs/server.crt:
Re-generated since the server name was wrong.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
38
46
    debug,
 
47
    delta,
39
48
    dirstate,
40
49
    errors,
41
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
55
    revision as _mod_revision,
44
56
    revisiontree,
 
57
    textui,
45
58
    trace,
46
59
    transform,
47
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
48
63
    )
49
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
50
66
import bzrlib.ui
51
67
""")
52
68
 
 
69
from bzrlib import symbol_versioning
53
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
56
 
from bzrlib.lock import LogicalLockResult
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
57
75
from bzrlib.mutabletree import needs_tree_write_lock
58
76
from bzrlib.osutils import (
59
77
    file_kind,
60
78
    isdir,
 
79
    normpath,
61
80
    pathjoin,
 
81
    rand_chars,
62
82
    realpath,
63
83
    safe_unicode,
 
84
    splitpath,
64
85
    )
65
 
from bzrlib.trace import mutter
 
86
from bzrlib.trace import mutter, note
66
87
from bzrlib.transport.local import LocalTransport
67
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
68
97
from bzrlib.tree import Tree
69
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
70
99
 
71
100
 
72
 
class DirStateWorkingTree(WorkingTree3):
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
73
120
    def __init__(self, basedir,
74
121
                 branch,
75
122
                 _control_files=None,
84
131
        """
85
132
        self._format = _format
86
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
87
136
        basedir = safe_unicode(basedir)
88
137
        mutter("opening working tree %r", basedir)
89
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
90
141
        self.basedir = realpath(basedir)
91
142
        # if branch is at our basedir and is a format 6 or less
92
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
93
146
        self._control_files = _control_files
94
 
        self._transport = self._control_files._transport
95
147
        self._dirty = None
96
148
        #-------------
97
149
        # during a read or write lock these objects are set, and are
101
153
        #-------------
102
154
        self._setup_directory_is_tree_reference()
103
155
        self._detect_case_handling()
104
 
        self._rules_searcher = None
105
 
        self.views = self._make_views()
106
 
        #--- allow tests to select the dirstate iter_changes implementation
107
 
        self._iter_changes = dirstate._process_entry
108
156
 
109
157
    @needs_tree_write_lock
110
158
    def _add(self, files, ids, kinds):
138
186
    @needs_tree_write_lock
139
187
    def add_reference(self, sub_tree):
140
188
        # use standard implementation, which calls back to self._add
141
 
        #
 
189
        # 
142
190
        # So we don't store the reference_revision in the working dirstate,
143
 
        # it's just recorded at the moment of commit.
 
191
        # it's just recorded at the moment of commit. 
144
192
        self._add_reference(sub_tree)
145
193
 
146
194
    def break_lock(self):
185
233
            WorkingTree3._comparison_data(self, entry, path)
186
234
        # it looks like a plain directory, but it's really a reference -- see
187
235
        # also kind()
188
 
        if (self._repo_supports_tree_reference and kind == 'directory'
189
 
            and entry is not None and entry.kind == 'tree-reference'):
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
190
239
            kind = 'tree-reference'
191
240
        return kind, executable, stat_value
192
241
 
218
267
            return self._dirstate
219
268
        local_path = self.bzrdir.get_workingtree_transport(None
220
269
            ).local_abspath('dirstate')
221
 
        self._dirstate = dirstate.DirState.on_file(local_path,
222
 
            self._sha1_provider())
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
223
271
        return self._dirstate
224
272
 
225
 
    def _sha1_provider(self):
226
 
        """A function that returns a SHA1Provider suitable for this tree.
227
 
 
228
 
        :return: None if content filtering is not supported by this tree.
229
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
230
 
          form of files, i.e. after read filters are applied.
231
 
        """
232
 
        if self.supports_content_filtering():
233
 
            return ContentFilterAwareSHA1Provider(self)
234
 
        else:
235
 
            return None
236
 
 
237
273
    def filter_unversioned_files(self, paths):
238
274
        """Filter out paths that are versioned.
239
275
 
271
307
 
272
308
    def _generate_inventory(self):
273
309
        """Create and set self.inventory from the dirstate object.
274
 
 
 
310
        
275
311
        This is relatively expensive: we have to walk the entire dirstate.
276
312
        Ideally we would not, and can deprecate this function.
277
313
        """
281
317
        state._read_dirblocks_if_needed()
282
318
        root_key, current_entry = self._get_entry(path='')
283
319
        current_id = root_key[2]
284
 
        if not (current_entry[0][0] == 'd'): # directory
285
 
            raise AssertionError(current_entry)
 
320
        assert current_entry[0][0] == 'd' # directory
286
321
        inv = Inventory(root_id=current_id)
287
322
        # Turn some things into local variables
288
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
321
356
                    # add this entry to the parent map.
322
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
323
358
                elif kind == 'tree-reference':
324
 
                    if not self._repo_supports_tree_reference:
325
 
                        raise errors.UnsupportedOperation(
326
 
                            self._generate_inventory,
327
 
                            self.branch.repository)
 
359
                    assert self._repo_supports_tree_reference, \
 
360
                        "repository of %r " \
 
361
                        "doesn't support tree references " \
 
362
                        "required by entry %r" \
 
363
                        % (self, name)
328
364
                    inv_entry.reference_revision = link_or_sha1 or None
329
365
                elif kind != 'symlink':
330
366
                    raise AssertionError("unknown kind %r" % kind)
331
367
                # These checks cost us around 40ms on a 55k entry tree
332
 
                if file_id in inv_byid:
333
 
                    raise AssertionError('file_id %s already in'
334
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
335
 
                if name_unicode in parent_ie.children:
336
 
                    raise AssertionError('name %r already in parent'
337
 
                        % (name_unicode,))
 
368
                assert file_id not in inv_byid, ('file_id %s already in'
 
369
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
370
                assert name_unicode not in parent_ie.children
338
371
                inv_byid[file_id] = inv_entry
339
372
                parent_ie.children[name_unicode] = inv_entry
340
373
        self._inventory = inv
345
378
        If either file_id or path is supplied, it is used as the key to lookup.
346
379
        If both are supplied, the fastest lookup is used, and an error is
347
380
        raised if they do not both point at the same row.
348
 
 
 
381
        
349
382
        :param file_id: An optional unicode file_id to be looked up.
350
383
        :param path: An optional unicode path to be looked up.
351
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
375
408
                    return None
376
409
                else:
377
410
                    raise
378
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
379
 
            stat_value=stat_value)
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
380
413
        if entry[1][0][0] == 'f':
381
 
            if link_or_sha1 is None:
382
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
383
 
                try:
384
 
                    sha1 = osutils.sha_file(file_obj)
385
 
                finally:
386
 
                    file_obj.close()
387
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
388
 
                return sha1
389
 
            else:
390
 
                return link_or_sha1
 
414
            return link_or_sha1
391
415
        return None
392
416
 
393
417
    def _get_inventory(self):
407
431
    @needs_read_lock
408
432
    def get_parent_ids(self):
409
433
        """See Tree.get_parent_ids.
410
 
 
 
434
        
411
435
        This implementation requests the ids list from the dirstate file.
412
436
        """
413
437
        return self.current_dirstate().get_parent_ids()
435
459
        return osutils.lexists(pathjoin(
436
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
437
461
 
438
 
    def has_or_had_id(self, file_id):
439
 
        state = self.current_dirstate()
440
 
        row, parents = self._get_entry(file_id=file_id)
441
 
        return row is not None
442
 
 
443
462
    @needs_read_lock
444
463
    def id2path(self, file_id):
445
464
        "Convert a file-id to a path."
475
494
 
476
495
            Note: The caller is expected to take a read-lock before calling this.
477
496
            """
478
 
            self._must_be_locked()
479
497
            if not path:
480
498
                path = self.id2path(file_id)
481
499
            mode = os.lstat(self.abspath(path)).st_mode
482
500
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
483
501
 
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
 
 
494
502
    @needs_read_lock
495
503
    def __iter__(self):
496
504
        """Iterate through file_ids for this tree.
509
517
        return iter(result)
510
518
 
511
519
    def iter_references(self):
512
 
        if not self._repo_supports_tree_reference:
513
 
            # When the repo doesn't support references, we will have nothing to
514
 
            # return
515
 
            return
516
520
        for key, tree_details in self.current_dirstate()._iter_entries():
517
521
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
518
522
                # not relevant to the working tree
520
524
            if not key[1]:
521
525
                # the root is not a reference.
522
526
                continue
523
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
527
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
524
528
            try:
525
 
                if self._kind(relpath) == 'tree-reference':
526
 
                    yield relpath, key[2]
 
529
                if self._kind(path) == 'tree-reference':
 
530
                    yield path, key[2]
527
531
            except errors.NoSuchFile:
528
532
                # path is missing on disk.
529
533
                continue
530
534
 
531
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
532
 
        """See MutableTree._observed_sha1."""
533
 
        state = self.current_dirstate()
534
 
        entry = self._get_entry(file_id=file_id, path=path)
535
 
        state._observed_sha1(entry, sha1, statvalue)
536
 
 
537
535
    def kind(self, file_id):
538
536
        """Return the kind of a file.
539
537
 
543
541
        Note: The caller is expected to take a read-lock before calling this.
544
542
        """
545
543
        relpath = self.id2path(file_id)
546
 
        if relpath is None:
547
 
            raise AssertionError(
548
 
                "path for id {%s} is None!" % file_id)
 
544
        assert relpath != None, \
 
545
            "path for id {%s} is None!" % file_id
549
546
        return self._kind(relpath)
550
547
 
551
548
    def _kind(self, relpath):
552
549
        abspath = self.abspath(relpath)
553
550
        kind = file_kind(abspath)
554
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
555
 
            entry = self._get_entry(path=relpath)
556
 
            if entry[1] is not None:
557
 
                if entry[1][0][0] == 't':
558
 
                    kind = 'tree-reference'
 
551
        if (self._repo_supports_tree_reference and
 
552
            kind == 'directory' and
 
553
            self._directory_is_tree_reference(relpath)):
 
554
            kind = 'tree-reference'
559
555
        return kind
560
556
 
561
557
    @needs_read_lock
568
564
            return _mod_revision.NULL_REVISION
569
565
 
570
566
    def lock_read(self):
571
 
        """See Branch.lock_read, and WorkingTree.unlock.
572
 
 
573
 
        :return: A bzrlib.lock.LogicalLockResult.
574
 
        """
 
567
        """See Branch.lock_read, and WorkingTree.unlock."""
575
568
        self.branch.lock_read()
576
569
        try:
577
570
            self._control_files.lock_read()
590
583
        except:
591
584
            self.branch.unlock()
592
585
            raise
593
 
        return LogicalLockResult(self.unlock)
594
586
 
595
587
    def _lock_self_write(self):
596
588
        """This should be called after the branch is locked."""
611
603
        except:
612
604
            self.branch.unlock()
613
605
            raise
614
 
        return LogicalLockResult(self.unlock)
615
606
 
616
607
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: A bzrlib.lock.LogicalLockResult.
620
 
        """
 
608
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
621
609
        self.branch.lock_read()
622
 
        return self._lock_self_write()
 
610
        self._lock_self_write()
623
611
 
624
612
    def lock_write(self):
625
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
626
 
 
627
 
        :return: A bzrlib.lock.LogicalLockResult.
628
 
        """
 
613
        """See MutableTree.lock_write, and WorkingTree.unlock."""
629
614
        self.branch.lock_write()
630
 
        return self._lock_self_write()
 
615
        self._lock_self_write()
631
616
 
632
617
    @needs_tree_write_lock
633
618
    def move(self, from_paths, to_dir, after=False):
635
620
        result = []
636
621
        if not from_paths:
637
622
            return result
 
623
 
638
624
        state = self.current_dirstate()
639
 
        if isinstance(from_paths, basestring):
640
 
            raise ValueError()
 
625
 
 
626
        assert not isinstance(from_paths, basestring)
641
627
        to_dir_utf8 = to_dir.encode('utf8')
642
628
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
643
629
        id_index = state._get_id_index()
665
651
        if self._inventory is not None:
666
652
            update_inventory = True
667
653
            inv = self.inventory
 
654
            to_dir_ie = inv[to_dir_id]
668
655
            to_dir_id = to_entry[0][2]
669
 
            to_dir_ie = inv[to_dir_id]
670
656
        else:
671
657
            update_inventory = False
672
658
 
703
689
            from_entry = self._get_entry(path=from_rel)
704
690
            if from_entry == (None, None):
705
691
                raise errors.BzrMoveFailedError(from_rel,to_dir,
706
 
                    errors.NotVersionedError(path=from_rel))
 
692
                    errors.NotVersionedError(path=str(from_rel)))
707
693
 
708
694
            from_id = from_entry[0][2]
709
695
            to_rel = pathjoin(to_dir, from_tail)
798
784
                if minikind == 'd':
799
785
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
800
786
                        """Recursively update all entries in this dirblock."""
801
 
                        if from_dir == '':
802
 
                            raise AssertionError("renaming root not supported")
 
787
                        assert from_dir != '', "renaming root not supported"
803
788
                        from_key = (from_dir, '')
804
789
                        from_block_idx, present = \
805
790
                            state._find_block_index_from_key(from_key)
818
803
 
819
804
                        # Grab a copy since move_one may update the list.
820
805
                        for entry in from_block[1][:]:
821
 
                            if not (entry[0][0] == from_dir):
822
 
                                raise AssertionError()
 
806
                            assert entry[0][0] == from_dir
823
807
                            cur_details = entry[1][0]
824
808
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
825
809
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
884
868
        for tree in trees:
885
869
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
886
870
                parents):
887
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
888
 
                    trees, require_versioned)
 
871
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
889
872
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
890
873
        # -- make all paths utf8 --
891
874
        paths_utf8 = set()
952
935
                raise errors.PathsNotVersionedError(paths)
953
936
        # -- remove redundancy in supplied paths to prevent over-scanning --
954
937
        search_paths = osutils.minimum_path_selection(paths)
955
 
        # sketch:
 
938
        # sketch: 
956
939
        # for all search_indexs in each path at or under each element of
957
940
        # search_paths, if the detail is relocated: add the id, and add the
958
941
        # relocated path as one to search if its not searched already. If the
1014
997
 
1015
998
    def read_working_inventory(self):
1016
999
        """Read the working inventory.
1017
 
 
 
1000
        
1018
1001
        This is a meaningless operation for dirstate, but we obey it anyhow.
1019
1002
        """
1020
1003
        return self.inventory
1038
1021
    def set_last_revision(self, new_revision):
1039
1022
        """Change the last revision in the working tree."""
1040
1023
        parents = self.get_parent_ids()
1041
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1042
 
            if len(parents) >= 2:
1043
 
                raise AssertionError(
1044
 
                    "setting the last parent to none with a pending merge is "
1045
 
                    "unsupported.")
 
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.")
1046
1028
            self.set_parent_ids([])
1047
1029
        else:
1048
1030
            self.set_parent_ids([new_revision] + parents[1:],
1051
1033
    @needs_tree_write_lock
1052
1034
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1053
1035
        """Set the parent ids to revision_ids.
1054
 
 
 
1036
        
1055
1037
        See also set_parent_trees. This api will try to retrieve the tree data
1056
1038
        for each element of revision_ids from the trees repository. If you have
1057
1039
        tree data already available, it is more efficient to use
1089
1071
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1090
1072
        real_trees = []
1091
1073
        ghosts = []
1092
 
 
1093
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1094
 
        graph = self.branch.repository.get_graph()
1095
 
        heads = graph.heads(parent_ids)
1096
 
        accepted_revisions = set()
1097
 
 
1098
1074
        # convert absent trees to the null tree, which we convert back to
1099
1075
        # missing on access.
1100
1076
        for rev_id, tree in parents_list:
1101
 
            if len(accepted_revisions) > 0:
1102
 
                # we always accept the first tree
1103
 
                if rev_id in accepted_revisions or rev_id not in heads:
1104
 
                    # We have already included either this tree, or its
1105
 
                    # descendent, so we skip it.
1106
 
                    continue
1107
1077
            _mod_revision.check_not_reserved_id(rev_id)
1108
1078
            if tree is not None:
1109
1079
                real_trees.append((rev_id, tree))
1110
1080
            else:
1111
1081
                real_trees.append((rev_id,
1112
 
                    self.branch.repository.revision_tree(
1113
 
                        _mod_revision.NULL_REVISION)))
 
1082
                    self.branch.repository.revision_tree(None)))
1114
1083
                ghosts.append(rev_id)
1115
 
            accepted_revisions.add(rev_id)
1116
1084
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1117
1085
        self._make_dirty(reset_inventory=False)
1118
1086
 
1211
1179
                # just forget the whole block.
1212
1180
                entry_index = 0
1213
1181
                while entry_index < len(block[1]):
 
1182
                    # Mark this file id as having been removed
1214
1183
                    entry = block[1][entry_index]
1215
 
                    if entry[1][0][0] in 'ar':
1216
 
                        # don't remove absent or renamed entries
 
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)):
1217
1188
                        entry_index += 1
1218
 
                    else:
1219
 
                        # Mark this file id as having been removed
1220
 
                        ids_to_unversion.discard(entry[0][2])
1221
 
                        if not state._make_absent(entry):
1222
 
                            # The block has not shrunk.
1223
 
                            entry_index += 1
1224
1189
                # go to the next block. (At the moment we dont delete empty
1225
1190
                # dirblocks)
1226
1191
                block_index += 1
1249
1214
            for file_id in file_ids:
1250
1215
                self._inventory.remove_recursive_id(file_id)
1251
1216
 
1252
 
    @needs_tree_write_lock
1253
 
    def rename_one(self, from_rel, to_rel, after=False):
1254
 
        """See WorkingTree.rename_one"""
1255
 
        self.flush()
1256
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1257
 
 
1258
 
    @needs_tree_write_lock
1259
 
    def apply_inventory_delta(self, changes):
1260
 
        """See MutableTree.apply_inventory_delta"""
1261
 
        state = self.current_dirstate()
1262
 
        state.update_by_delta(changes)
1263
 
        self._make_dirty(reset_inventory=True)
1264
 
 
1265
1217
    def update_basis_by_delta(self, new_revid, delta):
1266
1218
        """See MutableTree.update_basis_by_delta."""
1267
 
        if self.last_revision() == new_revid:
1268
 
            raise AssertionError()
 
1219
        assert self.last_revision() != new_revid
1269
1220
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1270
1221
 
1271
1222
    @needs_read_lock
1275
1226
    @needs_tree_write_lock
1276
1227
    def _write_inventory(self, inv):
1277
1228
        """Write inventory as the current inventory."""
1278
 
        if self._dirty:
1279
 
            raise AssertionError("attempting to write an inventory when the "
1280
 
                "dirstate is dirty will lose pending changes")
1281
 
        had_inventory = self._inventory is not None
1282
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1283
 
        # working inventory. We do this because self.inventory may be inv, or
1284
 
        # may have been modified, and either case would prevent a clean delta
1285
 
        # being created.
1286
 
        self._inventory = None
1287
 
        # generate a delta,
1288
 
        delta = inv._make_delta(self.inventory)
1289
 
        # and apply it.
1290
 
        self.apply_inventory_delta(delta)
1291
 
        if had_inventory:
 
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:
1292
1234
            self._inventory = inv
1293
1235
        self.flush()
1294
1236
 
1295
1237
 
1296
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1297
 
 
1298
 
    def __init__(self, tree):
1299
 
        self.tree = tree
1300
 
 
1301
 
    def sha1(self, abspath):
1302
 
        """See dirstate.SHA1Provider.sha1()."""
1303
 
        filters = self.tree._content_filter_stack(
1304
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1305
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1306
 
 
1307
 
    def stat_and_sha1(self, abspath):
1308
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1309
 
        filters = self.tree._content_filter_stack(
1310
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1311
 
        file_obj = file(abspath, 'rb', 65000)
1312
 
        try:
1313
 
            statvalue = os.fstat(file_obj.fileno())
1314
 
            if filters:
1315
 
                file_obj = filtered_input_file(file_obj, filters)
1316
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1317
 
        finally:
1318
 
            file_obj.close()
1319
 
        return statvalue, sha1
1320
 
 
1321
 
 
1322
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1323
 
    """Dirstate working tree that supports content filtering.
1324
 
 
1325
 
    The dirstate holds the hash and size of the canonical form of the file, 
1326
 
    and most methods must return that.
1327
 
    """
1328
 
 
1329
 
    def _file_content_summary(self, path, stat_result):
1330
 
        # This is to support the somewhat obsolete path_content_summary method
1331
 
        # with content filtering: see
1332
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1333
 
        #
1334
 
        # If the dirstate cache is up to date and knows the hash and size,
1335
 
        # return that.
1336
 
        # Otherwise if there are no content filters, return the on-disk size
1337
 
        # and leave the hash blank.
1338
 
        # Otherwise, read and filter the on-disk file and use its size and
1339
 
        # hash.
1340
 
        #
1341
 
        # The dirstate doesn't store the size of the canonical form so we
1342
 
        # can't trust it for content-filtered trees.  We just return None.
1343
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1344
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1345
 
        return ('file', None, executable, dirstate_sha1)
1346
 
 
1347
 
 
1348
 
class WorkingTree4(DirStateWorkingTree):
1349
 
    """This is the Format 4 working tree.
1350
 
 
1351
 
    This differs from WorkingTree3 by:
1352
 
     - Having a consolidated internal dirstate, stored in a
1353
 
       randomly-accessible sorted file on disk.
1354
 
     - Not having a regular inventory attribute.  One can be synthesized
1355
 
       on demand but this is expensive and should be avoided.
1356
 
 
1357
 
    This is new in bzr 0.15.
1358
 
    """
1359
 
 
1360
 
 
1361
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1362
 
    """This is the Format 5 working tree.
1363
 
 
1364
 
    This differs from WorkingTree4 by:
1365
 
     - Supporting content filtering.
1366
 
 
1367
 
    This is new in bzr 1.11.
1368
 
    """
1369
 
 
1370
 
 
1371
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1372
 
    """This is the Format 6 working tree.
1373
 
 
1374
 
    This differs from WorkingTree5 by:
1375
 
     - Supporting a current view that may mask the set of files in a tree
1376
 
       impacted by most user operations.
1377
 
 
1378
 
    This is new in bzr 1.14.
1379
 
    """
1380
 
 
1381
 
    def _make_views(self):
1382
 
        return views.PathBasedViews(self)
1383
 
 
1384
 
 
1385
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
 
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"
1386
1259
 
1387
1260
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1388
 
                   accelerator_tree=None, hardlink=False):
 
1261
                   accelerator_tree=None):
1389
1262
        """See WorkingTreeFormat.initialize().
1390
1263
 
1391
1264
        :param revision_id: allows creating a working tree at a different
1394
1267
            contents more quickly than the revision tree, i.e. a workingtree.
1395
1268
            The revision tree will be used for cases where accelerator_tree's
1396
1269
            content is different.
1397
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1398
 
            where possible.
1399
1270
 
1400
1271
        These trees get an initial random root id, if their repository supports
1401
1272
        rich root data, TREE_ROOT otherwise.
1406
1277
        control_files = self._open_control_files(a_bzrdir)
1407
1278
        control_files.create_lock()
1408
1279
        control_files.lock_write()
1409
 
        transport.put_bytes('format', self.get_format_string(),
1410
 
            mode=a_bzrdir._get_file_mode())
 
1280
        control_files.put_utf8('format', self.get_format_string())
1411
1281
        if from_branch is not None:
1412
1282
            branch = from_branch
1413
1283
        else:
1419
1289
        state = dirstate.DirState.initialize(local_path)
1420
1290
        state.unlock()
1421
1291
        del state
1422
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1292
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1423
1293
                         branch,
1424
1294
                         _format=self,
1425
1295
                         _bzrdir=a_bzrdir,
1427
1297
        wt._new_tree()
1428
1298
        wt.lock_tree_write()
1429
1299
        try:
1430
 
            self._init_custom_control_files(wt)
1431
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1300
            if revision_id in (None, NULL_REVISION):
1432
1301
                if branch.repository.supports_rich_root():
1433
1302
                    wt._set_root_id(generate_ids.gen_root_id())
1434
1303
                else:
1445
1314
                    pass
1446
1315
            if basis is None:
1447
1316
                basis = branch.repository.revision_tree(revision_id)
1448
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1317
            if revision_id == NULL_REVISION:
1449
1318
                parents_list = []
1450
1319
            else:
1451
1320
                parents_list = [(revision_id, basis)]
1452
1321
            basis.lock_read()
1453
 
            try:
1454
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
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)
1455
1329
                wt.flush()
1456
 
                # if the basis has a root id we have to use that; otherwise we
1457
 
                # use a new random one
1458
 
                basis_root_id = basis.get_root_id()
1459
 
                if basis_root_id is not None:
1460
 
                    wt._set_root_id(basis_root_id)
1461
 
                    wt.flush()
1462
 
                if wt.supports_content_filtering():
1463
 
                    # The original tree may not have the same content filters
1464
 
                    # applied so we can't safely build the inventory delta from
1465
 
                    # the source tree.
1466
 
                    delta_from_tree = False
1467
 
                else:
1468
 
                    delta_from_tree = True
1469
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1470
 
                # because wt4.apply_inventory_delta does not mutate the input
1471
 
                # inventory entries.
1472
 
                transform.build_tree(basis, wt, accelerator_tree,
1473
 
                                     hardlink=hardlink,
1474
 
                                     delta_from_tree=delta_from_tree)
1475
 
            finally:
1476
 
                basis.unlock()
 
1330
            transform.build_tree(basis, wt, accelerator_tree)
 
1331
            basis.unlock()
1477
1332
        finally:
1478
1333
            control_files.unlock()
1479
1334
            wt.unlock()
1480
1335
        return wt
1481
1336
 
1482
 
    def _init_custom_control_files(self, wt):
1483
 
        """Subclasses with custom control files should override this method.
1484
 
 
1485
 
        The working tree and control files are locked for writing when this
1486
 
        method is called.
1487
 
 
1488
 
        :param wt: the WorkingTree object
1489
 
        """
1490
 
 
1491
1337
    def _open(self, a_bzrdir, control_files):
1492
1338
        """Open the tree itself.
1493
1339
 
1494
1340
        :param a_bzrdir: the dir for the tree.
1495
1341
        :param control_files: the control files for the tree.
1496
1342
        """
1497
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1343
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1498
1344
                           branch=a_bzrdir.open_branch(),
1499
1345
                           _format=self,
1500
1346
                           _bzrdir=a_bzrdir,
1501
1347
                           _control_files=control_files)
1502
1348
 
1503
1349
    def __get_matchingbzrdir(self):
1504
 
        return self._get_matchingbzrdir()
1505
 
 
1506
 
    def _get_matchingbzrdir(self):
1507
 
        """Overrideable method to get a bzrdir for testing."""
1508
1350
        # please test against something that will let us do tree references
1509
1351
        return bzrdir.format_registry.make_bzrdir(
1510
1352
            'dirstate-with-subtree')
1512
1354
    _matchingbzrdir = property(__get_matchingbzrdir)
1513
1355
 
1514
1356
 
1515
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1516
 
    """The first consolidated dirstate working tree format.
1517
 
 
1518
 
    This format:
1519
 
        - exists within a metadir controlling .bzr
1520
 
        - includes an explicit version marker for the workingtree control
1521
 
          files, separate from the BzrDir format
1522
 
        - modifies the hash cache format
1523
 
        - is new in bzr 0.15
1524
 
        - uses a LockDir to guard access to it.
1525
 
    """
1526
 
 
1527
 
    upgrade_recommended = False
1528
 
 
1529
 
    _tree_class = WorkingTree4
1530
 
 
1531
 
    def get_format_string(self):
1532
 
        """See WorkingTreeFormat.get_format_string()."""
1533
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1534
 
 
1535
 
    def get_format_description(self):
1536
 
        """See WorkingTreeFormat.get_format_description()."""
1537
 
        return "Working tree format 4"
1538
 
 
1539
 
 
1540
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1541
 
    """WorkingTree format supporting content filtering.
1542
 
    """
1543
 
 
1544
 
    upgrade_recommended = False
1545
 
 
1546
 
    _tree_class = WorkingTree5
1547
 
 
1548
 
    def get_format_string(self):
1549
 
        """See WorkingTreeFormat.get_format_string()."""
1550
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1551
 
 
1552
 
    def get_format_description(self):
1553
 
        """See WorkingTreeFormat.get_format_description()."""
1554
 
        return "Working tree format 5"
1555
 
 
1556
 
    def supports_content_filtering(self):
1557
 
        return True
1558
 
 
1559
 
 
1560
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1561
 
    """WorkingTree format supporting views.
1562
 
    """
1563
 
 
1564
 
    upgrade_recommended = False
1565
 
 
1566
 
    _tree_class = WorkingTree6
1567
 
 
1568
 
    def get_format_string(self):
1569
 
        """See WorkingTreeFormat.get_format_string()."""
1570
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1571
 
 
1572
 
    def get_format_description(self):
1573
 
        """See WorkingTreeFormat.get_format_description()."""
1574
 
        return "Working tree format 6"
1575
 
 
1576
 
    def _init_custom_control_files(self, wt):
1577
 
        """Subclasses with custom control files should override this method."""
1578
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1579
 
 
1580
 
    def supports_content_filtering(self):
1581
 
        return True
1582
 
 
1583
 
    def supports_views(self):
1584
 
        return True
1585
 
 
1586
 
 
1587
1357
class DirStateRevisionTree(Tree):
1588
 
    """A revision tree pulling the inventory from a dirstate.
1589
 
    
1590
 
    Note that this is one of the historical (ie revision) trees cached in the
1591
 
    dirstate for easy access, not the workingtree.
1592
 
    """
 
1358
    """A revision tree pulling the inventory from a dirstate."""
1593
1359
 
1594
1360
    def __init__(self, dirstate, revision_id, repository):
1595
1361
        self._dirstate = dirstate
1598
1364
        self._inventory = None
1599
1365
        self._locked = 0
1600
1366
        self._dirstate_locked = False
1601
 
        self._repo_supports_tree_reference = getattr(
1602
 
            repository._format, "supports_tree_reference",
1603
 
            False)
1604
1367
 
1605
1368
    def __repr__(self):
1606
1369
        return "<%s of %s in %s>" % \
1609
1372
    def annotate_iter(self, file_id,
1610
1373
                      default_revision=_mod_revision.CURRENT_REVISION):
1611
1374
        """See Tree.annotate_iter"""
1612
 
        text_key = (file_id, self.inventory[file_id].revision)
1613
 
        annotations = self._repository.texts.annotate(text_key)
1614
 
        return [(key[-1], line) for (key, line) in annotations]
 
1375
        w = self._get_weave(file_id)
 
1376
        return w.annotate_iter(self.inventory[file_id].revision)
1615
1377
 
1616
1378
    def _get_ancestors(self, default_revision):
1617
1379
        return set(self._repository.get_ancestry(self._revision_id,
1646
1408
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1647
1409
        return path_utf8.decode('utf8')
1648
1410
 
1649
 
    def iter_references(self):
1650
 
        if not self._repo_supports_tree_reference:
1651
 
            # When the repo doesn't support references, we will have nothing to
1652
 
            # return
1653
 
            return iter([])
1654
 
        # Otherwise, fall back to the default implementation
1655
 
        return super(DirStateRevisionTree, self).iter_references()
1656
 
 
1657
1411
    def _get_parent_index(self):
1658
1412
        """Return the index in the dirstate referenced by this tree."""
1659
1413
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1664
1418
        If either file_id or path is supplied, it is used as the key to lookup.
1665
1419
        If both are supplied, the fastest lookup is used, and an error is
1666
1420
        raised if they do not both point at the same row.
1667
 
 
 
1421
        
1668
1422
        :param file_id: An optional unicode file_id to be looked up.
1669
1423
        :param path: An optional unicode path to be looked up.
1670
1424
        :return: The dirstate row tuple for path/file_id, or (None, None)
1684
1438
 
1685
1439
        This is relatively expensive: we have to walk the entire dirstate.
1686
1440
        """
1687
 
        if not self._locked:
1688
 
            raise AssertionError(
1689
 
                'cannot generate inventory of an unlocked '
1690
 
                'dirstate revision tree')
 
1441
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1442
            'dirstate revision tree'
1691
1443
        # separate call for profiling - makes it clear where the costs are.
1692
1444
        self._dirstate._read_dirblocks_if_needed()
1693
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1694
 
            raise AssertionError(
1695
 
                'parent %s has disappeared from %s' % (
1696
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
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())
1697
1448
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1698
1449
        # This is identical now to the WorkingTree _generate_inventory except
1699
1450
        # for the tree index use.
1700
1451
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1701
1452
        current_id = root_key[2]
1702
 
        if current_entry[parent_index][0] != 'd':
1703
 
            raise AssertionError()
 
1453
        assert current_entry[parent_index][0] == 'd'
1704
1454
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1705
1455
        inv.root.revision = current_entry[parent_index][4]
1706
1456
        # Turn some things into local variables
1746
1496
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1747
1497
                            % entry)
1748
1498
                # These checks cost us around 40ms on a 55k entry tree
1749
 
                if file_id in inv_byid:
1750
 
                    raise AssertionError('file_id %s already in'
1751
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1752
 
                if name_unicode in parent_ie.children:
1753
 
                    raise AssertionError('name %r already in parent'
1754
 
                        % (name_unicode,))
 
1499
                assert file_id not in inv_byid
 
1500
                assert name_unicode not in parent_ie.children
1755
1501
                inv_byid[file_id] = inv_entry
1756
1502
                parent_ie.children[name_unicode] = inv_entry
1757
1503
        self._inventory = inv
1767
1513
            return None
1768
1514
        parent_index = self._get_parent_index()
1769
1515
        last_changed_revision = entry[1][parent_index][4]
1770
 
        try:
1771
 
            rev = self._repository.get_revision(last_changed_revision)
1772
 
        except errors.NoSuchRevision:
1773
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1774
 
        return rev.timestamp
 
1516
        return self._repository.get_revision(last_changed_revision).timestamp
1775
1517
 
1776
1518
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1777
1519
        entry = self._get_entry(file_id=file_id, path=path)
1781
1523
            return parent_details[1]
1782
1524
        return None
1783
1525
 
 
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
 
1784
1534
    def get_file(self, file_id, path=None):
1785
1535
        return StringIO(self.get_file_text(file_id))
1786
1536
 
 
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
 
1787
1543
    def get_file_size(self, file_id):
1788
 
        """See Tree.get_file_size"""
1789
1544
        return self.inventory[file_id].text_size
1790
1545
 
1791
 
    def get_file_text(self, file_id, path=None):
1792
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1793
 
        return ''.join(content)
 
1546
    def get_file_text(self, file_id):
 
1547
        return ''.join(self.get_file_lines(file_id))
1794
1548
 
1795
1549
    def get_reference_revision(self, file_id, path=None):
1796
1550
        return self.inventory[file_id].reference_revision
1815
1569
        if entry[1][parent_index][0] != 'l':
1816
1570
            return None
1817
1571
        else:
1818
 
            target = entry[1][parent_index][1]
1819
 
            target = target.decode('utf8')
1820
 
            return target
 
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]
1821
1576
 
1822
1577
    def get_revision_id(self):
1823
1578
        """Return the revision id for this tree."""
1842
1597
 
1843
1598
    def kind(self, file_id):
1844
1599
        entry = self._get_entry(file_id=file_id)[1]
1845
 
        if entry is None:
 
1600
        if entry == None:
1846
1601
            raise errors.NoSuchId(tree=self, file_id=file_id)
1847
 
        parent_index = self._get_parent_index()
1848
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1849
 
 
1850
 
    def stored_kind(self, file_id):
1851
 
        """See Tree.stored_kind"""
1852
 
        return self.kind(file_id)
 
1602
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1853
1603
 
1854
1604
    def path_content_summary(self, path):
1855
1605
        """See Tree.path_content_summary."""
1871
1621
            return None
1872
1622
        return ie.executable
1873
1623
 
1874
 
    def is_locked(self):
1875
 
        return self._locked
1876
 
 
1877
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1624
    def list_files(self, include_root=False):
1878
1625
        # We use a standard implementation, because DirStateRevisionTree is
1879
1626
        # dealing with one of the parents of the current state
1880
1627
        inv = self._get_inventory()
1881
 
        if from_dir is None:
1882
 
            from_dir_id = None
1883
 
        else:
1884
 
            from_dir_id = inv.path2id(from_dir)
1885
 
            if from_dir_id is None:
1886
 
                # Directory not versioned
1887
 
                return
1888
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1889
 
        if inv.root is not None and not include_root and from_dir is None:
 
1628
        entries = inv.iter_entries()
 
1629
        if self.inventory.root is not None and not include_root:
1890
1630
            entries.next()
1891
1631
        for path, entry in entries:
1892
1632
            yield path, 'V', entry.kind, entry.file_id, entry
1893
1633
 
1894
1634
    def lock_read(self):
1895
 
        """Lock the tree for a set of operations.
1896
 
 
1897
 
        :return: A bzrlib.lock.LogicalLockResult.
1898
 
        """
 
1635
        """Lock the tree for a set of operations."""
1899
1636
        if not self._locked:
1900
1637
            self._repository.lock_read()
1901
1638
            if self._dirstate._lock_token is None:
1902
1639
                self._dirstate.lock_read()
1903
1640
                self._dirstate_locked = True
1904
1641
        self._locked += 1
1905
 
        return LogicalLockResult(self.unlock)
1906
1642
 
1907
1643
    def _must_be_locked(self):
1908
1644
        if not self._locked:
1929
1665
                self._dirstate_locked = False
1930
1666
            self._repository.unlock()
1931
1667
 
1932
 
    @needs_read_lock
1933
 
    def supports_tree_reference(self):
1934
 
        return self._repo_supports_tree_reference
1935
 
 
1936
1668
    def walkdirs(self, prefix=""):
1937
1669
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1938
 
        # implementation based on an inventory.
 
1670
        # implementation based on an inventory.  
1939
1671
        # This should be cleaned up to use the much faster Dirstate code
1940
1672
        # So for now, we just build up the parent inventory, and extract
1941
1673
        # it the same way RevisionTree does.
1970
1702
 
1971
1703
class InterDirStateTree(InterTree):
1972
1704
    """Fast path optimiser for changes_from with dirstate trees.
1973
 
 
1974
 
    This is used only when both trees are in the dirstate working file, and
1975
 
    the source is any parent within the dirstate, and the destination is
 
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 
1976
1708
    the current working tree of the same dirstate.
1977
1709
    """
1978
1710
    # this could be generalized to allow comparisons between any trees in the
1991
1723
        target.set_parent_ids([revid])
1992
1724
        return target.basis_tree(), target
1993
1725
 
1994
 
    @classmethod
1995
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1996
 
        result = klass.make_source_parent_tree(source, target)
1997
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1998
 
        return result
1999
 
 
2000
 
    @classmethod
2001
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2002
 
                                                  target):
2003
 
        from bzrlib.tests.test__dirstate_helpers import \
2004
 
            compiled_dirstate_helpers_feature
2005
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2006
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2007
 
        result = klass.make_source_parent_tree(source, target)
2008
 
        result[1]._iter_changes = ProcessEntryC
2009
 
        return result
2010
 
 
2011
1726
    _matching_from_tree_format = WorkingTreeFormat4()
2012
1727
    _matching_to_tree_format = WorkingTreeFormat4()
2013
 
 
2014
 
    @classmethod
2015
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2016
 
        # This method shouldn't be called, because we have python and C
2017
 
        # specific flavours.
2018
 
        raise NotImplementedError
2019
 
 
2020
 
    def iter_changes(self, include_unchanged=False,
 
1728
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1729
 
 
1730
    def _iter_changes(self, include_unchanged=False,
2021
1731
                      specific_files=None, pb=None, extra_trees=[],
2022
1732
                      require_versioned=True, want_unversioned=False):
2023
1733
        """Return the changes from source to target.
2024
1734
 
2025
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1735
        :return: An iterator that yields tuples. See InterTree._iter_changes
2026
1736
            for details.
2027
1737
        :param specific_files: An optional list of file paths to restrict the
2028
1738
            comparison to. When mapping filenames to ids, all matches in all
2039
1749
            output. An unversioned file is defined as one with (False, False)
2040
1750
            for the versioned pair.
2041
1751
        """
 
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
2042
1757
        # TODO: handle extra trees in the dirstate.
2043
1758
        if (extra_trees or specific_files == []):
2044
1759
            # we can't fast-path these cases (yet)
2045
 
            return super(InterDirStateTree, self).iter_changes(
 
1760
            for f in super(InterDirStateTree, self)._iter_changes(
2046
1761
                include_unchanged, specific_files, pb, extra_trees,
2047
 
                require_versioned, want_unversioned=want_unversioned)
 
1762
                require_versioned, want_unversioned=want_unversioned):
 
1763
                yield f
 
1764
            return
2048
1765
        parent_ids = self.target.get_parent_ids()
2049
 
        if not (self.source._revision_id in parent_ids
2050
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2051
 
            raise AssertionError(
2052
 
                "revision {%s} is not stored in {%s}, but %s "
2053
 
                "can only be used for trees stored in the dirstate"
2054
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
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)
2055
1771
        target_index = 0
2056
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1772
        if self.source._revision_id == NULL_REVISION:
2057
1773
            source_index = None
2058
1774
            indices = (target_index,)
2059
1775
        else:
2060
 
            if not (self.source._revision_id in parent_ids):
2061
 
                raise AssertionError(
2062
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2063
 
                    self.source._revision_id, parent_ids))
 
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)
2064
1779
            source_index = 1 + parent_ids.index(self.source._revision_id)
2065
1780
            indices = (source_index, target_index)
2066
1781
        # -- make all specific_files utf8 --
2067
1782
        if specific_files:
2068
1783
            specific_files_utf8 = set()
2069
1784
            for path in specific_files:
2070
 
                # Note, if there are many specific files, using cache_utf8
2071
 
                # would be good here.
2072
1785
                specific_files_utf8.add(path.encode('utf8'))
2073
1786
            specific_files = specific_files_utf8
2074
1787
        else:
2075
1788
            specific_files = set([''])
2076
1789
        # -- specific_files is now a utf8 path set --
2077
 
 
2078
1790
        # -- get the state object and prepare it.
2079
1791
        state = self.target.current_dirstate()
2080
1792
        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
2081
1811
        if require_versioned:
2082
1812
            # -- check all supplied paths are versioned in a search tree. --
2083
 
            not_versioned = []
 
1813
            all_versioned = True
2084
1814
            for path in specific_files:
2085
 
                path_entries = state._entries_for_path(path)
 
1815
                path_entries = _entries_for_path(path)
2086
1816
                if not path_entries:
2087
1817
                    # this specified path is not present at all: error
2088
 
                    not_versioned.append(path)
2089
 
                    continue
 
1818
                    all_versioned = False
 
1819
                    break
2090
1820
                found_versioned = False
2091
1821
                # for each id at this path
2092
1822
                for entry in path_entries:
2099
1829
                if not found_versioned:
2100
1830
                    # none of the indexes was not 'absent' at all ids for this
2101
1831
                    # path.
2102
 
                    not_versioned.append(path)
2103
 
            if len(not_versioned) > 0:
2104
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1832
                    all_versioned = False
 
1833
                    break
 
1834
            if not all_versioned:
 
1835
                raise errors.PathsNotVersionedError(specific_files)
2105
1836
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2106
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
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]
2107
1888
 
2108
1889
        use_filesystem_for_exec = (sys.platform != 'win32')
2109
 
        iter_changes = self.target._iter_changes(include_unchanged,
2110
 
            use_filesystem_for_exec, search_specific_files, state,
2111
 
            source_index, target_index, want_unversioned, self.target)
2112
 
        return iter_changes.iter_changes()
 
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
 
2113
2423
 
2114
2424
    @staticmethod
2115
2425
    def is_compatible(source, target):
2116
2426
        # the target must be a dirstate working tree
2117
 
        if not isinstance(target, DirStateWorkingTree):
 
2427
        if not isinstance(target, WorkingTree4):
2118
2428
            return False
2119
 
        # the source must be a revtree or dirstate rev tree.
 
2429
        # the source must be a revtreee or dirstate rev tree.
2120
2430
        if not isinstance(source,
2121
2431
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2122
2432
            return False
2123
2433
        # the source revid must be in the target dirstate
2124
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2434
        if not (source._revision_id == NULL_REVISION or
2125
2435
            source._revision_id in target.get_parent_ids()):
2126
 
            # TODO: what about ghosts? it may well need to
 
2436
            # TODO: what about ghosts? it may well need to 
2127
2437
            # check for them explicitly.
2128
2438
            return False
2129
2439
        return True
2139
2449
 
2140
2450
    def convert(self, tree):
2141
2451
        # lock the control files not the tree, so that we dont get tree
2142
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2452
        # on-unlock behaviours, and so that noone else diddles with the 
2143
2453
        # tree during upgrade.
2144
2454
        tree._control_files.lock_write()
2145
2455
        try:
2171
2481
 
2172
2482
    def update_format(self, tree):
2173
2483
        """Change the format marker."""
2174
 
        tree._transport.put_bytes('format',
2175
 
            self.target_format.get_format_string(),
2176
 
            mode=tree.bzrdir._get_file_mode())
2177
 
 
2178
 
 
2179
 
class Converter4to5(object):
2180
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2181
 
 
2182
 
    def __init__(self):
2183
 
        self.target_format = WorkingTreeFormat5()
2184
 
 
2185
 
    def convert(self, tree):
2186
 
        # lock the control files not the tree, so that we don't get tree
2187
 
        # on-unlock behaviours, and so that no-one else diddles with the
2188
 
        # tree during upgrade.
2189
 
        tree._control_files.lock_write()
2190
 
        try:
2191
 
            self.update_format(tree)
2192
 
        finally:
2193
 
            tree._control_files.unlock()
2194
 
 
2195
 
    def update_format(self, tree):
2196
 
        """Change the format marker."""
2197
 
        tree._transport.put_bytes('format',
2198
 
            self.target_format.get_format_string(),
2199
 
            mode=tree.bzrdir._get_file_mode())
2200
 
 
2201
 
 
2202
 
class Converter4or5to6(object):
2203
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2204
 
 
2205
 
    def __init__(self):
2206
 
        self.target_format = WorkingTreeFormat6()
2207
 
 
2208
 
    def convert(self, tree):
2209
 
        # lock the control files not the tree, so that we don't get tree
2210
 
        # on-unlock behaviours, and so that no-one else diddles with the
2211
 
        # tree during upgrade.
2212
 
        tree._control_files.lock_write()
2213
 
        try:
2214
 
            self.init_custom_control_files(tree)
2215
 
            self.update_format(tree)
2216
 
        finally:
2217
 
            tree._control_files.unlock()
2218
 
 
2219
 
    def init_custom_control_files(self, tree):
2220
 
        """Initialize custom control files."""
2221
 
        tree._transport.put_bytes('views', '',
2222
 
            mode=tree.bzrdir._get_file_mode())
2223
 
 
2224
 
    def update_format(self, tree):
2225
 
        """Change the format marker."""
2226
 
        tree._transport.put_bytes('format',
2227
 
            self.target_format.get_format_string(),
2228
 
            mode=tree.bzrdir._get_file_mode())
 
2484
        tree._control_files.put_utf8('format',
 
2485
            self.target_format.get_format_string())