/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: Robert Collins
  • Date: 2007-07-15 15:40:37 UTC
  • mto: (2592.3.33 repository)
  • mto: This revision was merged to the branch mainline in revision 2624.
  • Revision ID: robertc@robertcollins.net-20070715154037-3ar8g89decddc9su
Make GraphIndex accept nodes as key, value, references, so that the method
signature is closer to what a simple key->value index delivers. Also
change the behaviour when the reference list count is zero to accept
key, value as nodes, and emit key, value to make it identical in that case
to a simple key->value index. This may not be a good idea, but for now it
seems ok.

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,
38
 
    debug,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
39
47
    dirstate,
40
48
    errors,
41
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
55
    revision as _mod_revision,
44
56
    revisiontree,
45
 
    trace,
 
57
    textui,
46
58
    transform,
47
 
    views,
 
59
    urlutils,
 
60
    xml5,
 
61
    xml6,
48
62
    )
49
63
import bzrlib.branch
 
64
from bzrlib.transport import get_transport
50
65
import bzrlib.ui
51
66
""")
52
67
 
 
68
from bzrlib import symbol_versioning
53
69
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
 
70
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
71
from bzrlib.lockable_files import LockableFiles, TransportLock
 
72
from bzrlib.lockdir import LockDir
 
73
import bzrlib.mutabletree
57
74
from bzrlib.mutabletree import needs_tree_write_lock
58
75
from bzrlib.osutils import (
59
76
    file_kind,
60
77
    isdir,
 
78
    normpath,
61
79
    pathjoin,
 
80
    rand_chars,
62
81
    realpath,
63
82
    safe_unicode,
 
83
    splitpath,
64
84
    )
65
 
from bzrlib.trace import mutter
 
85
from bzrlib.trace import mutter, note
66
86
from bzrlib.transport.local import LocalTransport
67
87
from bzrlib.tree import InterTree
 
88
from bzrlib.progress import DummyProgress, ProgressPhase
 
89
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
90
from bzrlib.rio import RioReader, rio_file, Stanza
 
91
from bzrlib.symbol_versioning import (deprecated_passed,
 
92
        deprecated_method,
 
93
        deprecated_function,
 
94
        DEPRECATED_PARAMETER,
 
95
        )
68
96
from bzrlib.tree import Tree
69
97
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
70
98
 
71
99
 
72
 
class DirStateWorkingTree(WorkingTree3):
 
100
# This is the Windows equivalent of ENOTDIR
 
101
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
102
# just an error code.
 
103
ERROR_PATH_NOT_FOUND = 3
 
104
ERROR_DIRECTORY = 267
 
105
 
 
106
 
 
107
class WorkingTree4(WorkingTree3):
 
108
    """This is the Format 4 working tree.
 
109
 
 
110
    This differs from WorkingTree3 by:
 
111
     - Having a consolidated internal dirstate, stored in a
 
112
       randomly-accessible sorted file on disk.
 
113
     - Not having a regular inventory attribute.  One can be synthesized 
 
114
       on demand but this is expensive and should be avoided.
 
115
 
 
116
    This is new in bzr 0.15.
 
117
    """
 
118
 
73
119
    def __init__(self, basedir,
74
120
                 branch,
75
121
                 _control_files=None,
84
130
        """
85
131
        self._format = _format
86
132
        self.bzrdir = _bzrdir
 
133
        from bzrlib.trace import note, mutter
 
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
99
151
        self._dirstate = None
100
152
        self._inventory = None
101
153
        #-------------
102
 
        self._setup_directory_is_tree_reference()
103
 
        self._detect_case_handling()
104
 
        self._rules_searcher = None
105
 
        self.views = self._make_views()
106
 
        #--- allow tests to select the dirstate iter_changes implementation
107
 
        self._iter_changes = dirstate._process_entry
108
154
 
109
155
    @needs_tree_write_lock
110
156
    def _add(self, files, ids, kinds):
112
158
        state = self.current_dirstate()
113
159
        for f, file_id, kind in zip(files, ids, kinds):
114
160
            f = f.strip('/')
 
161
            assert '//' not in f
 
162
            assert '..' not in f
115
163
            if self.path2id(f):
116
164
                # special case tree root handling.
117
165
                if f == '' and self.path2id(f) == ROOT_ID:
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
 
273
    def _directory_is_tree_reference(self, relpath):
 
274
        # as a special case, if a directory contains control files then 
 
275
        # it's a tree reference, except that the root of the tree is not
 
276
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
277
        # TODO: We could ask all the control formats whether they
 
278
        # recognize this directory, but at the moment there's no cheap api
 
279
        # to do that.  Since we probably can only nest bzr checkouts and
 
280
        # they always use this name it's ok for now.  -- mbp 20060306
 
281
        #
 
282
        # FIXME: There is an unhandled case here of a subdirectory
 
283
        # containing .bzr but not a branch; that will probably blow up
 
284
        # when you try to commit it.  It might happen if there is a
 
285
        # checkout in a subdirectory.  This can be avoided by not adding
 
286
        # it.  mbp 20070306
236
287
 
237
288
    def filter_unversioned_files(self, paths):
238
289
        """Filter out paths that are versioned.
271
322
 
272
323
    def _generate_inventory(self):
273
324
        """Create and set self.inventory from the dirstate object.
274
 
 
 
325
        
275
326
        This is relatively expensive: we have to walk the entire dirstate.
276
327
        Ideally we would not, and can deprecate this function.
277
328
        """
281
332
        state._read_dirblocks_if_needed()
282
333
        root_key, current_entry = self._get_entry(path='')
283
334
        current_id = root_key[2]
284
 
        if not (current_entry[0][0] == 'd'): # directory
285
 
            raise AssertionError(current_entry)
 
335
        assert current_entry[0][0] == 'd' # directory
286
336
        inv = Inventory(root_id=current_id)
287
337
        # Turn some things into local variables
288
338
        minikind_to_kind = dirstate.DirState._minikind_to_kind
321
371
                    # add this entry to the parent map.
322
372
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
323
373
                elif kind == 'tree-reference':
324
 
                    if not self._repo_supports_tree_reference:
325
 
                        raise errors.UnsupportedOperation(
326
 
                            self._generate_inventory,
327
 
                            self.branch.repository)
 
374
                    assert self._repo_supports_tree_reference, \
 
375
                        "repository of %r " \
 
376
                        "doesn't support tree references " \
 
377
                        "required by entry %r" \
 
378
                        % (self, name)
328
379
                    inv_entry.reference_revision = link_or_sha1 or None
329
380
                elif kind != 'symlink':
330
381
                    raise AssertionError("unknown kind %r" % kind)
331
382
                # 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,))
 
383
                assert file_id not in inv_byid, ('file_id %s already in'
 
384
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
385
                assert name_unicode not in parent_ie.children
338
386
                inv_byid[file_id] = inv_entry
339
387
                parent_ie.children[name_unicode] = inv_entry
340
388
        self._inventory = inv
345
393
        If either file_id or path is supplied, it is used as the key to lookup.
346
394
        If both are supplied, the fastest lookup is used, and an error is
347
395
        raised if they do not both point at the same row.
348
 
 
 
396
        
349
397
        :param file_id: An optional unicode file_id to be looked up.
350
398
        :param path: An optional unicode path to be looked up.
351
399
        :return: The dirstate row tuple for path/file_id, or (None, None)
360
408
    def get_file_sha1(self, file_id, path=None, stat_value=None):
361
409
        # check file id is valid unconditionally.
362
410
        entry = self._get_entry(file_id=file_id, path=path)
363
 
        if entry[0] is None:
364
 
            raise errors.NoSuchId(self, file_id)
 
411
        assert entry[0] is not None, 'what error should this raise'
365
412
        if path is None:
366
413
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
367
414
 
375
422
                    return None
376
423
                else:
377
424
                    raise
378
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
379
 
            stat_value=stat_value)
 
425
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
426
                                          stat_value=stat_value)
380
427
        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
 
428
            return link_or_sha1
391
429
        return None
392
430
 
393
431
    def _get_inventory(self):
394
432
        """Get the inventory for the tree. This is only valid within a lock."""
395
 
        if 'evil' in debug.debug_flags:
396
 
            trace.mutter_callsite(2,
397
 
                "accessing .inventory forces a size of tree translation.")
398
433
        if self._inventory is not None:
399
434
            return self._inventory
400
435
        self._must_be_locked()
407
442
    @needs_read_lock
408
443
    def get_parent_ids(self):
409
444
        """See Tree.get_parent_ids.
410
 
 
 
445
        
411
446
        This implementation requests the ids list from the dirstate file.
412
447
        """
413
448
        return self.current_dirstate().get_parent_ids()
429
464
 
430
465
    def has_id(self, file_id):
431
466
        state = self.current_dirstate()
 
467
        file_id = osutils.safe_file_id(file_id)
432
468
        row, parents = self._get_entry(file_id=file_id)
433
469
        if row is None:
434
470
            return False
435
471
        return osutils.lexists(pathjoin(
436
472
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
437
473
 
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
474
    @needs_read_lock
444
475
    def id2path(self, file_id):
445
476
        "Convert a file-id to a path."
 
477
        file_id = osutils.safe_file_id(file_id)
446
478
        state = self.current_dirstate()
447
479
        entry = self._get_entry(file_id=file_id)
448
480
        if entry == (None, None):
450
482
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
451
483
        return path_utf8.decode('utf8')
452
484
 
453
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
454
 
        entry = self._get_entry(path=path)
455
 
        if entry == (None, None):
456
 
            return False # Missing entries are not executable
457
 
        return entry[1][0][3] # Executable?
458
 
 
459
485
    if not osutils.supports_executable():
460
486
        def is_executable(self, file_id, path=None):
461
487
            """Test if a file is executable or not.
462
488
 
463
489
            Note: The caller is expected to take a read-lock before calling this.
464
490
            """
 
491
            file_id = osutils.safe_file_id(file_id)
465
492
            entry = self._get_entry(file_id=file_id, path=path)
466
493
            if entry == (None, None):
467
494
                return False
468
495
            return entry[1][0][3]
469
 
 
470
 
        _is_executable_from_path_and_stat = \
471
 
            _is_executable_from_path_and_stat_from_basis
472
496
    else:
473
497
        def is_executable(self, file_id, path=None):
474
498
            """Test if a file is executable or not.
475
499
 
476
500
            Note: The caller is expected to take a read-lock before calling this.
477
501
            """
478
 
            self._must_be_locked()
479
502
            if not path:
 
503
                file_id = osutils.safe_file_id(file_id)
480
504
                path = self.id2path(file_id)
481
505
            mode = os.lstat(self.abspath(path)).st_mode
482
506
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
483
507
 
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
508
    @needs_read_lock
495
509
    def __iter__(self):
496
510
        """Iterate through file_ids for this tree.
509
523
        return iter(result)
510
524
 
511
525
    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
526
        for key, tree_details in self.current_dirstate()._iter_entries():
517
527
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
518
528
                # not relevant to the working tree
520
530
            if not key[1]:
521
531
                # the root is not a reference.
522
532
                continue
523
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
533
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
524
534
            try:
525
 
                if self._kind(relpath) == 'tree-reference':
526
 
                    yield relpath, key[2]
 
535
                if self._kind(path) == 'tree-reference':
 
536
                    yield path, key[2]
527
537
            except errors.NoSuchFile:
528
538
                # path is missing on disk.
529
539
                continue
530
540
 
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
541
    def kind(self, file_id):
538
542
        """Return the kind of a file.
539
543
 
543
547
        Note: The caller is expected to take a read-lock before calling this.
544
548
        """
545
549
        relpath = self.id2path(file_id)
546
 
        if relpath is None:
547
 
            raise AssertionError(
548
 
                "path for id {%s} is None!" % file_id)
 
550
        assert relpath != None, \
 
551
            "path for id {%s} is None!" % file_id
549
552
        return self._kind(relpath)
550
553
 
551
554
    def _kind(self, relpath):
552
555
        abspath = self.abspath(relpath)
553
556
        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'
 
557
        if (self._repo_supports_tree_reference and
 
558
            kind == 'directory' and
 
559
            self._directory_is_tree_reference(relpath)):
 
560
            kind = 'tree-reference'
559
561
        return kind
560
562
 
561
563
    @needs_read_lock
565
567
        if parent_ids:
566
568
            return parent_ids[0]
567
569
        else:
568
 
            return _mod_revision.NULL_REVISION
 
570
            return None
569
571
 
570
572
    def lock_read(self):
571
 
        """See Branch.lock_read, and WorkingTree.unlock.
572
 
 
573
 
        :return: A bzrlib.lock.LogicalLockResult.
574
 
        """
 
573
        """See Branch.lock_read, and WorkingTree.unlock."""
575
574
        self.branch.lock_read()
576
575
        try:
577
576
            self._control_files.lock_read()
590
589
        except:
591
590
            self.branch.unlock()
592
591
            raise
593
 
        return LogicalLockResult(self.unlock)
594
592
 
595
593
    def _lock_self_write(self):
596
594
        """This should be called after the branch is locked."""
611
609
        except:
612
610
            self.branch.unlock()
613
611
            raise
614
 
        return LogicalLockResult(self.unlock)
615
612
 
616
613
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: A bzrlib.lock.LogicalLockResult.
620
 
        """
 
614
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
621
615
        self.branch.lock_read()
622
 
        return self._lock_self_write()
 
616
        self._lock_self_write()
623
617
 
624
618
    def lock_write(self):
625
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
626
 
 
627
 
        :return: A bzrlib.lock.LogicalLockResult.
628
 
        """
 
619
        """See MutableTree.lock_write, and WorkingTree.unlock."""
629
620
        self.branch.lock_write()
630
 
        return self._lock_self_write()
 
621
        self._lock_self_write()
631
622
 
632
623
    @needs_tree_write_lock
633
624
    def move(self, from_paths, to_dir, after=False):
635
626
        result = []
636
627
        if not from_paths:
637
628
            return result
 
629
 
638
630
        state = self.current_dirstate()
639
 
        if isinstance(from_paths, basestring):
640
 
            raise ValueError()
 
631
 
 
632
        assert not isinstance(from_paths, basestring)
641
633
        to_dir_utf8 = to_dir.encode('utf8')
642
634
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
643
635
        id_index = state._get_id_index()
665
657
        if self._inventory is not None:
666
658
            update_inventory = True
667
659
            inv = self.inventory
 
660
            to_dir_ie = inv[to_dir_id]
668
661
            to_dir_id = to_entry[0][2]
669
 
            to_dir_ie = inv[to_dir_id]
670
662
        else:
671
663
            update_inventory = False
672
664
 
703
695
            from_entry = self._get_entry(path=from_rel)
704
696
            if from_entry == (None, None):
705
697
                raise errors.BzrMoveFailedError(from_rel,to_dir,
706
 
                    errors.NotVersionedError(path=from_rel))
 
698
                    errors.NotVersionedError(path=str(from_rel)))
707
699
 
708
700
            from_id = from_entry[0][2]
709
701
            to_rel = pathjoin(to_dir, from_tail)
736
728
                if from_missing: # implicitly just update our path mapping
737
729
                    move_file = False
738
730
                elif not after:
739
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
731
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
732
                        extra="(Use --after to update the Bazaar id)")
740
733
 
741
734
            rollbacks = []
742
735
            def rollback_rename():
798
791
                if minikind == 'd':
799
792
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
800
793
                        """Recursively update all entries in this dirblock."""
801
 
                        if from_dir == '':
802
 
                            raise AssertionError("renaming root not supported")
 
794
                        assert from_dir != '', "renaming root not supported"
803
795
                        from_key = (from_dir, '')
804
796
                        from_block_idx, present = \
805
797
                            state._find_block_index_from_key(from_key)
818
810
 
819
811
                        # Grab a copy since move_one may update the list.
820
812
                        for entry in from_block[1][:]:
821
 
                            if not (entry[0][0] == from_dir):
822
 
                                raise AssertionError()
 
813
                            assert entry[0][0] == from_dir
823
814
                            cur_details = entry[1][0]
824
815
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
825
816
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
884
875
        for tree in trees:
885
876
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
886
877
                parents):
887
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
888
 
                    trees, require_versioned)
 
878
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
889
879
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
890
880
        # -- make all paths utf8 --
891
881
        paths_utf8 = set()
951
941
            if not all_versioned:
952
942
                raise errors.PathsNotVersionedError(paths)
953
943
        # -- remove redundancy in supplied paths to prevent over-scanning --
954
 
        search_paths = osutils.minimum_path_selection(paths)
955
 
        # sketch:
 
944
        search_paths = set()
 
945
        for path in paths:
 
946
            other_paths = paths.difference(set([path]))
 
947
            if not osutils.is_inside_any(other_paths, path):
 
948
                # this is a top level path, we must check it.
 
949
                search_paths.add(path)
 
950
        # sketch: 
956
951
        # for all search_indexs in each path at or under each element of
957
952
        # search_paths, if the detail is relocated: add the id, and add the
958
953
        # relocated path as one to search if its not searched already. If the
1014
1009
 
1015
1010
    def read_working_inventory(self):
1016
1011
        """Read the working inventory.
1017
 
 
 
1012
        
1018
1013
        This is a meaningless operation for dirstate, but we obey it anyhow.
1019
1014
        """
1020
1015
        return self.inventory
1025
1020
 
1026
1021
        WorkingTree4 supplies revision_trees for any basis tree.
1027
1022
        """
 
1023
        revision_id = osutils.safe_revision_id(revision_id)
1028
1024
        dirstate = self.current_dirstate()
1029
1025
        parent_ids = dirstate.get_parent_ids()
1030
1026
        if revision_id not in parent_ids:
1037
1033
    @needs_tree_write_lock
1038
1034
    def set_last_revision(self, new_revision):
1039
1035
        """Change the last revision in the working tree."""
 
1036
        new_revision = osutils.safe_revision_id(new_revision)
1040
1037
        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.")
 
1038
        if new_revision in (NULL_REVISION, None):
 
1039
            assert len(parents) < 2, (
 
1040
                "setting the last parent to none with a pending merge is "
 
1041
                "unsupported.")
1046
1042
            self.set_parent_ids([])
1047
1043
        else:
1048
1044
            self.set_parent_ids([new_revision] + parents[1:],
1051
1047
    @needs_tree_write_lock
1052
1048
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1053
1049
        """Set the parent ids to revision_ids.
1054
 
 
 
1050
        
1055
1051
        See also set_parent_trees. This api will try to retrieve the tree data
1056
1052
        for each element of revision_ids from the trees repository. If you have
1057
1053
        tree data already available, it is more efficient to use
1061
1057
        :param revision_ids: The revision_ids to set as the parent ids of this
1062
1058
            working tree. Any of these may be ghosts.
1063
1059
        """
 
1060
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1064
1061
        trees = []
1065
1062
        for revision_id in revision_ids:
1066
1063
            try:
1089
1086
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1090
1087
        real_trees = []
1091
1088
        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
1089
        # convert absent trees to the null tree, which we convert back to
1099
1090
        # missing on access.
1100
1091
        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
 
1092
            rev_id = osutils.safe_revision_id(rev_id)
1107
1093
            _mod_revision.check_not_reserved_id(rev_id)
1108
1094
            if tree is not None:
1109
1095
                real_trees.append((rev_id, tree))
1110
1096
            else:
1111
1097
                real_trees.append((rev_id,
1112
 
                    self.branch.repository.revision_tree(
1113
 
                        _mod_revision.NULL_REVISION)))
 
1098
                    self.branch.repository.revision_tree(None)))
1114
1099
                ghosts.append(rev_id)
1115
 
            accepted_revisions.add(rev_id)
1116
1100
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1117
1101
        self._make_dirty(reset_inventory=False)
1118
1102
 
1123
1107
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1124
1108
            self._make_dirty(reset_inventory=True)
1125
1109
 
1126
 
    def _sha_from_stat(self, path, stat_result):
1127
 
        """Get a sha digest from the tree's stat cache.
1128
 
 
1129
 
        The default implementation assumes no stat cache is present.
1130
 
 
1131
 
        :param path: The path.
1132
 
        :param stat_result: The stat result being looked up.
1133
 
        """
1134
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1135
 
 
1136
1110
    @needs_read_lock
1137
1111
    def supports_tree_reference(self):
1138
1112
        return self._repo_supports_tree_reference
1139
1113
 
1140
1114
    def unlock(self):
1141
1115
        """Unlock in format 4 trees needs to write the entire dirstate."""
1142
 
        # do non-implementation specific cleanup
1143
 
        self._cleanup()
1144
 
 
1145
1116
        if self._control_files._lock_count == 1:
1146
1117
            # eventually we should do signature checking during read locks for
1147
1118
            # dirstate updates.
1178
1149
            return
1179
1150
        state = self.current_dirstate()
1180
1151
        state._read_dirblocks_if_needed()
1181
 
        ids_to_unversion = set(file_ids)
 
1152
        ids_to_unversion = set()
 
1153
        for file_id in file_ids:
 
1154
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1182
1155
        paths_to_unversion = set()
1183
1156
        # sketch:
1184
1157
        # check if the root is to be unversioned, if so, assert for now.
1211
1184
                # just forget the whole block.
1212
1185
                entry_index = 0
1213
1186
                while entry_index < len(block[1]):
 
1187
                    # Mark this file id as having been removed
1214
1188
                    entry = block[1][entry_index]
1215
 
                    if entry[1][0][0] in 'ar':
1216
 
                        # don't remove absent or renamed entries
 
1189
                    ids_to_unversion.discard(entry[0][2])
 
1190
                    if (entry[1][0][0] == 'a'
 
1191
                        or not state._make_absent(entry)):
1217
1192
                        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
1193
                # go to the next block. (At the moment we dont delete empty
1225
1194
                # dirblocks)
1226
1195
                block_index += 1
1249
1218
            for file_id in file_ids:
1250
1219
                self._inventory.remove_recursive_id(file_id)
1251
1220
 
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
 
    def update_basis_by_delta(self, new_revid, delta):
1266
 
        """See MutableTree.update_basis_by_delta."""
1267
 
        if self.last_revision() == new_revid:
1268
 
            raise AssertionError()
1269
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1270
 
 
1271
1221
    @needs_read_lock
1272
1222
    def _validate(self):
1273
1223
        self._dirstate._validate()
1275
1225
    @needs_tree_write_lock
1276
1226
    def _write_inventory(self, inv):
1277
1227
        """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:
 
1228
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1229
        self.current_dirstate().set_state_from_inventory(inv)
 
1230
        self._make_dirty(reset_inventory=False)
 
1231
        if self._inventory is not None:
1292
1232
            self._inventory = inv
1293
1233
        self.flush()
1294
1234
 
1295
1235
 
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):
1386
 
 
1387
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1388
 
                   accelerator_tree=None, hardlink=False):
 
1236
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1237
    """The first consolidated dirstate working tree format.
 
1238
 
 
1239
    This format:
 
1240
        - exists within a metadir controlling .bzr
 
1241
        - includes an explicit version marker for the workingtree control
 
1242
          files, separate from the BzrDir format
 
1243
        - modifies the hash cache format
 
1244
        - is new in bzr 0.15
 
1245
        - uses a LockDir to guard access to it.
 
1246
    """
 
1247
 
 
1248
    upgrade_recommended = False
 
1249
 
 
1250
    def get_format_string(self):
 
1251
        """See WorkingTreeFormat.get_format_string()."""
 
1252
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1253
 
 
1254
    def get_format_description(self):
 
1255
        """See WorkingTreeFormat.get_format_description()."""
 
1256
        return "Working tree format 4"
 
1257
 
 
1258
    def initialize(self, a_bzrdir, revision_id=None):
1389
1259
        """See WorkingTreeFormat.initialize().
1390
1260
 
1391
1261
        :param revision_id: allows creating a working tree at a different
1392
1262
        revision than the branch is at.
1393
 
        :param accelerator_tree: A tree which can be used for retrieving file
1394
 
            contents more quickly than the revision tree, i.e. a workingtree.
1395
 
            The revision tree will be used for cases where accelerator_tree's
1396
 
            content is different.
1397
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1398
 
            where possible.
1399
1263
 
1400
1264
        These trees get an initial random root id, if their repository supports
1401
1265
        rich root data, TREE_ROOT otherwise.
1402
1266
        """
 
1267
        revision_id = osutils.safe_revision_id(revision_id)
1403
1268
        if not isinstance(a_bzrdir.transport, LocalTransport):
1404
1269
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1405
1270
        transport = a_bzrdir.get_workingtree_transport(self)
1406
1271
        control_files = self._open_control_files(a_bzrdir)
1407
1272
        control_files.create_lock()
1408
1273
        control_files.lock_write()
1409
 
        transport.put_bytes('format', self.get_format_string(),
1410
 
            mode=a_bzrdir._get_file_mode())
1411
 
        if from_branch is not None:
1412
 
            branch = from_branch
1413
 
        else:
1414
 
            branch = a_bzrdir.open_branch()
 
1274
        control_files.put_utf8('format', self.get_format_string())
 
1275
        branch = a_bzrdir.open_branch()
1415
1276
        if revision_id is None:
1416
1277
            revision_id = branch.last_revision()
1417
1278
        local_path = transport.local_abspath('dirstate')
1419
1280
        state = dirstate.DirState.initialize(local_path)
1420
1281
        state.unlock()
1421
1282
        del state
1422
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1283
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1423
1284
                         branch,
1424
1285
                         _format=self,
1425
1286
                         _bzrdir=a_bzrdir,
1427
1288
        wt._new_tree()
1428
1289
        wt.lock_tree_write()
1429
1290
        try:
1430
 
            self._init_custom_control_files(wt)
1431
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1291
            if revision_id in (None, NULL_REVISION):
1432
1292
                if branch.repository.supports_rich_root():
1433
1293
                    wt._set_root_id(generate_ids.gen_root_id())
1434
1294
                else:
1435
1295
                    wt._set_root_id(ROOT_ID)
1436
1296
                wt.flush()
1437
 
            basis = None
1438
 
            # frequently, we will get here due to branching.  The accelerator
1439
 
            # tree will be the tree from the branch, so the desired basis
1440
 
            # tree will often be a parent of the accelerator tree.
1441
 
            if accelerator_tree is not None:
1442
 
                try:
1443
 
                    basis = accelerator_tree.revision_tree(revision_id)
1444
 
                except errors.NoSuchRevision:
1445
 
                    pass
1446
 
            if basis is None:
1447
 
                basis = branch.repository.revision_tree(revision_id)
1448
 
            if revision_id == _mod_revision.NULL_REVISION:
1449
 
                parents_list = []
1450
 
            else:
1451
 
                parents_list = [(revision_id, basis)]
 
1297
            wt.set_last_revision(revision_id)
 
1298
            wt.flush()
 
1299
            basis = wt.basis_tree()
1452
1300
            basis.lock_read()
1453
 
            try:
1454
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1301
            # if the basis has a root id we have to use that; otherwise we use
 
1302
            # a new random one
 
1303
            basis_root_id = basis.get_root_id()
 
1304
            if basis_root_id is not None:
 
1305
                wt._set_root_id(basis_root_id)
1455
1306
                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()
 
1307
            transform.build_tree(basis, wt)
 
1308
            basis.unlock()
1477
1309
        finally:
1478
1310
            control_files.unlock()
1479
1311
            wt.unlock()
1480
1312
        return wt
1481
1313
 
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
1314
    def _open(self, a_bzrdir, control_files):
1492
1315
        """Open the tree itself.
1493
1316
 
1494
1317
        :param a_bzrdir: the dir for the tree.
1495
1318
        :param control_files: the control files for the tree.
1496
1319
        """
1497
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1320
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1498
1321
                           branch=a_bzrdir.open_branch(),
1499
1322
                           _format=self,
1500
1323
                           _bzrdir=a_bzrdir,
1501
1324
                           _control_files=control_files)
1502
1325
 
1503
1326
    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
1327
        # please test against something that will let us do tree references
1509
1328
        return bzrdir.format_registry.make_bzrdir(
1510
1329
            'dirstate-with-subtree')
1512
1331
    _matchingbzrdir = property(__get_matchingbzrdir)
1513
1332
 
1514
1333
 
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
1334
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
 
    """
 
1335
    """A revision tree pulling the inventory from a dirstate."""
1593
1336
 
1594
1337
    def __init__(self, dirstate, revision_id, repository):
1595
1338
        self._dirstate = dirstate
1596
 
        self._revision_id = revision_id
 
1339
        self._revision_id = osutils.safe_revision_id(revision_id)
1597
1340
        self._repository = repository
1598
1341
        self._inventory = None
1599
1342
        self._locked = 0
1600
1343
        self._dirstate_locked = False
1601
 
        self._repo_supports_tree_reference = getattr(
1602
 
            repository._format, "supports_tree_reference",
1603
 
            False)
1604
1344
 
1605
1345
    def __repr__(self):
1606
1346
        return "<%s of %s in %s>" % \
1607
1347
            (self.__class__.__name__, self._revision_id, self._dirstate)
1608
1348
 
1609
 
    def annotate_iter(self, file_id,
1610
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1349
    def annotate_iter(self, file_id):
1611
1350
        """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]
 
1351
        w = self._repository.weave_store.get_weave(file_id,
 
1352
                           self._repository.get_transaction())
 
1353
        return w.annotate_iter(self.inventory[file_id].revision)
1615
1354
 
1616
 
    def _get_ancestors(self, default_revision):
1617
 
        return set(self._repository.get_ancestry(self._revision_id,
1618
 
                                                 topo_sorted=False))
1619
1355
    def _comparison_data(self, entry, path):
1620
1356
        """See Tree._comparison_data."""
1621
1357
        if entry is None:
1638
1374
    def get_root_id(self):
1639
1375
        return self.path2id('')
1640
1376
 
1641
 
    def id2path(self, file_id):
1642
 
        "Convert a file-id to a path."
1643
 
        entry = self._get_entry(file_id=file_id)
1644
 
        if entry == (None, None):
1645
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1646
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1647
 
        return path_utf8.decode('utf8')
1648
 
 
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
1377
    def _get_parent_index(self):
1658
1378
        """Return the index in the dirstate referenced by this tree."""
1659
1379
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1664
1384
        If either file_id or path is supplied, it is used as the key to lookup.
1665
1385
        If both are supplied, the fastest lookup is used, and an error is
1666
1386
        raised if they do not both point at the same row.
1667
 
 
 
1387
        
1668
1388
        :param file_id: An optional unicode file_id to be looked up.
1669
1389
        :param path: An optional unicode path to be looked up.
1670
1390
        :return: The dirstate row tuple for path/file_id, or (None, None)
1671
1391
        """
1672
1392
        if file_id is None and path is None:
1673
1393
            raise errors.BzrError('must supply file_id or path')
 
1394
        file_id = osutils.safe_file_id(file_id)
1674
1395
        if path is not None:
1675
1396
            path = path.encode('utf8')
1676
1397
        parent_index = self._get_parent_index()
1684
1405
 
1685
1406
        This is relatively expensive: we have to walk the entire dirstate.
1686
1407
        """
1687
 
        if not self._locked:
1688
 
            raise AssertionError(
1689
 
                'cannot generate inventory of an unlocked '
1690
 
                'dirstate revision tree')
 
1408
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1409
            'dirstate revision tree'
1691
1410
        # separate call for profiling - makes it clear where the costs are.
1692
1411
        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()))
 
1412
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1413
            'parent %s has disappeared from %s' % (
 
1414
            self._revision_id, self._dirstate.get_parent_ids())
1697
1415
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1698
1416
        # This is identical now to the WorkingTree _generate_inventory except
1699
1417
        # for the tree index use.
1700
1418
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1701
1419
        current_id = root_key[2]
1702
 
        if current_entry[parent_index][0] != 'd':
1703
 
            raise AssertionError()
 
1420
        assert current_entry[parent_index][0] == 'd'
1704
1421
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1705
1422
        inv.root.revision = current_entry[parent_index][4]
1706
1423
        # Turn some things into local variables
1746
1463
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1747
1464
                            % entry)
1748
1465
                # 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,))
 
1466
                assert file_id not in inv_byid
 
1467
                assert name_unicode not in parent_ie.children
1755
1468
                inv_byid[file_id] = inv_entry
1756
1469
                parent_ie.children[name_unicode] = inv_entry
1757
1470
        self._inventory = inv
1767
1480
            return None
1768
1481
        parent_index = self._get_parent_index()
1769
1482
        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
 
1483
        return self._repository.get_revision(last_changed_revision).timestamp
1775
1484
 
1776
1485
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1777
1486
        entry = self._get_entry(file_id=file_id, path=path)
1781
1490
            return parent_details[1]
1782
1491
        return None
1783
1492
 
1784
 
    def get_file(self, file_id, path=None):
 
1493
    def get_weave(self, file_id):
 
1494
        return self._repository.weave_store.get_weave(file_id,
 
1495
                self._repository.get_transaction())
 
1496
 
 
1497
    def get_file(self, file_id):
1785
1498
        return StringIO(self.get_file_text(file_id))
1786
1499
 
 
1500
    def get_file_lines(self, file_id):
 
1501
        ie = self.inventory[file_id]
 
1502
        return self._repository.weave_store.get_weave(file_id,
 
1503
                self._repository.get_transaction()).get_lines(ie.revision)
 
1504
 
1787
1505
    def get_file_size(self, file_id):
1788
 
        """See Tree.get_file_size"""
1789
1506
        return self.inventory[file_id].text_size
1790
1507
 
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)
 
1508
    def get_file_text(self, file_id):
 
1509
        return ''.join(self.get_file_lines(file_id))
1794
1510
 
1795
1511
    def get_reference_revision(self, file_id, path=None):
1796
1512
        return self.inventory[file_id].reference_revision
1797
1513
 
1798
 
    def iter_files_bytes(self, desired_files):
1799
 
        """See Tree.iter_files_bytes.
1800
 
 
1801
 
        This version is implemented on top of Repository.iter_files_bytes"""
1802
 
        parent_index = self._get_parent_index()
1803
 
        repo_desired_files = []
1804
 
        for file_id, identifier in desired_files:
1805
 
            entry = self._get_entry(file_id)
1806
 
            if entry == (None, None):
1807
 
                raise errors.NoSuchId(self, file_id)
1808
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1809
 
                                       identifier))
1810
 
        return self._repository.iter_files_bytes(repo_desired_files)
1811
 
 
1812
1514
    def get_symlink_target(self, file_id):
1813
1515
        entry = self._get_entry(file_id=file_id)
1814
1516
        parent_index = self._get_parent_index()
1815
1517
        if entry[1][parent_index][0] != 'l':
1816
1518
            return None
1817
1519
        else:
1818
 
            target = entry[1][parent_index][1]
1819
 
            target = target.decode('utf8')
1820
 
            return target
 
1520
            # At present, none of the tree implementations supports non-ascii
 
1521
            # symlink targets. So we will just assume that the dirstate path is
 
1522
            # correct.
 
1523
            return entry[1][parent_index][1]
1821
1524
 
1822
1525
    def get_revision_id(self):
1823
1526
        """Return the revision id for this tree."""
1841
1544
        return bool(self.path2id(filename))
1842
1545
 
1843
1546
    def kind(self, file_id):
1844
 
        entry = self._get_entry(file_id=file_id)[1]
1845
 
        if entry is None:
1846
 
            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)
1853
 
 
1854
 
    def path_content_summary(self, path):
1855
 
        """See Tree.path_content_summary."""
1856
 
        id = self.inventory.path2id(path)
1857
 
        if id is None:
1858
 
            return ('missing', None, None, None)
1859
 
        entry = self._inventory[id]
1860
 
        kind = entry.kind
1861
 
        if kind == 'file':
1862
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1863
 
        elif kind == 'symlink':
1864
 
            return (kind, None, None, entry.symlink_target)
1865
 
        else:
1866
 
            return (kind, None, None, None)
 
1547
        return self.inventory[file_id].kind
1867
1548
 
1868
1549
    def is_executable(self, file_id, path=None):
1869
1550
        ie = self.inventory[file_id]
1871
1552
            return None
1872
1553
        return ie.executable
1873
1554
 
1874
 
    def is_locked(self):
1875
 
        return self._locked
1876
 
 
1877
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1555
    def list_files(self, include_root=False):
1878
1556
        # We use a standard implementation, because DirStateRevisionTree is
1879
1557
        # dealing with one of the parents of the current state
1880
1558
        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:
 
1559
        entries = inv.iter_entries()
 
1560
        if self.inventory.root is not None and not include_root:
1890
1561
            entries.next()
1891
1562
        for path, entry in entries:
1892
1563
            yield path, 'V', entry.kind, entry.file_id, entry
1893
1564
 
1894
1565
    def lock_read(self):
1895
 
        """Lock the tree for a set of operations.
1896
 
 
1897
 
        :return: A bzrlib.lock.LogicalLockResult.
1898
 
        """
 
1566
        """Lock the tree for a set of operations."""
1899
1567
        if not self._locked:
1900
1568
            self._repository.lock_read()
1901
1569
            if self._dirstate._lock_token is None:
1902
1570
                self._dirstate.lock_read()
1903
1571
                self._dirstate_locked = True
1904
1572
        self._locked += 1
1905
 
        return LogicalLockResult(self.unlock)
1906
1573
 
1907
1574
    def _must_be_locked(self):
1908
1575
        if not self._locked:
1929
1596
                self._dirstate_locked = False
1930
1597
            self._repository.unlock()
1931
1598
 
1932
 
    @needs_read_lock
1933
 
    def supports_tree_reference(self):
1934
 
        return self._repo_supports_tree_reference
1935
 
 
1936
1599
    def walkdirs(self, prefix=""):
1937
1600
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1938
 
        # implementation based on an inventory.
 
1601
        # implementation based on an inventory.  
1939
1602
        # This should be cleaned up to use the much faster Dirstate code
1940
1603
        # So for now, we just build up the parent inventory, and extract
1941
1604
        # it the same way RevisionTree does.
1970
1633
 
1971
1634
class InterDirStateTree(InterTree):
1972
1635
    """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
 
1636
    
 
1637
    This is used only when both trees are in the dirstate working file, and 
 
1638
    the source is any parent within the dirstate, and the destination is 
1976
1639
    the current working tree of the same dirstate.
1977
1640
    """
1978
1641
    # this could be generalized to allow comparisons between any trees in the
1991
1654
        target.set_parent_ids([revid])
1992
1655
        return target.basis_tree(), target
1993
1656
 
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
1657
    _matching_from_tree_format = WorkingTreeFormat4()
2012
1658
    _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,
 
1659
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1660
 
 
1661
    def _iter_changes(self, include_unchanged=False,
2021
1662
                      specific_files=None, pb=None, extra_trees=[],
2022
1663
                      require_versioned=True, want_unversioned=False):
2023
1664
        """Return the changes from source to target.
2024
1665
 
2025
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1666
        :return: An iterator that yields tuples. See InterTree._iter_changes
2026
1667
            for details.
2027
1668
        :param specific_files: An optional list of file paths to restrict the
2028
1669
            comparison to. When mapping filenames to ids, all matches in all
2039
1680
            output. An unversioned file is defined as one with (False, False)
2040
1681
            for the versioned pair.
2041
1682
        """
 
1683
        utf8_decode = cache_utf8._utf8_decode
 
1684
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1685
        # NB: show_status depends on being able to pass in non-versioned files
 
1686
        # and report them as unknown
2042
1687
        # TODO: handle extra trees in the dirstate.
2043
 
        if (extra_trees or specific_files == []):
 
1688
        # TODO: handle comparisons as an empty tree as a different special
 
1689
        # case? mbp 20070226
 
1690
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2044
1691
            # we can't fast-path these cases (yet)
2045
 
            return super(InterDirStateTree, self).iter_changes(
 
1692
            for f in super(InterDirStateTree, self)._iter_changes(
2046
1693
                include_unchanged, specific_files, pb, extra_trees,
2047
 
                require_versioned, want_unversioned=want_unversioned)
 
1694
                require_versioned, want_unversioned=want_unversioned):
 
1695
                yield f
 
1696
            return
2048
1697
        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))
 
1698
        assert (self.source._revision_id in parent_ids), \
 
1699
                "revision {%s} is not stored in {%s}, but %s " \
 
1700
                "can only be used for trees stored in the dirstate" \
 
1701
                % (self.source._revision_id, self.target, self._iter_changes)
2055
1702
        target_index = 0
2056
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1703
        if self.source._revision_id == NULL_REVISION:
2057
1704
            source_index = None
2058
1705
            indices = (target_index,)
2059
1706
        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))
 
1707
            assert (self.source._revision_id in parent_ids), \
 
1708
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1709
                self.source._revision_id, parent_ids)
2064
1710
            source_index = 1 + parent_ids.index(self.source._revision_id)
2065
 
            indices = (source_index, target_index)
 
1711
            indices = (source_index,target_index)
2066
1712
        # -- make all specific_files utf8 --
2067
1713
        if specific_files:
2068
1714
            specific_files_utf8 = set()
2069
1715
            for path in specific_files:
2070
 
                # Note, if there are many specific files, using cache_utf8
2071
 
                # would be good here.
2072
1716
                specific_files_utf8.add(path.encode('utf8'))
2073
1717
            specific_files = specific_files_utf8
2074
1718
        else:
2075
1719
            specific_files = set([''])
2076
1720
        # -- specific_files is now a utf8 path set --
2077
 
 
2078
1721
        # -- get the state object and prepare it.
2079
1722
        state = self.target.current_dirstate()
2080
1723
        state._read_dirblocks_if_needed()
 
1724
        def _entries_for_path(path):
 
1725
            """Return a list with all the entries that match path for all ids.
 
1726
            """
 
1727
            dirname, basename = os.path.split(path)
 
1728
            key = (dirname, basename, '')
 
1729
            block_index, present = state._find_block_index_from_key(key)
 
1730
            if not present:
 
1731
                # the block which should contain path is absent.
 
1732
                return []
 
1733
            result = []
 
1734
            block = state._dirblocks[block_index][1]
 
1735
            entry_index, _ = state._find_entry_index(key, block)
 
1736
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1737
            while (entry_index < len(block) and
 
1738
                block[entry_index][0][0:2] == key[0:2]):
 
1739
                result.append(block[entry_index])
 
1740
                entry_index += 1
 
1741
            return result
2081
1742
        if require_versioned:
2082
1743
            # -- check all supplied paths are versioned in a search tree. --
2083
 
            not_versioned = []
 
1744
            all_versioned = True
2084
1745
            for path in specific_files:
2085
 
                path_entries = state._entries_for_path(path)
 
1746
                path_entries = _entries_for_path(path)
2086
1747
                if not path_entries:
2087
1748
                    # this specified path is not present at all: error
2088
 
                    not_versioned.append(path)
2089
 
                    continue
 
1749
                    all_versioned = False
 
1750
                    break
2090
1751
                found_versioned = False
2091
1752
                # for each id at this path
2092
1753
                for entry in path_entries:
2099
1760
                if not found_versioned:
2100
1761
                    # none of the indexes was not 'absent' at all ids for this
2101
1762
                    # path.
2102
 
                    not_versioned.append(path)
2103
 
            if len(not_versioned) > 0:
2104
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1763
                    all_versioned = False
 
1764
                    break
 
1765
            if not all_versioned:
 
1766
                raise errors.PathsNotVersionedError(specific_files)
2105
1767
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2106
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1768
        search_specific_files = set()
 
1769
        for path in specific_files:
 
1770
            other_specific_files = specific_files.difference(set([path]))
 
1771
            if not osutils.is_inside_any(other_specific_files, path):
 
1772
                # this is a top level path, we must check it.
 
1773
                search_specific_files.add(path)
 
1774
        # sketch: 
 
1775
        # compare source_index and target_index at or under each element of search_specific_files.
 
1776
        # follow the following comparison table. Note that we only want to do diff operations when
 
1777
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1778
        # for the target.
 
1779
        # cases:
 
1780
        # 
 
1781
        # Source | Target | disk | action
 
1782
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1783
        #        |        |      | diff check on source-target
 
1784
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1785
        #        |        |      | ???
 
1786
        #   r    |  a     |      | add source to search
 
1787
        #   r    |  a     |  a   | 
 
1788
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1789
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1790
        #   a    | fdlt   |      | add new id
 
1791
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1792
        #   a    |  a     |      | not present in either tree, skip
 
1793
        #   a    |  a     |  a   | not present in any tree, skip
 
1794
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1795
        #        |        |      | may not be selected by the users list of paths.
 
1796
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1797
        #        |        |      | may not be selected by the users list of paths.
 
1798
        #  fdlt  | fdlt   |      | content in both: diff them
 
1799
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1800
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1801
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1802
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1803
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1804
        #        |        |      | this id at the other path.
 
1805
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1806
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1807
        #        |        |      | this id at the other path.
 
1808
 
 
1809
        # for all search_indexs in each path at or under each element of
 
1810
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1811
        # relocated path as one to search if its not searched already. If the
 
1812
        # detail is not relocated, add the id.
 
1813
        searched_specific_files = set()
 
1814
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1815
        # Using a list so that we can access the values and change them in
 
1816
        # nested scope. Each one is [path, file_id, entry]
 
1817
        last_source_parent = [None, None]
 
1818
        last_target_parent = [None, None]
2107
1819
 
2108
1820
        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()
 
1821
 
 
1822
        # Just a sentry, so that _process_entry can say that this
 
1823
        # record is handled, but isn't interesting to process (unchanged)
 
1824
        uninteresting = object()
 
1825
 
 
1826
 
 
1827
        old_dirname_to_file_id = {}
 
1828
        new_dirname_to_file_id = {}
 
1829
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1830
        #       keeping a cache of directories that we have seen.
 
1831
 
 
1832
        def _process_entry(entry, path_info):
 
1833
            """Compare an entry and real disk to generate delta information.
 
1834
 
 
1835
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1836
                the path of entry. If None, then the path is considered absent.
 
1837
                (Perhaps we should pass in a concrete entry for this ?)
 
1838
                Basename is returned as a utf8 string because we expect this
 
1839
                tuple will be ignored, and don't want to take the time to
 
1840
                decode.
 
1841
            :return: None if these don't match
 
1842
                     A tuple of information about the change, or
 
1843
                     the object 'uninteresting' if these match, but are
 
1844
                     basically identical.
 
1845
            """
 
1846
            if source_index is None:
 
1847
                source_details = NULL_PARENT_DETAILS
 
1848
            else:
 
1849
                source_details = entry[1][source_index]
 
1850
            target_details = entry[1][target_index]
 
1851
            target_minikind = target_details[0]
 
1852
            if path_info is not None and target_minikind in 'fdlt':
 
1853
                assert target_index == 0
 
1854
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1855
                                                  stat_value=path_info[3])
 
1856
                # The entry may have been modified by update_entry
 
1857
                target_details = entry[1][target_index]
 
1858
                target_minikind = target_details[0]
 
1859
            else:
 
1860
                link_or_sha1 = None
 
1861
            file_id = entry[0][2]
 
1862
            source_minikind = source_details[0]
 
1863
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1864
                # claimed content in both: diff
 
1865
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1866
                #        |        |      | diff check on source-target
 
1867
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1868
                #        |        |      | ???
 
1869
                if source_minikind in 'r':
 
1870
                    # add the source to the search path to find any children it
 
1871
                    # has.  TODO ? : only add if it is a container ?
 
1872
                    if not osutils.is_inside_any(searched_specific_files,
 
1873
                                                 source_details[1]):
 
1874
                        search_specific_files.add(source_details[1])
 
1875
                    # generate the old path; this is needed for stating later
 
1876
                    # as well.
 
1877
                    old_path = source_details[1]
 
1878
                    old_dirname, old_basename = os.path.split(old_path)
 
1879
                    path = pathjoin(entry[0][0], entry[0][1])
 
1880
                    old_entry = state._get_entry(source_index,
 
1881
                                                 path_utf8=old_path)
 
1882
                    # update the source details variable to be the real
 
1883
                    # location.
 
1884
                    source_details = old_entry[1][source_index]
 
1885
                    source_minikind = source_details[0]
 
1886
                else:
 
1887
                    old_dirname = entry[0][0]
 
1888
                    old_basename = entry[0][1]
 
1889
                    old_path = path = None
 
1890
                if path_info is None:
 
1891
                    # the file is missing on disk, show as removed.
 
1892
                    content_change = True
 
1893
                    target_kind = None
 
1894
                    target_exec = False
 
1895
                else:
 
1896
                    # source and target are both versioned and disk file is present.
 
1897
                    target_kind = path_info[2]
 
1898
                    if target_kind == 'directory':
 
1899
                        if path is None:
 
1900
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1901
                        new_dirname_to_file_id[path] = file_id
 
1902
                        if source_minikind != 'd':
 
1903
                            content_change = True
 
1904
                        else:
 
1905
                            # directories have no fingerprint
 
1906
                            content_change = False
 
1907
                        target_exec = False
 
1908
                    elif target_kind == 'file':
 
1909
                        if source_minikind != 'f':
 
1910
                            content_change = True
 
1911
                        else:
 
1912
                            # We could check the size, but we already have the
 
1913
                            # sha1 hash.
 
1914
                            content_change = (link_or_sha1 != source_details[1])
 
1915
                        # Target details is updated at update_entry time
 
1916
                        if use_filesystem_for_exec:
 
1917
                            # We don't need S_ISREG here, because we are sure
 
1918
                            # we are dealing with a file.
 
1919
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1920
                        else:
 
1921
                            target_exec = target_details[3]
 
1922
                    elif target_kind == 'symlink':
 
1923
                        if source_minikind != 'l':
 
1924
                            content_change = True
 
1925
                        else:
 
1926
                            content_change = (link_or_sha1 != source_details[1])
 
1927
                        target_exec = False
 
1928
                    elif target_kind == 'tree-reference':
 
1929
                        if source_minikind != 't':
 
1930
                            content_change = True
 
1931
                        else:
 
1932
                            content_change = False
 
1933
                        target_exec = False
 
1934
                    else:
 
1935
                        raise Exception, "unknown kind %s" % path_info[2]
 
1936
                if source_minikind == 'd':
 
1937
                    if path is None:
 
1938
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1939
                    old_dirname_to_file_id[old_path] = file_id
 
1940
                # parent id is the entry for the path in the target tree
 
1941
                if old_dirname == last_source_parent[0]:
 
1942
                    source_parent_id = last_source_parent[1]
 
1943
                else:
 
1944
                    try:
 
1945
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1946
                    except KeyError:
 
1947
                        source_parent_entry = state._get_entry(source_index,
 
1948
                                                               path_utf8=old_dirname)
 
1949
                        source_parent_id = source_parent_entry[0][2]
 
1950
                    if source_parent_id == entry[0][2]:
 
1951
                        # This is the root, so the parent is None
 
1952
                        source_parent_id = None
 
1953
                    else:
 
1954
                        last_source_parent[0] = old_dirname
 
1955
                        last_source_parent[1] = source_parent_id
 
1956
                new_dirname = entry[0][0]
 
1957
                if new_dirname == last_target_parent[0]:
 
1958
                    target_parent_id = last_target_parent[1]
 
1959
                else:
 
1960
                    try:
 
1961
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1962
                    except KeyError:
 
1963
                        # TODO: We don't always need to do the lookup, because the
 
1964
                        #       parent entry will be the same as the source entry.
 
1965
                        target_parent_entry = state._get_entry(target_index,
 
1966
                                                               path_utf8=new_dirname)
 
1967
                        assert target_parent_entry != (None, None), (
 
1968
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1969
                            % (new_dirname, entry))
 
1970
                        target_parent_id = target_parent_entry[0][2]
 
1971
                    if target_parent_id == entry[0][2]:
 
1972
                        # This is the root, so the parent is None
 
1973
                        target_parent_id = None
 
1974
                    else:
 
1975
                        last_target_parent[0] = new_dirname
 
1976
                        last_target_parent[1] = target_parent_id
 
1977
 
 
1978
                source_exec = source_details[3]
 
1979
                if (include_unchanged
 
1980
                    or content_change
 
1981
                    or source_parent_id != target_parent_id
 
1982
                    or old_basename != entry[0][1]
 
1983
                    or source_exec != target_exec
 
1984
                    ):
 
1985
                    if old_path is None:
 
1986
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1987
                        old_path_u = utf8_decode(old_path)[0]
 
1988
                        path_u = old_path_u
 
1989
                    else:
 
1990
                        old_path_u = utf8_decode(old_path)[0]
 
1991
                        if old_path == path:
 
1992
                            path_u = old_path_u
 
1993
                        else:
 
1994
                            path_u = utf8_decode(path)[0]
 
1995
                    source_kind = _minikind_to_kind[source_minikind]
 
1996
                    return (entry[0][2],
 
1997
                           (old_path_u, path_u),
 
1998
                           content_change,
 
1999
                           (True, True),
 
2000
                           (source_parent_id, target_parent_id),
 
2001
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2002
                           (source_kind, target_kind),
 
2003
                           (source_exec, target_exec))
 
2004
                else:
 
2005
                    return uninteresting
 
2006
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2007
                # looks like a new file
 
2008
                if path_info is not None:
 
2009
                    path = pathjoin(entry[0][0], entry[0][1])
 
2010
                    # parent id is the entry for the path in the target tree
 
2011
                    # TODO: these are the same for an entire directory: cache em.
 
2012
                    parent_id = state._get_entry(target_index,
 
2013
                                                 path_utf8=entry[0][0])[0][2]
 
2014
                    if parent_id == entry[0][2]:
 
2015
                        parent_id = None
 
2016
                    if use_filesystem_for_exec:
 
2017
                        # We need S_ISREG here, because we aren't sure if this
 
2018
                        # is a file or not.
 
2019
                        target_exec = bool(
 
2020
                            stat.S_ISREG(path_info[3].st_mode)
 
2021
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2022
                    else:
 
2023
                        target_exec = target_details[3]
 
2024
                    return (entry[0][2],
 
2025
                           (None, utf8_decode(path)[0]),
 
2026
                           True,
 
2027
                           (False, True),
 
2028
                           (None, parent_id),
 
2029
                           (None, utf8_decode(entry[0][1])[0]),
 
2030
                           (None, path_info[2]),
 
2031
                           (None, target_exec))
 
2032
                else:
 
2033
                    # but its not on disk: we deliberately treat this as just
 
2034
                    # never-present. (Why ?! - RBC 20070224)
 
2035
                    pass
 
2036
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2037
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2038
                # if its still on disk, *and* theres no other entry at this
 
2039
                # path [we dont know this in this routine at the moment -
 
2040
                # perhaps we should change this - then it would be an unknown.
 
2041
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2042
                # parent id is the entry for the path in the target tree
 
2043
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2044
                if parent_id == entry[0][2]:
 
2045
                    parent_id = None
 
2046
                return (entry[0][2],
 
2047
                       (utf8_decode(old_path)[0], None),
 
2048
                       True,
 
2049
                       (True, False),
 
2050
                       (parent_id, None),
 
2051
                       (utf8_decode(entry[0][1])[0], None),
 
2052
                       (_minikind_to_kind[source_minikind], None),
 
2053
                       (source_details[3], None))
 
2054
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2055
                # a rename; could be a true rename, or a rename inherited from
 
2056
                # a renamed parent. TODO: handle this efficiently. Its not
 
2057
                # common case to rename dirs though, so a correct but slow
 
2058
                # implementation will do.
 
2059
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2060
                    search_specific_files.add(target_details[1])
 
2061
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2062
                # neither of the selected trees contain this file,
 
2063
                # so skip over it. This is not currently directly tested, but
 
2064
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2065
                pass
 
2066
            else:
 
2067
                raise AssertionError("don't know how to compare "
 
2068
                    "source_minikind=%r, target_minikind=%r"
 
2069
                    % (source_minikind, target_minikind))
 
2070
                ## import pdb;pdb.set_trace()
 
2071
            return None
 
2072
 
 
2073
        while search_specific_files:
 
2074
            # TODO: the pending list should be lexically sorted?  the
 
2075
            # interface doesn't require it.
 
2076
            current_root = search_specific_files.pop()
 
2077
            current_root_unicode = current_root.decode('utf8')
 
2078
            searched_specific_files.add(current_root)
 
2079
            # process the entries for this containing directory: the rest will be
 
2080
            # found by their parents recursively.
 
2081
            root_entries = _entries_for_path(current_root)
 
2082
            root_abspath = self.target.abspath(current_root_unicode)
 
2083
            try:
 
2084
                root_stat = os.lstat(root_abspath)
 
2085
            except OSError, e:
 
2086
                if e.errno == errno.ENOENT:
 
2087
                    # the path does not exist: let _process_entry know that.
 
2088
                    root_dir_info = None
 
2089
                else:
 
2090
                    # some other random error: hand it up.
 
2091
                    raise
 
2092
            else:
 
2093
                root_dir_info = ('', current_root,
 
2094
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2095
                    root_abspath)
 
2096
                if root_dir_info[2] == 'directory':
 
2097
                    if self.target._directory_is_tree_reference(
 
2098
                        current_root.decode('utf8')):
 
2099
                        root_dir_info = root_dir_info[:2] + \
 
2100
                            ('tree-reference',) + root_dir_info[3:]
 
2101
 
 
2102
            if not root_entries and not root_dir_info:
 
2103
                # this specified path is not present at all, skip it.
 
2104
                continue
 
2105
            path_handled = False
 
2106
            for entry in root_entries:
 
2107
                result = _process_entry(entry, root_dir_info)
 
2108
                if result is not None:
 
2109
                    path_handled = True
 
2110
                    if result is not uninteresting:
 
2111
                        yield result
 
2112
            if want_unversioned and not path_handled and root_dir_info:
 
2113
                new_executable = bool(
 
2114
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2115
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2116
                yield (None,
 
2117
                       (None, current_root_unicode),
 
2118
                       True,
 
2119
                       (False, False),
 
2120
                       (None, None),
 
2121
                       (None, splitpath(current_root_unicode)[-1]),
 
2122
                       (None, root_dir_info[2]),
 
2123
                       (None, new_executable)
 
2124
                      )
 
2125
            initial_key = (current_root, '', '')
 
2126
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2127
            if block_index == 0:
 
2128
                # we have processed the total root already, but because the
 
2129
                # initial key matched it we should skip it here.
 
2130
                block_index +=1
 
2131
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2132
                current_dir_info = None
 
2133
            else:
 
2134
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2135
                try:
 
2136
                    current_dir_info = dir_iterator.next()
 
2137
                except OSError, e:
 
2138
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2139
                    # python 2.5 has e.errno == EINVAL,
 
2140
                    #            and e.winerror == ERROR_DIRECTORY
 
2141
                    e_winerror = getattr(e, 'winerror', None)
 
2142
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2143
                    # there may be directories in the inventory even though
 
2144
                    # this path is not a file on disk: so mark it as end of
 
2145
                    # iterator
 
2146
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2147
                        current_dir_info = None
 
2148
                    elif (sys.platform == 'win32'
 
2149
                          and (e.errno in win_errors
 
2150
                               or e_winerror in win_errors)):
 
2151
                        current_dir_info = None
 
2152
                    else:
 
2153
                        raise
 
2154
                else:
 
2155
                    if current_dir_info[0][0] == '':
 
2156
                        # remove .bzr from iteration
 
2157
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2158
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2159
                        del current_dir_info[1][bzr_index]
 
2160
            # walk until both the directory listing and the versioned metadata
 
2161
            # are exhausted. 
 
2162
            if (block_index < len(state._dirblocks) and
 
2163
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2164
                current_block = state._dirblocks[block_index]
 
2165
            else:
 
2166
                current_block = None
 
2167
            while (current_dir_info is not None or
 
2168
                   current_block is not None):
 
2169
                if (current_dir_info and current_block
 
2170
                    and current_dir_info[0][0] != current_block[0]):
 
2171
                    if current_dir_info[0][0].split('/') < current_block[0].split('/'):
 
2172
                        # filesystem data refers to paths not covered by the dirblock.
 
2173
                        # this has two possibilities:
 
2174
                        # A) it is versioned but empty, so there is no block for it
 
2175
                        # B) it is not versioned.
 
2176
 
 
2177
                        # if (A) then we need to recurse into it to check for
 
2178
                        # new unknown files or directories.
 
2179
                        # if (B) then we should ignore it, because we don't
 
2180
                        # recurse into unknown directories.
 
2181
                        path_index = 0
 
2182
                        while path_index < len(current_dir_info[1]):
 
2183
                                current_path_info = current_dir_info[1][path_index]
 
2184
                                if want_unversioned:
 
2185
                                    if current_path_info[2] == 'directory':
 
2186
                                        if self.target._directory_is_tree_reference(
 
2187
                                            current_path_info[0].decode('utf8')):
 
2188
                                            current_path_info = current_path_info[:2] + \
 
2189
                                                ('tree-reference',) + current_path_info[3:]
 
2190
                                    new_executable = bool(
 
2191
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2192
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2193
                                    yield (None,
 
2194
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2195
                                        True,
 
2196
                                        (False, False),
 
2197
                                        (None, None),
 
2198
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2199
                                        (None, current_path_info[2]),
 
2200
                                        (None, new_executable))
 
2201
                                # dont descend into this unversioned path if it is
 
2202
                                # a dir
 
2203
                                if current_path_info[2] in ('directory',
 
2204
                                                            'tree-reference'):
 
2205
                                    del current_dir_info[1][path_index]
 
2206
                                    path_index -= 1
 
2207
                                path_index += 1
 
2208
 
 
2209
                        # This dir info has been handled, go to the next
 
2210
                        try:
 
2211
                            current_dir_info = dir_iterator.next()
 
2212
                        except StopIteration:
 
2213
                            current_dir_info = None
 
2214
                    else:
 
2215
                        # We have a dirblock entry for this location, but there
 
2216
                        # is no filesystem path for this. This is most likely
 
2217
                        # because a directory was removed from the disk.
 
2218
                        # We don't have to report the missing directory,
 
2219
                        # because that should have already been handled, but we
 
2220
                        # need to handle all of the files that are contained
 
2221
                        # within.
 
2222
                        for current_entry in current_block[1]:
 
2223
                            # entry referring to file not present on disk.
 
2224
                            # advance the entry only, after processing.
 
2225
                            result = _process_entry(current_entry, None)
 
2226
                            if result is not None:
 
2227
                                if result is not uninteresting:
 
2228
                                    yield result
 
2229
                        block_index +=1
 
2230
                        if (block_index < len(state._dirblocks) and
 
2231
                            osutils.is_inside(current_root,
 
2232
                                              state._dirblocks[block_index][0])):
 
2233
                            current_block = state._dirblocks[block_index]
 
2234
                        else:
 
2235
                            current_block = None
 
2236
                    continue
 
2237
                entry_index = 0
 
2238
                if current_block and entry_index < len(current_block[1]):
 
2239
                    current_entry = current_block[1][entry_index]
 
2240
                else:
 
2241
                    current_entry = None
 
2242
                advance_entry = True
 
2243
                path_index = 0
 
2244
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2245
                    current_path_info = current_dir_info[1][path_index]
 
2246
                    if current_path_info[2] == 'directory':
 
2247
                        if self.target._directory_is_tree_reference(
 
2248
                            current_path_info[0].decode('utf8')):
 
2249
                            current_path_info = current_path_info[:2] + \
 
2250
                                ('tree-reference',) + current_path_info[3:]
 
2251
                else:
 
2252
                    current_path_info = None
 
2253
                advance_path = True
 
2254
                path_handled = False
 
2255
                while (current_entry is not None or
 
2256
                    current_path_info is not None):
 
2257
                    if current_entry is None:
 
2258
                        # the check for path_handled when the path is adnvaced
 
2259
                        # will yield this path if needed.
 
2260
                        pass
 
2261
                    elif current_path_info is None:
 
2262
                        # no path is fine: the per entry code will handle it.
 
2263
                        result = _process_entry(current_entry, current_path_info)
 
2264
                        if result is not None:
 
2265
                            if result is not uninteresting:
 
2266
                                yield result
 
2267
                    elif (current_entry[0][1] != current_path_info[1]
 
2268
                          or current_entry[1][target_index][0] in 'ar'):
 
2269
                        # The current path on disk doesn't match the dirblock
 
2270
                        # record. Either the dirblock is marked as absent, or
 
2271
                        # the file on disk is not present at all in the
 
2272
                        # dirblock. Either way, report about the dirblock
 
2273
                        # entry, and let other code handle the filesystem one.
 
2274
 
 
2275
                        # Compare the basename for these files to determine
 
2276
                        # which comes first
 
2277
                        if current_path_info[1] < current_entry[0][1]:
 
2278
                            # extra file on disk: pass for now, but only
 
2279
                            # increment the path, not the entry
 
2280
                            advance_entry = False
 
2281
                        else:
 
2282
                            # entry referring to file not present on disk.
 
2283
                            # advance the entry only, after processing.
 
2284
                            result = _process_entry(current_entry, None)
 
2285
                            if result is not None:
 
2286
                                if result is not uninteresting:
 
2287
                                    yield result
 
2288
                            advance_path = False
 
2289
                    else:
 
2290
                        result = _process_entry(current_entry, current_path_info)
 
2291
                        if result is not None:
 
2292
                            path_handled = True
 
2293
                            if result is not uninteresting:
 
2294
                                yield result
 
2295
                    if advance_entry and current_entry is not None:
 
2296
                        entry_index += 1
 
2297
                        if entry_index < len(current_block[1]):
 
2298
                            current_entry = current_block[1][entry_index]
 
2299
                        else:
 
2300
                            current_entry = None
 
2301
                    else:
 
2302
                        advance_entry = True # reset the advance flaga
 
2303
                    if advance_path and current_path_info is not None:
 
2304
                        if not path_handled:
 
2305
                            # unversioned in all regards
 
2306
                            if want_unversioned:
 
2307
                                new_executable = bool(
 
2308
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2309
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2310
                                yield (None,
 
2311
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2312
                                    True,
 
2313
                                    (False, False),
 
2314
                                    (None, None),
 
2315
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2316
                                    (None, current_path_info[2]),
 
2317
                                    (None, new_executable))
 
2318
                            # dont descend into this unversioned path if it is
 
2319
                            # a dir
 
2320
                            if current_path_info[2] in ('directory'):
 
2321
                                del current_dir_info[1][path_index]
 
2322
                                path_index -= 1
 
2323
                        # dont descend the disk iterator into any tree 
 
2324
                        # paths.
 
2325
                        if current_path_info[2] == 'tree-reference':
 
2326
                            del current_dir_info[1][path_index]
 
2327
                            path_index -= 1
 
2328
                        path_index += 1
 
2329
                        if path_index < len(current_dir_info[1]):
 
2330
                            current_path_info = current_dir_info[1][path_index]
 
2331
                            if current_path_info[2] == 'directory':
 
2332
                                if self.target._directory_is_tree_reference(
 
2333
                                    current_path_info[0].decode('utf8')):
 
2334
                                    current_path_info = current_path_info[:2] + \
 
2335
                                        ('tree-reference',) + current_path_info[3:]
 
2336
                        else:
 
2337
                            current_path_info = None
 
2338
                        path_handled = False
 
2339
                    else:
 
2340
                        advance_path = True # reset the advance flagg.
 
2341
                if current_block is not None:
 
2342
                    block_index += 1
 
2343
                    if (block_index < len(state._dirblocks) and
 
2344
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2345
                        current_block = state._dirblocks[block_index]
 
2346
                    else:
 
2347
                        current_block = None
 
2348
                if current_dir_info is not None:
 
2349
                    try:
 
2350
                        current_dir_info = dir_iterator.next()
 
2351
                    except StopIteration:
 
2352
                        current_dir_info = None
 
2353
 
2113
2354
 
2114
2355
    @staticmethod
2115
2356
    def is_compatible(source, target):
2116
2357
        # the target must be a dirstate working tree
2117
 
        if not isinstance(target, DirStateWorkingTree):
 
2358
        if not isinstance(target, WorkingTree4):
2118
2359
            return False
2119
 
        # the source must be a revtree or dirstate rev tree.
 
2360
        # the source must be a revtreee or dirstate rev tree.
2120
2361
        if not isinstance(source,
2121
2362
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2122
2363
            return False
2123
2364
        # the source revid must be in the target dirstate
2124
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2365
        if not (source._revision_id == NULL_REVISION or
2125
2366
            source._revision_id in target.get_parent_ids()):
2126
 
            # TODO: what about ghosts? it may well need to
 
2367
            # TODO: what about ghosts? it may well need to 
2127
2368
            # check for them explicitly.
2128
2369
            return False
2129
2370
        return True
2139
2380
 
2140
2381
    def convert(self, tree):
2141
2382
        # 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
 
2383
        # on-unlock behaviours, and so that noone else diddles with the 
2143
2384
        # tree during upgrade.
2144
2385
        tree._control_files.lock_write()
2145
2386
        try:
2171
2412
 
2172
2413
    def update_format(self, tree):
2173
2414
        """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())
 
2415
        tree._control_files.put_utf8('format',
 
2416
            self.target_format.get_format_string())