~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2007-12-18 23:41:30 UTC
  • mfrom: (3099.3.7 graph_optimization)
  • Revision ID: pqm@pqm.ubuntu.com-20071218234130-061grgxsaf1g7bao
(jam) Implement ParentProviders.get_parent_map() and deprecate
        get_parents()

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2012 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
 
22
22
WorkingTree.open(dir).
23
23
"""
24
24
 
25
 
from __future__ import absolute_import
26
 
 
27
25
from cStringIO import StringIO
28
26
import os
29
27
import sys
30
28
 
31
29
from bzrlib.lazy_import import lazy_import
32
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
33
34
import errno
 
35
import itertools
 
36
import operator
34
37
import stat
 
38
from time import time
 
39
import warnings
35
40
 
 
41
import bzrlib
36
42
from bzrlib import (
37
43
    bzrdir,
38
44
    cache_utf8,
39
 
    config,
40
45
    conflicts as _mod_conflicts,
41
 
    controldir,
42
46
    debug,
 
47
    delta,
43
48
    dirstate,
44
49
    errors,
45
 
    filters as _mod_filters,
46
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
47
54
    osutils,
48
55
    revision as _mod_revision,
49
56
    revisiontree,
 
57
    textui,
50
58
    trace,
51
59
    transform,
52
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
53
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
54
67
""")
55
68
 
 
69
from bzrlib import symbol_versioning
56
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
57
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
58
 
from bzrlib.lock import LogicalLockResult
59
 
from bzrlib.lockable_files import LockableFiles
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
60
73
from bzrlib.lockdir import LockDir
61
 
from bzrlib.mutabletree import (
62
 
    MutableTree,
63
 
    needs_tree_write_lock,
64
 
    )
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
65
76
from bzrlib.osutils import (
66
77
    file_kind,
67
78
    isdir,
 
79
    normpath,
68
80
    pathjoin,
 
81
    rand_chars,
69
82
    realpath,
70
83
    safe_unicode,
71
 
    )
72
 
from bzrlib.symbol_versioning import (
73
 
    deprecated_in,
74
 
    deprecated_method,
75
 
    )
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
76
87
from bzrlib.transport.local import LocalTransport
77
 
from bzrlib.tree import (
78
 
    InterTree,
79
 
    InventoryTree,
80
 
    )
81
 
from bzrlib.workingtree import (
82
 
    InventoryWorkingTree,
83
 
    WorkingTree,
84
 
    WorkingTreeFormatMetaDir,
85
 
    )
86
 
 
87
 
 
88
 
class DirStateWorkingTree(InventoryWorkingTree):
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
89
119
 
90
120
    def __init__(self, basedir,
91
121
                 branch,
101
131
        """
102
132
        self._format = _format
103
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
104
136
        basedir = safe_unicode(basedir)
105
 
        trace.mutter("opening working tree %r", basedir)
 
137
        mutter("opening working tree %r", basedir)
106
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
107
141
        self.basedir = realpath(basedir)
108
142
        # if branch is at our basedir and is a format 6 or less
109
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
110
146
        self._control_files = _control_files
111
 
        self._transport = self._control_files._transport
112
147
        self._dirty = None
113
148
        #-------------
114
149
        # during a read or write lock these objects are set, and are
116
151
        self._dirstate = None
117
152
        self._inventory = None
118
153
        #-------------
119
 
        self._setup_directory_is_tree_reference()
120
154
        self._detect_case_handling()
121
 
        self._rules_searcher = None
122
 
        self.views = self._make_views()
123
 
        #--- allow tests to select the dirstate iter_changes implementation
124
 
        self._iter_changes = dirstate._process_entry
125
155
 
126
156
    @needs_tree_write_lock
127
157
    def _add(self, files, ids, kinds):
142
172
            state.add(f, file_id, kind, None, '')
143
173
        self._make_dirty(reset_inventory=True)
144
174
 
145
 
    def _get_check_refs(self):
146
 
        """Return the references needed to perform a check of this tree."""
147
 
        return [('trees', self.last_revision())]
148
 
 
149
175
    def _make_dirty(self, reset_inventory):
150
176
        """Make the tree state dirty.
151
177
 
159
185
    @needs_tree_write_lock
160
186
    def add_reference(self, sub_tree):
161
187
        # use standard implementation, which calls back to self._add
162
 
        #
 
188
        # 
163
189
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
190
        # it's just recorded at the moment of commit. 
165
191
        self._add_reference(sub_tree)
166
192
 
167
193
    def break_lock(self):
203
229
 
204
230
    def _comparison_data(self, entry, path):
205
231
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
232
            WorkingTree3._comparison_data(self, entry, path)
207
233
        # it looks like a plain directory, but it's really a reference -- see
208
234
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            and entry is not None and entry.kind == 'tree-reference'):
 
235
        if (self._repo_supports_tree_reference and
 
236
            kind == 'directory' and
 
237
            self._directory_is_tree_reference(path)):
211
238
            kind = 'tree-reference'
212
239
        return kind, executable, stat_value
213
240
 
215
242
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
243
        # mark the tree as dirty post commit - commit
217
244
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
245
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
246
        self._make_dirty(reset_inventory=True)
220
247
        return result
221
248
 
239
266
            return self._dirstate
240
267
        local_path = self.bzrdir.get_workingtree_transport(None
241
268
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
269
        self._dirstate = dirstate.DirState.on_file(local_path)
244
270
        return self._dirstate
245
271
 
246
 
    def _sha1_provider(self):
247
 
        """A function that returns a SHA1Provider suitable for this tree.
248
 
 
249
 
        :return: None if content filtering is not supported by this tree.
250
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
251
 
          form of files, i.e. after read filters are applied.
252
 
        """
253
 
        if self.supports_content_filtering():
254
 
            return ContentFilterAwareSHA1Provider(self)
255
 
        else:
256
 
            return None
257
 
 
258
 
    def _worth_saving_limit(self):
259
 
        """How many hash changes are ok before we must save the dirstate.
260
 
 
261
 
        :return: an integer. -1 means never save.
262
 
        """
263
 
        conf = self.get_config_stack()
264
 
        return conf.get('bzr.workingtree.worth_saving_limit')
265
 
 
266
272
    def filter_unversioned_files(self, paths):
267
273
        """Filter out paths that are versioned.
268
274
 
300
306
 
301
307
    def _generate_inventory(self):
302
308
        """Create and set self.inventory from the dirstate object.
303
 
 
 
309
        
304
310
        This is relatively expensive: we have to walk the entire dirstate.
305
311
        Ideally we would not, and can deprecate this function.
306
312
        """
310
316
        state._read_dirblocks_if_needed()
311
317
        root_key, current_entry = self._get_entry(path='')
312
318
        current_id = root_key[2]
313
 
        if not (current_entry[0][0] == 'd'): # directory
314
 
            raise AssertionError(current_entry)
 
319
        assert current_entry[0][0] == 'd' # directory
315
320
        inv = Inventory(root_id=current_id)
316
321
        # Turn some things into local variables
317
322
        minikind_to_kind = dirstate.DirState._minikind_to_kind
350
355
                    # add this entry to the parent map.
351
356
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
357
                elif kind == 'tree-reference':
353
 
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            self.branch.repository)
 
358
                    assert self._repo_supports_tree_reference, \
 
359
                        "repository of %r " \
 
360
                        "doesn't support tree references " \
 
361
                        "required by entry %r" \
 
362
                        % (self, name)
357
363
                    inv_entry.reference_revision = link_or_sha1 or None
358
364
                elif kind != 'symlink':
359
365
                    raise AssertionError("unknown kind %r" % kind)
360
366
                # These checks cost us around 40ms on a 55k entry tree
361
 
                if file_id in inv_byid:
362
 
                    raise AssertionError('file_id %s already in'
363
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
364
 
                if name_unicode in parent_ie.children:
365
 
                    raise AssertionError('name %r already in parent'
366
 
                        % (name_unicode,))
 
367
                assert file_id not in inv_byid, ('file_id %s already in'
 
368
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
369
                assert name_unicode not in parent_ie.children
367
370
                inv_byid[file_id] = inv_entry
368
371
                parent_ie.children[name_unicode] = inv_entry
369
372
        self._inventory = inv
374
377
        If either file_id or path is supplied, it is used as the key to lookup.
375
378
        If both are supplied, the fastest lookup is used, and an error is
376
379
        raised if they do not both point at the same row.
377
 
 
 
380
        
378
381
        :param file_id: An optional unicode file_id to be looked up.
379
382
        :param path: An optional unicode path to be looked up.
380
383
        :return: The dirstate row tuple for path/file_id, or (None, None)
398
401
        state = self.current_dirstate()
399
402
        if stat_value is None:
400
403
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
 
404
                stat_value = os.lstat(file_abspath)
402
405
            except OSError, e:
403
406
                if e.errno == errno.ENOENT:
404
407
                    return None
405
408
                else:
406
409
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
410
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
411
                                          stat_value=stat_value)
409
412
        if entry[1][0][0] == 'f':
410
 
            if link_or_sha1 is None:
411
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
412
 
                try:
413
 
                    sha1 = osutils.sha_file(file_obj)
414
 
                finally:
415
 
                    file_obj.close()
416
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
417
 
                return sha1
418
 
            else:
419
 
                return link_or_sha1
 
413
            return link_or_sha1
420
414
        return None
421
415
 
422
 
    def _get_root_inventory(self):
 
416
    def _get_inventory(self):
423
417
        """Get the inventory for the tree. This is only valid within a lock."""
424
418
        if 'evil' in debug.debug_flags:
425
419
            trace.mutter_callsite(2,
430
424
        self._generate_inventory()
431
425
        return self._inventory
432
426
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
427
    inventory = property(_get_inventory,
438
428
                         doc="Inventory of this Tree")
439
429
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
430
    @needs_read_lock
444
431
    def get_parent_ids(self):
445
432
        """See Tree.get_parent_ids.
446
 
 
 
433
        
447
434
        This implementation requests the ids list from the dirstate file.
448
435
        """
449
436
        return self.current_dirstate().get_parent_ids()
471
458
        return osutils.lexists(pathjoin(
472
459
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
460
 
474
 
    def has_or_had_id(self, file_id):
475
 
        state = self.current_dirstate()
476
 
        row, parents = self._get_entry(file_id=file_id)
477
 
        return row is not None
478
 
 
479
461
    @needs_read_lock
480
462
    def id2path(self, file_id):
481
463
        "Convert a file-id to a path."
492
474
            return False # Missing entries are not executable
493
475
        return entry[1][0][3] # Executable?
494
476
 
495
 
    def is_executable(self, file_id, path=None):
496
 
        """Test if a file is executable or not.
 
477
    if not osutils.supports_executable():
 
478
        def is_executable(self, file_id, path=None):
 
479
            """Test if a file is executable or not.
497
480
 
498
 
        Note: The caller is expected to take a read-lock before calling this.
499
 
        """
500
 
        if not self._supports_executable():
 
481
            Note: The caller is expected to take a read-lock before calling this.
 
482
            """
501
483
            entry = self._get_entry(file_id=file_id, path=path)
502
484
            if entry == (None, None):
503
485
                return False
504
486
            return entry[1][0][3]
505
 
        else:
506
 
            self._must_be_locked()
 
487
 
 
488
        _is_executable_from_path_and_stat = \
 
489
            _is_executable_from_path_and_stat_from_basis
 
490
    else:
 
491
        def is_executable(self, file_id, path=None):
 
492
            """Test if a file is executable or not.
 
493
 
 
494
            Note: The caller is expected to take a read-lock before calling this.
 
495
            """
507
496
            if not path:
508
497
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
498
            mode = os.lstat(self.abspath(path)).st_mode
510
499
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
500
 
512
 
    def all_file_ids(self):
513
 
        """See Tree.iter_all_file_ids"""
514
 
        self._must_be_locked()
515
 
        result = set()
516
 
        for key, tree_details in self.current_dirstate()._iter_entries():
517
 
            if tree_details[0][0] in ('a', 'r'): # relocated
518
 
                continue
519
 
            result.add(key[2])
520
 
        return result
521
 
 
522
501
    @needs_read_lock
523
502
    def __iter__(self):
524
503
        """Iterate through file_ids for this tree.
537
516
        return iter(result)
538
517
 
539
518
    def iter_references(self):
540
 
        if not self._repo_supports_tree_reference:
541
 
            # When the repo doesn't support references, we will have nothing to
542
 
            # return
543
 
            return
544
519
        for key, tree_details in self.current_dirstate()._iter_entries():
545
520
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
546
521
                # not relevant to the working tree
548
523
            if not key[1]:
549
524
                # the root is not a reference.
550
525
                continue
551
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
526
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
552
527
            try:
553
 
                if self._kind(relpath) == 'tree-reference':
554
 
                    yield relpath, key[2]
 
528
                if self._kind(path) == 'tree-reference':
 
529
                    yield path, key[2]
555
530
            except errors.NoSuchFile:
556
531
                # path is missing on disk.
557
532
                continue
558
533
 
559
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
560
 
        """See MutableTree._observed_sha1."""
561
 
        state = self.current_dirstate()
562
 
        entry = self._get_entry(file_id=file_id, path=path)
563
 
        state._observed_sha1(entry, sha1, statvalue)
564
 
 
565
534
    def kind(self, file_id):
566
535
        """Return the kind of a file.
567
536
 
571
540
        Note: The caller is expected to take a read-lock before calling this.
572
541
        """
573
542
        relpath = self.id2path(file_id)
574
 
        if relpath is None:
575
 
            raise AssertionError(
576
 
                "path for id {%s} is None!" % file_id)
 
543
        assert relpath != None, \
 
544
            "path for id {%s} is None!" % file_id
577
545
        return self._kind(relpath)
578
546
 
579
547
    def _kind(self, relpath):
580
548
        abspath = self.abspath(relpath)
581
549
        kind = file_kind(abspath)
582
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
583
 
            entry = self._get_entry(path=relpath)
584
 
            if entry[1] is not None:
585
 
                if entry[1][0][0] == 't':
586
 
                    kind = 'tree-reference'
 
550
        if (self._repo_supports_tree_reference and
 
551
            kind == 'directory' and
 
552
            self._directory_is_tree_reference(relpath)):
 
553
            kind = 'tree-reference'
587
554
        return kind
588
555
 
589
556
    @needs_read_lock
596
563
            return _mod_revision.NULL_REVISION
597
564
 
598
565
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
566
        """See Branch.lock_read, and WorkingTree.unlock."""
603
567
        self.branch.lock_read()
604
568
        try:
605
569
            self._control_files.lock_read()
618
582
        except:
619
583
            self.branch.unlock()
620
584
            raise
621
 
        return LogicalLockResult(self.unlock)
622
585
 
623
586
    def _lock_self_write(self):
624
587
        """This should be called after the branch is locked."""
639
602
        except:
640
603
            self.branch.unlock()
641
604
            raise
642
 
        return LogicalLockResult(self.unlock)
643
605
 
644
606
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
607
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
608
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
609
        self._lock_self_write()
651
610
 
652
611
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
612
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
613
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
614
        self._lock_self_write()
659
615
 
660
616
    @needs_tree_write_lock
661
617
    def move(self, from_paths, to_dir, after=False):
663
619
        result = []
664
620
        if not from_paths:
665
621
            return result
 
622
 
666
623
        state = self.current_dirstate()
667
 
        if isinstance(from_paths, basestring):
668
 
            raise ValueError()
 
624
 
 
625
        assert not isinstance(from_paths, basestring)
669
626
        to_dir_utf8 = to_dir.encode('utf8')
670
627
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
671
628
        id_index = state._get_id_index()
692
649
 
693
650
        if self._inventory is not None:
694
651
            update_inventory = True
695
 
            inv = self.root_inventory
 
652
            inv = self.inventory
 
653
            to_dir_ie = inv[to_dir_id]
696
654
            to_dir_id = to_entry[0][2]
697
 
            to_dir_ie = inv[to_dir_id]
698
655
        else:
699
656
            update_inventory = False
700
657
 
731
688
            from_entry = self._get_entry(path=from_rel)
732
689
            if from_entry == (None, None):
733
690
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
691
                    errors.NotVersionedError(path=str(from_rel)))
735
692
 
736
693
            from_id = from_entry[0][2]
737
694
            to_rel = pathjoin(to_dir, from_tail)
826
783
                if minikind == 'd':
827
784
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
828
785
                        """Recursively update all entries in this dirblock."""
829
 
                        if from_dir == '':
830
 
                            raise AssertionError("renaming root not supported")
 
786
                        assert from_dir != '', "renaming root not supported"
831
787
                        from_key = (from_dir, '')
832
788
                        from_block_idx, present = \
833
789
                            state._find_block_index_from_key(from_key)
846
802
 
847
803
                        # Grab a copy since move_one may update the list.
848
804
                        for entry in from_block[1][:]:
849
 
                            if not (entry[0][0] == from_dir):
850
 
                                raise AssertionError()
 
805
                            assert entry[0][0] == from_dir
851
806
                            cur_details = entry[1][0]
852
807
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
853
808
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
878
833
                rollback_rename()
879
834
                raise
880
835
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
836
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
837
            self._make_dirty(reset_inventory=False)
883
838
 
884
839
        return result
894
849
    @needs_read_lock
895
850
    def path2id(self, path):
896
851
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
852
        path = path.strip('/')
902
853
        entry = self._get_entry(path=path)
903
854
        if entry == (None, None):
916
867
        for tree in trees:
917
868
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
869
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
870
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
871
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
872
        # -- make all paths utf8 --
923
873
        paths_utf8 = set()
981
931
                    all_versioned = False
982
932
                    break
983
933
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
934
                raise errors.PathsNotVersionedError(paths)
986
935
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
936
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
937
        # sketch: 
989
938
        # for all search_indexs in each path at or under each element of
990
939
        # search_paths, if the detail is relocated: add the id, and add the
991
940
        # relocated path as one to search if its not searched already. If the
1037
986
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
987
            for dir_name in split_paths:
1039
988
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
989
                    raise errors.PathsNotVersionedError(paths)
1042
990
 
1043
991
        for dir_name_id, trees_info in found.iteritems():
1044
992
            for index in search_indexes:
1048
996
 
1049
997
    def read_working_inventory(self):
1050
998
        """Read the working inventory.
1051
 
 
 
999
        
1052
1000
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
1001
        """
1054
 
        return self.root_inventory
 
1002
        return self.inventory
1055
1003
 
1056
1004
    @needs_read_lock
1057
1005
    def revision_tree(self, revision_id):
1072
1020
    def set_last_revision(self, new_revision):
1073
1021
        """Change the last revision in the working tree."""
1074
1022
        parents = self.get_parent_ids()
1075
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1076
 
            if len(parents) >= 2:
1077
 
                raise AssertionError(
1078
 
                    "setting the last parent to none with a pending merge is "
1079
 
                    "unsupported.")
 
1023
        if new_revision in (NULL_REVISION, None):
 
1024
            assert len(parents) < 2, (
 
1025
                "setting the last parent to none with a pending merge is "
 
1026
                "unsupported.")
1080
1027
            self.set_parent_ids([])
1081
1028
        else:
1082
1029
            self.set_parent_ids([new_revision] + parents[1:],
1085
1032
    @needs_tree_write_lock
1086
1033
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
1034
        """Set the parent ids to revision_ids.
1088
 
 
 
1035
        
1089
1036
        See also set_parent_trees. This api will try to retrieve the tree data
1090
1037
        for each element of revision_ids from the trees repository. If you have
1091
1038
        tree data already available, it is more efficient to use
1123
1070
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1124
1071
        real_trees = []
1125
1072
        ghosts = []
1126
 
 
1127
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1128
 
        graph = self.branch.repository.get_graph()
1129
 
        heads = graph.heads(parent_ids)
1130
 
        accepted_revisions = set()
1131
 
 
1132
1073
        # convert absent trees to the null tree, which we convert back to
1133
1074
        # missing on access.
1134
1075
        for rev_id, tree in parents_list:
1135
 
            if len(accepted_revisions) > 0:
1136
 
                # we always accept the first tree
1137
 
                if rev_id in accepted_revisions or rev_id not in heads:
1138
 
                    # We have already included either this tree, or its
1139
 
                    # descendent, so we skip it.
1140
 
                    continue
1141
1076
            _mod_revision.check_not_reserved_id(rev_id)
1142
1077
            if tree is not None:
1143
1078
                real_trees.append((rev_id, tree))
1144
1079
            else:
1145
1080
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1081
                    self.branch.repository.revision_tree(None)))
1148
1082
                ghosts.append(rev_id)
1149
 
            accepted_revisions.add(rev_id)
1150
 
        updated = False
1151
 
        if (len(real_trees) == 1
1152
 
            and not ghosts
1153
 
            and self.branch.repository._format.fast_deltas
1154
 
            and isinstance(real_trees[0][1],
1155
 
                revisiontree.InventoryRevisionTree)
1156
 
            and self.get_parent_ids()):
1157
 
            rev_id, rev_tree = real_trees[0]
1158
 
            basis_id = self.get_parent_ids()[0]
1159
 
            # There are times when basis_tree won't be in
1160
 
            # self.branch.repository, (switch, for example)
1161
 
            try:
1162
 
                basis_tree = self.branch.repository.revision_tree(basis_id)
1163
 
            except errors.NoSuchRevision:
1164
 
                # Fall back to the set_parent_trees(), since we can't use
1165
 
                # _make_delta if we can't get the RevisionTree
1166
 
                pass
1167
 
            else:
1168
 
                delta = rev_tree.root_inventory._make_delta(
1169
 
                    basis_tree.root_inventory)
1170
 
                dirstate.update_basis_by_delta(delta, rev_id)
1171
 
                updated = True
1172
 
        if not updated:
1173
 
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1083
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1174
1084
        self._make_dirty(reset_inventory=False)
1175
1085
 
1176
1086
    def _set_root_id(self, file_id):
1196
1106
 
1197
1107
    def unlock(self):
1198
1108
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1109
        # do non-implementation specific cleanup
 
1110
        self._cleanup()
 
1111
 
1199
1112
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1113
            # eventually we should do signature checking during read locks for
1204
1114
            # dirstate updates.
1205
1115
            if self._control_files._lock_mode == 'w':
1268
1178
                # just forget the whole block.
1269
1179
                entry_index = 0
1270
1180
                while entry_index < len(block[1]):
 
1181
                    # Mark this file id as having been removed
1271
1182
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1183
                    ids_to_unversion.discard(entry[0][2])
 
1184
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1185
                                               # entries
 
1186
                        or not state._make_absent(entry)):
1274
1187
                        entry_index += 1
1275
 
                    else:
1276
 
                        # Mark this file id as having been removed
1277
 
                        ids_to_unversion.discard(entry[0][2])
1278
 
                        if not state._make_absent(entry):
1279
 
                            # The block has not shrunk.
1280
 
                            entry_index += 1
1281
1188
                # go to the next block. (At the moment we dont delete empty
1282
1189
                # dirblocks)
1283
1190
                block_index += 1
1304
1211
        # have to change the legacy inventory too.
1305
1212
        if self._inventory is not None:
1306
1213
            for file_id in file_ids:
1307
 
                if self._inventory.has_id(file_id):
1308
 
                    self._inventory.remove_recursive_id(file_id)
1309
 
 
1310
 
    @needs_tree_write_lock
1311
 
    def rename_one(self, from_rel, to_rel, after=False):
1312
 
        """See WorkingTree.rename_one"""
1313
 
        self.flush()
1314
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
1315
 
 
1316
 
    @needs_tree_write_lock
1317
 
    def apply_inventory_delta(self, changes):
1318
 
        """See MutableTree.apply_inventory_delta"""
1319
 
        state = self.current_dirstate()
1320
 
        state.update_by_delta(changes)
1321
 
        self._make_dirty(reset_inventory=True)
 
1214
                self._inventory.remove_recursive_id(file_id)
1322
1215
 
1323
1216
    def update_basis_by_delta(self, new_revid, delta):
1324
1217
        """See MutableTree.update_basis_by_delta."""
1325
 
        if self.last_revision() == new_revid:
1326
 
            raise AssertionError()
 
1218
        assert self.last_revision() != new_revid
1327
1219
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1328
1220
 
1329
1221
    @needs_read_lock
1333
1225
    @needs_tree_write_lock
1334
1226
    def _write_inventory(self, inv):
1335
1227
        """Write inventory as the current inventory."""
1336
 
        if self._dirty:
1337
 
            raise AssertionError("attempting to write an inventory when the "
1338
 
                "dirstate is dirty will lose pending changes")
1339
 
        had_inventory = self._inventory is not None
1340
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1341
 
        # working inventory. We do this because self.inventory may be inv, or
1342
 
        # may have been modified, and either case would prevent a clean delta
1343
 
        # being created.
1344
 
        self._inventory = None
1345
 
        # generate a delta,
1346
 
        delta = inv._make_delta(self.root_inventory)
1347
 
        # and apply it.
1348
 
        self.apply_inventory_delta(delta)
1349
 
        if had_inventory:
 
1228
        assert not self._dirty, ("attempting to write an inventory when the "
 
1229
            "dirstate is dirty will cause data loss")
 
1230
        self.current_dirstate().set_state_from_inventory(inv)
 
1231
        self._make_dirty(reset_inventory=False)
 
1232
        if self._inventory is not None:
1350
1233
            self._inventory = inv
1351
1234
        self.flush()
1352
1235
 
1353
 
    @needs_tree_write_lock
1354
 
    def reset_state(self, revision_ids=None):
1355
 
        """Reset the state of the working tree.
1356
 
 
1357
 
        This does a hard-reset to a last-known-good state. This is a way to
1358
 
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
1359
 
        """
1360
 
        if revision_ids is None:
1361
 
            revision_ids = self.get_parent_ids()
1362
 
        if not revision_ids:
1363
 
            base_tree = self.branch.repository.revision_tree(
1364
 
                _mod_revision.NULL_REVISION)
1365
 
            trees = []
1366
 
        else:
1367
 
            trees = zip(revision_ids,
1368
 
                        self.branch.repository.revision_trees(revision_ids))
1369
 
            base_tree = trees[0][1]
1370
 
        state = self.current_dirstate()
1371
 
        # We don't support ghosts yet
1372
 
        state.set_state_from_scratch(base_tree.root_inventory, trees, [])
1373
 
 
1374
 
 
1375
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1376
 
 
1377
 
    def __init__(self, tree):
1378
 
        self.tree = tree
1379
 
 
1380
 
    def sha1(self, abspath):
1381
 
        """See dirstate.SHA1Provider.sha1()."""
1382
 
        filters = self.tree._content_filter_stack(
1383
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1384
 
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
1385
 
 
1386
 
    def stat_and_sha1(self, abspath):
1387
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1388
 
        filters = self.tree._content_filter_stack(
1389
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1390
 
        file_obj = file(abspath, 'rb', 65000)
1391
 
        try:
1392
 
            statvalue = os.fstat(file_obj.fileno())
1393
 
            if filters:
1394
 
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
1395
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1396
 
        finally:
1397
 
            file_obj.close()
1398
 
        return statvalue, sha1
1399
 
 
1400
 
 
1401
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1402
 
    """Dirstate working tree that supports content filtering.
1403
 
 
1404
 
    The dirstate holds the hash and size of the canonical form of the file, 
1405
 
    and most methods must return that.
1406
 
    """
1407
 
 
1408
 
    def _file_content_summary(self, path, stat_result):
1409
 
        # This is to support the somewhat obsolete path_content_summary method
1410
 
        # with content filtering: see
1411
 
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
1412
 
        #
1413
 
        # If the dirstate cache is up to date and knows the hash and size,
1414
 
        # return that.
1415
 
        # Otherwise if there are no content filters, return the on-disk size
1416
 
        # and leave the hash blank.
1417
 
        # Otherwise, read and filter the on-disk file and use its size and
1418
 
        # hash.
1419
 
        #
1420
 
        # The dirstate doesn't store the size of the canonical form so we
1421
 
        # can't trust it for content-filtered trees.  We just return None.
1422
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1423
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1424
 
        return ('file', None, executable, dirstate_sha1)
1425
 
 
1426
 
 
1427
 
class WorkingTree4(DirStateWorkingTree):
1428
 
    """This is the Format 4 working tree.
1429
 
 
1430
 
    This differs from WorkingTree by:
1431
 
     - Having a consolidated internal dirstate, stored in a
1432
 
       randomly-accessible sorted file on disk.
1433
 
     - Not having a regular inventory attribute.  One can be synthesized
1434
 
       on demand but this is expensive and should be avoided.
1435
 
 
1436
 
    This is new in bzr 0.15.
1437
 
    """
1438
 
 
1439
 
 
1440
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1441
 
    """This is the Format 5 working tree.
1442
 
 
1443
 
    This differs from WorkingTree4 by:
1444
 
     - Supporting content filtering.
1445
 
 
1446
 
    This is new in bzr 1.11.
1447
 
    """
1448
 
 
1449
 
 
1450
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1451
 
    """This is the Format 6 working tree.
1452
 
 
1453
 
    This differs from WorkingTree5 by:
1454
 
     - Supporting a current view that may mask the set of files in a tree
1455
 
       impacted by most user operations.
1456
 
 
1457
 
    This is new in bzr 1.14.
1458
 
    """
1459
 
 
1460
 
    def _make_views(self):
1461
 
        return views.PathBasedViews(self)
1462
 
 
1463
 
 
1464
 
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
1465
 
 
1466
 
    missing_parent_conflicts = True
1467
 
 
1468
 
    supports_versioned_directories = True
1469
 
 
1470
 
    _lock_class = LockDir
1471
 
    _lock_file_name = 'lock'
1472
 
 
1473
 
    def _open_control_files(self, a_bzrdir):
1474
 
        transport = a_bzrdir.get_workingtree_transport(None)
1475
 
        return LockableFiles(transport, self._lock_file_name,
1476
 
                             self._lock_class)
1477
 
 
1478
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1479
 
                   accelerator_tree=None, hardlink=False):
 
1236
 
 
1237
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1238
    """The first consolidated dirstate working tree format.
 
1239
 
 
1240
    This format:
 
1241
        - exists within a metadir controlling .bzr
 
1242
        - includes an explicit version marker for the workingtree control
 
1243
          files, separate from the BzrDir format
 
1244
        - modifies the hash cache format
 
1245
        - is new in bzr 0.15
 
1246
        - uses a LockDir to guard access to it.
 
1247
    """
 
1248
 
 
1249
    upgrade_recommended = False
 
1250
 
 
1251
    def get_format_string(self):
 
1252
        """See WorkingTreeFormat.get_format_string()."""
 
1253
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1254
 
 
1255
    def get_format_description(self):
 
1256
        """See WorkingTreeFormat.get_format_description()."""
 
1257
        return "Working tree format 4"
 
1258
 
 
1259
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None):
1480
1260
        """See WorkingTreeFormat.initialize().
1481
1261
 
1482
1262
        :param revision_id: allows creating a working tree at a different
1483
 
            revision than the branch is at.
1484
 
        :param accelerator_tree: A tree which can be used for retrieving file
1485
 
            contents more quickly than the revision tree, i.e. a workingtree.
1486
 
            The revision tree will be used for cases where accelerator_tree's
1487
 
            content is different.
1488
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1489
 
            where possible.
 
1263
        revision than the branch is at.
1490
1264
 
1491
1265
        These trees get an initial random root id, if their repository supports
1492
1266
        rich root data, TREE_ROOT otherwise.
1497
1271
        control_files = self._open_control_files(a_bzrdir)
1498
1272
        control_files.create_lock()
1499
1273
        control_files.lock_write()
1500
 
        transport.put_bytes('format', self.as_string(),
1501
 
            mode=a_bzrdir._get_file_mode())
 
1274
        control_files.put_utf8('format', self.get_format_string())
1502
1275
        if from_branch is not None:
1503
1276
            branch = from_branch
1504
1277
        else:
1510
1283
        state = dirstate.DirState.initialize(local_path)
1511
1284
        state.unlock()
1512
1285
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1286
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1287
                         branch,
1515
1288
                         _format=self,
1516
1289
                         _bzrdir=a_bzrdir,
1518
1291
        wt._new_tree()
1519
1292
        wt.lock_tree_write()
1520
1293
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1294
            if revision_id in (None, NULL_REVISION):
1523
1295
                if branch.repository.supports_rich_root():
1524
1296
                    wt._set_root_id(generate_ids.gen_root_id())
1525
1297
                else:
1526
1298
                    wt._set_root_id(ROOT_ID)
1527
1299
                wt.flush()
1528
 
            basis = None
1529
 
            # frequently, we will get here due to branching.  The accelerator
1530
 
            # tree will be the tree from the branch, so the desired basis
1531
 
            # tree will often be a parent of the accelerator tree.
1532
 
            if accelerator_tree is not None:
1533
 
                try:
1534
 
                    basis = accelerator_tree.revision_tree(revision_id)
1535
 
                except errors.NoSuchRevision:
1536
 
                    pass
1537
 
            if basis is None:
1538
 
                basis = branch.repository.revision_tree(revision_id)
1539
 
            if revision_id == _mod_revision.NULL_REVISION:
1540
 
                parents_list = []
1541
 
            else:
1542
 
                parents_list = [(revision_id, basis)]
 
1300
            wt.set_last_revision(revision_id)
 
1301
            wt.flush()
 
1302
            basis = wt.basis_tree()
1543
1303
            basis.lock_read()
1544
 
            try:
1545
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1304
            # if the basis has a root id we have to use that; otherwise we use
 
1305
            # a new random one
 
1306
            basis_root_id = basis.get_root_id()
 
1307
            if basis_root_id is not None:
 
1308
                wt._set_root_id(basis_root_id)
1546
1309
                wt.flush()
1547
 
                # if the basis has a root id we have to use that; otherwise we
1548
 
                # use a new random one
1549
 
                basis_root_id = basis.get_root_id()
1550
 
                if basis_root_id is not None:
1551
 
                    wt._set_root_id(basis_root_id)
1552
 
                    wt.flush()
1553
 
                if wt.supports_content_filtering():
1554
 
                    # The original tree may not have the same content filters
1555
 
                    # applied so we can't safely build the inventory delta from
1556
 
                    # the source tree.
1557
 
                    delta_from_tree = False
1558
 
                else:
1559
 
                    delta_from_tree = True
1560
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1561
 
                # because wt4.apply_inventory_delta does not mutate the input
1562
 
                # inventory entries.
1563
 
                transform.build_tree(basis, wt, accelerator_tree,
1564
 
                                     hardlink=hardlink,
1565
 
                                     delta_from_tree=delta_from_tree)
1566
 
                for hook in MutableTree.hooks['post_build_tree']:
1567
 
                    hook(wt)
1568
 
            finally:
1569
 
                basis.unlock()
 
1310
            transform.build_tree(basis, wt)
 
1311
            basis.unlock()
1570
1312
        finally:
1571
1313
            control_files.unlock()
1572
1314
            wt.unlock()
1573
1315
        return wt
1574
1316
 
1575
 
    def _init_custom_control_files(self, wt):
1576
 
        """Subclasses with custom control files should override this method.
1577
 
 
1578
 
        The working tree and control files are locked for writing when this
1579
 
        method is called.
1580
 
 
1581
 
        :param wt: the WorkingTree object
1582
 
        """
1583
 
 
1584
 
    def open(self, a_bzrdir, _found=False):
1585
 
        """Return the WorkingTree object for a_bzrdir
1586
 
 
1587
 
        _found is a private parameter, do not use it. It is used to indicate
1588
 
               if format probing has already been done.
1589
 
        """
1590
 
        if not _found:
1591
 
            # we are being called directly and must probe.
1592
 
            raise NotImplementedError
1593
 
        if not isinstance(a_bzrdir.transport, LocalTransport):
1594
 
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1595
 
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
1596
 
        return wt
1597
 
 
1598
1317
    def _open(self, a_bzrdir, control_files):
1599
1318
        """Open the tree itself.
1600
1319
 
1601
1320
        :param a_bzrdir: the dir for the tree.
1602
1321
        :param control_files: the control files for the tree.
1603
1322
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1323
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1324
                           branch=a_bzrdir.open_branch(),
1606
1325
                           _format=self,
1607
1326
                           _bzrdir=a_bzrdir,
1608
1327
                           _control_files=control_files)
1609
1328
 
1610
1329
    def __get_matchingbzrdir(self):
1611
 
        return self._get_matchingbzrdir()
1612
 
 
1613
 
    def _get_matchingbzrdir(self):
1614
 
        """Overrideable method to get a bzrdir for testing."""
1615
1330
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1331
        return bzrdir.format_registry.make_bzrdir(
 
1332
            'dirstate-with-subtree')
1618
1333
 
1619
1334
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1335
 
1621
1336
 
1622
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1623
 
    """The first consolidated dirstate working tree format.
1624
 
 
1625
 
    This format:
1626
 
        - exists within a metadir controlling .bzr
1627
 
        - includes an explicit version marker for the workingtree control
1628
 
          files, separate from the ControlDir format
1629
 
        - modifies the hash cache format
1630
 
        - is new in bzr 0.15
1631
 
        - uses a LockDir to guard access to it.
1632
 
    """
1633
 
 
1634
 
    upgrade_recommended = False
1635
 
 
1636
 
    _tree_class = WorkingTree4
1637
 
 
1638
 
    @classmethod
1639
 
    def get_format_string(cls):
1640
 
        """See WorkingTreeFormat.get_format_string()."""
1641
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1642
 
 
1643
 
    def get_format_description(self):
1644
 
        """See WorkingTreeFormat.get_format_description()."""
1645
 
        return "Working tree format 4"
1646
 
 
1647
 
 
1648
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1649
 
    """WorkingTree format supporting content filtering.
1650
 
    """
1651
 
 
1652
 
    upgrade_recommended = False
1653
 
 
1654
 
    _tree_class = WorkingTree5
1655
 
 
1656
 
    @classmethod
1657
 
    def get_format_string(cls):
1658
 
        """See WorkingTreeFormat.get_format_string()."""
1659
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1660
 
 
1661
 
    def get_format_description(self):
1662
 
        """See WorkingTreeFormat.get_format_description()."""
1663
 
        return "Working tree format 5"
1664
 
 
1665
 
    def supports_content_filtering(self):
1666
 
        return True
1667
 
 
1668
 
 
1669
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1670
 
    """WorkingTree format supporting views.
1671
 
    """
1672
 
 
1673
 
    upgrade_recommended = False
1674
 
 
1675
 
    _tree_class = WorkingTree6
1676
 
 
1677
 
    @classmethod
1678
 
    def get_format_string(cls):
1679
 
        """See WorkingTreeFormat.get_format_string()."""
1680
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1681
 
 
1682
 
    def get_format_description(self):
1683
 
        """See WorkingTreeFormat.get_format_description()."""
1684
 
        return "Working tree format 6"
1685
 
 
1686
 
    def _init_custom_control_files(self, wt):
1687
 
        """Subclasses with custom control files should override this method."""
1688
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1689
 
 
1690
 
    def supports_content_filtering(self):
1691
 
        return True
1692
 
 
1693
 
    def supports_views(self):
1694
 
        return True
1695
 
 
1696
 
    def _get_matchingbzrdir(self):
1697
 
        """Overrideable method to get a bzrdir for testing."""
1698
 
        # We use 'development-subtree' instead of '2a', because we have a
1699
 
        # few tests that want to test tree references
1700
 
        return bzrdir.format_registry.make_bzrdir('development-subtree')
1701
 
 
1702
 
 
1703
 
class DirStateRevisionTree(InventoryTree):
1704
 
    """A revision tree pulling the inventory from a dirstate.
1705
 
    
1706
 
    Note that this is one of the historical (ie revision) trees cached in the
1707
 
    dirstate for easy access, not the workingtree.
1708
 
    """
 
1337
class DirStateRevisionTree(Tree):
 
1338
    """A revision tree pulling the inventory from a dirstate."""
1709
1339
 
1710
1340
    def __init__(self, dirstate, revision_id, repository):
1711
1341
        self._dirstate = dirstate
1714
1344
        self._inventory = None
1715
1345
        self._locked = 0
1716
1346
        self._dirstate_locked = False
1717
 
        self._repo_supports_tree_reference = getattr(
1718
 
            repository._format, "supports_tree_reference",
1719
 
            False)
1720
1347
 
1721
1348
    def __repr__(self):
1722
1349
        return "<%s of %s in %s>" % \
1725
1352
    def annotate_iter(self, file_id,
1726
1353
                      default_revision=_mod_revision.CURRENT_REVISION):
1727
1354
        """See Tree.annotate_iter"""
1728
 
        text_key = (file_id, self.get_file_revision(file_id))
1729
 
        annotations = self._repository.texts.annotate(text_key)
1730
 
        return [(key[-1], line) for (key, line) in annotations]
 
1355
        w = self._get_weave(file_id)
 
1356
        return w.annotate_iter(self.inventory[file_id].revision)
1731
1357
 
 
1358
    def _get_ancestors(self, default_revision):
 
1359
        return set(self._repository.get_ancestry(self._revision_id,
 
1360
                                                 topo_sorted=False))
1732
1361
    def _comparison_data(self, entry, path):
1733
1362
        """See Tree._comparison_data."""
1734
1363
        if entry is None:
1759
1388
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1760
1389
        return path_utf8.decode('utf8')
1761
1390
 
1762
 
    def iter_references(self):
1763
 
        if not self._repo_supports_tree_reference:
1764
 
            # When the repo doesn't support references, we will have nothing to
1765
 
            # return
1766
 
            return iter([])
1767
 
        # Otherwise, fall back to the default implementation
1768
 
        return super(DirStateRevisionTree, self).iter_references()
1769
 
 
1770
1391
    def _get_parent_index(self):
1771
1392
        """Return the index in the dirstate referenced by this tree."""
1772
1393
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1777
1398
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1399
        If both are supplied, the fastest lookup is used, and an error is
1779
1400
        raised if they do not both point at the same row.
1780
 
 
 
1401
        
1781
1402
        :param file_id: An optional unicode file_id to be looked up.
1782
1403
        :param path: An optional unicode path to be looked up.
1783
1404
        :return: The dirstate row tuple for path/file_id, or (None, None)
1787
1408
        if path is not None:
1788
1409
            path = path.encode('utf8')
1789
1410
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1411
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1412
 
1793
1413
    def _generate_inventory(self):
1794
1414
        """Create and set self.inventory from the dirstate object.
1798
1418
 
1799
1419
        This is relatively expensive: we have to walk the entire dirstate.
1800
1420
        """
1801
 
        if not self._locked:
1802
 
            raise AssertionError(
1803
 
                'cannot generate inventory of an unlocked '
1804
 
                'dirstate revision tree')
 
1421
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1422
            'dirstate revision tree'
1805
1423
        # separate call for profiling - makes it clear where the costs are.
1806
1424
        self._dirstate._read_dirblocks_if_needed()
1807
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1808
 
            raise AssertionError(
1809
 
                'parent %s has disappeared from %s' % (
1810
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1425
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1426
            'parent %s has disappeared from %s' % (
 
1427
            self._revision_id, self._dirstate.get_parent_ids())
1811
1428
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1812
1429
        # This is identical now to the WorkingTree _generate_inventory except
1813
1430
        # for the tree index use.
1814
1431
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1815
1432
        current_id = root_key[2]
1816
 
        if current_entry[parent_index][0] != 'd':
1817
 
            raise AssertionError()
 
1433
        assert current_entry[parent_index][0] == 'd'
1818
1434
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1819
1435
        inv.root.revision = current_entry[parent_index][4]
1820
1436
        # Turn some things into local variables
1851
1467
                elif kind == 'directory':
1852
1468
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1469
                elif kind == 'symlink':
 
1470
                    inv_entry.executable = False
 
1471
                    inv_entry.text_size = None
1854
1472
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1473
                elif kind == 'tree-reference':
1856
1474
                    inv_entry.reference_revision = fingerprint or None
1858
1476
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1859
1477
                            % entry)
1860
1478
                # These checks cost us around 40ms on a 55k entry tree
1861
 
                if file_id in inv_byid:
1862
 
                    raise AssertionError('file_id %s already in'
1863
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1864
 
                if name_unicode in parent_ie.children:
1865
 
                    raise AssertionError('name %r already in parent'
1866
 
                        % (name_unicode,))
 
1479
                assert file_id not in inv_byid
 
1480
                assert name_unicode not in parent_ie.children
1867
1481
                inv_byid[file_id] = inv_entry
1868
1482
                parent_ie.children[name_unicode] = inv_entry
1869
1483
        self._inventory = inv
1876
1490
        # Make sure the file exists
1877
1491
        entry = self._get_entry(file_id, path=path)
1878
1492
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1493
            return None
1880
1494
        parent_index = self._get_parent_index()
1881
1495
        last_changed_revision = entry[1][parent_index][4]
1882
 
        try:
1883
 
            rev = self._repository.get_revision(last_changed_revision)
1884
 
        except errors.NoSuchRevision:
1885
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1886
 
        return rev.timestamp
 
1496
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1497
 
1888
1498
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1499
        entry = self._get_entry(file_id=file_id, path=path)
1893
1503
            return parent_details[1]
1894
1504
        return None
1895
1505
 
1896
 
    @needs_read_lock
1897
 
    def get_file_revision(self, file_id):
1898
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1899
 
        return inv[inv_file_id].revision
 
1506
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1507
    def get_weave(self, file_id):
 
1508
        return self._get_weave(file_id)
 
1509
 
 
1510
    def _get_weave(self, file_id):
 
1511
        return self._repository.weave_store.get_weave(file_id,
 
1512
                self._repository.get_transaction())
1900
1513
 
1901
1514
    def get_file(self, file_id, path=None):
1902
1515
        return StringIO(self.get_file_text(file_id))
1903
1516
 
 
1517
    def get_file_lines(self, file_id):
 
1518
        entry = self._get_entry(file_id=file_id)[1]
 
1519
        if entry == None:
 
1520
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1521
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1522
 
1904
1523
    def get_file_size(self, file_id):
1905
 
        """See Tree.get_file_size"""
1906
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1907
 
        return inv[inv_file_id].text_size
 
1524
        return self.inventory[file_id].text_size
1908
1525
 
1909
 
    def get_file_text(self, file_id, path=None):
1910
 
        content = None
1911
 
        for _, content_iter in self.iter_files_bytes([(file_id, None)]):
1912
 
            if content is not None:
1913
 
                raise AssertionError('iter_files_bytes returned'
1914
 
                    ' too many entries')
1915
 
            # For each entry returned by iter_files_bytes, we must consume the
1916
 
            # content_iter before we step the files iterator.
1917
 
            content = ''.join(content_iter)
1918
 
        if content is None:
1919
 
            raise AssertionError('iter_files_bytes did not return'
1920
 
                ' the requested data')
1921
 
        return content
 
1526
    def get_file_text(self, file_id):
 
1527
        return ''.join(self.get_file_lines(file_id))
1922
1528
 
1923
1529
    def get_reference_revision(self, file_id, path=None):
1924
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1925
 
        return inv[inv_file_id].reference_revision
 
1530
        return self.inventory[file_id].reference_revision
1926
1531
 
1927
1532
    def iter_files_bytes(self, desired_files):
1928
1533
        """See Tree.iter_files_bytes.
1938
1543
                                       identifier))
1939
1544
        return self._repository.iter_files_bytes(repo_desired_files)
1940
1545
 
1941
 
    def get_symlink_target(self, file_id, path=None):
 
1546
    def get_symlink_target(self, file_id):
1942
1547
        entry = self._get_entry(file_id=file_id)
1943
1548
        parent_index = self._get_parent_index()
1944
1549
        if entry[1][parent_index][0] != 'l':
1945
1550
            return None
1946
1551
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1552
            # At present, none of the tree implementations supports non-ascii
 
1553
            # symlink targets. So we will just assume that the dirstate path is
 
1554
            # correct.
 
1555
            return entry[1][parent_index][1]
1950
1556
 
1951
1557
    def get_revision_id(self):
1952
1558
        """Return the revision id for this tree."""
1953
1559
        return self._revision_id
1954
1560
 
1955
 
    def _get_root_inventory(self):
 
1561
    def _get_inventory(self):
1956
1562
        if self._inventory is not None:
1957
1563
            return self._inventory
1958
1564
        self._must_be_locked()
1959
1565
        self._generate_inventory()
1960
1566
        return self._inventory
1961
1567
 
1962
 
    root_inventory = property(_get_root_inventory,
1963
 
                         doc="Inventory of this Tree")
1964
 
 
1965
 
    @deprecated_method(deprecated_in((2, 5, 0)))
1966
 
    def _get_inventory(self):
1967
 
        return self.root_inventory
1968
 
 
1969
1568
    inventory = property(_get_inventory,
1970
1569
                         doc="Inventory of this Tree")
1971
1570
 
1978
1577
 
1979
1578
    def kind(self, file_id):
1980
1579
        entry = self._get_entry(file_id=file_id)[1]
1981
 
        if entry is None:
 
1580
        if entry == None:
1982
1581
            raise errors.NoSuchId(tree=self, file_id=file_id)
1983
 
        parent_index = self._get_parent_index()
1984
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1985
 
 
1986
 
    def stored_kind(self, file_id):
1987
 
        """See Tree.stored_kind"""
1988
 
        return self.kind(file_id)
 
1582
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1989
1583
 
1990
1584
    def path_content_summary(self, path):
1991
1585
        """See Tree.path_content_summary."""
1992
 
        inv, inv_file_id = self._path2inv_file_id(path)
1993
 
        if inv_file_id is None:
 
1586
        id = self.inventory.path2id(path)
 
1587
        if id is None:
1994
1588
            return ('missing', None, None, None)
1995
 
        entry = inv[inv_file_id]
 
1589
        entry = self._inventory[id]
1996
1590
        kind = entry.kind
1997
1591
        if kind == 'file':
1998
1592
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
2002
1596
            return (kind, None, None, None)
2003
1597
 
2004
1598
    def is_executable(self, file_id, path=None):
2005
 
        inv, inv_file_id = self._unpack_file_id(file_id)
2006
 
        ie = inv[inv_file_id]
 
1599
        ie = self.inventory[file_id]
2007
1600
        if ie.kind != "file":
2008
 
            return False
 
1601
            return None
2009
1602
        return ie.executable
2010
1603
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1604
    def list_files(self, include_root=False):
2015
1605
        # We use a standard implementation, because DirStateRevisionTree is
2016
1606
        # dealing with one of the parents of the current state
2017
 
        if from_dir is None:
2018
 
            inv = self.root_inventory
2019
 
            from_dir_id = None
2020
 
        else:
2021
 
            inv, from_dir_id = self._path2inv_file_id(from_dir)
2022
 
            if from_dir_id is None:
2023
 
                # Directory not versioned
2024
 
                return
2025
 
        # FIXME: Support nested trees
2026
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
2027
 
        if inv.root is not None and not include_root and from_dir is None:
 
1607
        inv = self._get_inventory()
 
1608
        entries = inv.iter_entries()
 
1609
        if self.inventory.root is not None and not include_root:
2028
1610
            entries.next()
2029
1611
        for path, entry in entries:
2030
1612
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1613
 
2032
1614
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1615
        """Lock the tree for a set of operations."""
2037
1616
        if not self._locked:
2038
1617
            self._repository.lock_read()
2039
1618
            if self._dirstate._lock_token is None:
2040
1619
                self._dirstate.lock_read()
2041
1620
                self._dirstate_locked = True
2042
1621
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1622
 
2045
1623
    def _must_be_locked(self):
2046
1624
        if not self._locked:
2050
1628
    def path2id(self, path):
2051
1629
        """Return the id for path in this tree."""
2052
1630
        # lookup by path: faster than splitting and walking the ivnentory.
2053
 
        if isinstance(path, list):
2054
 
            if path == []:
2055
 
                path = [""]
2056
 
            path = osutils.pathjoin(*path)
2057
1631
        entry = self._get_entry(path=path)
2058
1632
        if entry == (None, None):
2059
1633
            return None
2071
1645
                self._dirstate_locked = False
2072
1646
            self._repository.unlock()
2073
1647
 
2074
 
    @needs_read_lock
2075
 
    def supports_tree_reference(self):
2076
 
        return self._repo_supports_tree_reference
2077
 
 
2078
1648
    def walkdirs(self, prefix=""):
2079
1649
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1650
        # implementation based on an inventory.  
2081
1651
        # This should be cleaned up to use the much faster Dirstate code
2082
1652
        # So for now, we just build up the parent inventory, and extract
2083
1653
        # it the same way RevisionTree does.
2084
1654
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1655
        inv = self._get_inventory()
2086
1656
        top_id = inv.path2id(prefix)
2087
1657
        if top_id is None:
2088
1658
            pending = []
2112
1682
 
2113
1683
class InterDirStateTree(InterTree):
2114
1684
    """Fast path optimiser for changes_from with dirstate trees.
2115
 
 
2116
 
    This is used only when both trees are in the dirstate working file, and
2117
 
    the source is any parent within the dirstate, and the destination is
 
1685
    
 
1686
    This is used only when both trees are in the dirstate working file, and 
 
1687
    the source is any parent within the dirstate, and the destination is 
2118
1688
    the current working tree of the same dirstate.
2119
1689
    """
2120
1690
    # this could be generalized to allow comparisons between any trees in the
2129
1699
    def make_source_parent_tree(source, target):
2130
1700
        """Change the source tree into a parent of the target."""
2131
1701
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1702
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1703
        target.set_parent_ids([revid])
2134
1704
        return target.basis_tree(), target
2135
1705
 
2136
 
    @classmethod
2137
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
2138
 
        result = klass.make_source_parent_tree(source, target)
2139
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2140
 
        return result
2141
 
 
2142
 
    @classmethod
2143
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2144
 
                                                  target):
2145
 
        from bzrlib.tests.test__dirstate_helpers import \
2146
 
            compiled_dirstate_helpers_feature
2147
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2148
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2149
 
        result = klass.make_source_parent_tree(source, target)
2150
 
        result[1]._iter_changes = ProcessEntryC
2151
 
        return result
2152
 
 
2153
1706
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1707
    _matching_to_tree_format = WorkingTreeFormat4()
2155
 
 
2156
 
    @classmethod
2157
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2158
 
        # This method shouldn't be called, because we have python and C
2159
 
        # specific flavours.
2160
 
        raise NotImplementedError
2161
 
 
2162
 
    def iter_changes(self, include_unchanged=False,
 
1708
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1709
 
 
1710
    def _iter_changes(self, include_unchanged=False,
2163
1711
                      specific_files=None, pb=None, extra_trees=[],
2164
1712
                      require_versioned=True, want_unversioned=False):
2165
1713
        """Return the changes from source to target.
2166
1714
 
2167
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1715
        :return: An iterator that yields tuples. See InterTree._iter_changes
2168
1716
            for details.
2169
1717
        :param specific_files: An optional list of file paths to restrict the
2170
1718
            comparison to. When mapping filenames to ids, all matches in all
2181
1729
            output. An unversioned file is defined as one with (False, False)
2182
1730
            for the versioned pair.
2183
1731
        """
 
1732
        utf8_decode = cache_utf8._utf8_decode
 
1733
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1734
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1735
        # NB: show_status depends on being able to pass in non-versioned files
 
1736
        # and report them as unknown
2184
1737
        # TODO: handle extra trees in the dirstate.
2185
 
        if (extra_trees or specific_files == []):
 
1738
        # TODO: handle comparisons as an empty tree as a different special
 
1739
        # case? mbp 20070226
 
1740
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1741
            or specific_files == []):
2186
1742
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1743
            for f in super(InterDirStateTree, self)._iter_changes(
2188
1744
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1745
                require_versioned, want_unversioned=want_unversioned):
 
1746
                yield f
 
1747
            return
2190
1748
        parent_ids = self.target.get_parent_ids()
2191
 
        if not (self.source._revision_id in parent_ids
2192
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2193
 
            raise AssertionError(
2194
 
                "revision {%s} is not stored in {%s}, but %s "
2195
 
                "can only be used for trees stored in the dirstate"
2196
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1749
        assert (self.source._revision_id in parent_ids), \
 
1750
                "revision {%s} is not stored in {%s}, but %s " \
 
1751
                "can only be used for trees stored in the dirstate" \
 
1752
                % (self.source._revision_id, self.target, self._iter_changes)
2197
1753
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1754
        if self.source._revision_id == NULL_REVISION:
2199
1755
            source_index = None
2200
1756
            indices = (target_index,)
2201
1757
        else:
2202
 
            if not (self.source._revision_id in parent_ids):
2203
 
                raise AssertionError(
2204
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2205
 
                    self.source._revision_id, parent_ids))
 
1758
            assert (self.source._revision_id in parent_ids), \
 
1759
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1760
                self.source._revision_id, parent_ids)
2206
1761
            source_index = 1 + parent_ids.index(self.source._revision_id)
2207
 
            indices = (source_index, target_index)
 
1762
            indices = (source_index,target_index)
2208
1763
        # -- make all specific_files utf8 --
2209
1764
        if specific_files:
2210
1765
            specific_files_utf8 = set()
2211
1766
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1767
                specific_files_utf8.add(path.encode('utf8'))
2215
1768
            specific_files = specific_files_utf8
2216
1769
        else:
2217
1770
            specific_files = set([''])
2218
1771
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1772
        # -- get the state object and prepare it.
2221
1773
        state = self.target.current_dirstate()
2222
1774
        state._read_dirblocks_if_needed()
 
1775
        def _entries_for_path(path):
 
1776
            """Return a list with all the entries that match path for all ids.
 
1777
            """
 
1778
            dirname, basename = os.path.split(path)
 
1779
            key = (dirname, basename, '')
 
1780
            block_index, present = state._find_block_index_from_key(key)
 
1781
            if not present:
 
1782
                # the block which should contain path is absent.
 
1783
                return []
 
1784
            result = []
 
1785
            block = state._dirblocks[block_index][1]
 
1786
            entry_index, _ = state._find_entry_index(key, block)
 
1787
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1788
            while (entry_index < len(block) and
 
1789
                block[entry_index][0][0:2] == key[0:2]):
 
1790
                result.append(block[entry_index])
 
1791
                entry_index += 1
 
1792
            return result
2223
1793
        if require_versioned:
2224
1794
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1795
            all_versioned = True
2226
1796
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1797
                path_entries = _entries_for_path(path)
2228
1798
                if not path_entries:
2229
1799
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1800
                    all_versioned = False
 
1801
                    break
2232
1802
                found_versioned = False
2233
1803
                # for each id at this path
2234
1804
                for entry in path_entries:
2241
1811
                if not found_versioned:
2242
1812
                    # none of the indexes was not 'absent' at all ids for this
2243
1813
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1814
                    all_versioned = False
 
1815
                    break
 
1816
            if not all_versioned:
 
1817
                raise errors.PathsNotVersionedError(specific_files)
2247
1818
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1819
        search_specific_files = set()
 
1820
        for path in specific_files:
 
1821
            other_specific_files = specific_files.difference(set([path]))
 
1822
            if not osutils.is_inside_any(other_specific_files, path):
 
1823
                # this is a top level path, we must check it.
 
1824
                search_specific_files.add(path)
 
1825
        # sketch: 
 
1826
        # compare source_index and target_index at or under each element of search_specific_files.
 
1827
        # follow the following comparison table. Note that we only want to do diff operations when
 
1828
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1829
        # for the target.
 
1830
        # cases:
 
1831
        # 
 
1832
        # Source | Target | disk | action
 
1833
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1834
        #        |        |      | diff check on source-target
 
1835
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1836
        #        |        |      | ???
 
1837
        #   r    |  a     |      | add source to search
 
1838
        #   r    |  a     |  a   | 
 
1839
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1840
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1841
        #   a    | fdlt   |      | add new id
 
1842
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1843
        #   a    |  a     |      | not present in either tree, skip
 
1844
        #   a    |  a     |  a   | not present in any tree, skip
 
1845
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1846
        #        |        |      | may not be selected by the users list of paths.
 
1847
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1848
        #        |        |      | may not be selected by the users list of paths.
 
1849
        #  fdlt  | fdlt   |      | content in both: diff them
 
1850
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1851
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1852
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1853
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1854
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1855
        #        |        |      | this id at the other path.
 
1856
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1857
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1858
        #        |        |      | this id at the other path.
 
1859
 
 
1860
        # for all search_indexs in each path at or under each element of
 
1861
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1862
        # relocated path as one to search if its not searched already. If the
 
1863
        # detail is not relocated, add the id.
 
1864
        searched_specific_files = set()
 
1865
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1866
        # Using a list so that we can access the values and change them in
 
1867
        # nested scope. Each one is [path, file_id, entry]
 
1868
        last_source_parent = [None, None]
 
1869
        last_target_parent = [None, None]
2249
1870
 
2250
1871
        use_filesystem_for_exec = (sys.platform != 'win32')
2251
 
        iter_changes = self.target._iter_changes(include_unchanged,
2252
 
            use_filesystem_for_exec, search_specific_files, state,
2253
 
            source_index, target_index, want_unversioned, self.target)
2254
 
        return iter_changes.iter_changes()
 
1872
 
 
1873
        # Just a sentry, so that _process_entry can say that this
 
1874
        # record is handled, but isn't interesting to process (unchanged)
 
1875
        uninteresting = object()
 
1876
 
 
1877
 
 
1878
        old_dirname_to_file_id = {}
 
1879
        new_dirname_to_file_id = {}
 
1880
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1881
        #       keeping a cache of directories that we have seen.
 
1882
 
 
1883
        def _process_entry(entry, path_info):
 
1884
            """Compare an entry and real disk to generate delta information.
 
1885
 
 
1886
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1887
                the path of entry. If None, then the path is considered absent.
 
1888
                (Perhaps we should pass in a concrete entry for this ?)
 
1889
                Basename is returned as a utf8 string because we expect this
 
1890
                tuple will be ignored, and don't want to take the time to
 
1891
                decode.
 
1892
            :return: None if these don't match
 
1893
                     A tuple of information about the change, or
 
1894
                     the object 'uninteresting' if these match, but are
 
1895
                     basically identical.
 
1896
            """
 
1897
            if source_index is None:
 
1898
                source_details = NULL_PARENT_DETAILS
 
1899
            else:
 
1900
                source_details = entry[1][source_index]
 
1901
            target_details = entry[1][target_index]
 
1902
            target_minikind = target_details[0]
 
1903
            if path_info is not None and target_minikind in 'fdlt':
 
1904
                assert target_index == 0
 
1905
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1906
                                                  stat_value=path_info[3])
 
1907
                # The entry may have been modified by update_entry
 
1908
                target_details = entry[1][target_index]
 
1909
                target_minikind = target_details[0]
 
1910
            else:
 
1911
                link_or_sha1 = None
 
1912
            file_id = entry[0][2]
 
1913
            source_minikind = source_details[0]
 
1914
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1915
                # claimed content in both: diff
 
1916
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1917
                #        |        |      | diff check on source-target
 
1918
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1919
                #        |        |      | ???
 
1920
                if source_minikind in 'r':
 
1921
                    # add the source to the search path to find any children it
 
1922
                    # has.  TODO ? : only add if it is a container ?
 
1923
                    if not osutils.is_inside_any(searched_specific_files,
 
1924
                                                 source_details[1]):
 
1925
                        search_specific_files.add(source_details[1])
 
1926
                    # generate the old path; this is needed for stating later
 
1927
                    # as well.
 
1928
                    old_path = source_details[1]
 
1929
                    old_dirname, old_basename = os.path.split(old_path)
 
1930
                    path = pathjoin(entry[0][0], entry[0][1])
 
1931
                    old_entry = state._get_entry(source_index,
 
1932
                                                 path_utf8=old_path)
 
1933
                    # update the source details variable to be the real
 
1934
                    # location.
 
1935
                    source_details = old_entry[1][source_index]
 
1936
                    source_minikind = source_details[0]
 
1937
                else:
 
1938
                    old_dirname = entry[0][0]
 
1939
                    old_basename = entry[0][1]
 
1940
                    old_path = path = None
 
1941
                if path_info is None:
 
1942
                    # the file is missing on disk, show as removed.
 
1943
                    content_change = True
 
1944
                    target_kind = None
 
1945
                    target_exec = False
 
1946
                else:
 
1947
                    # source and target are both versioned and disk file is present.
 
1948
                    target_kind = path_info[2]
 
1949
                    if target_kind == 'directory':
 
1950
                        if path is None:
 
1951
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1952
                        new_dirname_to_file_id[path] = file_id
 
1953
                        if source_minikind != 'd':
 
1954
                            content_change = True
 
1955
                        else:
 
1956
                            # directories have no fingerprint
 
1957
                            content_change = False
 
1958
                        target_exec = False
 
1959
                    elif target_kind == 'file':
 
1960
                        if source_minikind != 'f':
 
1961
                            content_change = True
 
1962
                        else:
 
1963
                            # We could check the size, but we already have the
 
1964
                            # sha1 hash.
 
1965
                            content_change = (link_or_sha1 != source_details[1])
 
1966
                        # Target details is updated at update_entry time
 
1967
                        if use_filesystem_for_exec:
 
1968
                            # We don't need S_ISREG here, because we are sure
 
1969
                            # we are dealing with a file.
 
1970
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1971
                        else:
 
1972
                            target_exec = target_details[3]
 
1973
                    elif target_kind == 'symlink':
 
1974
                        if source_minikind != 'l':
 
1975
                            content_change = True
 
1976
                        else:
 
1977
                            content_change = (link_or_sha1 != source_details[1])
 
1978
                        target_exec = False
 
1979
                    elif target_kind == 'tree-reference':
 
1980
                        if source_minikind != 't':
 
1981
                            content_change = True
 
1982
                        else:
 
1983
                            content_change = False
 
1984
                        target_exec = False
 
1985
                    else:
 
1986
                        raise Exception, "unknown kind %s" % path_info[2]
 
1987
                if source_minikind == 'd':
 
1988
                    if path is None:
 
1989
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1990
                    old_dirname_to_file_id[old_path] = file_id
 
1991
                # parent id is the entry for the path in the target tree
 
1992
                if old_dirname == last_source_parent[0]:
 
1993
                    source_parent_id = last_source_parent[1]
 
1994
                else:
 
1995
                    try:
 
1996
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1997
                    except KeyError:
 
1998
                        source_parent_entry = state._get_entry(source_index,
 
1999
                                                               path_utf8=old_dirname)
 
2000
                        source_parent_id = source_parent_entry[0][2]
 
2001
                    if source_parent_id == entry[0][2]:
 
2002
                        # This is the root, so the parent is None
 
2003
                        source_parent_id = None
 
2004
                    else:
 
2005
                        last_source_parent[0] = old_dirname
 
2006
                        last_source_parent[1] = source_parent_id
 
2007
                new_dirname = entry[0][0]
 
2008
                if new_dirname == last_target_parent[0]:
 
2009
                    target_parent_id = last_target_parent[1]
 
2010
                else:
 
2011
                    try:
 
2012
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2013
                    except KeyError:
 
2014
                        # TODO: We don't always need to do the lookup, because the
 
2015
                        #       parent entry will be the same as the source entry.
 
2016
                        target_parent_entry = state._get_entry(target_index,
 
2017
                                                               path_utf8=new_dirname)
 
2018
                        assert target_parent_entry != (None, None), (
 
2019
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2020
                            % (new_dirname, entry))
 
2021
                        target_parent_id = target_parent_entry[0][2]
 
2022
                    if target_parent_id == entry[0][2]:
 
2023
                        # This is the root, so the parent is None
 
2024
                        target_parent_id = None
 
2025
                    else:
 
2026
                        last_target_parent[0] = new_dirname
 
2027
                        last_target_parent[1] = target_parent_id
 
2028
 
 
2029
                source_exec = source_details[3]
 
2030
                if (include_unchanged
 
2031
                    or content_change
 
2032
                    or source_parent_id != target_parent_id
 
2033
                    or old_basename != entry[0][1]
 
2034
                    or source_exec != target_exec
 
2035
                    ):
 
2036
                    if old_path is None:
 
2037
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2038
                        old_path_u = utf8_decode(old_path)[0]
 
2039
                        path_u = old_path_u
 
2040
                    else:
 
2041
                        old_path_u = utf8_decode(old_path)[0]
 
2042
                        if old_path == path:
 
2043
                            path_u = old_path_u
 
2044
                        else:
 
2045
                            path_u = utf8_decode(path)[0]
 
2046
                    source_kind = _minikind_to_kind[source_minikind]
 
2047
                    return (entry[0][2],
 
2048
                           (old_path_u, path_u),
 
2049
                           content_change,
 
2050
                           (True, True),
 
2051
                           (source_parent_id, target_parent_id),
 
2052
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2053
                           (source_kind, target_kind),
 
2054
                           (source_exec, target_exec))
 
2055
                else:
 
2056
                    return uninteresting
 
2057
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2058
                # looks like a new file
 
2059
                if path_info is not None:
 
2060
                    path = pathjoin(entry[0][0], entry[0][1])
 
2061
                    # parent id is the entry for the path in the target tree
 
2062
                    # TODO: these are the same for an entire directory: cache em.
 
2063
                    parent_id = state._get_entry(target_index,
 
2064
                                                 path_utf8=entry[0][0])[0][2]
 
2065
                    if parent_id == entry[0][2]:
 
2066
                        parent_id = None
 
2067
                    if use_filesystem_for_exec:
 
2068
                        # We need S_ISREG here, because we aren't sure if this
 
2069
                        # is a file or not.
 
2070
                        target_exec = bool(
 
2071
                            stat.S_ISREG(path_info[3].st_mode)
 
2072
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2073
                    else:
 
2074
                        target_exec = target_details[3]
 
2075
                    return (entry[0][2],
 
2076
                           (None, utf8_decode(path)[0]),
 
2077
                           True,
 
2078
                           (False, True),
 
2079
                           (None, parent_id),
 
2080
                           (None, utf8_decode(entry[0][1])[0]),
 
2081
                           (None, path_info[2]),
 
2082
                           (None, target_exec))
 
2083
                else:
 
2084
                    # but its not on disk: we deliberately treat this as just
 
2085
                    # never-present. (Why ?! - RBC 20070224)
 
2086
                    pass
 
2087
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2088
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2089
                # if its still on disk, *and* theres no other entry at this
 
2090
                # path [we dont know this in this routine at the moment -
 
2091
                # perhaps we should change this - then it would be an unknown.
 
2092
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2093
                # parent id is the entry for the path in the target tree
 
2094
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2095
                if parent_id == entry[0][2]:
 
2096
                    parent_id = None
 
2097
                return (entry[0][2],
 
2098
                       (utf8_decode(old_path)[0], None),
 
2099
                       True,
 
2100
                       (True, False),
 
2101
                       (parent_id, None),
 
2102
                       (utf8_decode(entry[0][1])[0], None),
 
2103
                       (_minikind_to_kind[source_minikind], None),
 
2104
                       (source_details[3], None))
 
2105
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2106
                # a rename; could be a true rename, or a rename inherited from
 
2107
                # a renamed parent. TODO: handle this efficiently. Its not
 
2108
                # common case to rename dirs though, so a correct but slow
 
2109
                # implementation will do.
 
2110
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2111
                    search_specific_files.add(target_details[1])
 
2112
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2113
                # neither of the selected trees contain this file,
 
2114
                # so skip over it. This is not currently directly tested, but
 
2115
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2116
                pass
 
2117
            else:
 
2118
                raise AssertionError("don't know how to compare "
 
2119
                    "source_minikind=%r, target_minikind=%r"
 
2120
                    % (source_minikind, target_minikind))
 
2121
                ## import pdb;pdb.set_trace()
 
2122
            return None
 
2123
 
 
2124
        while search_specific_files:
 
2125
            # TODO: the pending list should be lexically sorted?  the
 
2126
            # interface doesn't require it.
 
2127
            current_root = search_specific_files.pop()
 
2128
            current_root_unicode = current_root.decode('utf8')
 
2129
            searched_specific_files.add(current_root)
 
2130
            # process the entries for this containing directory: the rest will be
 
2131
            # found by their parents recursively.
 
2132
            root_entries = _entries_for_path(current_root)
 
2133
            root_abspath = self.target.abspath(current_root_unicode)
 
2134
            try:
 
2135
                root_stat = os.lstat(root_abspath)
 
2136
            except OSError, e:
 
2137
                if e.errno == errno.ENOENT:
 
2138
                    # the path does not exist: let _process_entry know that.
 
2139
                    root_dir_info = None
 
2140
                else:
 
2141
                    # some other random error: hand it up.
 
2142
                    raise
 
2143
            else:
 
2144
                root_dir_info = ('', current_root,
 
2145
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2146
                    root_abspath)
 
2147
                if root_dir_info[2] == 'directory':
 
2148
                    if self.target._directory_is_tree_reference(
 
2149
                        current_root.decode('utf8')):
 
2150
                        root_dir_info = root_dir_info[:2] + \
 
2151
                            ('tree-reference',) + root_dir_info[3:]
 
2152
 
 
2153
            if not root_entries and not root_dir_info:
 
2154
                # this specified path is not present at all, skip it.
 
2155
                continue
 
2156
            path_handled = False
 
2157
            for entry in root_entries:
 
2158
                result = _process_entry(entry, root_dir_info)
 
2159
                if result is not None:
 
2160
                    path_handled = True
 
2161
                    if result is not uninteresting:
 
2162
                        yield result
 
2163
            if want_unversioned and not path_handled and root_dir_info:
 
2164
                new_executable = bool(
 
2165
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2166
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2167
                yield (None,
 
2168
                       (None, current_root_unicode),
 
2169
                       True,
 
2170
                       (False, False),
 
2171
                       (None, None),
 
2172
                       (None, splitpath(current_root_unicode)[-1]),
 
2173
                       (None, root_dir_info[2]),
 
2174
                       (None, new_executable)
 
2175
                      )
 
2176
            initial_key = (current_root, '', '')
 
2177
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2178
            if block_index == 0:
 
2179
                # we have processed the total root already, but because the
 
2180
                # initial key matched it we should skip it here.
 
2181
                block_index +=1
 
2182
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2183
                current_dir_info = None
 
2184
            else:
 
2185
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2186
                try:
 
2187
                    current_dir_info = dir_iterator.next()
 
2188
                except OSError, e:
 
2189
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2190
                    # python 2.5 has e.errno == EINVAL,
 
2191
                    #            and e.winerror == ERROR_DIRECTORY
 
2192
                    e_winerror = getattr(e, 'winerror', None)
 
2193
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2194
                    # there may be directories in the inventory even though
 
2195
                    # this path is not a file on disk: so mark it as end of
 
2196
                    # iterator
 
2197
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2198
                        current_dir_info = None
 
2199
                    elif (sys.platform == 'win32'
 
2200
                          and (e.errno in win_errors
 
2201
                               or e_winerror in win_errors)):
 
2202
                        current_dir_info = None
 
2203
                    else:
 
2204
                        raise
 
2205
                else:
 
2206
                    if current_dir_info[0][0] == '':
 
2207
                        # remove .bzr from iteration
 
2208
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2209
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2210
                        del current_dir_info[1][bzr_index]
 
2211
            # walk until both the directory listing and the versioned metadata
 
2212
            # are exhausted. 
 
2213
            if (block_index < len(state._dirblocks) and
 
2214
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2215
                current_block = state._dirblocks[block_index]
 
2216
            else:
 
2217
                current_block = None
 
2218
            while (current_dir_info is not None or
 
2219
                   current_block is not None):
 
2220
                if (current_dir_info and current_block
 
2221
                    and current_dir_info[0][0] != current_block[0]):
 
2222
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2223
                        # filesystem data refers to paths not covered by the dirblock.
 
2224
                        # this has two possibilities:
 
2225
                        # A) it is versioned but empty, so there is no block for it
 
2226
                        # B) it is not versioned.
 
2227
 
 
2228
                        # if (A) then we need to recurse into it to check for
 
2229
                        # new unknown files or directories.
 
2230
                        # if (B) then we should ignore it, because we don't
 
2231
                        # recurse into unknown directories.
 
2232
                        path_index = 0
 
2233
                        while path_index < len(current_dir_info[1]):
 
2234
                                current_path_info = current_dir_info[1][path_index]
 
2235
                                if want_unversioned:
 
2236
                                    if current_path_info[2] == 'directory':
 
2237
                                        if self.target._directory_is_tree_reference(
 
2238
                                            current_path_info[0].decode('utf8')):
 
2239
                                            current_path_info = current_path_info[:2] + \
 
2240
                                                ('tree-reference',) + current_path_info[3:]
 
2241
                                    new_executable = bool(
 
2242
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2243
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2244
                                    yield (None,
 
2245
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2246
                                        True,
 
2247
                                        (False, False),
 
2248
                                        (None, None),
 
2249
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2250
                                        (None, current_path_info[2]),
 
2251
                                        (None, new_executable))
 
2252
                                # dont descend into this unversioned path if it is
 
2253
                                # a dir
 
2254
                                if current_path_info[2] in ('directory',
 
2255
                                                            'tree-reference'):
 
2256
                                    del current_dir_info[1][path_index]
 
2257
                                    path_index -= 1
 
2258
                                path_index += 1
 
2259
 
 
2260
                        # This dir info has been handled, go to the next
 
2261
                        try:
 
2262
                            current_dir_info = dir_iterator.next()
 
2263
                        except StopIteration:
 
2264
                            current_dir_info = None
 
2265
                    else:
 
2266
                        # We have a dirblock entry for this location, but there
 
2267
                        # is no filesystem path for this. This is most likely
 
2268
                        # because a directory was removed from the disk.
 
2269
                        # We don't have to report the missing directory,
 
2270
                        # because that should have already been handled, but we
 
2271
                        # need to handle all of the files that are contained
 
2272
                        # within.
 
2273
                        for current_entry in current_block[1]:
 
2274
                            # entry referring to file not present on disk.
 
2275
                            # advance the entry only, after processing.
 
2276
                            result = _process_entry(current_entry, None)
 
2277
                            if result is not None:
 
2278
                                if result is not uninteresting:
 
2279
                                    yield result
 
2280
                        block_index +=1
 
2281
                        if (block_index < len(state._dirblocks) and
 
2282
                            osutils.is_inside(current_root,
 
2283
                                              state._dirblocks[block_index][0])):
 
2284
                            current_block = state._dirblocks[block_index]
 
2285
                        else:
 
2286
                            current_block = None
 
2287
                    continue
 
2288
                entry_index = 0
 
2289
                if current_block and entry_index < len(current_block[1]):
 
2290
                    current_entry = current_block[1][entry_index]
 
2291
                else:
 
2292
                    current_entry = None
 
2293
                advance_entry = True
 
2294
                path_index = 0
 
2295
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2296
                    current_path_info = current_dir_info[1][path_index]
 
2297
                    if current_path_info[2] == 'directory':
 
2298
                        if self.target._directory_is_tree_reference(
 
2299
                            current_path_info[0].decode('utf8')):
 
2300
                            current_path_info = current_path_info[:2] + \
 
2301
                                ('tree-reference',) + current_path_info[3:]
 
2302
                else:
 
2303
                    current_path_info = None
 
2304
                advance_path = True
 
2305
                path_handled = False
 
2306
                while (current_entry is not None or
 
2307
                    current_path_info is not None):
 
2308
                    if current_entry is None:
 
2309
                        # the check for path_handled when the path is adnvaced
 
2310
                        # will yield this path if needed.
 
2311
                        pass
 
2312
                    elif current_path_info is None:
 
2313
                        # no path is fine: the per entry code will handle it.
 
2314
                        result = _process_entry(current_entry, current_path_info)
 
2315
                        if result is not None:
 
2316
                            if result is not uninteresting:
 
2317
                                yield result
 
2318
                    elif (current_entry[0][1] != current_path_info[1]
 
2319
                          or current_entry[1][target_index][0] in 'ar'):
 
2320
                        # The current path on disk doesn't match the dirblock
 
2321
                        # record. Either the dirblock is marked as absent, or
 
2322
                        # the file on disk is not present at all in the
 
2323
                        # dirblock. Either way, report about the dirblock
 
2324
                        # entry, and let other code handle the filesystem one.
 
2325
 
 
2326
                        # Compare the basename for these files to determine
 
2327
                        # which comes first
 
2328
                        if current_path_info[1] < current_entry[0][1]:
 
2329
                            # extra file on disk: pass for now, but only
 
2330
                            # increment the path, not the entry
 
2331
                            advance_entry = False
 
2332
                        else:
 
2333
                            # entry referring to file not present on disk.
 
2334
                            # advance the entry only, after processing.
 
2335
                            result = _process_entry(current_entry, None)
 
2336
                            if result is not None:
 
2337
                                if result is not uninteresting:
 
2338
                                    yield result
 
2339
                            advance_path = False
 
2340
                    else:
 
2341
                        result = _process_entry(current_entry, current_path_info)
 
2342
                        if result is not None:
 
2343
                            path_handled = True
 
2344
                            if result is not uninteresting:
 
2345
                                yield result
 
2346
                    if advance_entry and current_entry is not None:
 
2347
                        entry_index += 1
 
2348
                        if entry_index < len(current_block[1]):
 
2349
                            current_entry = current_block[1][entry_index]
 
2350
                        else:
 
2351
                            current_entry = None
 
2352
                    else:
 
2353
                        advance_entry = True # reset the advance flaga
 
2354
                    if advance_path and current_path_info is not None:
 
2355
                        if not path_handled:
 
2356
                            # unversioned in all regards
 
2357
                            if want_unversioned:
 
2358
                                new_executable = bool(
 
2359
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2360
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2361
                                yield (None,
 
2362
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2363
                                    True,
 
2364
                                    (False, False),
 
2365
                                    (None, None),
 
2366
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2367
                                    (None, current_path_info[2]),
 
2368
                                    (None, new_executable))
 
2369
                            # dont descend into this unversioned path if it is
 
2370
                            # a dir
 
2371
                            if current_path_info[2] in ('directory'):
 
2372
                                del current_dir_info[1][path_index]
 
2373
                                path_index -= 1
 
2374
                        # dont descend the disk iterator into any tree 
 
2375
                        # paths.
 
2376
                        if current_path_info[2] == 'tree-reference':
 
2377
                            del current_dir_info[1][path_index]
 
2378
                            path_index -= 1
 
2379
                        path_index += 1
 
2380
                        if path_index < len(current_dir_info[1]):
 
2381
                            current_path_info = current_dir_info[1][path_index]
 
2382
                            if current_path_info[2] == 'directory':
 
2383
                                if self.target._directory_is_tree_reference(
 
2384
                                    current_path_info[0].decode('utf8')):
 
2385
                                    current_path_info = current_path_info[:2] + \
 
2386
                                        ('tree-reference',) + current_path_info[3:]
 
2387
                        else:
 
2388
                            current_path_info = None
 
2389
                        path_handled = False
 
2390
                    else:
 
2391
                        advance_path = True # reset the advance flagg.
 
2392
                if current_block is not None:
 
2393
                    block_index += 1
 
2394
                    if (block_index < len(state._dirblocks) and
 
2395
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2396
                        current_block = state._dirblocks[block_index]
 
2397
                    else:
 
2398
                        current_block = None
 
2399
                if current_dir_info is not None:
 
2400
                    try:
 
2401
                        current_dir_info = dir_iterator.next()
 
2402
                    except StopIteration:
 
2403
                        current_dir_info = None
 
2404
 
2255
2405
 
2256
2406
    @staticmethod
2257
2407
    def is_compatible(source, target):
2258
2408
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2409
        if not isinstance(target, WorkingTree4):
2260
2410
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2411
        # the source must be a revtreee or dirstate rev tree.
2262
2412
        if not isinstance(source,
2263
2413
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2414
            return False
2265
2415
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2416
        if not (source._revision_id == NULL_REVISION or
2267
2417
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2418
            # TODO: what about ghosts? it may well need to 
2269
2419
            # check for them explicitly.
2270
2420
            return False
2271
2421
        return True
2281
2431
 
2282
2432
    def convert(self, tree):
2283
2433
        # lock the control files not the tree, so that we dont get tree
2284
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2434
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2435
        # tree during upgrade.
2286
2436
        tree._control_files.lock_write()
2287
2437
        try:
2313
2463
 
2314
2464
    def update_format(self, tree):
2315
2465
        """Change the format marker."""
2316
 
        tree._transport.put_bytes('format',
2317
 
            self.target_format.as_string(),
2318
 
            mode=tree.bzrdir._get_file_mode())
2319
 
 
2320
 
 
2321
 
class Converter4to5(object):
2322
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2323
 
 
2324
 
    def __init__(self):
2325
 
        self.target_format = WorkingTreeFormat5()
2326
 
 
2327
 
    def convert(self, tree):
2328
 
        # lock the control files not the tree, so that we don't get tree
2329
 
        # on-unlock behaviours, and so that no-one else diddles with the
2330
 
        # tree during upgrade.
2331
 
        tree._control_files.lock_write()
2332
 
        try:
2333
 
            self.update_format(tree)
2334
 
        finally:
2335
 
            tree._control_files.unlock()
2336
 
 
2337
 
    def update_format(self, tree):
2338
 
        """Change the format marker."""
2339
 
        tree._transport.put_bytes('format',
2340
 
            self.target_format.as_string(),
2341
 
            mode=tree.bzrdir._get_file_mode())
2342
 
 
2343
 
 
2344
 
class Converter4or5to6(object):
2345
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2346
 
 
2347
 
    def __init__(self):
2348
 
        self.target_format = WorkingTreeFormat6()
2349
 
 
2350
 
    def convert(self, tree):
2351
 
        # lock the control files not the tree, so that we don't get tree
2352
 
        # on-unlock behaviours, and so that no-one else diddles with the
2353
 
        # tree during upgrade.
2354
 
        tree._control_files.lock_write()
2355
 
        try:
2356
 
            self.init_custom_control_files(tree)
2357
 
            self.update_format(tree)
2358
 
        finally:
2359
 
            tree._control_files.unlock()
2360
 
 
2361
 
    def init_custom_control_files(self, tree):
2362
 
        """Initialize custom control files."""
2363
 
        tree._transport.put_bytes('views', '',
2364
 
            mode=tree.bzrdir._get_file_mode())
2365
 
 
2366
 
    def update_format(self, tree):
2367
 
        """Change the format marker."""
2368
 
        tree._transport.put_bytes('format',
2369
 
            self.target_format.as_string(),
2370
 
            mode=tree.bzrdir._get_file_mode())
 
2466
        tree._control_files.put_utf8('format',
 
2467
            self.target_format.get_format_string())