~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

Merge bzr.dev

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
 
        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
154
 
126
155
    @needs_tree_write_lock
127
156
    def _add(self, files, ids, kinds):
142
171
            state.add(f, file_id, kind, None, '')
143
172
        self._make_dirty(reset_inventory=True)
144
173
 
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
174
    def _make_dirty(self, reset_inventory):
150
175
        """Make the tree state dirty.
151
176
 
159
184
    @needs_tree_write_lock
160
185
    def add_reference(self, sub_tree):
161
186
        # use standard implementation, which calls back to self._add
162
 
        #
 
187
        # 
163
188
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
189
        # it's just recorded at the moment of commit. 
165
190
        self._add_reference(sub_tree)
166
191
 
167
192
    def break_lock(self):
203
228
 
204
229
    def _comparison_data(self, entry, path):
205
230
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
231
            WorkingTree3._comparison_data(self, entry, path)
207
232
        # it looks like a plain directory, but it's really a reference -- see
208
233
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            and entry is not None and entry.kind == 'tree-reference'):
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
211
237
            kind = 'tree-reference'
212
238
        return kind, executable, stat_value
213
239
 
215
241
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
242
        # mark the tree as dirty post commit - commit
217
243
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
244
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
245
        self._make_dirty(reset_inventory=True)
220
246
        return result
221
247
 
239
265
            return self._dirstate
240
266
        local_path = self.bzrdir.get_workingtree_transport(None
241
267
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
244
269
        return self._dirstate
245
270
 
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
271
    def filter_unversioned_files(self, paths):
267
272
        """Filter out paths that are versioned.
268
273
 
300
305
 
301
306
    def _generate_inventory(self):
302
307
        """Create and set self.inventory from the dirstate object.
303
 
 
 
308
        
304
309
        This is relatively expensive: we have to walk the entire dirstate.
305
310
        Ideally we would not, and can deprecate this function.
306
311
        """
310
315
        state._read_dirblocks_if_needed()
311
316
        root_key, current_entry = self._get_entry(path='')
312
317
        current_id = root_key[2]
313
 
        if not (current_entry[0][0] == 'd'): # directory
314
 
            raise AssertionError(current_entry)
 
318
        assert current_entry[0][0] == 'd' # directory
315
319
        inv = Inventory(root_id=current_id)
316
320
        # Turn some things into local variables
317
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
350
354
                    # add this entry to the parent map.
351
355
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
356
                elif kind == 'tree-reference':
353
 
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            self.branch.repository)
 
357
                    assert self._repo_supports_tree_reference, \
 
358
                        "repository of %r " \
 
359
                        "doesn't support tree references " \
 
360
                        "required by entry %r" \
 
361
                        % (self, name)
357
362
                    inv_entry.reference_revision = link_or_sha1 or None
358
363
                elif kind != 'symlink':
359
364
                    raise AssertionError("unknown kind %r" % kind)
360
365
                # 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,))
 
366
                assert file_id not in inv_byid, ('file_id %s already in'
 
367
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                assert name_unicode not in parent_ie.children
367
369
                inv_byid[file_id] = inv_entry
368
370
                parent_ie.children[name_unicode] = inv_entry
369
371
        self._inventory = inv
374
376
        If either file_id or path is supplied, it is used as the key to lookup.
375
377
        If both are supplied, the fastest lookup is used, and an error is
376
378
        raised if they do not both point at the same row.
377
 
 
 
379
        
378
380
        :param file_id: An optional unicode file_id to be looked up.
379
381
        :param path: An optional unicode path to be looked up.
380
382
        :return: The dirstate row tuple for path/file_id, or (None, None)
398
400
        state = self.current_dirstate()
399
401
        if stat_value is None:
400
402
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
 
403
                stat_value = os.lstat(file_abspath)
402
404
            except OSError, e:
403
405
                if e.errno == errno.ENOENT:
404
406
                    return None
405
407
                else:
406
408
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
409
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
410
                                          stat_value=stat_value)
409
411
        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
 
412
            return link_or_sha1
420
413
        return None
421
414
 
422
 
    def _get_root_inventory(self):
 
415
    def _get_inventory(self):
423
416
        """Get the inventory for the tree. This is only valid within a lock."""
424
417
        if 'evil' in debug.debug_flags:
425
418
            trace.mutter_callsite(2,
430
423
        self._generate_inventory()
431
424
        return self._inventory
432
425
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
426
    inventory = property(_get_inventory,
438
427
                         doc="Inventory of this Tree")
439
428
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
429
    @needs_read_lock
444
430
    def get_parent_ids(self):
445
431
        """See Tree.get_parent_ids.
446
 
 
 
432
        
447
433
        This implementation requests the ids list from the dirstate file.
448
434
        """
449
435
        return self.current_dirstate().get_parent_ids()
471
457
        return osutils.lexists(pathjoin(
472
458
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
459
 
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
460
    @needs_read_lock
480
461
    def id2path(self, file_id):
481
462
        "Convert a file-id to a path."
492
473
            return False # Missing entries are not executable
493
474
        return entry[1][0][3] # Executable?
494
475
 
495
 
    def is_executable(self, file_id, path=None):
496
 
        """Test if a file is executable or not.
 
476
    if not osutils.supports_executable():
 
477
        def is_executable(self, file_id, path=None):
 
478
            """Test if a file is executable or not.
497
479
 
498
 
        Note: The caller is expected to take a read-lock before calling this.
499
 
        """
500
 
        if not self._supports_executable():
 
480
            Note: The caller is expected to take a read-lock before calling this.
 
481
            """
501
482
            entry = self._get_entry(file_id=file_id, path=path)
502
483
            if entry == (None, None):
503
484
                return False
504
485
            return entry[1][0][3]
505
 
        else:
506
 
            self._must_be_locked()
 
486
 
 
487
        _is_executable_from_path_and_stat = \
 
488
            _is_executable_from_path_and_stat_from_basis
 
489
    else:
 
490
        def is_executable(self, file_id, path=None):
 
491
            """Test if a file is executable or not.
 
492
 
 
493
            Note: The caller is expected to take a read-lock before calling this.
 
494
            """
507
495
            if not path:
508
496
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
497
            mode = os.lstat(self.abspath(path)).st_mode
510
498
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
499
 
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
500
    @needs_read_lock
523
501
    def __iter__(self):
524
502
        """Iterate through file_ids for this tree.
537
515
        return iter(result)
538
516
 
539
517
    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
518
        for key, tree_details in self.current_dirstate()._iter_entries():
545
519
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
546
520
                # not relevant to the working tree
548
522
            if not key[1]:
549
523
                # the root is not a reference.
550
524
                continue
551
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
525
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
552
526
            try:
553
 
                if self._kind(relpath) == 'tree-reference':
554
 
                    yield relpath, key[2]
 
527
                if self._kind(path) == 'tree-reference':
 
528
                    yield path, key[2]
555
529
            except errors.NoSuchFile:
556
530
                # path is missing on disk.
557
531
                continue
558
532
 
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
533
    def kind(self, file_id):
566
534
        """Return the kind of a file.
567
535
 
571
539
        Note: The caller is expected to take a read-lock before calling this.
572
540
        """
573
541
        relpath = self.id2path(file_id)
574
 
        if relpath is None:
575
 
            raise AssertionError(
576
 
                "path for id {%s} is None!" % file_id)
 
542
        assert relpath != None, \
 
543
            "path for id {%s} is None!" % file_id
577
544
        return self._kind(relpath)
578
545
 
579
546
    def _kind(self, relpath):
580
547
        abspath = self.abspath(relpath)
581
548
        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'
 
549
        if (self._repo_supports_tree_reference and
 
550
            kind == 'directory' and
 
551
            self._directory_is_tree_reference(relpath)):
 
552
            kind = 'tree-reference'
587
553
        return kind
588
554
 
589
555
    @needs_read_lock
596
562
            return _mod_revision.NULL_REVISION
597
563
 
598
564
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
565
        """See Branch.lock_read, and WorkingTree.unlock."""
603
566
        self.branch.lock_read()
604
567
        try:
605
568
            self._control_files.lock_read()
618
581
        except:
619
582
            self.branch.unlock()
620
583
            raise
621
 
        return LogicalLockResult(self.unlock)
622
584
 
623
585
    def _lock_self_write(self):
624
586
        """This should be called after the branch is locked."""
639
601
        except:
640
602
            self.branch.unlock()
641
603
            raise
642
 
        return LogicalLockResult(self.unlock)
643
604
 
644
605
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
606
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
607
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
608
        self._lock_self_write()
651
609
 
652
610
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
611
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
612
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
613
        self._lock_self_write()
659
614
 
660
615
    @needs_tree_write_lock
661
616
    def move(self, from_paths, to_dir, after=False):
663
618
        result = []
664
619
        if not from_paths:
665
620
            return result
 
621
 
666
622
        state = self.current_dirstate()
667
 
        if isinstance(from_paths, basestring):
668
 
            raise ValueError()
 
623
 
 
624
        assert not isinstance(from_paths, basestring)
669
625
        to_dir_utf8 = to_dir.encode('utf8')
670
626
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
671
627
        id_index = state._get_id_index()
692
648
 
693
649
        if self._inventory is not None:
694
650
            update_inventory = True
695
 
            inv = self.root_inventory
 
651
            inv = self.inventory
 
652
            to_dir_ie = inv[to_dir_id]
696
653
            to_dir_id = to_entry[0][2]
697
 
            to_dir_ie = inv[to_dir_id]
698
654
        else:
699
655
            update_inventory = False
700
656
 
731
687
            from_entry = self._get_entry(path=from_rel)
732
688
            if from_entry == (None, None):
733
689
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
690
                    errors.NotVersionedError(path=str(from_rel)))
735
691
 
736
692
            from_id = from_entry[0][2]
737
693
            to_rel = pathjoin(to_dir, from_tail)
826
782
                if minikind == 'd':
827
783
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
828
784
                        """Recursively update all entries in this dirblock."""
829
 
                        if from_dir == '':
830
 
                            raise AssertionError("renaming root not supported")
 
785
                        assert from_dir != '', "renaming root not supported"
831
786
                        from_key = (from_dir, '')
832
787
                        from_block_idx, present = \
833
788
                            state._find_block_index_from_key(from_key)
846
801
 
847
802
                        # Grab a copy since move_one may update the list.
848
803
                        for entry in from_block[1][:]:
849
 
                            if not (entry[0][0] == from_dir):
850
 
                                raise AssertionError()
 
804
                            assert entry[0][0] == from_dir
851
805
                            cur_details = entry[1][0]
852
806
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
853
807
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
878
832
                rollback_rename()
879
833
                raise
880
834
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
835
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
836
            self._make_dirty(reset_inventory=False)
883
837
 
884
838
        return result
894
848
    @needs_read_lock
895
849
    def path2id(self, path):
896
850
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
851
        path = path.strip('/')
902
852
        entry = self._get_entry(path=path)
903
853
        if entry == (None, None):
916
866
        for tree in trees:
917
867
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
868
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
869
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
870
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
871
        # -- make all paths utf8 --
923
872
        paths_utf8 = set()
981
930
                    all_versioned = False
982
931
                    break
983
932
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
933
                raise errors.PathsNotVersionedError(paths)
986
934
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
935
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
936
        # sketch: 
989
937
        # for all search_indexs in each path at or under each element of
990
938
        # search_paths, if the detail is relocated: add the id, and add the
991
939
        # relocated path as one to search if its not searched already. If the
1037
985
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
986
            for dir_name in split_paths:
1039
987
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
988
                    raise errors.PathsNotVersionedError(paths)
1042
989
 
1043
990
        for dir_name_id, trees_info in found.iteritems():
1044
991
            for index in search_indexes:
1048
995
 
1049
996
    def read_working_inventory(self):
1050
997
        """Read the working inventory.
1051
 
 
 
998
        
1052
999
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
1000
        """
1054
 
        return self.root_inventory
 
1001
        return self.inventory
1055
1002
 
1056
1003
    @needs_read_lock
1057
1004
    def revision_tree(self, revision_id):
1072
1019
    def set_last_revision(self, new_revision):
1073
1020
        """Change the last revision in the working tree."""
1074
1021
        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.")
 
1022
        if new_revision in (NULL_REVISION, None):
 
1023
            assert len(parents) < 2, (
 
1024
                "setting the last parent to none with a pending merge is "
 
1025
                "unsupported.")
1080
1026
            self.set_parent_ids([])
1081
1027
        else:
1082
1028
            self.set_parent_ids([new_revision] + parents[1:],
1085
1031
    @needs_tree_write_lock
1086
1032
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
1033
        """Set the parent ids to revision_ids.
1088
 
 
 
1034
        
1089
1035
        See also set_parent_trees. This api will try to retrieve the tree data
1090
1036
        for each element of revision_ids from the trees repository. If you have
1091
1037
        tree data already available, it is more efficient to use
1123
1069
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1124
1070
        real_trees = []
1125
1071
        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
1072
        # convert absent trees to the null tree, which we convert back to
1133
1073
        # missing on access.
1134
1074
        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
1075
            _mod_revision.check_not_reserved_id(rev_id)
1142
1076
            if tree is not None:
1143
1077
                real_trees.append((rev_id, tree))
1144
1078
            else:
1145
1079
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1080
                    self.branch.repository.revision_tree(None)))
1148
1081
                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)
 
1082
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1174
1083
        self._make_dirty(reset_inventory=False)
1175
1084
 
1176
1085
    def _set_root_id(self, file_id):
1196
1105
 
1197
1106
    def unlock(self):
1198
1107
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1108
        # do non-implementation specific cleanup
 
1109
        self._cleanup()
 
1110
 
1199
1111
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1112
            # eventually we should do signature checking during read locks for
1204
1113
            # dirstate updates.
1205
1114
            if self._control_files._lock_mode == 'w':
1268
1177
                # just forget the whole block.
1269
1178
                entry_index = 0
1270
1179
                while entry_index < len(block[1]):
 
1180
                    # Mark this file id as having been removed
1271
1181
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1182
                    ids_to_unversion.discard(entry[0][2])
 
1183
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1184
                                               # entries
 
1185
                        or not state._make_absent(entry)):
1274
1186
                        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
1187
                # go to the next block. (At the moment we dont delete empty
1282
1188
                # dirblocks)
1283
1189
                block_index += 1
1304
1210
        # have to change the legacy inventory too.
1305
1211
        if self._inventory is not None:
1306
1212
            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)
 
1213
                self._inventory.remove_recursive_id(file_id)
1322
1214
 
1323
1215
    def update_basis_by_delta(self, new_revid, delta):
1324
1216
        """See MutableTree.update_basis_by_delta."""
1325
 
        if self.last_revision() == new_revid:
1326
 
            raise AssertionError()
 
1217
        assert self.last_revision() != new_revid
1327
1218
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1328
1219
 
1329
1220
    @needs_read_lock
1333
1224
    @needs_tree_write_lock
1334
1225
    def _write_inventory(self, inv):
1335
1226
        """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:
 
1227
        assert not self._dirty, ("attempting to write an inventory when the "
 
1228
            "dirstate is dirty will cause data loss")
 
1229
        self.current_dirstate().set_state_from_inventory(inv)
 
1230
        self._make_dirty(reset_inventory=False)
 
1231
        if self._inventory is not None:
1350
1232
            self._inventory = inv
1351
1233
        self.flush()
1352
1234
 
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):
 
1235
 
 
1236
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1237
    """The first consolidated dirstate working tree format.
 
1238
 
 
1239
    This format:
 
1240
        - exists within a metadir controlling .bzr
 
1241
        - includes an explicit version marker for the workingtree control
 
1242
          files, separate from the BzrDir format
 
1243
        - modifies the hash cache format
 
1244
        - is new in bzr 0.15
 
1245
        - uses a LockDir to guard access to it.
 
1246
    """
 
1247
 
 
1248
    upgrade_recommended = False
 
1249
 
 
1250
    def get_format_string(self):
 
1251
        """See WorkingTreeFormat.get_format_string()."""
 
1252
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1253
 
 
1254
    def get_format_description(self):
 
1255
        """See WorkingTreeFormat.get_format_description()."""
 
1256
        return "Working tree format 4"
 
1257
 
 
1258
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None):
1480
1259
        """See WorkingTreeFormat.initialize().
1481
1260
 
1482
1261
        :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.
 
1262
        revision than the branch is at.
1490
1263
 
1491
1264
        These trees get an initial random root id, if their repository supports
1492
1265
        rich root data, TREE_ROOT otherwise.
1497
1270
        control_files = self._open_control_files(a_bzrdir)
1498
1271
        control_files.create_lock()
1499
1272
        control_files.lock_write()
1500
 
        transport.put_bytes('format', self.as_string(),
1501
 
            mode=a_bzrdir._get_file_mode())
 
1273
        control_files.put_utf8('format', self.get_format_string())
1502
1274
        if from_branch is not None:
1503
1275
            branch = from_branch
1504
1276
        else:
1510
1282
        state = dirstate.DirState.initialize(local_path)
1511
1283
        state.unlock()
1512
1284
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1285
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1286
                         branch,
1515
1287
                         _format=self,
1516
1288
                         _bzrdir=a_bzrdir,
1518
1290
        wt._new_tree()
1519
1291
        wt.lock_tree_write()
1520
1292
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1293
            if revision_id in (None, NULL_REVISION):
1523
1294
                if branch.repository.supports_rich_root():
1524
1295
                    wt._set_root_id(generate_ids.gen_root_id())
1525
1296
                else:
1526
1297
                    wt._set_root_id(ROOT_ID)
1527
1298
                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)]
 
1299
            wt.set_last_revision(revision_id)
 
1300
            wt.flush()
 
1301
            basis = wt.basis_tree()
1543
1302
            basis.lock_read()
1544
 
            try:
1545
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1303
            # if the basis has a root id we have to use that; otherwise we use
 
1304
            # a new random one
 
1305
            basis_root_id = basis.get_root_id()
 
1306
            if basis_root_id is not None:
 
1307
                wt._set_root_id(basis_root_id)
1546
1308
                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()
 
1309
            transform.build_tree(basis, wt)
 
1310
            basis.unlock()
1570
1311
        finally:
1571
1312
            control_files.unlock()
1572
1313
            wt.unlock()
1573
1314
        return wt
1574
1315
 
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
1316
    def _open(self, a_bzrdir, control_files):
1599
1317
        """Open the tree itself.
1600
1318
 
1601
1319
        :param a_bzrdir: the dir for the tree.
1602
1320
        :param control_files: the control files for the tree.
1603
1321
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1322
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1323
                           branch=a_bzrdir.open_branch(),
1606
1324
                           _format=self,
1607
1325
                           _bzrdir=a_bzrdir,
1608
1326
                           _control_files=control_files)
1609
1327
 
1610
1328
    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
1329
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1330
        return bzrdir.format_registry.make_bzrdir(
 
1331
            'dirstate-with-subtree')
1618
1332
 
1619
1333
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1334
 
1621
1335
 
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
 
    """
 
1336
class DirStateRevisionTree(Tree):
 
1337
    """A revision tree pulling the inventory from a dirstate."""
1709
1338
 
1710
1339
    def __init__(self, dirstate, revision_id, repository):
1711
1340
        self._dirstate = dirstate
1714
1343
        self._inventory = None
1715
1344
        self._locked = 0
1716
1345
        self._dirstate_locked = False
1717
 
        self._repo_supports_tree_reference = getattr(
1718
 
            repository._format, "supports_tree_reference",
1719
 
            False)
1720
1346
 
1721
1347
    def __repr__(self):
1722
1348
        return "<%s of %s in %s>" % \
1725
1351
    def annotate_iter(self, file_id,
1726
1352
                      default_revision=_mod_revision.CURRENT_REVISION):
1727
1353
        """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]
 
1354
        w = self._get_weave(file_id)
 
1355
        return w.annotate_iter(self.inventory[file_id].revision)
1731
1356
 
 
1357
    def _get_ancestors(self, default_revision):
 
1358
        return set(self._repository.get_ancestry(self._revision_id,
 
1359
                                                 topo_sorted=False))
1732
1360
    def _comparison_data(self, entry, path):
1733
1361
        """See Tree._comparison_data."""
1734
1362
        if entry is None:
1751
1379
    def get_root_id(self):
1752
1380
        return self.path2id('')
1753
1381
 
1754
 
    def id2path(self, file_id):
1755
 
        "Convert a file-id to a path."
1756
 
        entry = self._get_entry(file_id=file_id)
1757
 
        if entry == (None, None):
1758
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1759
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1760
 
        return path_utf8.decode('utf8')
1761
 
 
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
1382
    def _get_parent_index(self):
1771
1383
        """Return the index in the dirstate referenced by this tree."""
1772
1384
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1777
1389
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1390
        If both are supplied, the fastest lookup is used, and an error is
1779
1391
        raised if they do not both point at the same row.
1780
 
 
 
1392
        
1781
1393
        :param file_id: An optional unicode file_id to be looked up.
1782
1394
        :param path: An optional unicode path to be looked up.
1783
1395
        :return: The dirstate row tuple for path/file_id, or (None, None)
1787
1399
        if path is not None:
1788
1400
            path = path.encode('utf8')
1789
1401
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1402
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1403
 
1793
1404
    def _generate_inventory(self):
1794
1405
        """Create and set self.inventory from the dirstate object.
1798
1409
 
1799
1410
        This is relatively expensive: we have to walk the entire dirstate.
1800
1411
        """
1801
 
        if not self._locked:
1802
 
            raise AssertionError(
1803
 
                'cannot generate inventory of an unlocked '
1804
 
                'dirstate revision tree')
 
1412
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1413
            'dirstate revision tree'
1805
1414
        # separate call for profiling - makes it clear where the costs are.
1806
1415
        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()))
 
1416
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1417
            'parent %s has disappeared from %s' % (
 
1418
            self._revision_id, self._dirstate.get_parent_ids())
1811
1419
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1812
1420
        # This is identical now to the WorkingTree _generate_inventory except
1813
1421
        # for the tree index use.
1814
1422
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1815
1423
        current_id = root_key[2]
1816
 
        if current_entry[parent_index][0] != 'd':
1817
 
            raise AssertionError()
 
1424
        assert current_entry[parent_index][0] == 'd'
1818
1425
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1819
1426
        inv.root.revision = current_entry[parent_index][4]
1820
1427
        # Turn some things into local variables
1851
1458
                elif kind == 'directory':
1852
1459
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1460
                elif kind == 'symlink':
 
1461
                    inv_entry.executable = False
 
1462
                    inv_entry.text_size = None
1854
1463
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1464
                elif kind == 'tree-reference':
1856
1465
                    inv_entry.reference_revision = fingerprint or None
1858
1467
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1859
1468
                            % entry)
1860
1469
                # 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,))
 
1470
                assert file_id not in inv_byid
 
1471
                assert name_unicode not in parent_ie.children
1867
1472
                inv_byid[file_id] = inv_entry
1868
1473
                parent_ie.children[name_unicode] = inv_entry
1869
1474
        self._inventory = inv
1876
1481
        # Make sure the file exists
1877
1482
        entry = self._get_entry(file_id, path=path)
1878
1483
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1484
            return None
1880
1485
        parent_index = self._get_parent_index()
1881
1486
        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
 
1487
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1488
 
1888
1489
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1490
        entry = self._get_entry(file_id=file_id, path=path)
1893
1494
            return parent_details[1]
1894
1495
        return None
1895
1496
 
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
 
1497
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1498
    def get_weave(self, file_id):
 
1499
        return self._get_weave(file_id)
 
1500
 
 
1501
    def _get_weave(self, file_id):
 
1502
        return self._repository.weave_store.get_weave(file_id,
 
1503
                self._repository.get_transaction())
1900
1504
 
1901
1505
    def get_file(self, file_id, path=None):
1902
1506
        return StringIO(self.get_file_text(file_id))
1903
1507
 
 
1508
    def get_file_lines(self, file_id):
 
1509
        ie = self.inventory[file_id]
 
1510
        return self._get_weave(file_id).get_lines(ie.revision)
 
1511
 
1904
1512
    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
 
1513
        return self.inventory[file_id].text_size
1908
1514
 
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
 
1515
    def get_file_text(self, file_id):
 
1516
        return ''.join(self.get_file_lines(file_id))
1922
1517
 
1923
1518
    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
 
1519
        return self.inventory[file_id].reference_revision
1926
1520
 
1927
1521
    def iter_files_bytes(self, desired_files):
1928
1522
        """See Tree.iter_files_bytes.
1938
1532
                                       identifier))
1939
1533
        return self._repository.iter_files_bytes(repo_desired_files)
1940
1534
 
1941
 
    def get_symlink_target(self, file_id, path=None):
 
1535
    def get_symlink_target(self, file_id):
1942
1536
        entry = self._get_entry(file_id=file_id)
1943
1537
        parent_index = self._get_parent_index()
1944
1538
        if entry[1][parent_index][0] != 'l':
1945
1539
            return None
1946
1540
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1541
            # At present, none of the tree implementations supports non-ascii
 
1542
            # symlink targets. So we will just assume that the dirstate path is
 
1543
            # correct.
 
1544
            return entry[1][parent_index][1]
1950
1545
 
1951
1546
    def get_revision_id(self):
1952
1547
        """Return the revision id for this tree."""
1953
1548
        return self._revision_id
1954
1549
 
1955
 
    def _get_root_inventory(self):
 
1550
    def _get_inventory(self):
1956
1551
        if self._inventory is not None:
1957
1552
            return self._inventory
1958
1553
        self._must_be_locked()
1959
1554
        self._generate_inventory()
1960
1555
        return self._inventory
1961
1556
 
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
1557
    inventory = property(_get_inventory,
1970
1558
                         doc="Inventory of this Tree")
1971
1559
 
1977
1565
        return bool(self.path2id(filename))
1978
1566
 
1979
1567
    def kind(self, file_id):
1980
 
        entry = self._get_entry(file_id=file_id)[1]
1981
 
        if entry is None:
1982
 
            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)
 
1568
        return self.inventory[file_id].kind
1989
1569
 
1990
1570
    def path_content_summary(self, path):
1991
1571
        """See Tree.path_content_summary."""
1992
 
        inv, inv_file_id = self._path2inv_file_id(path)
1993
 
        if inv_file_id is None:
 
1572
        id = self.inventory.path2id(path)
 
1573
        if id is None:
1994
1574
            return ('missing', None, None, None)
1995
 
        entry = inv[inv_file_id]
 
1575
        entry = self._inventory[id]
1996
1576
        kind = entry.kind
1997
1577
        if kind == 'file':
1998
1578
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
2002
1582
            return (kind, None, None, None)
2003
1583
 
2004
1584
    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]
 
1585
        ie = self.inventory[file_id]
2007
1586
        if ie.kind != "file":
2008
 
            return False
 
1587
            return None
2009
1588
        return ie.executable
2010
1589
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1590
    def list_files(self, include_root=False):
2015
1591
        # We use a standard implementation, because DirStateRevisionTree is
2016
1592
        # 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:
 
1593
        inv = self._get_inventory()
 
1594
        entries = inv.iter_entries()
 
1595
        if self.inventory.root is not None and not include_root:
2028
1596
            entries.next()
2029
1597
        for path, entry in entries:
2030
1598
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1599
 
2032
1600
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1601
        """Lock the tree for a set of operations."""
2037
1602
        if not self._locked:
2038
1603
            self._repository.lock_read()
2039
1604
            if self._dirstate._lock_token is None:
2040
1605
                self._dirstate.lock_read()
2041
1606
                self._dirstate_locked = True
2042
1607
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1608
 
2045
1609
    def _must_be_locked(self):
2046
1610
        if not self._locked:
2050
1614
    def path2id(self, path):
2051
1615
        """Return the id for path in this tree."""
2052
1616
        # 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
1617
        entry = self._get_entry(path=path)
2058
1618
        if entry == (None, None):
2059
1619
            return None
2071
1631
                self._dirstate_locked = False
2072
1632
            self._repository.unlock()
2073
1633
 
2074
 
    @needs_read_lock
2075
 
    def supports_tree_reference(self):
2076
 
        return self._repo_supports_tree_reference
2077
 
 
2078
1634
    def walkdirs(self, prefix=""):
2079
1635
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1636
        # implementation based on an inventory.  
2081
1637
        # This should be cleaned up to use the much faster Dirstate code
2082
1638
        # So for now, we just build up the parent inventory, and extract
2083
1639
        # it the same way RevisionTree does.
2084
1640
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1641
        inv = self._get_inventory()
2086
1642
        top_id = inv.path2id(prefix)
2087
1643
        if top_id is None:
2088
1644
            pending = []
2112
1668
 
2113
1669
class InterDirStateTree(InterTree):
2114
1670
    """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
 
1671
    
 
1672
    This is used only when both trees are in the dirstate working file, and 
 
1673
    the source is any parent within the dirstate, and the destination is 
2118
1674
    the current working tree of the same dirstate.
2119
1675
    """
2120
1676
    # this could be generalized to allow comparisons between any trees in the
2129
1685
    def make_source_parent_tree(source, target):
2130
1686
        """Change the source tree into a parent of the target."""
2131
1687
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1688
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1689
        target.set_parent_ids([revid])
2134
1690
        return target.basis_tree(), target
2135
1691
 
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
1692
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1693
    _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,
 
1694
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1695
 
 
1696
    def _iter_changes(self, include_unchanged=False,
2163
1697
                      specific_files=None, pb=None, extra_trees=[],
2164
1698
                      require_versioned=True, want_unversioned=False):
2165
1699
        """Return the changes from source to target.
2166
1700
 
2167
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1701
        :return: An iterator that yields tuples. See InterTree._iter_changes
2168
1702
            for details.
2169
1703
        :param specific_files: An optional list of file paths to restrict the
2170
1704
            comparison to. When mapping filenames to ids, all matches in all
2181
1715
            output. An unversioned file is defined as one with (False, False)
2182
1716
            for the versioned pair.
2183
1717
        """
 
1718
        utf8_decode = cache_utf8._utf8_decode
 
1719
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1720
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1721
        # NB: show_status depends on being able to pass in non-versioned files
 
1722
        # and report them as unknown
2184
1723
        # TODO: handle extra trees in the dirstate.
2185
 
        if (extra_trees or specific_files == []):
 
1724
        # TODO: handle comparisons as an empty tree as a different special
 
1725
        # case? mbp 20070226
 
1726
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1727
            or specific_files == []):
2186
1728
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1729
            for f in super(InterDirStateTree, self)._iter_changes(
2188
1730
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1731
                require_versioned, want_unversioned=want_unversioned):
 
1732
                yield f
 
1733
            return
2190
1734
        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))
 
1735
        assert (self.source._revision_id in parent_ids), \
 
1736
                "revision {%s} is not stored in {%s}, but %s " \
 
1737
                "can only be used for trees stored in the dirstate" \
 
1738
                % (self.source._revision_id, self.target, self._iter_changes)
2197
1739
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1740
        if self.source._revision_id == NULL_REVISION:
2199
1741
            source_index = None
2200
1742
            indices = (target_index,)
2201
1743
        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))
 
1744
            assert (self.source._revision_id in parent_ids), \
 
1745
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1746
                self.source._revision_id, parent_ids)
2206
1747
            source_index = 1 + parent_ids.index(self.source._revision_id)
2207
 
            indices = (source_index, target_index)
 
1748
            indices = (source_index,target_index)
2208
1749
        # -- make all specific_files utf8 --
2209
1750
        if specific_files:
2210
1751
            specific_files_utf8 = set()
2211
1752
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1753
                specific_files_utf8.add(path.encode('utf8'))
2215
1754
            specific_files = specific_files_utf8
2216
1755
        else:
2217
1756
            specific_files = set([''])
2218
1757
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1758
        # -- get the state object and prepare it.
2221
1759
        state = self.target.current_dirstate()
2222
1760
        state._read_dirblocks_if_needed()
 
1761
        def _entries_for_path(path):
 
1762
            """Return a list with all the entries that match path for all ids.
 
1763
            """
 
1764
            dirname, basename = os.path.split(path)
 
1765
            key = (dirname, basename, '')
 
1766
            block_index, present = state._find_block_index_from_key(key)
 
1767
            if not present:
 
1768
                # the block which should contain path is absent.
 
1769
                return []
 
1770
            result = []
 
1771
            block = state._dirblocks[block_index][1]
 
1772
            entry_index, _ = state._find_entry_index(key, block)
 
1773
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1774
            while (entry_index < len(block) and
 
1775
                block[entry_index][0][0:2] == key[0:2]):
 
1776
                result.append(block[entry_index])
 
1777
                entry_index += 1
 
1778
            return result
2223
1779
        if require_versioned:
2224
1780
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1781
            all_versioned = True
2226
1782
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1783
                path_entries = _entries_for_path(path)
2228
1784
                if not path_entries:
2229
1785
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1786
                    all_versioned = False
 
1787
                    break
2232
1788
                found_versioned = False
2233
1789
                # for each id at this path
2234
1790
                for entry in path_entries:
2241
1797
                if not found_versioned:
2242
1798
                    # none of the indexes was not 'absent' at all ids for this
2243
1799
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1800
                    all_versioned = False
 
1801
                    break
 
1802
            if not all_versioned:
 
1803
                raise errors.PathsNotVersionedError(specific_files)
2247
1804
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1805
        search_specific_files = set()
 
1806
        for path in specific_files:
 
1807
            other_specific_files = specific_files.difference(set([path]))
 
1808
            if not osutils.is_inside_any(other_specific_files, path):
 
1809
                # this is a top level path, we must check it.
 
1810
                search_specific_files.add(path)
 
1811
        # sketch: 
 
1812
        # compare source_index and target_index at or under each element of search_specific_files.
 
1813
        # follow the following comparison table. Note that we only want to do diff operations when
 
1814
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1815
        # for the target.
 
1816
        # cases:
 
1817
        # 
 
1818
        # Source | Target | disk | action
 
1819
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1820
        #        |        |      | diff check on source-target
 
1821
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1822
        #        |        |      | ???
 
1823
        #   r    |  a     |      | add source to search
 
1824
        #   r    |  a     |  a   | 
 
1825
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1826
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1827
        #   a    | fdlt   |      | add new id
 
1828
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1829
        #   a    |  a     |      | not present in either tree, skip
 
1830
        #   a    |  a     |  a   | not present in any tree, skip
 
1831
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1832
        #        |        |      | may not be selected by the users list of paths.
 
1833
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1834
        #        |        |      | may not be selected by the users list of paths.
 
1835
        #  fdlt  | fdlt   |      | content in both: diff them
 
1836
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1837
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1838
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1839
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1840
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1841
        #        |        |      | this id at the other path.
 
1842
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1843
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1844
        #        |        |      | this id at the other path.
 
1845
 
 
1846
        # for all search_indexs in each path at or under each element of
 
1847
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1848
        # relocated path as one to search if its not searched already. If the
 
1849
        # detail is not relocated, add the id.
 
1850
        searched_specific_files = set()
 
1851
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1852
        # Using a list so that we can access the values and change them in
 
1853
        # nested scope. Each one is [path, file_id, entry]
 
1854
        last_source_parent = [None, None]
 
1855
        last_target_parent = [None, None]
2249
1856
 
2250
1857
        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()
 
1858
 
 
1859
        # Just a sentry, so that _process_entry can say that this
 
1860
        # record is handled, but isn't interesting to process (unchanged)
 
1861
        uninteresting = object()
 
1862
 
 
1863
 
 
1864
        old_dirname_to_file_id = {}
 
1865
        new_dirname_to_file_id = {}
 
1866
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1867
        #       keeping a cache of directories that we have seen.
 
1868
 
 
1869
        def _process_entry(entry, path_info):
 
1870
            """Compare an entry and real disk to generate delta information.
 
1871
 
 
1872
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1873
                the path of entry. If None, then the path is considered absent.
 
1874
                (Perhaps we should pass in a concrete entry for this ?)
 
1875
                Basename is returned as a utf8 string because we expect this
 
1876
                tuple will be ignored, and don't want to take the time to
 
1877
                decode.
 
1878
            :return: None if these don't match
 
1879
                     A tuple of information about the change, or
 
1880
                     the object 'uninteresting' if these match, but are
 
1881
                     basically identical.
 
1882
            """
 
1883
            if source_index is None:
 
1884
                source_details = NULL_PARENT_DETAILS
 
1885
            else:
 
1886
                source_details = entry[1][source_index]
 
1887
            target_details = entry[1][target_index]
 
1888
            target_minikind = target_details[0]
 
1889
            if path_info is not None and target_minikind in 'fdlt':
 
1890
                assert target_index == 0
 
1891
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1892
                                                  stat_value=path_info[3])
 
1893
                # The entry may have been modified by update_entry
 
1894
                target_details = entry[1][target_index]
 
1895
                target_minikind = target_details[0]
 
1896
            else:
 
1897
                link_or_sha1 = None
 
1898
            file_id = entry[0][2]
 
1899
            source_minikind = source_details[0]
 
1900
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1901
                # claimed content in both: diff
 
1902
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1903
                #        |        |      | diff check on source-target
 
1904
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1905
                #        |        |      | ???
 
1906
                if source_minikind in 'r':
 
1907
                    # add the source to the search path to find any children it
 
1908
                    # has.  TODO ? : only add if it is a container ?
 
1909
                    if not osutils.is_inside_any(searched_specific_files,
 
1910
                                                 source_details[1]):
 
1911
                        search_specific_files.add(source_details[1])
 
1912
                    # generate the old path; this is needed for stating later
 
1913
                    # as well.
 
1914
                    old_path = source_details[1]
 
1915
                    old_dirname, old_basename = os.path.split(old_path)
 
1916
                    path = pathjoin(entry[0][0], entry[0][1])
 
1917
                    old_entry = state._get_entry(source_index,
 
1918
                                                 path_utf8=old_path)
 
1919
                    # update the source details variable to be the real
 
1920
                    # location.
 
1921
                    source_details = old_entry[1][source_index]
 
1922
                    source_minikind = source_details[0]
 
1923
                else:
 
1924
                    old_dirname = entry[0][0]
 
1925
                    old_basename = entry[0][1]
 
1926
                    old_path = path = None
 
1927
                if path_info is None:
 
1928
                    # the file is missing on disk, show as removed.
 
1929
                    content_change = True
 
1930
                    target_kind = None
 
1931
                    target_exec = False
 
1932
                else:
 
1933
                    # source and target are both versioned and disk file is present.
 
1934
                    target_kind = path_info[2]
 
1935
                    if target_kind == 'directory':
 
1936
                        if path is None:
 
1937
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1938
                        new_dirname_to_file_id[path] = file_id
 
1939
                        if source_minikind != 'd':
 
1940
                            content_change = True
 
1941
                        else:
 
1942
                            # directories have no fingerprint
 
1943
                            content_change = False
 
1944
                        target_exec = False
 
1945
                    elif target_kind == 'file':
 
1946
                        if source_minikind != 'f':
 
1947
                            content_change = True
 
1948
                        else:
 
1949
                            # We could check the size, but we already have the
 
1950
                            # sha1 hash.
 
1951
                            content_change = (link_or_sha1 != source_details[1])
 
1952
                        # Target details is updated at update_entry time
 
1953
                        if use_filesystem_for_exec:
 
1954
                            # We don't need S_ISREG here, because we are sure
 
1955
                            # we are dealing with a file.
 
1956
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1957
                        else:
 
1958
                            target_exec = target_details[3]
 
1959
                    elif target_kind == 'symlink':
 
1960
                        if source_minikind != 'l':
 
1961
                            content_change = True
 
1962
                        else:
 
1963
                            content_change = (link_or_sha1 != source_details[1])
 
1964
                        target_exec = False
 
1965
                    elif target_kind == 'tree-reference':
 
1966
                        if source_minikind != 't':
 
1967
                            content_change = True
 
1968
                        else:
 
1969
                            content_change = False
 
1970
                        target_exec = False
 
1971
                    else:
 
1972
                        raise Exception, "unknown kind %s" % path_info[2]
 
1973
                if source_minikind == 'd':
 
1974
                    if path is None:
 
1975
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1976
                    old_dirname_to_file_id[old_path] = file_id
 
1977
                # parent id is the entry for the path in the target tree
 
1978
                if old_dirname == last_source_parent[0]:
 
1979
                    source_parent_id = last_source_parent[1]
 
1980
                else:
 
1981
                    try:
 
1982
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1983
                    except KeyError:
 
1984
                        source_parent_entry = state._get_entry(source_index,
 
1985
                                                               path_utf8=old_dirname)
 
1986
                        source_parent_id = source_parent_entry[0][2]
 
1987
                    if source_parent_id == entry[0][2]:
 
1988
                        # This is the root, so the parent is None
 
1989
                        source_parent_id = None
 
1990
                    else:
 
1991
                        last_source_parent[0] = old_dirname
 
1992
                        last_source_parent[1] = source_parent_id
 
1993
                new_dirname = entry[0][0]
 
1994
                if new_dirname == last_target_parent[0]:
 
1995
                    target_parent_id = last_target_parent[1]
 
1996
                else:
 
1997
                    try:
 
1998
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1999
                    except KeyError:
 
2000
                        # TODO: We don't always need to do the lookup, because the
 
2001
                        #       parent entry will be the same as the source entry.
 
2002
                        target_parent_entry = state._get_entry(target_index,
 
2003
                                                               path_utf8=new_dirname)
 
2004
                        assert target_parent_entry != (None, None), (
 
2005
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2006
                            % (new_dirname, entry))
 
2007
                        target_parent_id = target_parent_entry[0][2]
 
2008
                    if target_parent_id == entry[0][2]:
 
2009
                        # This is the root, so the parent is None
 
2010
                        target_parent_id = None
 
2011
                    else:
 
2012
                        last_target_parent[0] = new_dirname
 
2013
                        last_target_parent[1] = target_parent_id
 
2014
 
 
2015
                source_exec = source_details[3]
 
2016
                if (include_unchanged
 
2017
                    or content_change
 
2018
                    or source_parent_id != target_parent_id
 
2019
                    or old_basename != entry[0][1]
 
2020
                    or source_exec != target_exec
 
2021
                    ):
 
2022
                    if old_path is None:
 
2023
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2024
                        old_path_u = utf8_decode(old_path)[0]
 
2025
                        path_u = old_path_u
 
2026
                    else:
 
2027
                        old_path_u = utf8_decode(old_path)[0]
 
2028
                        if old_path == path:
 
2029
                            path_u = old_path_u
 
2030
                        else:
 
2031
                            path_u = utf8_decode(path)[0]
 
2032
                    source_kind = _minikind_to_kind[source_minikind]
 
2033
                    return (entry[0][2],
 
2034
                           (old_path_u, path_u),
 
2035
                           content_change,
 
2036
                           (True, True),
 
2037
                           (source_parent_id, target_parent_id),
 
2038
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2039
                           (source_kind, target_kind),
 
2040
                           (source_exec, target_exec))
 
2041
                else:
 
2042
                    return uninteresting
 
2043
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2044
                # looks like a new file
 
2045
                if path_info is not None:
 
2046
                    path = pathjoin(entry[0][0], entry[0][1])
 
2047
                    # parent id is the entry for the path in the target tree
 
2048
                    # TODO: these are the same for an entire directory: cache em.
 
2049
                    parent_id = state._get_entry(target_index,
 
2050
                                                 path_utf8=entry[0][0])[0][2]
 
2051
                    if parent_id == entry[0][2]:
 
2052
                        parent_id = None
 
2053
                    if use_filesystem_for_exec:
 
2054
                        # We need S_ISREG here, because we aren't sure if this
 
2055
                        # is a file or not.
 
2056
                        target_exec = bool(
 
2057
                            stat.S_ISREG(path_info[3].st_mode)
 
2058
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2059
                    else:
 
2060
                        target_exec = target_details[3]
 
2061
                    return (entry[0][2],
 
2062
                           (None, utf8_decode(path)[0]),
 
2063
                           True,
 
2064
                           (False, True),
 
2065
                           (None, parent_id),
 
2066
                           (None, utf8_decode(entry[0][1])[0]),
 
2067
                           (None, path_info[2]),
 
2068
                           (None, target_exec))
 
2069
                else:
 
2070
                    # but its not on disk: we deliberately treat this as just
 
2071
                    # never-present. (Why ?! - RBC 20070224)
 
2072
                    pass
 
2073
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2074
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2075
                # if its still on disk, *and* theres no other entry at this
 
2076
                # path [we dont know this in this routine at the moment -
 
2077
                # perhaps we should change this - then it would be an unknown.
 
2078
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2079
                # parent id is the entry for the path in the target tree
 
2080
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2081
                if parent_id == entry[0][2]:
 
2082
                    parent_id = None
 
2083
                return (entry[0][2],
 
2084
                       (utf8_decode(old_path)[0], None),
 
2085
                       True,
 
2086
                       (True, False),
 
2087
                       (parent_id, None),
 
2088
                       (utf8_decode(entry[0][1])[0], None),
 
2089
                       (_minikind_to_kind[source_minikind], None),
 
2090
                       (source_details[3], None))
 
2091
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2092
                # a rename; could be a true rename, or a rename inherited from
 
2093
                # a renamed parent. TODO: handle this efficiently. Its not
 
2094
                # common case to rename dirs though, so a correct but slow
 
2095
                # implementation will do.
 
2096
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2097
                    search_specific_files.add(target_details[1])
 
2098
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2099
                # neither of the selected trees contain this file,
 
2100
                # so skip over it. This is not currently directly tested, but
 
2101
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2102
                pass
 
2103
            else:
 
2104
                raise AssertionError("don't know how to compare "
 
2105
                    "source_minikind=%r, target_minikind=%r"
 
2106
                    % (source_minikind, target_minikind))
 
2107
                ## import pdb;pdb.set_trace()
 
2108
            return None
 
2109
 
 
2110
        while search_specific_files:
 
2111
            # TODO: the pending list should be lexically sorted?  the
 
2112
            # interface doesn't require it.
 
2113
            current_root = search_specific_files.pop()
 
2114
            current_root_unicode = current_root.decode('utf8')
 
2115
            searched_specific_files.add(current_root)
 
2116
            # process the entries for this containing directory: the rest will be
 
2117
            # found by their parents recursively.
 
2118
            root_entries = _entries_for_path(current_root)
 
2119
            root_abspath = self.target.abspath(current_root_unicode)
 
2120
            try:
 
2121
                root_stat = os.lstat(root_abspath)
 
2122
            except OSError, e:
 
2123
                if e.errno == errno.ENOENT:
 
2124
                    # the path does not exist: let _process_entry know that.
 
2125
                    root_dir_info = None
 
2126
                else:
 
2127
                    # some other random error: hand it up.
 
2128
                    raise
 
2129
            else:
 
2130
                root_dir_info = ('', current_root,
 
2131
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2132
                    root_abspath)
 
2133
                if root_dir_info[2] == 'directory':
 
2134
                    if self.target._directory_is_tree_reference(
 
2135
                        current_root.decode('utf8')):
 
2136
                        root_dir_info = root_dir_info[:2] + \
 
2137
                            ('tree-reference',) + root_dir_info[3:]
 
2138
 
 
2139
            if not root_entries and not root_dir_info:
 
2140
                # this specified path is not present at all, skip it.
 
2141
                continue
 
2142
            path_handled = False
 
2143
            for entry in root_entries:
 
2144
                result = _process_entry(entry, root_dir_info)
 
2145
                if result is not None:
 
2146
                    path_handled = True
 
2147
                    if result is not uninteresting:
 
2148
                        yield result
 
2149
            if want_unversioned and not path_handled and root_dir_info:
 
2150
                new_executable = bool(
 
2151
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2152
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2153
                yield (None,
 
2154
                       (None, current_root_unicode),
 
2155
                       True,
 
2156
                       (False, False),
 
2157
                       (None, None),
 
2158
                       (None, splitpath(current_root_unicode)[-1]),
 
2159
                       (None, root_dir_info[2]),
 
2160
                       (None, new_executable)
 
2161
                      )
 
2162
            initial_key = (current_root, '', '')
 
2163
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2164
            if block_index == 0:
 
2165
                # we have processed the total root already, but because the
 
2166
                # initial key matched it we should skip it here.
 
2167
                block_index +=1
 
2168
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2169
                current_dir_info = None
 
2170
            else:
 
2171
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2172
                try:
 
2173
                    current_dir_info = dir_iterator.next()
 
2174
                except OSError, e:
 
2175
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2176
                    # python 2.5 has e.errno == EINVAL,
 
2177
                    #            and e.winerror == ERROR_DIRECTORY
 
2178
                    e_winerror = getattr(e, 'winerror', None)
 
2179
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2180
                    # there may be directories in the inventory even though
 
2181
                    # this path is not a file on disk: so mark it as end of
 
2182
                    # iterator
 
2183
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2184
                        current_dir_info = None
 
2185
                    elif (sys.platform == 'win32'
 
2186
                          and (e.errno in win_errors
 
2187
                               or e_winerror in win_errors)):
 
2188
                        current_dir_info = None
 
2189
                    else:
 
2190
                        raise
 
2191
                else:
 
2192
                    if current_dir_info[0][0] == '':
 
2193
                        # remove .bzr from iteration
 
2194
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2195
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2196
                        del current_dir_info[1][bzr_index]
 
2197
            # walk until both the directory listing and the versioned metadata
 
2198
            # are exhausted. 
 
2199
            if (block_index < len(state._dirblocks) and
 
2200
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2201
                current_block = state._dirblocks[block_index]
 
2202
            else:
 
2203
                current_block = None
 
2204
            while (current_dir_info is not None or
 
2205
                   current_block is not None):
 
2206
                if (current_dir_info and current_block
 
2207
                    and current_dir_info[0][0] != current_block[0]):
 
2208
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2209
                        # filesystem data refers to paths not covered by the dirblock.
 
2210
                        # this has two possibilities:
 
2211
                        # A) it is versioned but empty, so there is no block for it
 
2212
                        # B) it is not versioned.
 
2213
 
 
2214
                        # if (A) then we need to recurse into it to check for
 
2215
                        # new unknown files or directories.
 
2216
                        # if (B) then we should ignore it, because we don't
 
2217
                        # recurse into unknown directories.
 
2218
                        path_index = 0
 
2219
                        while path_index < len(current_dir_info[1]):
 
2220
                                current_path_info = current_dir_info[1][path_index]
 
2221
                                if want_unversioned:
 
2222
                                    if current_path_info[2] == 'directory':
 
2223
                                        if self.target._directory_is_tree_reference(
 
2224
                                            current_path_info[0].decode('utf8')):
 
2225
                                            current_path_info = current_path_info[:2] + \
 
2226
                                                ('tree-reference',) + current_path_info[3:]
 
2227
                                    new_executable = bool(
 
2228
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2229
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2230
                                    yield (None,
 
2231
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2232
                                        True,
 
2233
                                        (False, False),
 
2234
                                        (None, None),
 
2235
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2236
                                        (None, current_path_info[2]),
 
2237
                                        (None, new_executable))
 
2238
                                # dont descend into this unversioned path if it is
 
2239
                                # a dir
 
2240
                                if current_path_info[2] in ('directory',
 
2241
                                                            'tree-reference'):
 
2242
                                    del current_dir_info[1][path_index]
 
2243
                                    path_index -= 1
 
2244
                                path_index += 1
 
2245
 
 
2246
                        # This dir info has been handled, go to the next
 
2247
                        try:
 
2248
                            current_dir_info = dir_iterator.next()
 
2249
                        except StopIteration:
 
2250
                            current_dir_info = None
 
2251
                    else:
 
2252
                        # We have a dirblock entry for this location, but there
 
2253
                        # is no filesystem path for this. This is most likely
 
2254
                        # because a directory was removed from the disk.
 
2255
                        # We don't have to report the missing directory,
 
2256
                        # because that should have already been handled, but we
 
2257
                        # need to handle all of the files that are contained
 
2258
                        # within.
 
2259
                        for current_entry in current_block[1]:
 
2260
                            # entry referring to file not present on disk.
 
2261
                            # advance the entry only, after processing.
 
2262
                            result = _process_entry(current_entry, None)
 
2263
                            if result is not None:
 
2264
                                if result is not uninteresting:
 
2265
                                    yield result
 
2266
                        block_index +=1
 
2267
                        if (block_index < len(state._dirblocks) and
 
2268
                            osutils.is_inside(current_root,
 
2269
                                              state._dirblocks[block_index][0])):
 
2270
                            current_block = state._dirblocks[block_index]
 
2271
                        else:
 
2272
                            current_block = None
 
2273
                    continue
 
2274
                entry_index = 0
 
2275
                if current_block and entry_index < len(current_block[1]):
 
2276
                    current_entry = current_block[1][entry_index]
 
2277
                else:
 
2278
                    current_entry = None
 
2279
                advance_entry = True
 
2280
                path_index = 0
 
2281
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2282
                    current_path_info = current_dir_info[1][path_index]
 
2283
                    if current_path_info[2] == 'directory':
 
2284
                        if self.target._directory_is_tree_reference(
 
2285
                            current_path_info[0].decode('utf8')):
 
2286
                            current_path_info = current_path_info[:2] + \
 
2287
                                ('tree-reference',) + current_path_info[3:]
 
2288
                else:
 
2289
                    current_path_info = None
 
2290
                advance_path = True
 
2291
                path_handled = False
 
2292
                while (current_entry is not None or
 
2293
                    current_path_info is not None):
 
2294
                    if current_entry is None:
 
2295
                        # the check for path_handled when the path is adnvaced
 
2296
                        # will yield this path if needed.
 
2297
                        pass
 
2298
                    elif current_path_info is None:
 
2299
                        # no path is fine: the per entry code will handle it.
 
2300
                        result = _process_entry(current_entry, current_path_info)
 
2301
                        if result is not None:
 
2302
                            if result is not uninteresting:
 
2303
                                yield result
 
2304
                    elif (current_entry[0][1] != current_path_info[1]
 
2305
                          or current_entry[1][target_index][0] in 'ar'):
 
2306
                        # The current path on disk doesn't match the dirblock
 
2307
                        # record. Either the dirblock is marked as absent, or
 
2308
                        # the file on disk is not present at all in the
 
2309
                        # dirblock. Either way, report about the dirblock
 
2310
                        # entry, and let other code handle the filesystem one.
 
2311
 
 
2312
                        # Compare the basename for these files to determine
 
2313
                        # which comes first
 
2314
                        if current_path_info[1] < current_entry[0][1]:
 
2315
                            # extra file on disk: pass for now, but only
 
2316
                            # increment the path, not the entry
 
2317
                            advance_entry = False
 
2318
                        else:
 
2319
                            # entry referring to file not present on disk.
 
2320
                            # advance the entry only, after processing.
 
2321
                            result = _process_entry(current_entry, None)
 
2322
                            if result is not None:
 
2323
                                if result is not uninteresting:
 
2324
                                    yield result
 
2325
                            advance_path = False
 
2326
                    else:
 
2327
                        result = _process_entry(current_entry, current_path_info)
 
2328
                        if result is not None:
 
2329
                            path_handled = True
 
2330
                            if result is not uninteresting:
 
2331
                                yield result
 
2332
                    if advance_entry and current_entry is not None:
 
2333
                        entry_index += 1
 
2334
                        if entry_index < len(current_block[1]):
 
2335
                            current_entry = current_block[1][entry_index]
 
2336
                        else:
 
2337
                            current_entry = None
 
2338
                    else:
 
2339
                        advance_entry = True # reset the advance flaga
 
2340
                    if advance_path and current_path_info is not None:
 
2341
                        if not path_handled:
 
2342
                            # unversioned in all regards
 
2343
                            if want_unversioned:
 
2344
                                new_executable = bool(
 
2345
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2346
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2347
                                yield (None,
 
2348
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2349
                                    True,
 
2350
                                    (False, False),
 
2351
                                    (None, None),
 
2352
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2353
                                    (None, current_path_info[2]),
 
2354
                                    (None, new_executable))
 
2355
                            # dont descend into this unversioned path if it is
 
2356
                            # a dir
 
2357
                            if current_path_info[2] in ('directory'):
 
2358
                                del current_dir_info[1][path_index]
 
2359
                                path_index -= 1
 
2360
                        # dont descend the disk iterator into any tree 
 
2361
                        # paths.
 
2362
                        if current_path_info[2] == 'tree-reference':
 
2363
                            del current_dir_info[1][path_index]
 
2364
                            path_index -= 1
 
2365
                        path_index += 1
 
2366
                        if path_index < len(current_dir_info[1]):
 
2367
                            current_path_info = current_dir_info[1][path_index]
 
2368
                            if current_path_info[2] == 'directory':
 
2369
                                if self.target._directory_is_tree_reference(
 
2370
                                    current_path_info[0].decode('utf8')):
 
2371
                                    current_path_info = current_path_info[:2] + \
 
2372
                                        ('tree-reference',) + current_path_info[3:]
 
2373
                        else:
 
2374
                            current_path_info = None
 
2375
                        path_handled = False
 
2376
                    else:
 
2377
                        advance_path = True # reset the advance flagg.
 
2378
                if current_block is not None:
 
2379
                    block_index += 1
 
2380
                    if (block_index < len(state._dirblocks) and
 
2381
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2382
                        current_block = state._dirblocks[block_index]
 
2383
                    else:
 
2384
                        current_block = None
 
2385
                if current_dir_info is not None:
 
2386
                    try:
 
2387
                        current_dir_info = dir_iterator.next()
 
2388
                    except StopIteration:
 
2389
                        current_dir_info = None
 
2390
 
2255
2391
 
2256
2392
    @staticmethod
2257
2393
    def is_compatible(source, target):
2258
2394
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2395
        if not isinstance(target, WorkingTree4):
2260
2396
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2397
        # the source must be a revtreee or dirstate rev tree.
2262
2398
        if not isinstance(source,
2263
2399
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2400
            return False
2265
2401
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2402
        if not (source._revision_id == NULL_REVISION or
2267
2403
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2404
            # TODO: what about ghosts? it may well need to 
2269
2405
            # check for them explicitly.
2270
2406
            return False
2271
2407
        return True
2281
2417
 
2282
2418
    def convert(self, tree):
2283
2419
        # 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
 
2420
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2421
        # tree during upgrade.
2286
2422
        tree._control_files.lock_write()
2287
2423
        try:
2313
2449
 
2314
2450
    def update_format(self, tree):
2315
2451
        """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())
 
2452
        tree._control_files.put_utf8('format',
 
2453
            self.target_format.get_format_string())