~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2007-07-04 01:39:50 UTC
  • mto: This revision was merged to the branch mainline in revision 2581.
  • Revision ID: robertc@robertcollins.net-20070704013950-7pp23plwyqjvgkxg
Review feedback.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2011 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
 
41
import bzrlib
34
42
from bzrlib import (
35
43
    bzrdir,
36
44
    cache_utf8,
37
 
    config,
38
45
    conflicts as _mod_conflicts,
39
 
    debug,
 
46
    delta,
40
47
    dirstate,
41
48
    errors,
42
 
    filters as _mod_filters,
43
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
44
54
    osutils,
45
 
    revision as _mod_revision,
46
55
    revisiontree,
47
 
    trace,
 
56
    textui,
48
57
    transform,
49
 
    views,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
50
61
    )
 
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
 
64
import bzrlib.ui
51
65
""")
52
66
 
 
67
from bzrlib import symbol_versioning
53
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
55
 
from bzrlib.lock import LogicalLockResult
56
 
from bzrlib.lockable_files import LockableFiles
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
57
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
58
73
from bzrlib.mutabletree import needs_tree_write_lock
59
74
from bzrlib.osutils import (
60
75
    file_kind,
61
76
    isdir,
 
77
    normpath,
62
78
    pathjoin,
 
79
    rand_chars,
63
80
    realpath,
64
81
    safe_unicode,
 
82
    splitpath,
65
83
    )
 
84
from bzrlib.trace import mutter, note
66
85
from bzrlib.transport.local import LocalTransport
67
 
from bzrlib.tree import (
68
 
    InterTree,
69
 
    InventoryTree,
70
 
    )
71
 
from bzrlib.workingtree import (
72
 
    InventoryWorkingTree,
73
 
    WorkingTree,
74
 
    WorkingTreeFormat,
75
 
    )
76
 
 
77
 
 
78
 
class DirStateWorkingTree(InventoryWorkingTree):
 
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
 
95
from bzrlib.tree import Tree
 
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
97
 
 
98
 
 
99
# This is the Windows equivalent of ENOTDIR
 
100
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
101
# just an error code.
 
102
ERROR_PATH_NOT_FOUND = 3
 
103
ERROR_DIRECTORY = 267
 
104
 
 
105
 
 
106
class WorkingTree4(WorkingTree3):
 
107
    """This is the Format 4 working tree.
 
108
 
 
109
    This differs from WorkingTree3 by:
 
110
     - Having a consolidated internal dirstate, stored in a
 
111
       randomly-accessible sorted file on disk.
 
112
     - Not having a regular inventory attribute.  One can be synthesized 
 
113
       on demand but this is expensive and should be avoided.
 
114
 
 
115
    This is new in bzr 0.15.
 
116
    """
79
117
 
80
118
    def __init__(self, basedir,
81
119
                 branch,
91
129
        """
92
130
        self._format = _format
93
131
        self.bzrdir = _bzrdir
 
132
        from bzrlib.trace import note, mutter
 
133
        assert isinstance(basedir, basestring), \
 
134
            "base directory %r is not a string" % basedir
94
135
        basedir = safe_unicode(basedir)
95
 
        trace.mutter("opening working tree %r", basedir)
 
136
        mutter("opening working tree %r", basedir)
96
137
        self._branch = branch
 
138
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
139
            "branch %r is not a Branch" % self.branch
97
140
        self.basedir = realpath(basedir)
98
141
        # if branch is at our basedir and is a format 6 or less
99
142
        # assume all other formats have their own control files.
 
143
        assert isinstance(_control_files, LockableFiles), \
 
144
            "_control_files must be a LockableFiles, not %r" % _control_files
100
145
        self._control_files = _control_files
101
 
        self._transport = self._control_files._transport
102
146
        self._dirty = None
103
147
        #-------------
104
148
        # during a read or write lock these objects are set, and are
106
150
        self._dirstate = None
107
151
        self._inventory = None
108
152
        #-------------
109
 
        self._setup_directory_is_tree_reference()
110
 
        self._detect_case_handling()
111
 
        self._rules_searcher = None
112
 
        self.views = self._make_views()
113
 
        #--- allow tests to select the dirstate iter_changes implementation
114
 
        self._iter_changes = dirstate._process_entry
115
153
 
116
154
    @needs_tree_write_lock
117
155
    def _add(self, files, ids, kinds):
119
157
        state = self.current_dirstate()
120
158
        for f, file_id, kind in zip(files, ids, kinds):
121
159
            f = f.strip('/')
 
160
            assert '//' not in f
 
161
            assert '..' not in f
122
162
            if self.path2id(f):
123
163
                # special case tree root handling.
124
164
                if f == '' and self.path2id(f) == ROOT_ID:
132
172
            state.add(f, file_id, kind, None, '')
133
173
        self._make_dirty(reset_inventory=True)
134
174
 
135
 
    def _get_check_refs(self):
136
 
        """Return the references needed to perform a check of this tree."""
137
 
        return [('trees', self.last_revision())]
138
 
 
139
175
    def _make_dirty(self, reset_inventory):
140
176
        """Make the tree state dirty.
141
177
 
149
185
    @needs_tree_write_lock
150
186
    def add_reference(self, sub_tree):
151
187
        # use standard implementation, which calls back to self._add
152
 
        #
 
188
        # 
153
189
        # So we don't store the reference_revision in the working dirstate,
154
 
        # it's just recorded at the moment of commit.
 
190
        # it's just recorded at the moment of commit. 
155
191
        self._add_reference(sub_tree)
156
192
 
157
193
    def break_lock(self):
193
229
 
194
230
    def _comparison_data(self, entry, path):
195
231
        kind, executable, stat_value = \
196
 
            WorkingTree._comparison_data(self, entry, path)
 
232
            WorkingTree3._comparison_data(self, entry, path)
197
233
        # it looks like a plain directory, but it's really a reference -- see
198
234
        # also kind()
199
 
        if (self._repo_supports_tree_reference and kind == 'directory'
200
 
            and entry is not None and entry.kind == 'tree-reference'):
 
235
        if (self._repo_supports_tree_reference and
 
236
            kind == 'directory' and
 
237
            self._directory_is_tree_reference(path)):
201
238
            kind = 'tree-reference'
202
239
        return kind, executable, stat_value
203
240
 
205
242
    def commit(self, message=None, revprops=None, *args, **kwargs):
206
243
        # mark the tree as dirty post commit - commit
207
244
        # can change the current versioned list by doing deletes.
208
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
245
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
209
246
        self._make_dirty(reset_inventory=True)
210
247
        return result
211
248
 
229
266
            return self._dirstate
230
267
        local_path = self.bzrdir.get_workingtree_transport(None
231
268
            ).local_abspath('dirstate')
232
 
        self._dirstate = dirstate.DirState.on_file(local_path,
233
 
            self._sha1_provider(), self._worth_saving_limit())
 
269
        self._dirstate = dirstate.DirState.on_file(local_path)
234
270
        return self._dirstate
235
271
 
236
 
    def _sha1_provider(self):
237
 
        """A function that returns a SHA1Provider suitable for this tree.
238
 
 
239
 
        :return: None if content filtering is not supported by this tree.
240
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
241
 
          form of files, i.e. after read filters are applied.
242
 
        """
243
 
        if self.supports_content_filtering():
244
 
            return ContentFilterAwareSHA1Provider(self)
245
 
        else:
246
 
            return None
247
 
 
248
 
    def _worth_saving_limit(self):
249
 
        """How many hash changes are ok before we must save the dirstate.
250
 
 
251
 
        :return: an integer. -1 means never save.
252
 
        """
253
 
        # FIXME: We want a WorkingTreeStack here -- vila 20110812
254
 
        conf = config.BranchStack(self.branch)
255
 
        return conf.get('bzr.workingtree.worth_saving_limit')
 
272
    def _directory_is_tree_reference(self, relpath):
 
273
        # as a special case, if a directory contains control files then 
 
274
        # it's a tree reference, except that the root of the tree is not
 
275
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
276
        # TODO: We could ask all the control formats whether they
 
277
        # recognize this directory, but at the moment there's no cheap api
 
278
        # to do that.  Since we probably can only nest bzr checkouts and
 
279
        # they always use this name it's ok for now.  -- mbp 20060306
 
280
        #
 
281
        # FIXME: There is an unhandled case here of a subdirectory
 
282
        # containing .bzr but not a branch; that will probably blow up
 
283
        # when you try to commit it.  It might happen if there is a
 
284
        # checkout in a subdirectory.  This can be avoided by not adding
 
285
        # it.  mbp 20070306
256
286
 
257
287
    def filter_unversioned_files(self, paths):
258
288
        """Filter out paths that are versioned.
291
321
 
292
322
    def _generate_inventory(self):
293
323
        """Create and set self.inventory from the dirstate object.
294
 
 
 
324
        
295
325
        This is relatively expensive: we have to walk the entire dirstate.
296
326
        Ideally we would not, and can deprecate this function.
297
327
        """
301
331
        state._read_dirblocks_if_needed()
302
332
        root_key, current_entry = self._get_entry(path='')
303
333
        current_id = root_key[2]
304
 
        if not (current_entry[0][0] == 'd'): # directory
305
 
            raise AssertionError(current_entry)
 
334
        assert current_entry[0][0] == 'd' # directory
306
335
        inv = Inventory(root_id=current_id)
307
336
        # Turn some things into local variables
308
337
        minikind_to_kind = dirstate.DirState._minikind_to_kind
341
370
                    # add this entry to the parent map.
342
371
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
343
372
                elif kind == 'tree-reference':
344
 
                    if not self._repo_supports_tree_reference:
345
 
                        raise errors.UnsupportedOperation(
346
 
                            self._generate_inventory,
347
 
                            self.branch.repository)
 
373
                    assert self._repo_supports_tree_reference, \
 
374
                        "repository of %r " \
 
375
                        "doesn't support tree references " \
 
376
                        "required by entry %r" \
 
377
                        % (self, name)
348
378
                    inv_entry.reference_revision = link_or_sha1 or None
349
379
                elif kind != 'symlink':
350
380
                    raise AssertionError("unknown kind %r" % kind)
351
381
                # These checks cost us around 40ms on a 55k entry tree
352
 
                if file_id in inv_byid:
353
 
                    raise AssertionError('file_id %s already in'
354
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
355
 
                if name_unicode in parent_ie.children:
356
 
                    raise AssertionError('name %r already in parent'
357
 
                        % (name_unicode,))
 
382
                assert file_id not in inv_byid, ('file_id %s already in'
 
383
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
384
                assert name_unicode not in parent_ie.children
358
385
                inv_byid[file_id] = inv_entry
359
386
                parent_ie.children[name_unicode] = inv_entry
360
387
        self._inventory = inv
365
392
        If either file_id or path is supplied, it is used as the key to lookup.
366
393
        If both are supplied, the fastest lookup is used, and an error is
367
394
        raised if they do not both point at the same row.
368
 
 
 
395
        
369
396
        :param file_id: An optional unicode file_id to be looked up.
370
397
        :param path: An optional unicode path to be looked up.
371
398
        :return: The dirstate row tuple for path/file_id, or (None, None)
380
407
    def get_file_sha1(self, file_id, path=None, stat_value=None):
381
408
        # check file id is valid unconditionally.
382
409
        entry = self._get_entry(file_id=file_id, path=path)
383
 
        if entry[0] is None:
384
 
            raise errors.NoSuchId(self, file_id)
 
410
        assert entry[0] is not None, 'what error should this raise'
385
411
        if path is None:
386
412
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
387
413
 
389
415
        state = self.current_dirstate()
390
416
        if stat_value is None:
391
417
            try:
392
 
                stat_value = osutils.lstat(file_abspath)
 
418
                stat_value = os.lstat(file_abspath)
393
419
            except OSError, e:
394
420
                if e.errno == errno.ENOENT:
395
421
                    return None
396
422
                else:
397
423
                    raise
398
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
399
 
            stat_value=stat_value)
 
424
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
425
                                          stat_value=stat_value)
400
426
        if entry[1][0][0] == 'f':
401
 
            if link_or_sha1 is None:
402
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
403
 
                try:
404
 
                    sha1 = osutils.sha_file(file_obj)
405
 
                finally:
406
 
                    file_obj.close()
407
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
408
 
                return sha1
409
 
            else:
410
 
                return link_or_sha1
 
427
            return link_or_sha1
411
428
        return None
412
429
 
413
430
    def _get_inventory(self):
414
431
        """Get the inventory for the tree. This is only valid within a lock."""
415
 
        if 'evil' in debug.debug_flags:
416
 
            trace.mutter_callsite(2,
417
 
                "accessing .inventory forces a size of tree translation.")
418
432
        if self._inventory is not None:
419
433
            return self._inventory
420
434
        self._must_be_locked()
427
441
    @needs_read_lock
428
442
    def get_parent_ids(self):
429
443
        """See Tree.get_parent_ids.
430
 
 
 
444
        
431
445
        This implementation requests the ids list from the dirstate file.
432
446
        """
433
447
        return self.current_dirstate().get_parent_ids()
449
463
 
450
464
    def has_id(self, file_id):
451
465
        state = self.current_dirstate()
 
466
        file_id = osutils.safe_file_id(file_id)
452
467
        row, parents = self._get_entry(file_id=file_id)
453
468
        if row is None:
454
469
            return False
455
470
        return osutils.lexists(pathjoin(
456
471
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
457
472
 
458
 
    def has_or_had_id(self, file_id):
459
 
        state = self.current_dirstate()
460
 
        row, parents = self._get_entry(file_id=file_id)
461
 
        return row is not None
462
 
 
463
473
    @needs_read_lock
464
474
    def id2path(self, file_id):
465
475
        "Convert a file-id to a path."
 
476
        file_id = osutils.safe_file_id(file_id)
466
477
        state = self.current_dirstate()
467
478
        entry = self._get_entry(file_id=file_id)
468
479
        if entry == (None, None):
470
481
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
471
482
        return path_utf8.decode('utf8')
472
483
 
473
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
474
 
        entry = self._get_entry(path=path)
475
 
        if entry == (None, None):
476
 
            return False # Missing entries are not executable
477
 
        return entry[1][0][3] # Executable?
478
 
 
479
484
    if not osutils.supports_executable():
480
485
        def is_executable(self, file_id, path=None):
481
486
            """Test if a file is executable or not.
482
487
 
483
488
            Note: The caller is expected to take a read-lock before calling this.
484
489
            """
 
490
            file_id = osutils.safe_file_id(file_id)
485
491
            entry = self._get_entry(file_id=file_id, path=path)
486
492
            if entry == (None, None):
487
493
                return False
488
494
            return entry[1][0][3]
489
 
 
490
 
        _is_executable_from_path_and_stat = \
491
 
            _is_executable_from_path_and_stat_from_basis
492
495
    else:
493
496
        def is_executable(self, file_id, path=None):
494
497
            """Test if a file is executable or not.
495
498
 
496
499
            Note: The caller is expected to take a read-lock before calling this.
497
500
            """
498
 
            self._must_be_locked()
499
501
            if not path:
 
502
                file_id = osutils.safe_file_id(file_id)
500
503
                path = self.id2path(file_id)
501
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
504
            mode = os.lstat(self.abspath(path)).st_mode
502
505
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
503
506
 
504
 
    def all_file_ids(self):
505
 
        """See Tree.iter_all_file_ids"""
506
 
        self._must_be_locked()
507
 
        result = set()
508
 
        for key, tree_details in self.current_dirstate()._iter_entries():
509
 
            if tree_details[0][0] in ('a', 'r'): # relocated
510
 
                continue
511
 
            result.add(key[2])
512
 
        return result
513
 
 
514
507
    @needs_read_lock
515
508
    def __iter__(self):
516
509
        """Iterate through file_ids for this tree.
529
522
        return iter(result)
530
523
 
531
524
    def iter_references(self):
532
 
        if not self._repo_supports_tree_reference:
533
 
            # When the repo doesn't support references, we will have nothing to
534
 
            # return
535
 
            return
536
525
        for key, tree_details in self.current_dirstate()._iter_entries():
537
526
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
538
527
                # not relevant to the working tree
540
529
            if not key[1]:
541
530
                # the root is not a reference.
542
531
                continue
543
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
532
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
544
533
            try:
545
 
                if self._kind(relpath) == 'tree-reference':
546
 
                    yield relpath, key[2]
 
534
                if self._kind(path) == 'tree-reference':
 
535
                    yield path, key[2]
547
536
            except errors.NoSuchFile:
548
537
                # path is missing on disk.
549
538
                continue
550
539
 
551
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
552
 
        """See MutableTree._observed_sha1."""
553
 
        state = self.current_dirstate()
554
 
        entry = self._get_entry(file_id=file_id, path=path)
555
 
        state._observed_sha1(entry, sha1, statvalue)
556
 
 
557
540
    def kind(self, file_id):
558
541
        """Return the kind of a file.
559
542
 
563
546
        Note: The caller is expected to take a read-lock before calling this.
564
547
        """
565
548
        relpath = self.id2path(file_id)
566
 
        if relpath is None:
567
 
            raise AssertionError(
568
 
                "path for id {%s} is None!" % file_id)
 
549
        assert relpath != None, \
 
550
            "path for id {%s} is None!" % file_id
569
551
        return self._kind(relpath)
570
552
 
571
553
    def _kind(self, relpath):
572
554
        abspath = self.abspath(relpath)
573
555
        kind = file_kind(abspath)
574
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
575
 
            entry = self._get_entry(path=relpath)
576
 
            if entry[1] is not None:
577
 
                if entry[1][0][0] == 't':
578
 
                    kind = 'tree-reference'
 
556
        if (self._repo_supports_tree_reference and
 
557
            kind == 'directory' and
 
558
            self._directory_is_tree_reference(relpath)):
 
559
            kind = 'tree-reference'
579
560
        return kind
580
561
 
581
562
    @needs_read_lock
585
566
        if parent_ids:
586
567
            return parent_ids[0]
587
568
        else:
588
 
            return _mod_revision.NULL_REVISION
 
569
            return None
589
570
 
590
571
    def lock_read(self):
591
 
        """See Branch.lock_read, and WorkingTree.unlock.
592
 
 
593
 
        :return: A bzrlib.lock.LogicalLockResult.
594
 
        """
 
572
        """See Branch.lock_read, and WorkingTree.unlock."""
595
573
        self.branch.lock_read()
596
574
        try:
597
575
            self._control_files.lock_read()
610
588
        except:
611
589
            self.branch.unlock()
612
590
            raise
613
 
        return LogicalLockResult(self.unlock)
614
591
 
615
592
    def _lock_self_write(self):
616
593
        """This should be called after the branch is locked."""
631
608
        except:
632
609
            self.branch.unlock()
633
610
            raise
634
 
        return LogicalLockResult(self.unlock)
635
611
 
636
612
    def lock_tree_write(self):
637
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
638
 
 
639
 
        :return: A bzrlib.lock.LogicalLockResult.
640
 
        """
 
613
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
641
614
        self.branch.lock_read()
642
 
        return self._lock_self_write()
 
615
        self._lock_self_write()
643
616
 
644
617
    def lock_write(self):
645
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
618
        """See MutableTree.lock_write, and WorkingTree.unlock."""
649
619
        self.branch.lock_write()
650
 
        return self._lock_self_write()
 
620
        self._lock_self_write()
651
621
 
652
622
    @needs_tree_write_lock
653
623
    def move(self, from_paths, to_dir, after=False):
655
625
        result = []
656
626
        if not from_paths:
657
627
            return result
 
628
 
658
629
        state = self.current_dirstate()
659
 
        if isinstance(from_paths, basestring):
660
 
            raise ValueError()
 
630
 
 
631
        assert not isinstance(from_paths, basestring)
661
632
        to_dir_utf8 = to_dir.encode('utf8')
662
633
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
663
634
        id_index = state._get_id_index()
685
656
        if self._inventory is not None:
686
657
            update_inventory = True
687
658
            inv = self.inventory
 
659
            to_dir_ie = inv[to_dir_id]
688
660
            to_dir_id = to_entry[0][2]
689
 
            to_dir_ie = inv[to_dir_id]
690
661
        else:
691
662
            update_inventory = False
692
663
 
723
694
            from_entry = self._get_entry(path=from_rel)
724
695
            if from_entry == (None, None):
725
696
                raise errors.BzrMoveFailedError(from_rel,to_dir,
726
 
                    errors.NotVersionedError(path=from_rel))
 
697
                    errors.NotVersionedError(path=str(from_rel)))
727
698
 
728
699
            from_id = from_entry[0][2]
729
700
            to_rel = pathjoin(to_dir, from_tail)
756
727
                if from_missing: # implicitly just update our path mapping
757
728
                    move_file = False
758
729
                elif not after:
759
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
730
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
731
                        extra="(Use --after to update the Bazaar id)")
760
732
 
761
733
            rollbacks = []
762
734
            def rollback_rename():
818
790
                if minikind == 'd':
819
791
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
820
792
                        """Recursively update all entries in this dirblock."""
821
 
                        if from_dir == '':
822
 
                            raise AssertionError("renaming root not supported")
 
793
                        assert from_dir != '', "renaming root not supported"
823
794
                        from_key = (from_dir, '')
824
795
                        from_block_idx, present = \
825
796
                            state._find_block_index_from_key(from_key)
838
809
 
839
810
                        # Grab a copy since move_one may update the list.
840
811
                        for entry in from_block[1][:]:
841
 
                            if not (entry[0][0] == from_dir):
842
 
                                raise AssertionError()
 
812
                            assert entry[0][0] == from_dir
843
813
                            cur_details = entry[1][0]
844
814
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
845
815
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
870
840
                rollback_rename()
871
841
                raise
872
842
            result.append((from_rel, to_rel))
873
 
            state._mark_modified()
 
843
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
874
844
            self._make_dirty(reset_inventory=False)
875
845
 
876
846
        return result
904
874
        for tree in trees:
905
875
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
906
876
                parents):
907
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
908
 
                    trees, require_versioned)
 
877
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
909
878
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
910
879
        # -- make all paths utf8 --
911
880
        paths_utf8 = set()
971
940
            if not all_versioned:
972
941
                raise errors.PathsNotVersionedError(paths)
973
942
        # -- remove redundancy in supplied paths to prevent over-scanning --
974
 
        search_paths = osutils.minimum_path_selection(paths)
975
 
        # sketch:
 
943
        search_paths = set()
 
944
        for path in paths:
 
945
            other_paths = paths.difference(set([path]))
 
946
            if not osutils.is_inside_any(other_paths, path):
 
947
                # this is a top level path, we must check it.
 
948
                search_paths.add(path)
 
949
        # sketch: 
976
950
        # for all search_indexs in each path at or under each element of
977
951
        # search_paths, if the detail is relocated: add the id, and add the
978
952
        # relocated path as one to search if its not searched already. If the
1034
1008
 
1035
1009
    def read_working_inventory(self):
1036
1010
        """Read the working inventory.
1037
 
 
 
1011
        
1038
1012
        This is a meaningless operation for dirstate, but we obey it anyhow.
1039
1013
        """
1040
1014
        return self.inventory
1045
1019
 
1046
1020
        WorkingTree4 supplies revision_trees for any basis tree.
1047
1021
        """
 
1022
        revision_id = osutils.safe_revision_id(revision_id)
1048
1023
        dirstate = self.current_dirstate()
1049
1024
        parent_ids = dirstate.get_parent_ids()
1050
1025
        if revision_id not in parent_ids:
1057
1032
    @needs_tree_write_lock
1058
1033
    def set_last_revision(self, new_revision):
1059
1034
        """Change the last revision in the working tree."""
 
1035
        new_revision = osutils.safe_revision_id(new_revision)
1060
1036
        parents = self.get_parent_ids()
1061
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1062
 
            if len(parents) >= 2:
1063
 
                raise AssertionError(
1064
 
                    "setting the last parent to none with a pending merge is "
1065
 
                    "unsupported.")
 
1037
        if new_revision in (NULL_REVISION, None):
 
1038
            assert len(parents) < 2, (
 
1039
                "setting the last parent to none with a pending merge is "
 
1040
                "unsupported.")
1066
1041
            self.set_parent_ids([])
1067
1042
        else:
1068
1043
            self.set_parent_ids([new_revision] + parents[1:],
1071
1046
    @needs_tree_write_lock
1072
1047
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1073
1048
        """Set the parent ids to revision_ids.
1074
 
 
 
1049
        
1075
1050
        See also set_parent_trees. This api will try to retrieve the tree data
1076
1051
        for each element of revision_ids from the trees repository. If you have
1077
1052
        tree data already available, it is more efficient to use
1081
1056
        :param revision_ids: The revision_ids to set as the parent ids of this
1082
1057
            working tree. Any of these may be ghosts.
1083
1058
        """
 
1059
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1084
1060
        trees = []
1085
1061
        for revision_id in revision_ids:
1086
1062
            try:
1109
1085
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1110
1086
        real_trees = []
1111
1087
        ghosts = []
1112
 
 
1113
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1114
 
        graph = self.branch.repository.get_graph()
1115
 
        heads = graph.heads(parent_ids)
1116
 
        accepted_revisions = set()
1117
 
 
1118
1088
        # convert absent trees to the null tree, which we convert back to
1119
1089
        # missing on access.
1120
1090
        for rev_id, tree in parents_list:
1121
 
            if len(accepted_revisions) > 0:
1122
 
                # we always accept the first tree
1123
 
                if rev_id in accepted_revisions or rev_id not in heads:
1124
 
                    # We have already included either this tree, or its
1125
 
                    # descendent, so we skip it.
1126
 
                    continue
1127
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1091
            rev_id = osutils.safe_revision_id(rev_id)
1128
1092
            if tree is not None:
1129
1093
                real_trees.append((rev_id, tree))
1130
1094
            else:
1131
1095
                real_trees.append((rev_id,
1132
 
                    self.branch.repository.revision_tree(
1133
 
                        _mod_revision.NULL_REVISION)))
 
1096
                    self.branch.repository.revision_tree(None)))
1134
1097
                ghosts.append(rev_id)
1135
 
            accepted_revisions.add(rev_id)
1136
 
        updated = False
1137
 
        if (len(real_trees) == 1
1138
 
            and not ghosts
1139
 
            and self.branch.repository._format.fast_deltas
1140
 
            and isinstance(real_trees[0][1],
1141
 
                revisiontree.InventoryRevisionTree)
1142
 
            and self.get_parent_ids()):
1143
 
            rev_id, rev_tree = real_trees[0]
1144
 
            basis_id = self.get_parent_ids()[0]
1145
 
            # There are times when basis_tree won't be in
1146
 
            # self.branch.repository, (switch, for example)
1147
 
            try:
1148
 
                basis_tree = self.branch.repository.revision_tree(basis_id)
1149
 
            except errors.NoSuchRevision:
1150
 
                # Fall back to the set_parent_trees(), since we can't use
1151
 
                # _make_delta if we can't get the RevisionTree
1152
 
                pass
1153
 
            else:
1154
 
                delta = rev_tree.inventory._make_delta(basis_tree.inventory)
1155
 
                dirstate.update_basis_by_delta(delta, rev_id)
1156
 
                updated = True
1157
 
        if not updated:
1158
 
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1098
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1159
1099
        self._make_dirty(reset_inventory=False)
1160
1100
 
1161
1101
    def _set_root_id(self, file_id):
1165
1105
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1166
1106
            self._make_dirty(reset_inventory=True)
1167
1107
 
1168
 
    def _sha_from_stat(self, path, stat_result):
1169
 
        """Get a sha digest from the tree's stat cache.
1170
 
 
1171
 
        The default implementation assumes no stat cache is present.
1172
 
 
1173
 
        :param path: The path.
1174
 
        :param stat_result: The stat result being looked up.
1175
 
        """
1176
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1177
 
 
1178
1108
    @needs_read_lock
1179
1109
    def supports_tree_reference(self):
1180
1110
        return self._repo_supports_tree_reference
1182
1112
    def unlock(self):
1183
1113
        """Unlock in format 4 trees needs to write the entire dirstate."""
1184
1114
        if self._control_files._lock_count == 1:
1185
 
            # do non-implementation specific cleanup
1186
 
            self._cleanup()
1187
 
 
1188
1115
            # eventually we should do signature checking during read locks for
1189
1116
            # dirstate updates.
1190
1117
            if self._control_files._lock_mode == 'w':
1220
1147
            return
1221
1148
        state = self.current_dirstate()
1222
1149
        state._read_dirblocks_if_needed()
1223
 
        ids_to_unversion = set(file_ids)
 
1150
        ids_to_unversion = set()
 
1151
        for file_id in file_ids:
 
1152
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1224
1153
        paths_to_unversion = set()
1225
1154
        # sketch:
1226
1155
        # check if the root is to be unversioned, if so, assert for now.
1253
1182
                # just forget the whole block.
1254
1183
                entry_index = 0
1255
1184
                while entry_index < len(block[1]):
 
1185
                    # Mark this file id as having been removed
1256
1186
                    entry = block[1][entry_index]
1257
 
                    if entry[1][0][0] in 'ar':
1258
 
                        # don't remove absent or renamed entries
 
1187
                    ids_to_unversion.discard(entry[0][2])
 
1188
                    if (entry[1][0][0] == 'a'
 
1189
                        or not state._make_absent(entry)):
1259
1190
                        entry_index += 1
1260
 
                    else:
1261
 
                        # Mark this file id as having been removed
1262
 
                        ids_to_unversion.discard(entry[0][2])
1263
 
                        if not state._make_absent(entry):
1264
 
                            # The block has not shrunk.
1265
 
                            entry_index += 1
1266
1191
                # go to the next block. (At the moment we dont delete empty
1267
1192
                # dirblocks)
1268
1193
                block_index += 1
1289
1214
        # have to change the legacy inventory too.
1290
1215
        if self._inventory is not None:
1291
1216
            for file_id in file_ids:
1292
 
                if self._inventory.has_id(file_id):
1293
 
                    self._inventory.remove_recursive_id(file_id)
1294
 
 
1295
 
    @needs_tree_write_lock
1296
 
    def rename_one(self, from_rel, to_rel, after=False):
1297
 
        """See WorkingTree.rename_one"""
1298
 
        self.flush()
1299
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
1300
 
 
1301
 
    @needs_tree_write_lock
1302
 
    def apply_inventory_delta(self, changes):
1303
 
        """See MutableTree.apply_inventory_delta"""
1304
 
        state = self.current_dirstate()
1305
 
        state.update_by_delta(changes)
1306
 
        self._make_dirty(reset_inventory=True)
1307
 
 
1308
 
    def update_basis_by_delta(self, new_revid, delta):
1309
 
        """See MutableTree.update_basis_by_delta."""
1310
 
        if self.last_revision() == new_revid:
1311
 
            raise AssertionError()
1312
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1217
                self._inventory.remove_recursive_id(file_id)
1313
1218
 
1314
1219
    @needs_read_lock
1315
1220
    def _validate(self):
1318
1223
    @needs_tree_write_lock
1319
1224
    def _write_inventory(self, inv):
1320
1225
        """Write inventory as the current inventory."""
1321
 
        if self._dirty:
1322
 
            raise AssertionError("attempting to write an inventory when the "
1323
 
                "dirstate is dirty will lose pending changes")
1324
 
        had_inventory = self._inventory is not None
1325
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1326
 
        # working inventory. We do this because self.inventory may be inv, or
1327
 
        # may have been modified, and either case would prevent a clean delta
1328
 
        # being created.
1329
 
        self._inventory = None
1330
 
        # generate a delta,
1331
 
        delta = inv._make_delta(self.inventory)
1332
 
        # and apply it.
1333
 
        self.apply_inventory_delta(delta)
1334
 
        if had_inventory:
 
1226
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1227
        self.current_dirstate().set_state_from_inventory(inv)
 
1228
        self._make_dirty(reset_inventory=False)
 
1229
        if self._inventory is not None:
1335
1230
            self._inventory = inv
1336
1231
        self.flush()
1337
1232
 
1338
 
    @needs_tree_write_lock
1339
 
    def reset_state(self, revision_ids=None):
1340
 
        """Reset the state of the working tree.
1341
 
 
1342
 
        This does a hard-reset to a last-known-good state. This is a way to
1343
 
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
1344
 
        """
1345
 
        if revision_ids is None:
1346
 
            revision_ids = self.get_parent_ids()
1347
 
        if not revision_ids:
1348
 
            base_tree = self.branch.repository.revision_tree(
1349
 
                _mod_revision.NULL_REVISION)
1350
 
            trees = []
1351
 
        else:
1352
 
            trees = zip(revision_ids,
1353
 
                        self.branch.repository.revision_trees(revision_ids))
1354
 
            base_tree = trees[0][1]
1355
 
        state = self.current_dirstate()
1356
 
        # We don't support ghosts yet
1357
 
        state.set_state_from_scratch(base_tree.inventory, trees, [])
1358
 
 
1359
 
 
1360
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1361
 
 
1362
 
    def __init__(self, tree):
1363
 
        self.tree = tree
1364
 
 
1365
 
    def sha1(self, abspath):
1366
 
        """See dirstate.SHA1Provider.sha1()."""
1367
 
        filters = self.tree._content_filter_stack(
1368
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1369
 
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
1370
 
 
1371
 
    def stat_and_sha1(self, abspath):
1372
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1373
 
        filters = self.tree._content_filter_stack(
1374
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1375
 
        file_obj = file(abspath, 'rb', 65000)
1376
 
        try:
1377
 
            statvalue = os.fstat(file_obj.fileno())
1378
 
            if filters:
1379
 
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
1380
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1381
 
        finally:
1382
 
            file_obj.close()
1383
 
        return statvalue, sha1
1384
 
 
1385
 
 
1386
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1387
 
    """Dirstate working tree that supports content filtering.
1388
 
 
1389
 
    The dirstate holds the hash and size of the canonical form of the file, 
1390
 
    and most methods must return that.
1391
 
    """
1392
 
 
1393
 
    def _file_content_summary(self, path, stat_result):
1394
 
        # This is to support the somewhat obsolete path_content_summary method
1395
 
        # with content filtering: see
1396
 
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
1397
 
        #
1398
 
        # If the dirstate cache is up to date and knows the hash and size,
1399
 
        # return that.
1400
 
        # Otherwise if there are no content filters, return the on-disk size
1401
 
        # and leave the hash blank.
1402
 
        # Otherwise, read and filter the on-disk file and use its size and
1403
 
        # hash.
1404
 
        #
1405
 
        # The dirstate doesn't store the size of the canonical form so we
1406
 
        # can't trust it for content-filtered trees.  We just return None.
1407
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1408
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1409
 
        return ('file', None, executable, dirstate_sha1)
1410
 
 
1411
 
 
1412
 
class WorkingTree4(DirStateWorkingTree):
1413
 
    """This is the Format 4 working tree.
1414
 
 
1415
 
    This differs from WorkingTree by:
1416
 
     - Having a consolidated internal dirstate, stored in a
1417
 
       randomly-accessible sorted file on disk.
1418
 
     - Not having a regular inventory attribute.  One can be synthesized
1419
 
       on demand but this is expensive and should be avoided.
1420
 
 
1421
 
    This is new in bzr 0.15.
1422
 
    """
1423
 
 
1424
 
 
1425
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1426
 
    """This is the Format 5 working tree.
1427
 
 
1428
 
    This differs from WorkingTree4 by:
1429
 
     - Supporting content filtering.
1430
 
 
1431
 
    This is new in bzr 1.11.
1432
 
    """
1433
 
 
1434
 
 
1435
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1436
 
    """This is the Format 6 working tree.
1437
 
 
1438
 
    This differs from WorkingTree5 by:
1439
 
     - Supporting a current view that may mask the set of files in a tree
1440
 
       impacted by most user operations.
1441
 
 
1442
 
    This is new in bzr 1.14.
1443
 
    """
1444
 
 
1445
 
    def _make_views(self):
1446
 
        return views.PathBasedViews(self)
1447
 
 
1448
 
 
1449
 
class DirStateWorkingTreeFormat(WorkingTreeFormat):
1450
 
 
1451
 
    missing_parent_conflicts = True
1452
 
 
1453
 
    supports_versioned_directories = True
1454
 
 
1455
 
    _lock_class = LockDir
1456
 
    _lock_file_name = 'lock'
1457
 
 
1458
 
    def _open_control_files(self, a_bzrdir):
1459
 
        transport = a_bzrdir.get_workingtree_transport(None)
1460
 
        return LockableFiles(transport, self._lock_file_name,
1461
 
                             self._lock_class)
1462
 
 
1463
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1464
 
                   accelerator_tree=None, hardlink=False):
 
1233
 
 
1234
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1235
    """The first consolidated dirstate working tree format.
 
1236
 
 
1237
    This format:
 
1238
        - exists within a metadir controlling .bzr
 
1239
        - includes an explicit version marker for the workingtree control
 
1240
          files, separate from the BzrDir format
 
1241
        - modifies the hash cache format
 
1242
        - is new in bzr 0.15
 
1243
        - uses a LockDir to guard access to it.
 
1244
    """
 
1245
 
 
1246
    upgrade_recommended = False
 
1247
 
 
1248
    def get_format_string(self):
 
1249
        """See WorkingTreeFormat.get_format_string()."""
 
1250
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1251
 
 
1252
    def get_format_description(self):
 
1253
        """See WorkingTreeFormat.get_format_description()."""
 
1254
        return "Working tree format 4"
 
1255
 
 
1256
    def initialize(self, a_bzrdir, revision_id=None):
1465
1257
        """See WorkingTreeFormat.initialize().
1466
1258
 
1467
1259
        :param revision_id: allows creating a working tree at a different
1468
 
            revision than the branch is at.
1469
 
        :param accelerator_tree: A tree which can be used for retrieving file
1470
 
            contents more quickly than the revision tree, i.e. a workingtree.
1471
 
            The revision tree will be used for cases where accelerator_tree's
1472
 
            content is different.
1473
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1474
 
            where possible.
 
1260
        revision than the branch is at.
1475
1261
 
1476
1262
        These trees get an initial random root id, if their repository supports
1477
1263
        rich root data, TREE_ROOT otherwise.
1478
1264
        """
 
1265
        revision_id = osutils.safe_revision_id(revision_id)
1479
1266
        if not isinstance(a_bzrdir.transport, LocalTransport):
1480
1267
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1481
1268
        transport = a_bzrdir.get_workingtree_transport(self)
1482
1269
        control_files = self._open_control_files(a_bzrdir)
1483
1270
        control_files.create_lock()
1484
1271
        control_files.lock_write()
1485
 
        transport.put_bytes('format', self.get_format_string(),
1486
 
            mode=a_bzrdir._get_file_mode())
1487
 
        if from_branch is not None:
1488
 
            branch = from_branch
1489
 
        else:
1490
 
            branch = a_bzrdir.open_branch()
 
1272
        control_files.put_utf8('format', self.get_format_string())
 
1273
        branch = a_bzrdir.open_branch()
1491
1274
        if revision_id is None:
1492
1275
            revision_id = branch.last_revision()
1493
1276
        local_path = transport.local_abspath('dirstate')
1495
1278
        state = dirstate.DirState.initialize(local_path)
1496
1279
        state.unlock()
1497
1280
        del state
1498
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1281
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1499
1282
                         branch,
1500
1283
                         _format=self,
1501
1284
                         _bzrdir=a_bzrdir,
1503
1286
        wt._new_tree()
1504
1287
        wt.lock_tree_write()
1505
1288
        try:
1506
 
            self._init_custom_control_files(wt)
1507
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1289
            if revision_id in (None, NULL_REVISION):
1508
1290
                if branch.repository.supports_rich_root():
1509
1291
                    wt._set_root_id(generate_ids.gen_root_id())
1510
1292
                else:
1511
1293
                    wt._set_root_id(ROOT_ID)
1512
1294
                wt.flush()
1513
 
            basis = None
1514
 
            # frequently, we will get here due to branching.  The accelerator
1515
 
            # tree will be the tree from the branch, so the desired basis
1516
 
            # tree will often be a parent of the accelerator tree.
1517
 
            if accelerator_tree is not None:
1518
 
                try:
1519
 
                    basis = accelerator_tree.revision_tree(revision_id)
1520
 
                except errors.NoSuchRevision:
1521
 
                    pass
1522
 
            if basis is None:
1523
 
                basis = branch.repository.revision_tree(revision_id)
1524
 
            if revision_id == _mod_revision.NULL_REVISION:
1525
 
                parents_list = []
1526
 
            else:
1527
 
                parents_list = [(revision_id, basis)]
 
1295
            wt.set_last_revision(revision_id)
 
1296
            wt.flush()
 
1297
            basis = wt.basis_tree()
1528
1298
            basis.lock_read()
1529
 
            try:
1530
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1299
            # if the basis has a root id we have to use that; otherwise we use
 
1300
            # a new random one
 
1301
            basis_root_id = basis.get_root_id()
 
1302
            if basis_root_id is not None:
 
1303
                wt._set_root_id(basis_root_id)
1531
1304
                wt.flush()
1532
 
                # if the basis has a root id we have to use that; otherwise we
1533
 
                # use a new random one
1534
 
                basis_root_id = basis.get_root_id()
1535
 
                if basis_root_id is not None:
1536
 
                    wt._set_root_id(basis_root_id)
1537
 
                    wt.flush()
1538
 
                if wt.supports_content_filtering():
1539
 
                    # The original tree may not have the same content filters
1540
 
                    # applied so we can't safely build the inventory delta from
1541
 
                    # the source tree.
1542
 
                    delta_from_tree = False
1543
 
                else:
1544
 
                    delta_from_tree = True
1545
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1546
 
                # because wt4.apply_inventory_delta does not mutate the input
1547
 
                # inventory entries.
1548
 
                transform.build_tree(basis, wt, accelerator_tree,
1549
 
                                     hardlink=hardlink,
1550
 
                                     delta_from_tree=delta_from_tree)
1551
 
            finally:
1552
 
                basis.unlock()
 
1305
            transform.build_tree(basis, wt)
 
1306
            basis.unlock()
1553
1307
        finally:
1554
1308
            control_files.unlock()
1555
1309
            wt.unlock()
1556
1310
        return wt
1557
1311
 
1558
 
    def _init_custom_control_files(self, wt):
1559
 
        """Subclasses with custom control files should override this method.
1560
 
 
1561
 
        The working tree and control files are locked for writing when this
1562
 
        method is called.
1563
 
 
1564
 
        :param wt: the WorkingTree object
1565
 
        """
1566
 
 
1567
 
    def open(self, a_bzrdir, _found=False):
1568
 
        """Return the WorkingTree object for a_bzrdir
1569
 
 
1570
 
        _found is a private parameter, do not use it. It is used to indicate
1571
 
               if format probing has already been done.
1572
 
        """
1573
 
        if not _found:
1574
 
            # we are being called directly and must probe.
1575
 
            raise NotImplementedError
1576
 
        if not isinstance(a_bzrdir.transport, LocalTransport):
1577
 
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1578
 
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
1579
 
        return wt
1580
 
 
1581
1312
    def _open(self, a_bzrdir, control_files):
1582
1313
        """Open the tree itself.
1583
1314
 
1584
1315
        :param a_bzrdir: the dir for the tree.
1585
1316
        :param control_files: the control files for the tree.
1586
1317
        """
1587
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1318
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1588
1319
                           branch=a_bzrdir.open_branch(),
1589
1320
                           _format=self,
1590
1321
                           _bzrdir=a_bzrdir,
1591
1322
                           _control_files=control_files)
1592
1323
 
1593
1324
    def __get_matchingbzrdir(self):
1594
 
        return self._get_matchingbzrdir()
1595
 
 
1596
 
    def _get_matchingbzrdir(self):
1597
 
        """Overrideable method to get a bzrdir for testing."""
1598
1325
        # please test against something that will let us do tree references
1599
1326
        return bzrdir.format_registry.make_bzrdir(
1600
1327
            'dirstate-with-subtree')
1602
1329
    _matchingbzrdir = property(__get_matchingbzrdir)
1603
1330
 
1604
1331
 
1605
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1606
 
    """The first consolidated dirstate working tree format.
1607
 
 
1608
 
    This format:
1609
 
        - exists within a metadir controlling .bzr
1610
 
        - includes an explicit version marker for the workingtree control
1611
 
          files, separate from the BzrDir format
1612
 
        - modifies the hash cache format
1613
 
        - is new in bzr 0.15
1614
 
        - uses a LockDir to guard access to it.
1615
 
    """
1616
 
 
1617
 
    upgrade_recommended = False
1618
 
 
1619
 
    _tree_class = WorkingTree4
1620
 
 
1621
 
    def get_format_string(self):
1622
 
        """See WorkingTreeFormat.get_format_string()."""
1623
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1624
 
 
1625
 
    def get_format_description(self):
1626
 
        """See WorkingTreeFormat.get_format_description()."""
1627
 
        return "Working tree format 4"
1628
 
 
1629
 
 
1630
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1631
 
    """WorkingTree format supporting content filtering.
1632
 
    """
1633
 
 
1634
 
    upgrade_recommended = False
1635
 
 
1636
 
    _tree_class = WorkingTree5
1637
 
 
1638
 
    def get_format_string(self):
1639
 
        """See WorkingTreeFormat.get_format_string()."""
1640
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1641
 
 
1642
 
    def get_format_description(self):
1643
 
        """See WorkingTreeFormat.get_format_description()."""
1644
 
        return "Working tree format 5"
1645
 
 
1646
 
    def supports_content_filtering(self):
1647
 
        return True
1648
 
 
1649
 
 
1650
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1651
 
    """WorkingTree format supporting views.
1652
 
    """
1653
 
 
1654
 
    upgrade_recommended = False
1655
 
 
1656
 
    _tree_class = WorkingTree6
1657
 
 
1658
 
    def get_format_string(self):
1659
 
        """See WorkingTreeFormat.get_format_string()."""
1660
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1661
 
 
1662
 
    def get_format_description(self):
1663
 
        """See WorkingTreeFormat.get_format_description()."""
1664
 
        return "Working tree format 6"
1665
 
 
1666
 
    def _init_custom_control_files(self, wt):
1667
 
        """Subclasses with custom control files should override this method."""
1668
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1669
 
 
1670
 
    def supports_content_filtering(self):
1671
 
        return True
1672
 
 
1673
 
    def supports_views(self):
1674
 
        return True
1675
 
 
1676
 
 
1677
 
class DirStateRevisionTree(InventoryTree):
1678
 
    """A revision tree pulling the inventory from a dirstate.
1679
 
    
1680
 
    Note that this is one of the historical (ie revision) trees cached in the
1681
 
    dirstate for easy access, not the workingtree.
1682
 
    """
 
1332
class DirStateRevisionTree(Tree):
 
1333
    """A revision tree pulling the inventory from a dirstate."""
1683
1334
 
1684
1335
    def __init__(self, dirstate, revision_id, repository):
1685
1336
        self._dirstate = dirstate
1686
 
        self._revision_id = revision_id
 
1337
        self._revision_id = osutils.safe_revision_id(revision_id)
1687
1338
        self._repository = repository
1688
1339
        self._inventory = None
1689
1340
        self._locked = 0
1690
1341
        self._dirstate_locked = False
1691
 
        self._repo_supports_tree_reference = getattr(
1692
 
            repository._format, "supports_tree_reference",
1693
 
            False)
1694
1342
 
1695
1343
    def __repr__(self):
1696
1344
        return "<%s of %s in %s>" % \
1697
1345
            (self.__class__.__name__, self._revision_id, self._dirstate)
1698
1346
 
1699
 
    def annotate_iter(self, file_id,
1700
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1347
    def annotate_iter(self, file_id):
1701
1348
        """See Tree.annotate_iter"""
1702
 
        text_key = (file_id, self.get_file_revision(file_id))
1703
 
        annotations = self._repository.texts.annotate(text_key)
1704
 
        return [(key[-1], line) for (key, line) in annotations]
 
1349
        w = self._repository.weave_store.get_weave(file_id,
 
1350
                           self._repository.get_transaction())
 
1351
        return w.annotate_iter(self.inventory[file_id].revision)
1705
1352
 
1706
1353
    def _comparison_data(self, entry, path):
1707
1354
        """See Tree._comparison_data."""
1725
1372
    def get_root_id(self):
1726
1373
        return self.path2id('')
1727
1374
 
1728
 
    def id2path(self, file_id):
1729
 
        "Convert a file-id to a path."
1730
 
        entry = self._get_entry(file_id=file_id)
1731
 
        if entry == (None, None):
1732
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1733
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1734
 
        return path_utf8.decode('utf8')
1735
 
 
1736
 
    def iter_references(self):
1737
 
        if not self._repo_supports_tree_reference:
1738
 
            # When the repo doesn't support references, we will have nothing to
1739
 
            # return
1740
 
            return iter([])
1741
 
        # Otherwise, fall back to the default implementation
1742
 
        return super(DirStateRevisionTree, self).iter_references()
1743
 
 
1744
1375
    def _get_parent_index(self):
1745
1376
        """Return the index in the dirstate referenced by this tree."""
1746
1377
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1751
1382
        If either file_id or path is supplied, it is used as the key to lookup.
1752
1383
        If both are supplied, the fastest lookup is used, and an error is
1753
1384
        raised if they do not both point at the same row.
1754
 
 
 
1385
        
1755
1386
        :param file_id: An optional unicode file_id to be looked up.
1756
1387
        :param path: An optional unicode path to be looked up.
1757
1388
        :return: The dirstate row tuple for path/file_id, or (None, None)
1758
1389
        """
1759
1390
        if file_id is None and path is None:
1760
1391
            raise errors.BzrError('must supply file_id or path')
 
1392
        file_id = osutils.safe_file_id(file_id)
1761
1393
        if path is not None:
1762
1394
            path = path.encode('utf8')
1763
1395
        parent_index = self._get_parent_index()
1771
1403
 
1772
1404
        This is relatively expensive: we have to walk the entire dirstate.
1773
1405
        """
1774
 
        if not self._locked:
1775
 
            raise AssertionError(
1776
 
                'cannot generate inventory of an unlocked '
1777
 
                'dirstate revision tree')
 
1406
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1407
            'dirstate revision tree'
1778
1408
        # separate call for profiling - makes it clear where the costs are.
1779
1409
        self._dirstate._read_dirblocks_if_needed()
1780
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1781
 
            raise AssertionError(
1782
 
                'parent %s has disappeared from %s' % (
1783
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1410
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1411
            'parent %s has disappeared from %s' % (
 
1412
            self._revision_id, self._dirstate.get_parent_ids())
1784
1413
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1785
1414
        # This is identical now to the WorkingTree _generate_inventory except
1786
1415
        # for the tree index use.
1787
1416
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1788
1417
        current_id = root_key[2]
1789
 
        if current_entry[parent_index][0] != 'd':
1790
 
            raise AssertionError()
 
1418
        assert current_entry[parent_index][0] == 'd'
1791
1419
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1792
1420
        inv.root.revision = current_entry[parent_index][4]
1793
1421
        # Turn some things into local variables
1824
1452
                elif kind == 'directory':
1825
1453
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1826
1454
                elif kind == 'symlink':
 
1455
                    inv_entry.executable = False
 
1456
                    inv_entry.text_size = None
1827
1457
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1828
1458
                elif kind == 'tree-reference':
1829
1459
                    inv_entry.reference_revision = fingerprint or None
1831
1461
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1832
1462
                            % entry)
1833
1463
                # These checks cost us around 40ms on a 55k entry tree
1834
 
                if file_id in inv_byid:
1835
 
                    raise AssertionError('file_id %s already in'
1836
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1837
 
                if name_unicode in parent_ie.children:
1838
 
                    raise AssertionError('name %r already in parent'
1839
 
                        % (name_unicode,))
 
1464
                assert file_id not in inv_byid
 
1465
                assert name_unicode not in parent_ie.children
1840
1466
                inv_byid[file_id] = inv_entry
1841
1467
                parent_ie.children[name_unicode] = inv_entry
1842
1468
        self._inventory = inv
1852
1478
            return None
1853
1479
        parent_index = self._get_parent_index()
1854
1480
        last_changed_revision = entry[1][parent_index][4]
1855
 
        try:
1856
 
            rev = self._repository.get_revision(last_changed_revision)
1857
 
        except errors.NoSuchRevision:
1858
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1859
 
        return rev.timestamp
 
1481
        return self._repository.get_revision(last_changed_revision).timestamp
1860
1482
 
1861
1483
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1862
1484
        entry = self._get_entry(file_id=file_id, path=path)
1866
1488
            return parent_details[1]
1867
1489
        return None
1868
1490
 
1869
 
    @needs_read_lock
1870
 
    def get_file_revision(self, file_id):
1871
 
        return self.inventory[file_id].revision
1872
 
 
1873
 
    def get_file(self, file_id, path=None):
 
1491
    def get_file(self, file_id):
1874
1492
        return StringIO(self.get_file_text(file_id))
1875
1493
 
 
1494
    def get_file_lines(self, file_id):
 
1495
        ie = self.inventory[file_id]
 
1496
        return self._repository.weave_store.get_weave(file_id,
 
1497
                self._repository.get_transaction()).get_lines(ie.revision)
 
1498
 
1876
1499
    def get_file_size(self, file_id):
1877
 
        """See Tree.get_file_size"""
1878
1500
        return self.inventory[file_id].text_size
1879
1501
 
1880
 
    def get_file_text(self, file_id, path=None):
1881
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1882
 
        return ''.join(content)
 
1502
    def get_file_text(self, file_id):
 
1503
        return ''.join(self.get_file_lines(file_id))
1883
1504
 
1884
1505
    def get_reference_revision(self, file_id, path=None):
1885
1506
        return self.inventory[file_id].reference_revision
1886
1507
 
1887
 
    def iter_files_bytes(self, desired_files):
1888
 
        """See Tree.iter_files_bytes.
1889
 
 
1890
 
        This version is implemented on top of Repository.iter_files_bytes"""
1891
 
        parent_index = self._get_parent_index()
1892
 
        repo_desired_files = []
1893
 
        for file_id, identifier in desired_files:
1894
 
            entry = self._get_entry(file_id)
1895
 
            if entry == (None, None):
1896
 
                raise errors.NoSuchId(self, file_id)
1897
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1898
 
                                       identifier))
1899
 
        return self._repository.iter_files_bytes(repo_desired_files)
1900
 
 
1901
 
    def get_symlink_target(self, file_id, path=None):
 
1508
    def get_symlink_target(self, file_id):
1902
1509
        entry = self._get_entry(file_id=file_id)
1903
1510
        parent_index = self._get_parent_index()
1904
1511
        if entry[1][parent_index][0] != 'l':
1905
1512
            return None
1906
1513
        else:
1907
 
            target = entry[1][parent_index][1]
1908
 
            target = target.decode('utf8')
1909
 
            return target
 
1514
            # At present, none of the tree implementations supports non-ascii
 
1515
            # symlink targets. So we will just assume that the dirstate path is
 
1516
            # correct.
 
1517
            return entry[1][parent_index][1]
1910
1518
 
1911
1519
    def get_revision_id(self):
1912
1520
        """Return the revision id for this tree."""
1930
1538
        return bool(self.path2id(filename))
1931
1539
 
1932
1540
    def kind(self, file_id):
1933
 
        entry = self._get_entry(file_id=file_id)[1]
1934
 
        if entry is None:
1935
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1936
 
        parent_index = self._get_parent_index()
1937
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1938
 
 
1939
 
    def stored_kind(self, file_id):
1940
 
        """See Tree.stored_kind"""
1941
 
        return self.kind(file_id)
1942
 
 
1943
 
    def path_content_summary(self, path):
1944
 
        """See Tree.path_content_summary."""
1945
 
        id = self.inventory.path2id(path)
1946
 
        if id is None:
1947
 
            return ('missing', None, None, None)
1948
 
        entry = self._inventory[id]
1949
 
        kind = entry.kind
1950
 
        if kind == 'file':
1951
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1952
 
        elif kind == 'symlink':
1953
 
            return (kind, None, None, entry.symlink_target)
1954
 
        else:
1955
 
            return (kind, None, None, None)
 
1541
        return self.inventory[file_id].kind
1956
1542
 
1957
1543
    def is_executable(self, file_id, path=None):
1958
1544
        ie = self.inventory[file_id]
1959
1545
        if ie.kind != "file":
1960
 
            return False
 
1546
            return None
1961
1547
        return ie.executable
1962
1548
 
1963
 
    def is_locked(self):
1964
 
        return self._locked
1965
 
 
1966
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1549
    def list_files(self, include_root=False):
1967
1550
        # We use a standard implementation, because DirStateRevisionTree is
1968
1551
        # dealing with one of the parents of the current state
1969
1552
        inv = self._get_inventory()
1970
 
        if from_dir is None:
1971
 
            from_dir_id = None
1972
 
        else:
1973
 
            from_dir_id = inv.path2id(from_dir)
1974
 
            if from_dir_id is None:
1975
 
                # Directory not versioned
1976
 
                return
1977
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1978
 
        if inv.root is not None and not include_root and from_dir is None:
 
1553
        entries = inv.iter_entries()
 
1554
        if self.inventory.root is not None and not include_root:
1979
1555
            entries.next()
1980
1556
        for path, entry in entries:
1981
1557
            yield path, 'V', entry.kind, entry.file_id, entry
1982
1558
 
1983
1559
    def lock_read(self):
1984
 
        """Lock the tree for a set of operations.
1985
 
 
1986
 
        :return: A bzrlib.lock.LogicalLockResult.
1987
 
        """
 
1560
        """Lock the tree for a set of operations."""
1988
1561
        if not self._locked:
1989
1562
            self._repository.lock_read()
1990
1563
            if self._dirstate._lock_token is None:
1991
1564
                self._dirstate.lock_read()
1992
1565
                self._dirstate_locked = True
1993
1566
        self._locked += 1
1994
 
        return LogicalLockResult(self.unlock)
1995
1567
 
1996
1568
    def _must_be_locked(self):
1997
1569
        if not self._locked:
2018
1590
                self._dirstate_locked = False
2019
1591
            self._repository.unlock()
2020
1592
 
2021
 
    @needs_read_lock
2022
 
    def supports_tree_reference(self):
2023
 
        return self._repo_supports_tree_reference
2024
 
 
2025
1593
    def walkdirs(self, prefix=""):
2026
1594
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2027
 
        # implementation based on an inventory.
 
1595
        # implementation based on an inventory.  
2028
1596
        # This should be cleaned up to use the much faster Dirstate code
2029
1597
        # So for now, we just build up the parent inventory, and extract
2030
1598
        # it the same way RevisionTree does.
2059
1627
 
2060
1628
class InterDirStateTree(InterTree):
2061
1629
    """Fast path optimiser for changes_from with dirstate trees.
2062
 
 
2063
 
    This is used only when both trees are in the dirstate working file, and
2064
 
    the source is any parent within the dirstate, and the destination is
 
1630
    
 
1631
    This is used only when both trees are in the dirstate working file, and 
 
1632
    the source is any parent within the dirstate, and the destination is 
2065
1633
    the current working tree of the same dirstate.
2066
1634
    """
2067
1635
    # this could be generalized to allow comparisons between any trees in the
2076
1644
    def make_source_parent_tree(source, target):
2077
1645
        """Change the source tree into a parent of the target."""
2078
1646
        revid = source.commit('record tree')
2079
 
        target.branch.fetch(source.branch, revid)
 
1647
        target.branch.repository.fetch(source.branch.repository, revid)
2080
1648
        target.set_parent_ids([revid])
2081
1649
        return target.basis_tree(), target
2082
1650
 
2083
 
    @classmethod
2084
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
2085
 
        result = klass.make_source_parent_tree(source, target)
2086
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2087
 
        return result
2088
 
 
2089
 
    @classmethod
2090
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2091
 
                                                  target):
2092
 
        from bzrlib.tests.test__dirstate_helpers import \
2093
 
            compiled_dirstate_helpers_feature
2094
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2095
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2096
 
        result = klass.make_source_parent_tree(source, target)
2097
 
        result[1]._iter_changes = ProcessEntryC
2098
 
        return result
2099
 
 
2100
1651
    _matching_from_tree_format = WorkingTreeFormat4()
2101
1652
    _matching_to_tree_format = WorkingTreeFormat4()
2102
 
 
2103
 
    @classmethod
2104
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2105
 
        # This method shouldn't be called, because we have python and C
2106
 
        # specific flavours.
2107
 
        raise NotImplementedError
2108
 
 
2109
 
    def iter_changes(self, include_unchanged=False,
 
1653
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1654
 
 
1655
    def _iter_changes(self, include_unchanged=False,
2110
1656
                      specific_files=None, pb=None, extra_trees=[],
2111
1657
                      require_versioned=True, want_unversioned=False):
2112
1658
        """Return the changes from source to target.
2113
1659
 
2114
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1660
        :return: An iterator that yields tuples. See InterTree._iter_changes
2115
1661
            for details.
2116
1662
        :param specific_files: An optional list of file paths to restrict the
2117
1663
            comparison to. When mapping filenames to ids, all matches in all
2128
1674
            output. An unversioned file is defined as one with (False, False)
2129
1675
            for the versioned pair.
2130
1676
        """
 
1677
        utf8_decode = cache_utf8._utf8_decode
 
1678
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1679
        # NB: show_status depends on being able to pass in non-versioned files
 
1680
        # and report them as unknown
2131
1681
        # TODO: handle extra trees in the dirstate.
2132
 
        if (extra_trees or specific_files == []):
 
1682
        # TODO: handle comparisons as an empty tree as a different special
 
1683
        # case? mbp 20070226
 
1684
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2133
1685
            # we can't fast-path these cases (yet)
2134
 
            return super(InterDirStateTree, self).iter_changes(
 
1686
            for f in super(InterDirStateTree, self)._iter_changes(
2135
1687
                include_unchanged, specific_files, pb, extra_trees,
2136
 
                require_versioned, want_unversioned=want_unversioned)
 
1688
                require_versioned, want_unversioned=want_unversioned):
 
1689
                yield f
 
1690
            return
2137
1691
        parent_ids = self.target.get_parent_ids()
2138
 
        if not (self.source._revision_id in parent_ids
2139
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2140
 
            raise AssertionError(
2141
 
                "revision {%s} is not stored in {%s}, but %s "
2142
 
                "can only be used for trees stored in the dirstate"
2143
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1692
        assert (self.source._revision_id in parent_ids), \
 
1693
                "revision {%s} is not stored in {%s}, but %s " \
 
1694
                "can only be used for trees stored in the dirstate" \
 
1695
                % (self.source._revision_id, self.target, self._iter_changes)
2144
1696
        target_index = 0
2145
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1697
        if self.source._revision_id == NULL_REVISION:
2146
1698
            source_index = None
2147
1699
            indices = (target_index,)
2148
1700
        else:
2149
 
            if not (self.source._revision_id in parent_ids):
2150
 
                raise AssertionError(
2151
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2152
 
                    self.source._revision_id, parent_ids))
 
1701
            assert (self.source._revision_id in parent_ids), \
 
1702
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1703
                self.source._revision_id, parent_ids)
2153
1704
            source_index = 1 + parent_ids.index(self.source._revision_id)
2154
 
            indices = (source_index, target_index)
 
1705
            indices = (source_index,target_index)
2155
1706
        # -- make all specific_files utf8 --
2156
1707
        if specific_files:
2157
1708
            specific_files_utf8 = set()
2158
1709
            for path in specific_files:
2159
 
                # Note, if there are many specific files, using cache_utf8
2160
 
                # would be good here.
2161
1710
                specific_files_utf8.add(path.encode('utf8'))
2162
1711
            specific_files = specific_files_utf8
2163
1712
        else:
2164
1713
            specific_files = set([''])
2165
1714
        # -- specific_files is now a utf8 path set --
2166
 
 
2167
1715
        # -- get the state object and prepare it.
2168
1716
        state = self.target.current_dirstate()
2169
1717
        state._read_dirblocks_if_needed()
 
1718
        def _entries_for_path(path):
 
1719
            """Return a list with all the entries that match path for all ids.
 
1720
            """
 
1721
            dirname, basename = os.path.split(path)
 
1722
            key = (dirname, basename, '')
 
1723
            block_index, present = state._find_block_index_from_key(key)
 
1724
            if not present:
 
1725
                # the block which should contain path is absent.
 
1726
                return []
 
1727
            result = []
 
1728
            block = state._dirblocks[block_index][1]
 
1729
            entry_index, _ = state._find_entry_index(key, block)
 
1730
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1731
            while (entry_index < len(block) and
 
1732
                block[entry_index][0][0:2] == key[0:2]):
 
1733
                result.append(block[entry_index])
 
1734
                entry_index += 1
 
1735
            return result
2170
1736
        if require_versioned:
2171
1737
            # -- check all supplied paths are versioned in a search tree. --
2172
 
            not_versioned = []
 
1738
            all_versioned = True
2173
1739
            for path in specific_files:
2174
 
                path_entries = state._entries_for_path(path)
 
1740
                path_entries = _entries_for_path(path)
2175
1741
                if not path_entries:
2176
1742
                    # this specified path is not present at all: error
2177
 
                    not_versioned.append(path)
2178
 
                    continue
 
1743
                    all_versioned = False
 
1744
                    break
2179
1745
                found_versioned = False
2180
1746
                # for each id at this path
2181
1747
                for entry in path_entries:
2188
1754
                if not found_versioned:
2189
1755
                    # none of the indexes was not 'absent' at all ids for this
2190
1756
                    # path.
2191
 
                    not_versioned.append(path)
2192
 
            if len(not_versioned) > 0:
2193
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1757
                    all_versioned = False
 
1758
                    break
 
1759
            if not all_versioned:
 
1760
                raise errors.PathsNotVersionedError(specific_files)
2194
1761
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2195
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1762
        search_specific_files = set()
 
1763
        for path in specific_files:
 
1764
            other_specific_files = specific_files.difference(set([path]))
 
1765
            if not osutils.is_inside_any(other_specific_files, path):
 
1766
                # this is a top level path, we must check it.
 
1767
                search_specific_files.add(path)
 
1768
        # sketch: 
 
1769
        # compare source_index and target_index at or under each element of search_specific_files.
 
1770
        # follow the following comparison table. Note that we only want to do diff operations when
 
1771
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1772
        # for the target.
 
1773
        # cases:
 
1774
        # 
 
1775
        # Source | Target | disk | action
 
1776
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1777
        #        |        |      | diff check on source-target
 
1778
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1779
        #        |        |      | ???
 
1780
        #   r    |  a     |      | add source to search
 
1781
        #   r    |  a     |  a   | 
 
1782
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1783
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1784
        #   a    | fdlt   |      | add new id
 
1785
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1786
        #   a    |  a     |      | not present in either tree, skip
 
1787
        #   a    |  a     |  a   | not present in any tree, skip
 
1788
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1789
        #        |        |      | may not be selected by the users list of paths.
 
1790
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1791
        #        |        |      | may not be selected by the users list of paths.
 
1792
        #  fdlt  | fdlt   |      | content in both: diff them
 
1793
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1794
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1795
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1796
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1797
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1798
        #        |        |      | this id at the other path.
 
1799
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1800
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1801
        #        |        |      | this id at the other path.
 
1802
 
 
1803
        # for all search_indexs in each path at or under each element of
 
1804
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1805
        # relocated path as one to search if its not searched already. If the
 
1806
        # detail is not relocated, add the id.
 
1807
        searched_specific_files = set()
 
1808
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1809
        # Using a list so that we can access the values and change them in
 
1810
        # nested scope. Each one is [path, file_id, entry]
 
1811
        last_source_parent = [None, None]
 
1812
        last_target_parent = [None, None]
2196
1813
 
2197
1814
        use_filesystem_for_exec = (sys.platform != 'win32')
2198
 
        iter_changes = self.target._iter_changes(include_unchanged,
2199
 
            use_filesystem_for_exec, search_specific_files, state,
2200
 
            source_index, target_index, want_unversioned, self.target)
2201
 
        return iter_changes.iter_changes()
 
1815
 
 
1816
        # Just a sentry, so that _process_entry can say that this
 
1817
        # record is handled, but isn't interesting to process (unchanged)
 
1818
        uninteresting = object()
 
1819
 
 
1820
 
 
1821
        old_dirname_to_file_id = {}
 
1822
        new_dirname_to_file_id = {}
 
1823
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1824
        #       keeping a cache of directories that we have seen.
 
1825
 
 
1826
        def _process_entry(entry, path_info):
 
1827
            """Compare an entry and real disk to generate delta information.
 
1828
 
 
1829
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1830
                the path of entry. If None, then the path is considered absent.
 
1831
                (Perhaps we should pass in a concrete entry for this ?)
 
1832
                Basename is returned as a utf8 string because we expect this
 
1833
                tuple will be ignored, and don't want to take the time to
 
1834
                decode.
 
1835
            :return: None if these don't match
 
1836
                     A tuple of information about the change, or
 
1837
                     the object 'uninteresting' if these match, but are
 
1838
                     basically identical.
 
1839
            """
 
1840
            if source_index is None:
 
1841
                source_details = NULL_PARENT_DETAILS
 
1842
            else:
 
1843
                source_details = entry[1][source_index]
 
1844
            target_details = entry[1][target_index]
 
1845
            target_minikind = target_details[0]
 
1846
            if path_info is not None and target_minikind in 'fdlt':
 
1847
                assert target_index == 0
 
1848
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1849
                                                  stat_value=path_info[3])
 
1850
                # The entry may have been modified by update_entry
 
1851
                target_details = entry[1][target_index]
 
1852
                target_minikind = target_details[0]
 
1853
            else:
 
1854
                link_or_sha1 = None
 
1855
            file_id = entry[0][2]
 
1856
            source_minikind = source_details[0]
 
1857
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1858
                # claimed content in both: diff
 
1859
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1860
                #        |        |      | diff check on source-target
 
1861
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1862
                #        |        |      | ???
 
1863
                if source_minikind in 'r':
 
1864
                    # add the source to the search path to find any children it
 
1865
                    # has.  TODO ? : only add if it is a container ?
 
1866
                    if not osutils.is_inside_any(searched_specific_files,
 
1867
                                                 source_details[1]):
 
1868
                        search_specific_files.add(source_details[1])
 
1869
                    # generate the old path; this is needed for stating later
 
1870
                    # as well.
 
1871
                    old_path = source_details[1]
 
1872
                    old_dirname, old_basename = os.path.split(old_path)
 
1873
                    path = pathjoin(entry[0][0], entry[0][1])
 
1874
                    old_entry = state._get_entry(source_index,
 
1875
                                                 path_utf8=old_path)
 
1876
                    # update the source details variable to be the real
 
1877
                    # location.
 
1878
                    source_details = old_entry[1][source_index]
 
1879
                    source_minikind = source_details[0]
 
1880
                else:
 
1881
                    old_dirname = entry[0][0]
 
1882
                    old_basename = entry[0][1]
 
1883
                    old_path = path = None
 
1884
                if path_info is None:
 
1885
                    # the file is missing on disk, show as removed.
 
1886
                    content_change = True
 
1887
                    target_kind = None
 
1888
                    target_exec = False
 
1889
                else:
 
1890
                    # source and target are both versioned and disk file is present.
 
1891
                    target_kind = path_info[2]
 
1892
                    if target_kind == 'directory':
 
1893
                        if path is None:
 
1894
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1895
                        new_dirname_to_file_id[path] = file_id
 
1896
                        if source_minikind != 'd':
 
1897
                            content_change = True
 
1898
                        else:
 
1899
                            # directories have no fingerprint
 
1900
                            content_change = False
 
1901
                        target_exec = False
 
1902
                    elif target_kind == 'file':
 
1903
                        if source_minikind != 'f':
 
1904
                            content_change = True
 
1905
                        else:
 
1906
                            # We could check the size, but we already have the
 
1907
                            # sha1 hash.
 
1908
                            content_change = (link_or_sha1 != source_details[1])
 
1909
                        # Target details is updated at update_entry time
 
1910
                        if use_filesystem_for_exec:
 
1911
                            # We don't need S_ISREG here, because we are sure
 
1912
                            # we are dealing with a file.
 
1913
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1914
                        else:
 
1915
                            target_exec = target_details[3]
 
1916
                    elif target_kind == 'symlink':
 
1917
                        if source_minikind != 'l':
 
1918
                            content_change = True
 
1919
                        else:
 
1920
                            content_change = (link_or_sha1 != source_details[1])
 
1921
                        target_exec = False
 
1922
                    elif target_kind == 'tree-reference':
 
1923
                        if source_minikind != 't':
 
1924
                            content_change = True
 
1925
                        else:
 
1926
                            content_change = False
 
1927
                        target_exec = False
 
1928
                    else:
 
1929
                        raise Exception, "unknown kind %s" % path_info[2]
 
1930
                if source_minikind == 'd':
 
1931
                    if path is None:
 
1932
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1933
                    old_dirname_to_file_id[old_path] = file_id
 
1934
                # parent id is the entry for the path in the target tree
 
1935
                if old_dirname == last_source_parent[0]:
 
1936
                    source_parent_id = last_source_parent[1]
 
1937
                else:
 
1938
                    try:
 
1939
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1940
                    except KeyError:
 
1941
                        source_parent_entry = state._get_entry(source_index,
 
1942
                                                               path_utf8=old_dirname)
 
1943
                        source_parent_id = source_parent_entry[0][2]
 
1944
                    if source_parent_id == entry[0][2]:
 
1945
                        # This is the root, so the parent is None
 
1946
                        source_parent_id = None
 
1947
                    else:
 
1948
                        last_source_parent[0] = old_dirname
 
1949
                        last_source_parent[1] = source_parent_id
 
1950
                new_dirname = entry[0][0]
 
1951
                if new_dirname == last_target_parent[0]:
 
1952
                    target_parent_id = last_target_parent[1]
 
1953
                else:
 
1954
                    try:
 
1955
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1956
                    except KeyError:
 
1957
                        # TODO: We don't always need to do the lookup, because the
 
1958
                        #       parent entry will be the same as the source entry.
 
1959
                        target_parent_entry = state._get_entry(target_index,
 
1960
                                                               path_utf8=new_dirname)
 
1961
                        assert target_parent_entry != (None, None), (
 
1962
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1963
                            % (new_dirname, entry))
 
1964
                        target_parent_id = target_parent_entry[0][2]
 
1965
                    if target_parent_id == entry[0][2]:
 
1966
                        # This is the root, so the parent is None
 
1967
                        target_parent_id = None
 
1968
                    else:
 
1969
                        last_target_parent[0] = new_dirname
 
1970
                        last_target_parent[1] = target_parent_id
 
1971
 
 
1972
                source_exec = source_details[3]
 
1973
                if (include_unchanged
 
1974
                    or content_change
 
1975
                    or source_parent_id != target_parent_id
 
1976
                    or old_basename != entry[0][1]
 
1977
                    or source_exec != target_exec
 
1978
                    ):
 
1979
                    if old_path is None:
 
1980
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1981
                        old_path_u = utf8_decode(old_path)[0]
 
1982
                        path_u = old_path_u
 
1983
                    else:
 
1984
                        old_path_u = utf8_decode(old_path)[0]
 
1985
                        if old_path == path:
 
1986
                            path_u = old_path_u
 
1987
                        else:
 
1988
                            path_u = utf8_decode(path)[0]
 
1989
                    source_kind = _minikind_to_kind[source_minikind]
 
1990
                    return (entry[0][2],
 
1991
                           (old_path_u, path_u),
 
1992
                           content_change,
 
1993
                           (True, True),
 
1994
                           (source_parent_id, target_parent_id),
 
1995
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
1996
                           (source_kind, target_kind),
 
1997
                           (source_exec, target_exec))
 
1998
                else:
 
1999
                    return uninteresting
 
2000
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2001
                # looks like a new file
 
2002
                if path_info is not None:
 
2003
                    path = pathjoin(entry[0][0], entry[0][1])
 
2004
                    # parent id is the entry for the path in the target tree
 
2005
                    # TODO: these are the same for an entire directory: cache em.
 
2006
                    parent_id = state._get_entry(target_index,
 
2007
                                                 path_utf8=entry[0][0])[0][2]
 
2008
                    if parent_id == entry[0][2]:
 
2009
                        parent_id = None
 
2010
                    if use_filesystem_for_exec:
 
2011
                        # We need S_ISREG here, because we aren't sure if this
 
2012
                        # is a file or not.
 
2013
                        target_exec = bool(
 
2014
                            stat.S_ISREG(path_info[3].st_mode)
 
2015
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2016
                    else:
 
2017
                        target_exec = target_details[3]
 
2018
                    return (entry[0][2],
 
2019
                           (None, utf8_decode(path)[0]),
 
2020
                           True,
 
2021
                           (False, True),
 
2022
                           (None, parent_id),
 
2023
                           (None, utf8_decode(entry[0][1])[0]),
 
2024
                           (None, path_info[2]),
 
2025
                           (None, target_exec))
 
2026
                else:
 
2027
                    # but its not on disk: we deliberately treat this as just
 
2028
                    # never-present. (Why ?! - RBC 20070224)
 
2029
                    pass
 
2030
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2031
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2032
                # if its still on disk, *and* theres no other entry at this
 
2033
                # path [we dont know this in this routine at the moment -
 
2034
                # perhaps we should change this - then it would be an unknown.
 
2035
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2036
                # parent id is the entry for the path in the target tree
 
2037
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2038
                if parent_id == entry[0][2]:
 
2039
                    parent_id = None
 
2040
                return (entry[0][2],
 
2041
                       (utf8_decode(old_path)[0], None),
 
2042
                       True,
 
2043
                       (True, False),
 
2044
                       (parent_id, None),
 
2045
                       (utf8_decode(entry[0][1])[0], None),
 
2046
                       (_minikind_to_kind[source_minikind], None),
 
2047
                       (source_details[3], None))
 
2048
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2049
                # a rename; could be a true rename, or a rename inherited from
 
2050
                # a renamed parent. TODO: handle this efficiently. Its not
 
2051
                # common case to rename dirs though, so a correct but slow
 
2052
                # implementation will do.
 
2053
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2054
                    search_specific_files.add(target_details[1])
 
2055
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2056
                # neither of the selected trees contain this file,
 
2057
                # so skip over it. This is not currently directly tested, but
 
2058
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2059
                pass
 
2060
            else:
 
2061
                raise AssertionError("don't know how to compare "
 
2062
                    "source_minikind=%r, target_minikind=%r"
 
2063
                    % (source_minikind, target_minikind))
 
2064
                ## import pdb;pdb.set_trace()
 
2065
            return None
 
2066
 
 
2067
        while search_specific_files:
 
2068
            # TODO: the pending list should be lexically sorted?  the
 
2069
            # interface doesn't require it.
 
2070
            current_root = search_specific_files.pop()
 
2071
            current_root_unicode = current_root.decode('utf8')
 
2072
            searched_specific_files.add(current_root)
 
2073
            # process the entries for this containing directory: the rest will be
 
2074
            # found by their parents recursively.
 
2075
            root_entries = _entries_for_path(current_root)
 
2076
            root_abspath = self.target.abspath(current_root_unicode)
 
2077
            try:
 
2078
                root_stat = os.lstat(root_abspath)
 
2079
            except OSError, e:
 
2080
                if e.errno == errno.ENOENT:
 
2081
                    # the path does not exist: let _process_entry know that.
 
2082
                    root_dir_info = None
 
2083
                else:
 
2084
                    # some other random error: hand it up.
 
2085
                    raise
 
2086
            else:
 
2087
                root_dir_info = ('', current_root,
 
2088
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2089
                    root_abspath)
 
2090
                if root_dir_info[2] == 'directory':
 
2091
                    if self.target._directory_is_tree_reference(
 
2092
                        current_root.decode('utf8')):
 
2093
                        root_dir_info = root_dir_info[:2] + \
 
2094
                            ('tree-reference',) + root_dir_info[3:]
 
2095
 
 
2096
            if not root_entries and not root_dir_info:
 
2097
                # this specified path is not present at all, skip it.
 
2098
                continue
 
2099
            path_handled = False
 
2100
            for entry in root_entries:
 
2101
                result = _process_entry(entry, root_dir_info)
 
2102
                if result is not None:
 
2103
                    path_handled = True
 
2104
                    if result is not uninteresting:
 
2105
                        yield result
 
2106
            if want_unversioned and not path_handled and root_dir_info:
 
2107
                new_executable = bool(
 
2108
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2109
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2110
                yield (None,
 
2111
                       (None, current_root_unicode),
 
2112
                       True,
 
2113
                       (False, False),
 
2114
                       (None, None),
 
2115
                       (None, splitpath(current_root_unicode)[-1]),
 
2116
                       (None, root_dir_info[2]),
 
2117
                       (None, new_executable)
 
2118
                      )
 
2119
            initial_key = (current_root, '', '')
 
2120
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2121
            if block_index == 0:
 
2122
                # we have processed the total root already, but because the
 
2123
                # initial key matched it we should skip it here.
 
2124
                block_index +=1
 
2125
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2126
                current_dir_info = None
 
2127
            else:
 
2128
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2129
                try:
 
2130
                    current_dir_info = dir_iterator.next()
 
2131
                except OSError, e:
 
2132
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2133
                    # python 2.5 has e.errno == EINVAL,
 
2134
                    #            and e.winerror == ERROR_DIRECTORY
 
2135
                    e_winerror = getattr(e, 'winerror', None)
 
2136
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2137
                    # there may be directories in the inventory even though
 
2138
                    # this path is not a file on disk: so mark it as end of
 
2139
                    # iterator
 
2140
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2141
                        current_dir_info = None
 
2142
                    elif (sys.platform == 'win32'
 
2143
                          and (e.errno in win_errors
 
2144
                               or e_winerror in win_errors)):
 
2145
                        current_dir_info = None
 
2146
                    else:
 
2147
                        raise
 
2148
                else:
 
2149
                    if current_dir_info[0][0] == '':
 
2150
                        # remove .bzr from iteration
 
2151
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2152
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2153
                        del current_dir_info[1][bzr_index]
 
2154
            # walk until both the directory listing and the versioned metadata
 
2155
            # are exhausted. 
 
2156
            if (block_index < len(state._dirblocks) and
 
2157
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2158
                current_block = state._dirblocks[block_index]
 
2159
            else:
 
2160
                current_block = None
 
2161
            while (current_dir_info is not None or
 
2162
                   current_block is not None):
 
2163
                if (current_dir_info and current_block
 
2164
                    and current_dir_info[0][0] != current_block[0]):
 
2165
                    if current_dir_info[0][0].split('/') < current_block[0].split('/'):
 
2166
                        # filesystem data refers to paths not covered by the dirblock.
 
2167
                        # this has two possibilities:
 
2168
                        # A) it is versioned but empty, so there is no block for it
 
2169
                        # B) it is not versioned.
 
2170
 
 
2171
                        # if (A) then we need to recurse into it to check for
 
2172
                        # new unknown files or directories.
 
2173
                        # if (B) then we should ignore it, because we don't
 
2174
                        # recurse into unknown directories.
 
2175
                        path_index = 0
 
2176
                        while path_index < len(current_dir_info[1]):
 
2177
                                current_path_info = current_dir_info[1][path_index]
 
2178
                                if want_unversioned:
 
2179
                                    if current_path_info[2] == 'directory':
 
2180
                                        if self.target._directory_is_tree_reference(
 
2181
                                            current_path_info[0].decode('utf8')):
 
2182
                                            current_path_info = current_path_info[:2] + \
 
2183
                                                ('tree-reference',) + current_path_info[3:]
 
2184
                                    new_executable = bool(
 
2185
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2186
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2187
                                    yield (None,
 
2188
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2189
                                        True,
 
2190
                                        (False, False),
 
2191
                                        (None, None),
 
2192
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2193
                                        (None, current_path_info[2]),
 
2194
                                        (None, new_executable))
 
2195
                                # dont descend into this unversioned path if it is
 
2196
                                # a dir
 
2197
                                if current_path_info[2] in ('directory',
 
2198
                                                            'tree-reference'):
 
2199
                                    del current_dir_info[1][path_index]
 
2200
                                    path_index -= 1
 
2201
                                path_index += 1
 
2202
 
 
2203
                        # This dir info has been handled, go to the next
 
2204
                        try:
 
2205
                            current_dir_info = dir_iterator.next()
 
2206
                        except StopIteration:
 
2207
                            current_dir_info = None
 
2208
                    else:
 
2209
                        # We have a dirblock entry for this location, but there
 
2210
                        # is no filesystem path for this. This is most likely
 
2211
                        # because a directory was removed from the disk.
 
2212
                        # We don't have to report the missing directory,
 
2213
                        # because that should have already been handled, but we
 
2214
                        # need to handle all of the files that are contained
 
2215
                        # within.
 
2216
                        for current_entry in current_block[1]:
 
2217
                            # entry referring to file not present on disk.
 
2218
                            # advance the entry only, after processing.
 
2219
                            result = _process_entry(current_entry, None)
 
2220
                            if result is not None:
 
2221
                                if result is not uninteresting:
 
2222
                                    yield result
 
2223
                        block_index +=1
 
2224
                        if (block_index < len(state._dirblocks) and
 
2225
                            osutils.is_inside(current_root,
 
2226
                                              state._dirblocks[block_index][0])):
 
2227
                            current_block = state._dirblocks[block_index]
 
2228
                        else:
 
2229
                            current_block = None
 
2230
                    continue
 
2231
                entry_index = 0
 
2232
                if current_block and entry_index < len(current_block[1]):
 
2233
                    current_entry = current_block[1][entry_index]
 
2234
                else:
 
2235
                    current_entry = None
 
2236
                advance_entry = True
 
2237
                path_index = 0
 
2238
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2239
                    current_path_info = current_dir_info[1][path_index]
 
2240
                    if current_path_info[2] == 'directory':
 
2241
                        if self.target._directory_is_tree_reference(
 
2242
                            current_path_info[0].decode('utf8')):
 
2243
                            current_path_info = current_path_info[:2] + \
 
2244
                                ('tree-reference',) + current_path_info[3:]
 
2245
                else:
 
2246
                    current_path_info = None
 
2247
                advance_path = True
 
2248
                path_handled = False
 
2249
                while (current_entry is not None or
 
2250
                    current_path_info is not None):
 
2251
                    if current_entry is None:
 
2252
                        # the check for path_handled when the path is adnvaced
 
2253
                        # will yield this path if needed.
 
2254
                        pass
 
2255
                    elif current_path_info is None:
 
2256
                        # no path is fine: the per entry code will handle it.
 
2257
                        result = _process_entry(current_entry, current_path_info)
 
2258
                        if result is not None:
 
2259
                            if result is not uninteresting:
 
2260
                                yield result
 
2261
                    elif (current_entry[0][1] != current_path_info[1]
 
2262
                          or current_entry[1][target_index][0] in 'ar'):
 
2263
                        # The current path on disk doesn't match the dirblock
 
2264
                        # record. Either the dirblock is marked as absent, or
 
2265
                        # the file on disk is not present at all in the
 
2266
                        # dirblock. Either way, report about the dirblock
 
2267
                        # entry, and let other code handle the filesystem one.
 
2268
                        if current_path_info[1].split('/') < current_entry[0][1].split('/'):
 
2269
                            # extra file on disk: pass for now, but only
 
2270
                            # increment the path, not the entry
 
2271
                            advance_entry = False
 
2272
                        else:
 
2273
                            # entry referring to file not present on disk.
 
2274
                            # advance the entry only, after processing.
 
2275
                            result = _process_entry(current_entry, None)
 
2276
                            if result is not None:
 
2277
                                if result is not uninteresting:
 
2278
                                    yield result
 
2279
                            advance_path = False
 
2280
                    else:
 
2281
                        result = _process_entry(current_entry, current_path_info)
 
2282
                        if result is not None:
 
2283
                            path_handled = True
 
2284
                            if result is not uninteresting:
 
2285
                                yield result
 
2286
                    if advance_entry and current_entry is not None:
 
2287
                        entry_index += 1
 
2288
                        if entry_index < len(current_block[1]):
 
2289
                            current_entry = current_block[1][entry_index]
 
2290
                        else:
 
2291
                            current_entry = None
 
2292
                    else:
 
2293
                        advance_entry = True # reset the advance flaga
 
2294
                    if advance_path and current_path_info is not None:
 
2295
                        if not path_handled:
 
2296
                            # unversioned in all regards
 
2297
                            if want_unversioned:
 
2298
                                new_executable = bool(
 
2299
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2300
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2301
                                yield (None,
 
2302
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2303
                                    True,
 
2304
                                    (False, False),
 
2305
                                    (None, None),
 
2306
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2307
                                    (None, current_path_info[2]),
 
2308
                                    (None, new_executable))
 
2309
                            # dont descend into this unversioned path if it is
 
2310
                            # a dir
 
2311
                            if current_path_info[2] in ('directory'):
 
2312
                                del current_dir_info[1][path_index]
 
2313
                                path_index -= 1
 
2314
                        # dont descend the disk iterator into any tree 
 
2315
                        # paths.
 
2316
                        if current_path_info[2] == 'tree-reference':
 
2317
                            del current_dir_info[1][path_index]
 
2318
                            path_index -= 1
 
2319
                        path_index += 1
 
2320
                        if path_index < len(current_dir_info[1]):
 
2321
                            current_path_info = current_dir_info[1][path_index]
 
2322
                            if current_path_info[2] == 'directory':
 
2323
                                if self.target._directory_is_tree_reference(
 
2324
                                    current_path_info[0].decode('utf8')):
 
2325
                                    current_path_info = current_path_info[:2] + \
 
2326
                                        ('tree-reference',) + current_path_info[3:]
 
2327
                        else:
 
2328
                            current_path_info = None
 
2329
                        path_handled = False
 
2330
                    else:
 
2331
                        advance_path = True # reset the advance flagg.
 
2332
                if current_block is not None:
 
2333
                    block_index += 1
 
2334
                    if (block_index < len(state._dirblocks) and
 
2335
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2336
                        current_block = state._dirblocks[block_index]
 
2337
                    else:
 
2338
                        current_block = None
 
2339
                if current_dir_info is not None:
 
2340
                    try:
 
2341
                        current_dir_info = dir_iterator.next()
 
2342
                    except StopIteration:
 
2343
                        current_dir_info = None
 
2344
 
2202
2345
 
2203
2346
    @staticmethod
2204
2347
    def is_compatible(source, target):
2205
2348
        # the target must be a dirstate working tree
2206
 
        if not isinstance(target, DirStateWorkingTree):
 
2349
        if not isinstance(target, WorkingTree4):
2207
2350
            return False
2208
 
        # the source must be a revtree or dirstate rev tree.
 
2351
        # the source must be a revtreee or dirstate rev tree.
2209
2352
        if not isinstance(source,
2210
2353
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2211
2354
            return False
2212
2355
        # the source revid must be in the target dirstate
2213
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2356
        if not (source._revision_id == NULL_REVISION or
2214
2357
            source._revision_id in target.get_parent_ids()):
2215
 
            # TODO: what about ghosts? it may well need to
 
2358
            # TODO: what about ghosts? it may well need to 
2216
2359
            # check for them explicitly.
2217
2360
            return False
2218
2361
        return True
2228
2371
 
2229
2372
    def convert(self, tree):
2230
2373
        # lock the control files not the tree, so that we dont get tree
2231
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2374
        # on-unlock behaviours, and so that noone else diddles with the 
2232
2375
        # tree during upgrade.
2233
2376
        tree._control_files.lock_write()
2234
2377
        try:
2260
2403
 
2261
2404
    def update_format(self, tree):
2262
2405
        """Change the format marker."""
2263
 
        tree._transport.put_bytes('format',
2264
 
            self.target_format.get_format_string(),
2265
 
            mode=tree.bzrdir._get_file_mode())
2266
 
 
2267
 
 
2268
 
class Converter4to5(object):
2269
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2270
 
 
2271
 
    def __init__(self):
2272
 
        self.target_format = WorkingTreeFormat5()
2273
 
 
2274
 
    def convert(self, tree):
2275
 
        # lock the control files not the tree, so that we don't get tree
2276
 
        # on-unlock behaviours, and so that no-one else diddles with the
2277
 
        # tree during upgrade.
2278
 
        tree._control_files.lock_write()
2279
 
        try:
2280
 
            self.update_format(tree)
2281
 
        finally:
2282
 
            tree._control_files.unlock()
2283
 
 
2284
 
    def update_format(self, tree):
2285
 
        """Change the format marker."""
2286
 
        tree._transport.put_bytes('format',
2287
 
            self.target_format.get_format_string(),
2288
 
            mode=tree.bzrdir._get_file_mode())
2289
 
 
2290
 
 
2291
 
class Converter4or5to6(object):
2292
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2293
 
 
2294
 
    def __init__(self):
2295
 
        self.target_format = WorkingTreeFormat6()
2296
 
 
2297
 
    def convert(self, tree):
2298
 
        # lock the control files not the tree, so that we don't get tree
2299
 
        # on-unlock behaviours, and so that no-one else diddles with the
2300
 
        # tree during upgrade.
2301
 
        tree._control_files.lock_write()
2302
 
        try:
2303
 
            self.init_custom_control_files(tree)
2304
 
            self.update_format(tree)
2305
 
        finally:
2306
 
            tree._control_files.unlock()
2307
 
 
2308
 
    def init_custom_control_files(self, tree):
2309
 
        """Initialize custom control files."""
2310
 
        tree._transport.put_bytes('views', '',
2311
 
            mode=tree.bzrdir._get_file_mode())
2312
 
 
2313
 
    def update_format(self, tree):
2314
 
        """Change the format marker."""
2315
 
        tree._transport.put_bytes('format',
2316
 
            self.target_format.get_format_string(),
2317
 
            mode=tree.bzrdir._get_file_mode())
 
2406
        tree._control_files.put_utf8('format',
 
2407
            self.target_format.get_format_string())