/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2007-04-13 05:06:23 UTC
  • mfrom: (2279.7.13 bzr.dev.hpss.api.changes)
  • Revision ID: pqm@pqm.ubuntu.com-20070413050623-10v4wozs1tu04kcu
(Andrew Bennetts, Robert Collins) Add a 'token' argument to lock_write.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
38
 
    debug,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
39
47
    dirstate,
40
48
    errors,
41
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
 
    revision as _mod_revision,
44
55
    revisiontree,
45
 
    trace,
 
56
    textui,
46
57
    transform,
47
 
    views,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
48
61
    )
49
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
50
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.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
56
73
from bzrlib.mutabletree import needs_tree_write_lock
57
74
from bzrlib.osutils import (
58
75
    file_kind,
59
76
    isdir,
 
77
    normpath,
60
78
    pathjoin,
 
79
    rand_chars,
61
80
    realpath,
62
81
    safe_unicode,
 
82
    splitpath,
63
83
    )
64
 
from bzrlib.trace import mutter
 
84
from bzrlib.trace import mutter, note
65
85
from bzrlib.transport.local import LocalTransport
66
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
        )
67
95
from bzrlib.tree import Tree
68
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
97
 
70
98
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
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_DIRECTORY = 267
 
103
 
 
104
 
 
105
class WorkingTree4(WorkingTree3):
 
106
    """This is the Format 4 working tree.
 
107
 
 
108
    This differs from WorkingTree3 by:
 
109
     - Having a consolidated internal dirstate, stored in a
 
110
       randomly-accessible sorted file on disk.
 
111
     - Not having a regular inventory attribute.  One can be synthesized 
 
112
       on demand but this is expensive and should be avoided.
 
113
 
 
114
    This is new in bzr 0.15.
 
115
    """
 
116
 
72
117
    def __init__(self, basedir,
73
118
                 branch,
74
119
                 _control_files=None,
83
128
        """
84
129
        self._format = _format
85
130
        self.bzrdir = _bzrdir
 
131
        from bzrlib.trace import note, mutter
 
132
        assert isinstance(basedir, basestring), \
 
133
            "base directory %r is not a string" % basedir
86
134
        basedir = safe_unicode(basedir)
87
135
        mutter("opening working tree %r", basedir)
88
136
        self._branch = branch
 
137
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
138
            "branch %r is not a Branch" % self.branch
89
139
        self.basedir = realpath(basedir)
90
140
        # if branch is at our basedir and is a format 6 or less
91
141
        # assume all other formats have their own control files.
 
142
        assert isinstance(_control_files, LockableFiles), \
 
143
            "_control_files must be a LockableFiles, not %r" % _control_files
92
144
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
145
        self._dirty = None
95
146
        #-------------
96
147
        # during a read or write lock these objects are set, and are
98
149
        self._dirstate = None
99
150
        self._inventory = None
100
151
        #-------------
101
 
        self._setup_directory_is_tree_reference()
102
 
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
152
 
108
153
    @needs_tree_write_lock
109
154
    def _add(self, files, ids, kinds):
111
156
        state = self.current_dirstate()
112
157
        for f, file_id, kind in zip(files, ids, kinds):
113
158
            f = f.strip('/')
 
159
            assert '//' not in f
 
160
            assert '..' not in f
114
161
            if self.path2id(f):
115
162
                # special case tree root handling.
116
163
                if f == '' and self.path2id(f) == ROOT_ID:
137
184
    @needs_tree_write_lock
138
185
    def add_reference(self, sub_tree):
139
186
        # use standard implementation, which calls back to self._add
140
 
        #
 
187
        # 
141
188
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
189
        # it's just recorded at the moment of commit. 
143
190
        self._add_reference(sub_tree)
144
191
 
145
192
    def break_lock(self):
184
231
            WorkingTree3._comparison_data(self, entry, path)
185
232
        # it looks like a plain directory, but it's really a reference -- see
186
233
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
189
237
            kind = 'tree-reference'
190
238
        return kind, executable, stat_value
191
239
 
217
265
            return self._dirstate
218
266
        local_path = self.bzrdir.get_workingtree_transport(None
219
267
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
222
269
        return self._dirstate
223
270
 
224
 
    def _sha1_provider(self):
225
 
        """A function that returns a SHA1Provider suitable for this tree.
226
 
 
227
 
        :return: None if content filtering is not supported by this tree.
228
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
229
 
          form of files, i.e. after read filters are applied.
230
 
        """
231
 
        if self.supports_content_filtering():
232
 
            return ContentFilterAwareSHA1Provider(self)
233
 
        else:
234
 
            return None
 
271
    def _directory_is_tree_reference(self, relpath):
 
272
        # as a special case, if a directory contains control files then 
 
273
        # it's a tree reference, except that the root of the tree is not
 
274
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
275
        # TODO: We could ask all the control formats whether they
 
276
        # recognize this directory, but at the moment there's no cheap api
 
277
        # to do that.  Since we probably can only nest bzr checkouts and
 
278
        # they always use this name it's ok for now.  -- mbp 20060306
 
279
        #
 
280
        # FIXME: There is an unhandled case here of a subdirectory
 
281
        # containing .bzr but not a branch; that will probably blow up
 
282
        # when you try to commit it.  It might happen if there is a
 
283
        # checkout in a subdirectory.  This can be avoided by not adding
 
284
        # it.  mbp 20070306
235
285
 
236
286
    def filter_unversioned_files(self, paths):
237
287
        """Filter out paths that are versioned.
270
320
 
271
321
    def _generate_inventory(self):
272
322
        """Create and set self.inventory from the dirstate object.
273
 
 
 
323
        
274
324
        This is relatively expensive: we have to walk the entire dirstate.
275
325
        Ideally we would not, and can deprecate this function.
276
326
        """
280
330
        state._read_dirblocks_if_needed()
281
331
        root_key, current_entry = self._get_entry(path='')
282
332
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
333
        assert current_entry[0][0] == 'd' # directory
285
334
        inv = Inventory(root_id=current_id)
286
335
        # Turn some things into local variables
287
336
        minikind_to_kind = dirstate.DirState._minikind_to_kind
320
369
                    # add this entry to the parent map.
321
370
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
371
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
 
372
                    assert self._repo_supports_tree_reference, \
 
373
                        "repository of %r " \
 
374
                        "doesn't support tree references " \
 
375
                        "required by entry %r" \
 
376
                        % (self, name)
327
377
                    inv_entry.reference_revision = link_or_sha1 or None
328
378
                elif kind != 'symlink':
329
379
                    raise AssertionError("unknown kind %r" % kind)
330
380
                # These checks cost us around 40ms on a 55k entry tree
331
 
                if file_id in inv_byid:
332
 
                    raise AssertionError('file_id %s already in'
333
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
334
 
                if name_unicode in parent_ie.children:
335
 
                    raise AssertionError('name %r already in parent'
336
 
                        % (name_unicode,))
 
381
                assert file_id not in inv_byid, ('file_id %s already in'
 
382
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
383
                assert name_unicode not in parent_ie.children
337
384
                inv_byid[file_id] = inv_entry
338
385
                parent_ie.children[name_unicode] = inv_entry
339
386
        self._inventory = inv
344
391
        If either file_id or path is supplied, it is used as the key to lookup.
345
392
        If both are supplied, the fastest lookup is used, and an error is
346
393
        raised if they do not both point at the same row.
347
 
 
 
394
        
348
395
        :param file_id: An optional unicode file_id to be looked up.
349
396
        :param path: An optional unicode path to be looked up.
350
397
        :return: The dirstate row tuple for path/file_id, or (None, None)
359
406
    def get_file_sha1(self, file_id, path=None, stat_value=None):
360
407
        # check file id is valid unconditionally.
361
408
        entry = self._get_entry(file_id=file_id, path=path)
362
 
        if entry[0] is None:
363
 
            raise errors.NoSuchId(self, file_id)
 
409
        assert entry[0] is not None, 'what error should this raise'
364
410
        if path is None:
365
411
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
366
412
 
367
413
        file_abspath = self.abspath(path)
368
414
        state = self.current_dirstate()
369
 
        if stat_value is None:
370
 
            try:
371
 
                stat_value = os.lstat(file_abspath)
372
 
            except OSError, e:
373
 
                if e.errno == errno.ENOENT:
374
 
                    return None
375
 
                else:
376
 
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
415
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
416
                                          stat_value=stat_value)
379
417
        if entry[1][0][0] == 'f':
380
 
            if link_or_sha1 is None:
381
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
382
 
                try:
383
 
                    sha1 = osutils.sha_file(file_obj)
384
 
                finally:
385
 
                    file_obj.close()
386
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
387
 
                return sha1
388
 
            else:
389
 
                return link_or_sha1
 
418
            return link_or_sha1
390
419
        return None
391
420
 
392
421
    def _get_inventory(self):
393
422
        """Get the inventory for the tree. This is only valid within a lock."""
394
 
        if 'evil' in debug.debug_flags:
395
 
            trace.mutter_callsite(2,
396
 
                "accessing .inventory forces a size of tree translation.")
397
423
        if self._inventory is not None:
398
424
            return self._inventory
399
425
        self._must_be_locked()
406
432
    @needs_read_lock
407
433
    def get_parent_ids(self):
408
434
        """See Tree.get_parent_ids.
409
 
 
 
435
        
410
436
        This implementation requests the ids list from the dirstate file.
411
437
        """
412
438
        return self.current_dirstate().get_parent_ids()
428
454
 
429
455
    def has_id(self, file_id):
430
456
        state = self.current_dirstate()
 
457
        file_id = osutils.safe_file_id(file_id)
431
458
        row, parents = self._get_entry(file_id=file_id)
432
459
        if row is None:
433
460
            return False
434
461
        return osutils.lexists(pathjoin(
435
462
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
463
 
437
 
    def has_or_had_id(self, file_id):
438
 
        state = self.current_dirstate()
439
 
        row, parents = self._get_entry(file_id=file_id)
440
 
        return row is not None
441
 
 
442
464
    @needs_read_lock
443
465
    def id2path(self, file_id):
444
 
        "Convert a file-id to a path."
 
466
        file_id = osutils.safe_file_id(file_id)
445
467
        state = self.current_dirstate()
446
468
        entry = self._get_entry(file_id=file_id)
447
469
        if entry == (None, None):
449
471
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
472
        return path_utf8.decode('utf8')
451
473
 
452
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
453
 
        entry = self._get_entry(path=path)
454
 
        if entry == (None, None):
455
 
            return False # Missing entries are not executable
456
 
        return entry[1][0][3] # Executable?
457
 
 
458
474
    if not osutils.supports_executable():
 
475
        @needs_read_lock
459
476
        def is_executable(self, file_id, path=None):
460
 
            """Test if a file is executable or not.
461
 
 
462
 
            Note: The caller is expected to take a read-lock before calling this.
463
 
            """
 
477
            file_id = osutils.safe_file_id(file_id)
464
478
            entry = self._get_entry(file_id=file_id, path=path)
465
479
            if entry == (None, None):
466
480
                return False
467
481
            return entry[1][0][3]
468
 
 
469
 
        _is_executable_from_path_and_stat = \
470
 
            _is_executable_from_path_and_stat_from_basis
471
482
    else:
 
483
        @needs_read_lock
472
484
        def is_executable(self, file_id, path=None):
473
 
            """Test if a file is executable or not.
474
 
 
475
 
            Note: The caller is expected to take a read-lock before calling this.
476
 
            """
477
 
            self._must_be_locked()
478
485
            if not path:
 
486
                file_id = osutils.safe_file_id(file_id)
479
487
                path = self.id2path(file_id)
480
488
            mode = os.lstat(self.abspath(path)).st_mode
481
489
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
490
 
483
 
    def all_file_ids(self):
484
 
        """See Tree.iter_all_file_ids"""
485
 
        self._must_be_locked()
486
 
        result = set()
487
 
        for key, tree_details in self.current_dirstate()._iter_entries():
488
 
            if tree_details[0][0] in ('a', 'r'): # relocated
489
 
                continue
490
 
            result.add(key[2])
491
 
        return result
492
 
 
493
491
    @needs_read_lock
494
492
    def __iter__(self):
495
493
        """Iterate through file_ids for this tree.
508
506
        return iter(result)
509
507
 
510
508
    def iter_references(self):
511
 
        if not self._repo_supports_tree_reference:
512
 
            # When the repo doesn't support references, we will have nothing to
513
 
            # return
514
 
            return
515
509
        for key, tree_details in self.current_dirstate()._iter_entries():
516
510
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
511
                # not relevant to the working tree
519
513
            if not key[1]:
520
514
                # the root is not a reference.
521
515
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
516
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
517
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
518
                if self._kind(path) == 'tree-reference':
 
519
                    yield path, key[2]
526
520
            except errors.NoSuchFile:
527
521
                # path is missing on disk.
528
522
                continue
529
523
 
530
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
531
 
        """See MutableTree._observed_sha1."""
532
 
        state = self.current_dirstate()
533
 
        entry = self._get_entry(file_id=file_id, path=path)
534
 
        state._observed_sha1(entry, sha1, statvalue)
535
 
 
 
524
    @needs_read_lock
536
525
    def kind(self, file_id):
537
526
        """Return the kind of a file.
538
527
 
539
528
        This is always the actual kind that's on disk, regardless of what it
540
529
        was added as.
541
 
 
542
 
        Note: The caller is expected to take a read-lock before calling this.
543
530
        """
544
531
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
532
        assert relpath != None, \
 
533
            "path for id {%s} is None!" % file_id
548
534
        return self._kind(relpath)
549
535
 
550
536
    def _kind(self, relpath):
551
537
        abspath = self.abspath(relpath)
552
538
        kind = file_kind(abspath)
553
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
554
 
            entry = self._get_entry(path=relpath)
555
 
            if entry[1] is not None:
556
 
                if entry[1][0][0] == 't':
557
 
                    kind = 'tree-reference'
 
539
        if (self._repo_supports_tree_reference and
 
540
            kind == 'directory' and
 
541
            self._directory_is_tree_reference(relpath)):
 
542
            kind = 'tree-reference'
558
543
        return kind
559
544
 
560
545
    @needs_read_lock
564
549
        if parent_ids:
565
550
            return parent_ids[0]
566
551
        else:
567
 
            return _mod_revision.NULL_REVISION
 
552
            return None
568
553
 
569
554
    def lock_read(self):
570
 
        """See Branch.lock_read, and WorkingTree.unlock.
571
 
 
572
 
        :return: An object with an unlock method which will release the lock
573
 
            obtained.
574
 
        """
 
555
        """See Branch.lock_read, and WorkingTree.unlock."""
575
556
        self.branch.lock_read()
576
557
        try:
577
558
            self._control_files.lock_read()
590
571
        except:
591
572
            self.branch.unlock()
592
573
            raise
593
 
        return self
594
574
 
595
575
    def _lock_self_write(self):
596
576
        """This should be called after the branch is locked."""
611
591
        except:
612
592
            self.branch.unlock()
613
593
            raise
614
 
        return self
615
594
 
616
595
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: An object with an unlock method which will release the lock
620
 
            obtained.
621
 
        """
 
596
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
622
597
        self.branch.lock_read()
623
 
        return self._lock_self_write()
 
598
        self._lock_self_write()
624
599
 
625
600
    def lock_write(self):
626
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
627
 
 
628
 
        :return: An object with an unlock method which will release the lock
629
 
            obtained.
630
 
        """
 
601
        """See MutableTree.lock_write, and WorkingTree.unlock."""
631
602
        self.branch.lock_write()
632
 
        return self._lock_self_write()
 
603
        self._lock_self_write()
633
604
 
634
605
    @needs_tree_write_lock
635
606
    def move(self, from_paths, to_dir, after=False):
637
608
        result = []
638
609
        if not from_paths:
639
610
            return result
 
611
 
640
612
        state = self.current_dirstate()
641
 
        if isinstance(from_paths, basestring):
642
 
            raise ValueError()
 
613
 
 
614
        assert not isinstance(from_paths, basestring)
643
615
        to_dir_utf8 = to_dir.encode('utf8')
644
616
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
645
617
        id_index = state._get_id_index()
667
639
        if self._inventory is not None:
668
640
            update_inventory = True
669
641
            inv = self.inventory
 
642
            to_dir_ie = inv[to_dir_id]
670
643
            to_dir_id = to_entry[0][2]
671
 
            to_dir_ie = inv[to_dir_id]
672
644
        else:
673
645
            update_inventory = False
674
646
 
697
669
            new_entry = to_block[1][added_entry_index]
698
670
            rollbacks.append(lambda:state._make_absent(new_entry))
699
671
 
 
672
        # create rename entries and tuples
700
673
        for from_rel in from_paths:
701
674
            # from_rel is 'pathinroot/foo/bar'
702
675
            from_rel_utf8 = from_rel.encode('utf8')
705
678
            from_entry = self._get_entry(path=from_rel)
706
679
            if from_entry == (None, None):
707
680
                raise errors.BzrMoveFailedError(from_rel,to_dir,
708
 
                    errors.NotVersionedError(path=from_rel))
 
681
                    errors.NotVersionedError(path=str(from_rel)))
709
682
 
710
683
            from_id = from_entry[0][2]
711
684
            to_rel = pathjoin(to_dir, from_tail)
738
711
                if from_missing: # implicitly just update our path mapping
739
712
                    move_file = False
740
713
                elif not after:
741
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
714
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
715
                        extra="(Use --after to update the Bazaar id)")
742
716
 
743
717
            rollbacks = []
744
718
            def rollback_rename():
799
773
 
800
774
                if minikind == 'd':
801
775
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
802
 
                        """Recursively update all entries in this dirblock."""
803
 
                        if from_dir == '':
804
 
                            raise AssertionError("renaming root not supported")
 
776
                        """all entries in this block need updating.
 
777
 
 
778
                        TODO: This is pretty ugly, and doesn't support
 
779
                        reverting, but it works.
 
780
                        """
 
781
                        assert from_dir != '', "renaming root not supported"
805
782
                        from_key = (from_dir, '')
806
783
                        from_block_idx, present = \
807
784
                            state._find_block_index_from_key(from_key)
817
794
                        to_block_index = state._ensure_block(
818
795
                            to_block_index, to_entry_index, to_dir_utf8)
819
796
                        to_block = state._dirblocks[to_block_index]
820
 
 
821
 
                        # Grab a copy since move_one may update the list.
822
 
                        for entry in from_block[1][:]:
823
 
                            if not (entry[0][0] == from_dir):
824
 
                                raise AssertionError()
 
797
                        for entry in from_block[1]:
 
798
                            assert entry[0][0] == from_dir
825
799
                            cur_details = entry[1][0]
826
800
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
827
801
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
828
802
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
829
803
                            minikind = cur_details[0]
830
 
                            if minikind in 'ar':
831
 
                                # Deleted children of a renamed directory
832
 
                                # Do not need to be updated.
833
 
                                # Children that have been renamed out of this
834
 
                                # directory should also not be updated
835
 
                                continue
836
804
                            move_one(entry, from_path_utf8=from_path_utf8,
837
805
                                     minikind=minikind,
838
806
                                     executable=cur_details[3],
886
854
        for tree in trees:
887
855
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
888
856
                parents):
889
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
890
 
                    trees, require_versioned)
 
857
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
891
858
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
892
859
        # -- make all paths utf8 --
893
860
        paths_utf8 = set()
953
920
            if not all_versioned:
954
921
                raise errors.PathsNotVersionedError(paths)
955
922
        # -- remove redundancy in supplied paths to prevent over-scanning --
956
 
        search_paths = osutils.minimum_path_selection(paths)
957
 
        # sketch:
 
923
        search_paths = set()
 
924
        for path in paths:
 
925
            other_paths = paths.difference(set([path]))
 
926
            if not osutils.is_inside_any(other_paths, path):
 
927
                # this is a top level path, we must check it.
 
928
                search_paths.add(path)
 
929
        # sketch: 
958
930
        # for all search_indexs in each path at or under each element of
959
931
        # search_paths, if the detail is relocated: add the id, and add the
960
932
        # relocated path as one to search if its not searched already. If the
1016
988
 
1017
989
    def read_working_inventory(self):
1018
990
        """Read the working inventory.
1019
 
 
 
991
        
1020
992
        This is a meaningless operation for dirstate, but we obey it anyhow.
1021
993
        """
1022
994
        return self.inventory
1027
999
 
1028
1000
        WorkingTree4 supplies revision_trees for any basis tree.
1029
1001
        """
 
1002
        revision_id = osutils.safe_revision_id(revision_id)
1030
1003
        dirstate = self.current_dirstate()
1031
1004
        parent_ids = dirstate.get_parent_ids()
1032
1005
        if revision_id not in parent_ids:
1039
1012
    @needs_tree_write_lock
1040
1013
    def set_last_revision(self, new_revision):
1041
1014
        """Change the last revision in the working tree."""
 
1015
        new_revision = osutils.safe_revision_id(new_revision)
1042
1016
        parents = self.get_parent_ids()
1043
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1044
 
            if len(parents) >= 2:
1045
 
                raise AssertionError(
1046
 
                    "setting the last parent to none with a pending merge is "
1047
 
                    "unsupported.")
 
1017
        if new_revision in (NULL_REVISION, None):
 
1018
            assert len(parents) < 2, (
 
1019
                "setting the last parent to none with a pending merge is "
 
1020
                "unsupported.")
1048
1021
            self.set_parent_ids([])
1049
1022
        else:
1050
1023
            self.set_parent_ids([new_revision] + parents[1:],
1053
1026
    @needs_tree_write_lock
1054
1027
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1055
1028
        """Set the parent ids to revision_ids.
1056
 
 
 
1029
        
1057
1030
        See also set_parent_trees. This api will try to retrieve the tree data
1058
1031
        for each element of revision_ids from the trees repository. If you have
1059
1032
        tree data already available, it is more efficient to use
1063
1036
        :param revision_ids: The revision_ids to set as the parent ids of this
1064
1037
            working tree. Any of these may be ghosts.
1065
1038
        """
 
1039
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1066
1040
        trees = []
1067
1041
        for revision_id in revision_ids:
1068
1042
            try:
1074
1048
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1075
1049
                revtree = None
1076
1050
            trees.append((revision_id, revtree))
 
1051
        self.current_dirstate()._validate()
1077
1052
        self.set_parent_trees(trees,
1078
1053
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1054
        self.current_dirstate()._validate()
1079
1055
 
1080
1056
    @needs_tree_write_lock
1081
1057
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1086
1062
            parent tree - i.e. a ghost.
1087
1063
        """
1088
1064
        dirstate = self.current_dirstate()
 
1065
        dirstate._validate()
1089
1066
        if len(parents_list) > 0:
1090
1067
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1091
1068
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1092
1069
        real_trees = []
1093
1070
        ghosts = []
1094
 
 
1095
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1096
 
        graph = self.branch.repository.get_graph()
1097
 
        heads = graph.heads(parent_ids)
1098
 
        accepted_revisions = set()
1099
 
 
1100
1071
        # convert absent trees to the null tree, which we convert back to
1101
1072
        # missing on access.
1102
1073
        for rev_id, tree in parents_list:
1103
 
            if len(accepted_revisions) > 0:
1104
 
                # we always accept the first tree
1105
 
                if rev_id in accepted_revisions or rev_id not in heads:
1106
 
                    # We have already included either this tree, or its
1107
 
                    # descendent, so we skip it.
1108
 
                    continue
1109
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1074
            rev_id = osutils.safe_revision_id(rev_id)
1110
1075
            if tree is not None:
1111
1076
                real_trees.append((rev_id, tree))
1112
1077
            else:
1113
1078
                real_trees.append((rev_id,
1114
 
                    self.branch.repository.revision_tree(
1115
 
                        _mod_revision.NULL_REVISION)))
 
1079
                    self.branch.repository.revision_tree(None)))
1116
1080
                ghosts.append(rev_id)
1117
 
            accepted_revisions.add(rev_id)
 
1081
        dirstate._validate()
1118
1082
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1083
        dirstate._validate()
1119
1084
        self._make_dirty(reset_inventory=False)
 
1085
        dirstate._validate()
1120
1086
 
1121
1087
    def _set_root_id(self, file_id):
1122
1088
        """See WorkingTree.set_root_id."""
1125
1091
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1126
1092
            self._make_dirty(reset_inventory=True)
1127
1093
 
1128
 
    def _sha_from_stat(self, path, stat_result):
1129
 
        """Get a sha digest from the tree's stat cache.
1130
 
 
1131
 
        The default implementation assumes no stat cache is present.
1132
 
 
1133
 
        :param path: The path.
1134
 
        :param stat_result: The stat result being looked up.
1135
 
        """
1136
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1137
 
 
1138
1094
    @needs_read_lock
1139
1095
    def supports_tree_reference(self):
1140
1096
        return self._repo_supports_tree_reference
1141
1097
 
1142
1098
    def unlock(self):
1143
1099
        """Unlock in format 4 trees needs to write the entire dirstate."""
1144
 
        # do non-implementation specific cleanup
1145
 
        self._cleanup()
1146
 
 
1147
1100
        if self._control_files._lock_count == 1:
1148
1101
            # eventually we should do signature checking during read locks for
1149
1102
            # dirstate updates.
1180
1133
            return
1181
1134
        state = self.current_dirstate()
1182
1135
        state._read_dirblocks_if_needed()
1183
 
        ids_to_unversion = set(file_ids)
 
1136
        ids_to_unversion = set()
 
1137
        for file_id in file_ids:
 
1138
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1184
1139
        paths_to_unversion = set()
1185
1140
        # sketch:
1186
1141
        # check if the root is to be unversioned, if so, assert for now.
1213
1168
                # just forget the whole block.
1214
1169
                entry_index = 0
1215
1170
                while entry_index < len(block[1]):
 
1171
                    # Mark this file id as having been removed
1216
1172
                    entry = block[1][entry_index]
1217
 
                    if entry[1][0][0] in 'ar':
1218
 
                        # don't remove absent or renamed entries
 
1173
                    ids_to_unversion.discard(entry[0][2])
 
1174
                    if (entry[1][0][0] == 'a'
 
1175
                        or not state._make_absent(entry)):
1219
1176
                        entry_index += 1
1220
 
                    else:
1221
 
                        # Mark this file id as having been removed
1222
 
                        ids_to_unversion.discard(entry[0][2])
1223
 
                        if not state._make_absent(entry):
1224
 
                            # The block has not shrunk.
1225
 
                            entry_index += 1
1226
1177
                # go to the next block. (At the moment we dont delete empty
1227
1178
                # dirblocks)
1228
1179
                block_index += 1
1251
1202
            for file_id in file_ids:
1252
1203
                self._inventory.remove_recursive_id(file_id)
1253
1204
 
1254
 
    @needs_tree_write_lock
1255
 
    def rename_one(self, from_rel, to_rel, after=False):
1256
 
        """See WorkingTree.rename_one"""
1257
 
        self.flush()
1258
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1259
 
 
1260
 
    @needs_tree_write_lock
1261
 
    def apply_inventory_delta(self, changes):
1262
 
        """See MutableTree.apply_inventory_delta"""
1263
 
        state = self.current_dirstate()
1264
 
        state.update_by_delta(changes)
1265
 
        self._make_dirty(reset_inventory=True)
1266
 
 
1267
 
    def update_basis_by_delta(self, new_revid, delta):
1268
 
        """See MutableTree.update_basis_by_delta."""
1269
 
        if self.last_revision() == new_revid:
1270
 
            raise AssertionError()
1271
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1272
 
 
1273
1205
    @needs_read_lock
1274
1206
    def _validate(self):
1275
1207
        self._dirstate._validate()
1277
1209
    @needs_tree_write_lock
1278
1210
    def _write_inventory(self, inv):
1279
1211
        """Write inventory as the current inventory."""
1280
 
        if self._dirty:
1281
 
            raise AssertionError("attempting to write an inventory when the "
1282
 
                "dirstate is dirty will lose pending changes")
1283
 
        had_inventory = self._inventory is not None
1284
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1285
 
        # working inventory. We do this because self.inventory may be inv, or
1286
 
        # may have been modified, and either case would prevent a clean delta
1287
 
        # being created.
1288
 
        self._inventory = None
1289
 
        # generate a delta,
1290
 
        delta = inv._make_delta(self.inventory)
1291
 
        # and apply it.
1292
 
        self.apply_inventory_delta(delta)
1293
 
        if had_inventory:
 
1212
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1213
        self.current_dirstate().set_state_from_inventory(inv)
 
1214
        self._make_dirty(reset_inventory=False)
 
1215
        if self._inventory is not None:
1294
1216
            self._inventory = inv
1295
1217
        self.flush()
1296
1218
 
1297
1219
 
1298
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1299
 
 
1300
 
    def __init__(self, tree):
1301
 
        self.tree = tree
1302
 
 
1303
 
    def sha1(self, abspath):
1304
 
        """See dirstate.SHA1Provider.sha1()."""
1305
 
        filters = self.tree._content_filter_stack(
1306
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1307
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1308
 
 
1309
 
    def stat_and_sha1(self, abspath):
1310
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1311
 
        filters = self.tree._content_filter_stack(
1312
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1313
 
        file_obj = file(abspath, 'rb', 65000)
1314
 
        try:
1315
 
            statvalue = os.fstat(file_obj.fileno())
1316
 
            if filters:
1317
 
                file_obj = filtered_input_file(file_obj, filters)
1318
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1319
 
        finally:
1320
 
            file_obj.close()
1321
 
        return statvalue, sha1
1322
 
 
1323
 
 
1324
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1325
 
    """Dirstate working tree that supports content filtering.
1326
 
 
1327
 
    The dirstate holds the hash and size of the canonical form of the file, 
1328
 
    and most methods must return that.
1329
 
    """
1330
 
 
1331
 
    def _file_content_summary(self, path, stat_result):
1332
 
        # This is to support the somewhat obsolete path_content_summary method
1333
 
        # with content filtering: see
1334
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1335
 
        #
1336
 
        # If the dirstate cache is up to date and knows the hash and size,
1337
 
        # return that.
1338
 
        # Otherwise if there are no content filters, return the on-disk size
1339
 
        # and leave the hash blank.
1340
 
        # Otherwise, read and filter the on-disk file and use its size and
1341
 
        # hash.
1342
 
        #
1343
 
        # The dirstate doesn't store the size of the canonical form so we
1344
 
        # can't trust it for content-filtered trees.  We just return None.
1345
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1346
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1347
 
        return ('file', None, executable, dirstate_sha1)
1348
 
 
1349
 
 
1350
 
class WorkingTree4(DirStateWorkingTree):
1351
 
    """This is the Format 4 working tree.
1352
 
 
1353
 
    This differs from WorkingTree3 by:
1354
 
     - Having a consolidated internal dirstate, stored in a
1355
 
       randomly-accessible sorted file on disk.
1356
 
     - Not having a regular inventory attribute.  One can be synthesized
1357
 
       on demand but this is expensive and should be avoided.
1358
 
 
1359
 
    This is new in bzr 0.15.
1360
 
    """
1361
 
 
1362
 
 
1363
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1364
 
    """This is the Format 5 working tree.
1365
 
 
1366
 
    This differs from WorkingTree4 by:
1367
 
     - Supporting content filtering.
1368
 
 
1369
 
    This is new in bzr 1.11.
1370
 
    """
1371
 
 
1372
 
 
1373
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1374
 
    """This is the Format 6 working tree.
1375
 
 
1376
 
    This differs from WorkingTree5 by:
1377
 
     - Supporting a current view that may mask the set of files in a tree
1378
 
       impacted by most user operations.
1379
 
 
1380
 
    This is new in bzr 1.14.
1381
 
    """
1382
 
 
1383
 
    def _make_views(self):
1384
 
        return views.PathBasedViews(self)
1385
 
 
1386
 
 
1387
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1388
 
 
1389
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1390
 
                   accelerator_tree=None, hardlink=False):
 
1220
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1221
    """The first consolidated dirstate working tree format.
 
1222
 
 
1223
    This format:
 
1224
        - exists within a metadir controlling .bzr
 
1225
        - includes an explicit version marker for the workingtree control
 
1226
          files, separate from the BzrDir format
 
1227
        - modifies the hash cache format
 
1228
        - is new in bzr 0.15
 
1229
        - uses a LockDir to guard access to it.
 
1230
    """
 
1231
 
 
1232
    upgrade_recommended = False
 
1233
 
 
1234
    def get_format_string(self):
 
1235
        """See WorkingTreeFormat.get_format_string()."""
 
1236
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1237
 
 
1238
    def get_format_description(self):
 
1239
        """See WorkingTreeFormat.get_format_description()."""
 
1240
        return "Working tree format 4"
 
1241
 
 
1242
    def initialize(self, a_bzrdir, revision_id=None):
1391
1243
        """See WorkingTreeFormat.initialize().
1392
1244
 
1393
1245
        :param revision_id: allows creating a working tree at a different
1394
1246
        revision than the branch is at.
1395
 
        :param accelerator_tree: A tree which can be used for retrieving file
1396
 
            contents more quickly than the revision tree, i.e. a workingtree.
1397
 
            The revision tree will be used for cases where accelerator_tree's
1398
 
            content is different.
1399
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1400
 
            where possible.
1401
1247
 
1402
 
        These trees get an initial random root id, if their repository supports
1403
 
        rich root data, TREE_ROOT otherwise.
 
1248
        These trees get an initial random root id.
1404
1249
        """
 
1250
        revision_id = osutils.safe_revision_id(revision_id)
1405
1251
        if not isinstance(a_bzrdir.transport, LocalTransport):
1406
1252
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1407
1253
        transport = a_bzrdir.get_workingtree_transport(self)
1408
1254
        control_files = self._open_control_files(a_bzrdir)
1409
1255
        control_files.create_lock()
1410
1256
        control_files.lock_write()
1411
 
        transport.put_bytes('format', self.get_format_string(),
1412
 
            mode=a_bzrdir._get_file_mode())
1413
 
        if from_branch is not None:
1414
 
            branch = from_branch
1415
 
        else:
1416
 
            branch = a_bzrdir.open_branch()
 
1257
        control_files.put_utf8('format', self.get_format_string())
 
1258
        branch = a_bzrdir.open_branch()
1417
1259
        if revision_id is None:
1418
1260
            revision_id = branch.last_revision()
1419
1261
        local_path = transport.local_abspath('dirstate')
1421
1263
        state = dirstate.DirState.initialize(local_path)
1422
1264
        state.unlock()
1423
1265
        del state
1424
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1266
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1425
1267
                         branch,
1426
1268
                         _format=self,
1427
1269
                         _bzrdir=a_bzrdir,
1428
1270
                         _control_files=control_files)
1429
1271
        wt._new_tree()
1430
1272
        wt.lock_tree_write()
 
1273
        wt.current_dirstate()._validate()
1431
1274
        try:
1432
 
            self._init_custom_control_files(wt)
1433
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
1434
 
                if branch.repository.supports_rich_root():
1435
 
                    wt._set_root_id(generate_ids.gen_root_id())
1436
 
                else:
1437
 
                    wt._set_root_id(ROOT_ID)
 
1275
            if revision_id in (None, NULL_REVISION):
 
1276
                wt._set_root_id(generate_ids.gen_root_id())
1438
1277
                wt.flush()
1439
 
            basis = None
1440
 
            # frequently, we will get here due to branching.  The accelerator
1441
 
            # tree will be the tree from the branch, so the desired basis
1442
 
            # tree will often be a parent of the accelerator tree.
1443
 
            if accelerator_tree is not None:
1444
 
                try:
1445
 
                    basis = accelerator_tree.revision_tree(revision_id)
1446
 
                except errors.NoSuchRevision:
1447
 
                    pass
1448
 
            if basis is None:
1449
 
                basis = branch.repository.revision_tree(revision_id)
1450
 
            if revision_id == _mod_revision.NULL_REVISION:
1451
 
                parents_list = []
1452
 
            else:
1453
 
                parents_list = [(revision_id, basis)]
 
1278
                wt.current_dirstate()._validate()
 
1279
            wt.set_last_revision(revision_id)
 
1280
            wt.flush()
 
1281
            basis = wt.basis_tree()
1454
1282
            basis.lock_read()
1455
 
            try:
1456
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1283
            # if the basis has a root id we have to use that; otherwise we use
 
1284
            # a new random one
 
1285
            basis_root_id = basis.get_root_id()
 
1286
            if basis_root_id is not None:
 
1287
                wt._set_root_id(basis_root_id)
1457
1288
                wt.flush()
1458
 
                # if the basis has a root id we have to use that; otherwise we
1459
 
                # use a new random one
1460
 
                basis_root_id = basis.get_root_id()
1461
 
                if basis_root_id is not None:
1462
 
                    wt._set_root_id(basis_root_id)
1463
 
                    wt.flush()
1464
 
                if wt.supports_content_filtering():
1465
 
                    # The original tree may not have the same content filters
1466
 
                    # applied so we can't safely build the inventory delta from
1467
 
                    # the source tree.
1468
 
                    delta_from_tree = False
1469
 
                else:
1470
 
                    delta_from_tree = True
1471
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1472
 
                # because wt4.apply_inventory_delta does not mutate the input
1473
 
                # inventory entries.
1474
 
                transform.build_tree(basis, wt, accelerator_tree,
1475
 
                                     hardlink=hardlink,
1476
 
                                     delta_from_tree=delta_from_tree)
1477
 
            finally:
1478
 
                basis.unlock()
 
1289
            transform.build_tree(basis, wt)
 
1290
            basis.unlock()
1479
1291
        finally:
1480
1292
            control_files.unlock()
1481
1293
            wt.unlock()
1482
1294
        return wt
1483
1295
 
1484
 
    def _init_custom_control_files(self, wt):
1485
 
        """Subclasses with custom control files should override this method.
1486
 
 
1487
 
        The working tree and control files are locked for writing when this
1488
 
        method is called.
1489
 
 
1490
 
        :param wt: the WorkingTree object
1491
 
        """
1492
 
 
1493
1296
    def _open(self, a_bzrdir, control_files):
1494
1297
        """Open the tree itself.
1495
1298
 
1496
1299
        :param a_bzrdir: the dir for the tree.
1497
1300
        :param control_files: the control files for the tree.
1498
1301
        """
1499
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1302
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1500
1303
                           branch=a_bzrdir.open_branch(),
1501
1304
                           _format=self,
1502
1305
                           _bzrdir=a_bzrdir,
1503
1306
                           _control_files=control_files)
1504
1307
 
1505
1308
    def __get_matchingbzrdir(self):
1506
 
        return self._get_matchingbzrdir()
1507
 
 
1508
 
    def _get_matchingbzrdir(self):
1509
 
        """Overrideable method to get a bzrdir for testing."""
1510
1309
        # please test against something that will let us do tree references
1511
1310
        return bzrdir.format_registry.make_bzrdir(
1512
1311
            'dirstate-with-subtree')
1514
1313
    _matchingbzrdir = property(__get_matchingbzrdir)
1515
1314
 
1516
1315
 
1517
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1518
 
    """The first consolidated dirstate working tree format.
1519
 
 
1520
 
    This format:
1521
 
        - exists within a metadir controlling .bzr
1522
 
        - includes an explicit version marker for the workingtree control
1523
 
          files, separate from the BzrDir format
1524
 
        - modifies the hash cache format
1525
 
        - is new in bzr 0.15
1526
 
        - uses a LockDir to guard access to it.
1527
 
    """
1528
 
 
1529
 
    upgrade_recommended = False
1530
 
 
1531
 
    _tree_class = WorkingTree4
1532
 
 
1533
 
    def get_format_string(self):
1534
 
        """See WorkingTreeFormat.get_format_string()."""
1535
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1536
 
 
1537
 
    def get_format_description(self):
1538
 
        """See WorkingTreeFormat.get_format_description()."""
1539
 
        return "Working tree format 4"
1540
 
 
1541
 
 
1542
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1543
 
    """WorkingTree format supporting content filtering.
1544
 
    """
1545
 
 
1546
 
    upgrade_recommended = False
1547
 
 
1548
 
    _tree_class = WorkingTree5
1549
 
 
1550
 
    def get_format_string(self):
1551
 
        """See WorkingTreeFormat.get_format_string()."""
1552
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1553
 
 
1554
 
    def get_format_description(self):
1555
 
        """See WorkingTreeFormat.get_format_description()."""
1556
 
        return "Working tree format 5"
1557
 
 
1558
 
    def supports_content_filtering(self):
1559
 
        return True
1560
 
 
1561
 
 
1562
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1563
 
    """WorkingTree format supporting views.
1564
 
    """
1565
 
 
1566
 
    upgrade_recommended = False
1567
 
 
1568
 
    _tree_class = WorkingTree6
1569
 
 
1570
 
    def get_format_string(self):
1571
 
        """See WorkingTreeFormat.get_format_string()."""
1572
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1573
 
 
1574
 
    def get_format_description(self):
1575
 
        """See WorkingTreeFormat.get_format_description()."""
1576
 
        return "Working tree format 6"
1577
 
 
1578
 
    def _init_custom_control_files(self, wt):
1579
 
        """Subclasses with custom control files should override this method."""
1580
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1581
 
 
1582
 
    def supports_content_filtering(self):
1583
 
        return True
1584
 
 
1585
 
    def supports_views(self):
1586
 
        return True
1587
 
 
1588
 
 
1589
1316
class DirStateRevisionTree(Tree):
1590
 
    """A revision tree pulling the inventory from a dirstate.
1591
 
    
1592
 
    Note that this is one of the historical (ie revision) trees cached in the
1593
 
    dirstate for easy access, not the workingtree.
1594
 
    """
 
1317
    """A revision tree pulling the inventory from a dirstate."""
1595
1318
 
1596
1319
    def __init__(self, dirstate, revision_id, repository):
1597
1320
        self._dirstate = dirstate
1598
 
        self._revision_id = revision_id
 
1321
        self._revision_id = osutils.safe_revision_id(revision_id)
1599
1322
        self._repository = repository
1600
1323
        self._inventory = None
1601
1324
        self._locked = 0
1602
1325
        self._dirstate_locked = False
1603
 
        self._repo_supports_tree_reference = getattr(
1604
 
            repository._format, "supports_tree_reference",
1605
 
            False)
1606
1326
 
1607
1327
    def __repr__(self):
1608
1328
        return "<%s of %s in %s>" % \
1609
1329
            (self.__class__.__name__, self._revision_id, self._dirstate)
1610
1330
 
1611
 
    def annotate_iter(self, file_id,
1612
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1331
    def annotate_iter(self, file_id):
1613
1332
        """See Tree.annotate_iter"""
1614
 
        text_key = (file_id, self.inventory[file_id].revision)
1615
 
        annotations = self._repository.texts.annotate(text_key)
1616
 
        return [(key[-1], line) for (key, line) in annotations]
 
1333
        w = self._repository.weave_store.get_weave(file_id,
 
1334
                           self._repository.get_transaction())
 
1335
        return w.annotate_iter(self.inventory[file_id].revision)
1617
1336
 
1618
 
    def _get_ancestors(self, default_revision):
1619
 
        return set(self._repository.get_ancestry(self._revision_id,
1620
 
                                                 topo_sorted=False))
1621
1337
    def _comparison_data(self, entry, path):
1622
1338
        """See Tree._comparison_data."""
1623
1339
        if entry is None:
1640
1356
    def get_root_id(self):
1641
1357
        return self.path2id('')
1642
1358
 
1643
 
    def id2path(self, file_id):
1644
 
        "Convert a file-id to a path."
1645
 
        entry = self._get_entry(file_id=file_id)
1646
 
        if entry == (None, None):
1647
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1648
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1649
 
        return path_utf8.decode('utf8')
1650
 
 
1651
 
    def iter_references(self):
1652
 
        if not self._repo_supports_tree_reference:
1653
 
            # When the repo doesn't support references, we will have nothing to
1654
 
            # return
1655
 
            return iter([])
1656
 
        # Otherwise, fall back to the default implementation
1657
 
        return super(DirStateRevisionTree, self).iter_references()
1658
 
 
1659
1359
    def _get_parent_index(self):
1660
1360
        """Return the index in the dirstate referenced by this tree."""
1661
1361
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1666
1366
        If either file_id or path is supplied, it is used as the key to lookup.
1667
1367
        If both are supplied, the fastest lookup is used, and an error is
1668
1368
        raised if they do not both point at the same row.
1669
 
 
 
1369
        
1670
1370
        :param file_id: An optional unicode file_id to be looked up.
1671
1371
        :param path: An optional unicode path to be looked up.
1672
1372
        :return: The dirstate row tuple for path/file_id, or (None, None)
1673
1373
        """
1674
1374
        if file_id is None and path is None:
1675
1375
            raise errors.BzrError('must supply file_id or path')
 
1376
        file_id = osutils.safe_file_id(file_id)
1676
1377
        if path is not None:
1677
1378
            path = path.encode('utf8')
1678
1379
        parent_index = self._get_parent_index()
1686
1387
 
1687
1388
        This is relatively expensive: we have to walk the entire dirstate.
1688
1389
        """
1689
 
        if not self._locked:
1690
 
            raise AssertionError(
1691
 
                'cannot generate inventory of an unlocked '
1692
 
                'dirstate revision tree')
 
1390
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1391
            'dirstate revision tree'
1693
1392
        # separate call for profiling - makes it clear where the costs are.
1694
1393
        self._dirstate._read_dirblocks_if_needed()
1695
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1696
 
            raise AssertionError(
1697
 
                'parent %s has disappeared from %s' % (
1698
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1394
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1395
            'parent %s has disappeared from %s' % (
 
1396
            self._revision_id, self._dirstate.get_parent_ids())
1699
1397
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1700
1398
        # This is identical now to the WorkingTree _generate_inventory except
1701
1399
        # for the tree index use.
1702
1400
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1703
1401
        current_id = root_key[2]
1704
 
        if current_entry[parent_index][0] != 'd':
1705
 
            raise AssertionError()
 
1402
        assert current_entry[parent_index][0] == 'd'
1706
1403
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1707
1404
        inv.root.revision = current_entry[parent_index][4]
1708
1405
        # Turn some things into local variables
1748
1445
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1749
1446
                            % entry)
1750
1447
                # These checks cost us around 40ms on a 55k entry tree
1751
 
                if file_id in inv_byid:
1752
 
                    raise AssertionError('file_id %s already in'
1753
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1754
 
                if name_unicode in parent_ie.children:
1755
 
                    raise AssertionError('name %r already in parent'
1756
 
                        % (name_unicode,))
 
1448
                assert file_id not in inv_byid
 
1449
                assert name_unicode not in parent_ie.children
1757
1450
                inv_byid[file_id] = inv_entry
1758
1451
                parent_ie.children[name_unicode] = inv_entry
1759
1452
        self._inventory = inv
1769
1462
            return None
1770
1463
        parent_index = self._get_parent_index()
1771
1464
        last_changed_revision = entry[1][parent_index][4]
1772
 
        try:
1773
 
            rev = self._repository.get_revision(last_changed_revision)
1774
 
        except errors.NoSuchRevision:
1775
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1776
 
        return rev.timestamp
 
1465
        return self._repository.get_revision(last_changed_revision).timestamp
1777
1466
 
1778
1467
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1779
1468
        entry = self._get_entry(file_id=file_id, path=path)
1783
1472
            return parent_details[1]
1784
1473
        return None
1785
1474
 
1786
 
    def get_file(self, file_id, path=None):
 
1475
    def get_file(self, file_id):
1787
1476
        return StringIO(self.get_file_text(file_id))
1788
1477
 
 
1478
    def get_file_lines(self, file_id):
 
1479
        ie = self.inventory[file_id]
 
1480
        return self._repository.weave_store.get_weave(file_id,
 
1481
                self._repository.get_transaction()).get_lines(ie.revision)
 
1482
 
1789
1483
    def get_file_size(self, file_id):
1790
 
        """See Tree.get_file_size"""
1791
1484
        return self.inventory[file_id].text_size
1792
1485
 
1793
 
    def get_file_text(self, file_id, path=None):
1794
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1795
 
        return ''.join(content)
 
1486
    def get_file_text(self, file_id):
 
1487
        return ''.join(self.get_file_lines(file_id))
1796
1488
 
1797
1489
    def get_reference_revision(self, file_id, path=None):
1798
1490
        return self.inventory[file_id].reference_revision
1799
1491
 
1800
 
    def iter_files_bytes(self, desired_files):
1801
 
        """See Tree.iter_files_bytes.
1802
 
 
1803
 
        This version is implemented on top of Repository.iter_files_bytes"""
1804
 
        parent_index = self._get_parent_index()
1805
 
        repo_desired_files = []
1806
 
        for file_id, identifier in desired_files:
1807
 
            entry = self._get_entry(file_id)
1808
 
            if entry == (None, None):
1809
 
                raise errors.NoSuchId(self, file_id)
1810
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1811
 
                                       identifier))
1812
 
        return self._repository.iter_files_bytes(repo_desired_files)
1813
 
 
1814
1492
    def get_symlink_target(self, file_id):
1815
1493
        entry = self._get_entry(file_id=file_id)
1816
1494
        parent_index = self._get_parent_index()
1817
1495
        if entry[1][parent_index][0] != 'l':
1818
1496
            return None
1819
1497
        else:
1820
 
            target = entry[1][parent_index][1]
1821
 
            target = target.decode('utf8')
1822
 
            return target
 
1498
            # At present, none of the tree implementations supports non-ascii
 
1499
            # symlink targets. So we will just assume that the dirstate path is
 
1500
            # correct.
 
1501
            return entry[1][parent_index][1]
1823
1502
 
1824
1503
    def get_revision_id(self):
1825
1504
        """Return the revision id for this tree."""
1843
1522
        return bool(self.path2id(filename))
1844
1523
 
1845
1524
    def kind(self, file_id):
1846
 
        entry = self._get_entry(file_id=file_id)[1]
1847
 
        if entry is None:
1848
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1849
 
        parent_index = self._get_parent_index()
1850
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1851
 
 
1852
 
    def stored_kind(self, file_id):
1853
 
        """See Tree.stored_kind"""
1854
 
        return self.kind(file_id)
1855
 
 
1856
 
    def path_content_summary(self, path):
1857
 
        """See Tree.path_content_summary."""
1858
 
        id = self.inventory.path2id(path)
1859
 
        if id is None:
1860
 
            return ('missing', None, None, None)
1861
 
        entry = self._inventory[id]
1862
 
        kind = entry.kind
1863
 
        if kind == 'file':
1864
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1865
 
        elif kind == 'symlink':
1866
 
            return (kind, None, None, entry.symlink_target)
1867
 
        else:
1868
 
            return (kind, None, None, None)
 
1525
        return self.inventory[file_id].kind
1869
1526
 
1870
1527
    def is_executable(self, file_id, path=None):
1871
1528
        ie = self.inventory[file_id]
1873
1530
            return None
1874
1531
        return ie.executable
1875
1532
 
1876
 
    def is_locked(self):
1877
 
        return self._locked
1878
 
 
1879
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1533
    def list_files(self, include_root=False):
1880
1534
        # We use a standard implementation, because DirStateRevisionTree is
1881
1535
        # dealing with one of the parents of the current state
1882
1536
        inv = self._get_inventory()
1883
 
        if from_dir is None:
1884
 
            from_dir_id = None
1885
 
        else:
1886
 
            from_dir_id = inv.path2id(from_dir)
1887
 
            if from_dir_id is None:
1888
 
                # Directory not versioned
1889
 
                return
1890
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1891
 
        if inv.root is not None and not include_root and from_dir is None:
 
1537
        entries = inv.iter_entries()
 
1538
        if self.inventory.root is not None and not include_root:
1892
1539
            entries.next()
1893
1540
        for path, entry in entries:
1894
1541
            yield path, 'V', entry.kind, entry.file_id, entry
1895
1542
 
1896
1543
    def lock_read(self):
1897
 
        """Lock the tree for a set of operations.
1898
 
 
1899
 
        :return: An object with an unlock method which will release the lock
1900
 
            obtained.
1901
 
        """
 
1544
        """Lock the tree for a set of operations."""
1902
1545
        if not self._locked:
1903
1546
            self._repository.lock_read()
1904
1547
            if self._dirstate._lock_token is None:
1905
1548
                self._dirstate.lock_read()
1906
1549
                self._dirstate_locked = True
1907
1550
        self._locked += 1
1908
 
        return self
1909
1551
 
1910
1552
    def _must_be_locked(self):
1911
1553
        if not self._locked:
1932
1574
                self._dirstate_locked = False
1933
1575
            self._repository.unlock()
1934
1576
 
1935
 
    @needs_read_lock
1936
 
    def supports_tree_reference(self):
1937
 
        return self._repo_supports_tree_reference
1938
 
 
1939
1577
    def walkdirs(self, prefix=""):
1940
1578
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1941
 
        # implementation based on an inventory.
 
1579
        # implementation based on an inventory.  
1942
1580
        # This should be cleaned up to use the much faster Dirstate code
1943
1581
        # So for now, we just build up the parent inventory, and extract
1944
1582
        # it the same way RevisionTree does.
1973
1611
 
1974
1612
class InterDirStateTree(InterTree):
1975
1613
    """Fast path optimiser for changes_from with dirstate trees.
1976
 
 
1977
 
    This is used only when both trees are in the dirstate working file, and
1978
 
    the source is any parent within the dirstate, and the destination is
 
1614
    
 
1615
    This is used only when both trees are in the dirstate working file, and 
 
1616
    the source is any parent within the dirstate, and the destination is 
1979
1617
    the current working tree of the same dirstate.
1980
1618
    """
1981
1619
    # this could be generalized to allow comparisons between any trees in the
1994
1632
        target.set_parent_ids([revid])
1995
1633
        return target.basis_tree(), target
1996
1634
 
1997
 
    @classmethod
1998
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1999
 
        result = klass.make_source_parent_tree(source, target)
2000
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2001
 
        return result
2002
 
 
2003
 
    @classmethod
2004
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2005
 
                                                  target):
2006
 
        from bzrlib.tests.test__dirstate_helpers import \
2007
 
            compiled_dirstate_helpers_feature
2008
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2009
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2010
 
        result = klass.make_source_parent_tree(source, target)
2011
 
        result[1]._iter_changes = ProcessEntryC
2012
 
        return result
2013
 
 
2014
1635
    _matching_from_tree_format = WorkingTreeFormat4()
2015
1636
    _matching_to_tree_format = WorkingTreeFormat4()
2016
 
 
2017
 
    @classmethod
2018
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2019
 
        # This method shouldn't be called, because we have python and C
2020
 
        # specific flavours.
2021
 
        raise NotImplementedError
2022
 
 
2023
 
    def iter_changes(self, include_unchanged=False,
 
1637
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1638
 
 
1639
    def _iter_changes(self, include_unchanged=False,
2024
1640
                      specific_files=None, pb=None, extra_trees=[],
2025
1641
                      require_versioned=True, want_unversioned=False):
2026
1642
        """Return the changes from source to target.
2027
1643
 
2028
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1644
        :return: An iterator that yields tuples. See InterTree._iter_changes
2029
1645
            for details.
2030
1646
        :param specific_files: An optional list of file paths to restrict the
2031
1647
            comparison to. When mapping filenames to ids, all matches in all
2042
1658
            output. An unversioned file is defined as one with (False, False)
2043
1659
            for the versioned pair.
2044
1660
        """
 
1661
        utf8_decode_or_none = cache_utf8._utf8_decode_with_None
 
1662
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1663
        # NB: show_status depends on being able to pass in non-versioned files
 
1664
        # and report them as unknown
2045
1665
        # TODO: handle extra trees in the dirstate.
2046
 
        if (extra_trees or specific_files == []):
 
1666
        # TODO: handle comparisons as an empty tree as a different special
 
1667
        # case? mbp 20070226
 
1668
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2047
1669
            # we can't fast-path these cases (yet)
2048
 
            return super(InterDirStateTree, self).iter_changes(
 
1670
            for f in super(InterDirStateTree, self)._iter_changes(
2049
1671
                include_unchanged, specific_files, pb, extra_trees,
2050
 
                require_versioned, want_unversioned=want_unversioned)
 
1672
                require_versioned, want_unversioned=want_unversioned):
 
1673
                yield f
 
1674
            return
2051
1675
        parent_ids = self.target.get_parent_ids()
2052
 
        if not (self.source._revision_id in parent_ids
2053
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2054
 
            raise AssertionError(
2055
 
                "revision {%s} is not stored in {%s}, but %s "
2056
 
                "can only be used for trees stored in the dirstate"
2057
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1676
        assert (self.source._revision_id in parent_ids), \
 
1677
                "revision {%s} is not stored in {%s}, but %s " \
 
1678
                "can only be used for trees stored in the dirstate" \
 
1679
                % (self.source._revision_id, self.target, self._iter_changes)
2058
1680
        target_index = 0
2059
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1681
        if self.source._revision_id == NULL_REVISION:
2060
1682
            source_index = None
2061
1683
            indices = (target_index,)
2062
1684
        else:
2063
 
            if not (self.source._revision_id in parent_ids):
2064
 
                raise AssertionError(
2065
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2066
 
                    self.source._revision_id, parent_ids))
 
1685
            assert (self.source._revision_id in parent_ids), \
 
1686
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1687
                self.source._revision_id, parent_ids)
2067
1688
            source_index = 1 + parent_ids.index(self.source._revision_id)
2068
 
            indices = (source_index, target_index)
 
1689
            indices = (source_index,target_index)
2069
1690
        # -- make all specific_files utf8 --
2070
1691
        if specific_files:
2071
1692
            specific_files_utf8 = set()
2072
1693
            for path in specific_files:
2073
 
                # Note, if there are many specific files, using cache_utf8
2074
 
                # would be good here.
2075
1694
                specific_files_utf8.add(path.encode('utf8'))
2076
1695
            specific_files = specific_files_utf8
2077
1696
        else:
2078
1697
            specific_files = set([''])
2079
1698
        # -- specific_files is now a utf8 path set --
2080
 
 
2081
1699
        # -- get the state object and prepare it.
2082
1700
        state = self.target.current_dirstate()
2083
1701
        state._read_dirblocks_if_needed()
 
1702
        def _entries_for_path(path):
 
1703
            """Return a list with all the entries that match path for all ids.
 
1704
            """
 
1705
            dirname, basename = os.path.split(path)
 
1706
            key = (dirname, basename, '')
 
1707
            block_index, present = state._find_block_index_from_key(key)
 
1708
            if not present:
 
1709
                # the block which should contain path is absent.
 
1710
                return []
 
1711
            result = []
 
1712
            block = state._dirblocks[block_index][1]
 
1713
            entry_index, _ = state._find_entry_index(key, block)
 
1714
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1715
            while (entry_index < len(block) and
 
1716
                block[entry_index][0][0:2] == key[0:2]):
 
1717
                result.append(block[entry_index])
 
1718
                entry_index += 1
 
1719
            return result
2084
1720
        if require_versioned:
2085
1721
            # -- check all supplied paths are versioned in a search tree. --
2086
 
            not_versioned = []
 
1722
            all_versioned = True
2087
1723
            for path in specific_files:
2088
 
                path_entries = state._entries_for_path(path)
 
1724
                path_entries = _entries_for_path(path)
2089
1725
                if not path_entries:
2090
1726
                    # this specified path is not present at all: error
2091
 
                    not_versioned.append(path)
2092
 
                    continue
 
1727
                    all_versioned = False
 
1728
                    break
2093
1729
                found_versioned = False
2094
1730
                # for each id at this path
2095
1731
                for entry in path_entries:
2102
1738
                if not found_versioned:
2103
1739
                    # none of the indexes was not 'absent' at all ids for this
2104
1740
                    # path.
2105
 
                    not_versioned.append(path)
2106
 
            if len(not_versioned) > 0:
2107
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1741
                    all_versioned = False
 
1742
                    break
 
1743
            if not all_versioned:
 
1744
                raise errors.PathsNotVersionedError(specific_files)
2108
1745
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2109
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1746
        search_specific_files = set()
 
1747
        for path in specific_files:
 
1748
            other_specific_files = specific_files.difference(set([path]))
 
1749
            if not osutils.is_inside_any(other_specific_files, path):
 
1750
                # this is a top level path, we must check it.
 
1751
                search_specific_files.add(path)
 
1752
        # sketch: 
 
1753
        # compare source_index and target_index at or under each element of search_specific_files.
 
1754
        # follow the following comparison table. Note that we only want to do diff operations when
 
1755
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1756
        # for the target.
 
1757
        # cases:
 
1758
        # 
 
1759
        # Source | Target | disk | action
 
1760
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1761
        #        |        |      | diff check on source-target
 
1762
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1763
        #        |        |      | ???
 
1764
        #   r    |  a     |      | add source to search
 
1765
        #   r    |  a     |  a   | 
 
1766
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1767
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1768
        #   a    | fdlt   |      | add new id
 
1769
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1770
        #   a    |  a     |      | not present in either tree, skip
 
1771
        #   a    |  a     |  a   | not present in any tree, skip
 
1772
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1773
        #        |        |      | may not be selected by the users list of paths.
 
1774
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1775
        #        |        |      | may not be selected by the users list of paths.
 
1776
        #  fdlt  | fdlt   |      | content in both: diff them
 
1777
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1778
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1779
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1780
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1781
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1782
        #        |        |      | this id at the other path.
 
1783
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1784
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1785
        #        |        |      | this id at the other path.
 
1786
 
 
1787
        # for all search_indexs in each path at or under each element of
 
1788
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1789
        # relocated path as one to search if its not searched already. If the
 
1790
        # detail is not relocated, add the id.
 
1791
        searched_specific_files = set()
 
1792
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1793
        # Using a list so that we can access the values and change them in
 
1794
        # nested scope. Each one is [path, file_id, entry]
 
1795
        last_source_parent = [None, None, None]
 
1796
        last_target_parent = [None, None, None]
2110
1797
 
2111
1798
        use_filesystem_for_exec = (sys.platform != 'win32')
2112
 
        iter_changes = self.target._iter_changes(include_unchanged,
2113
 
            use_filesystem_for_exec, search_specific_files, state,
2114
 
            source_index, target_index, want_unversioned, self.target)
2115
 
        return iter_changes.iter_changes()
 
1799
 
 
1800
        def _process_entry(entry, path_info):
 
1801
            """Compare an entry and real disk to generate delta information.
 
1802
 
 
1803
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1804
                the path of entry. If None, then the path is considered absent.
 
1805
                (Perhaps we should pass in a concrete entry for this ?)
 
1806
                Basename is returned as a utf8 string because we expect this
 
1807
                tuple will be ignored, and don't want to take the time to
 
1808
                decode.
 
1809
            """
 
1810
            if source_index is None:
 
1811
                source_details = NULL_PARENT_DETAILS
 
1812
            else:
 
1813
                source_details = entry[1][source_index]
 
1814
            target_details = entry[1][target_index]
 
1815
            target_minikind = target_details[0]
 
1816
            if path_info is not None and target_minikind in 'fdlt':
 
1817
                assert target_index == 0
 
1818
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1819
                                                  stat_value=path_info[3])
 
1820
                # The entry may have been modified by update_entry
 
1821
                target_details = entry[1][target_index]
 
1822
                target_minikind = target_details[0]
 
1823
            else:
 
1824
                link_or_sha1 = None
 
1825
            source_minikind = source_details[0]
 
1826
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1827
                # claimed content in both: diff
 
1828
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1829
                #        |        |      | diff check on source-target
 
1830
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1831
                #        |        |      | ???
 
1832
                if source_minikind in 'r':
 
1833
                    # add the source to the search path to find any children it
 
1834
                    # has.  TODO ? : only add if it is a container ?
 
1835
                    if not osutils.is_inside_any(searched_specific_files,
 
1836
                                                 source_details[1]):
 
1837
                        search_specific_files.add(source_details[1])
 
1838
                    # generate the old path; this is needed for stating later
 
1839
                    # as well.
 
1840
                    old_path = source_details[1]
 
1841
                    old_dirname, old_basename = os.path.split(old_path)
 
1842
                    path = pathjoin(entry[0][0], entry[0][1])
 
1843
                    old_entry = state._get_entry(source_index,
 
1844
                                                 path_utf8=old_path)
 
1845
                    # update the source details variable to be the real
 
1846
                    # location.
 
1847
                    source_details = old_entry[1][source_index]
 
1848
                    source_minikind = source_details[0]
 
1849
                else:
 
1850
                    old_dirname = entry[0][0]
 
1851
                    old_basename = entry[0][1]
 
1852
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1853
                if path_info is None:
 
1854
                    # the file is missing on disk, show as removed.
 
1855
                    content_change = True
 
1856
                    target_kind = None
 
1857
                    target_exec = False
 
1858
                else:
 
1859
                    # source and target are both versioned and disk file is present.
 
1860
                    target_kind = path_info[2]
 
1861
                    if target_kind == 'directory':
 
1862
                        if source_minikind != 'd':
 
1863
                            content_change = True
 
1864
                        else:
 
1865
                            # directories have no fingerprint
 
1866
                            content_change = False
 
1867
                        target_exec = False
 
1868
                    elif target_kind == 'file':
 
1869
                        if source_minikind != 'f':
 
1870
                            content_change = True
 
1871
                        else:
 
1872
                            # We could check the size, but we already have the
 
1873
                            # sha1 hash.
 
1874
                            content_change = (link_or_sha1 != source_details[1])
 
1875
                        # Target details is updated at update_entry time
 
1876
                        if use_filesystem_for_exec:
 
1877
                            # We don't need S_ISREG here, because we are sure
 
1878
                            # we are dealing with a file.
 
1879
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1880
                        else:
 
1881
                            target_exec = target_details[3]
 
1882
                    elif target_kind == 'symlink':
 
1883
                        if source_minikind != 'l':
 
1884
                            content_change = True
 
1885
                        else:
 
1886
                            content_change = (link_or_sha1 != source_details[1])
 
1887
                        target_exec = False
 
1888
                    elif target_kind == 'tree-reference':
 
1889
                        if source_minikind != 't':
 
1890
                            content_change = True
 
1891
                        else:
 
1892
                            content_change = False
 
1893
                        target_exec = False
 
1894
                    else:
 
1895
                        raise Exception, "unknown kind %s" % path_info[2]
 
1896
                # parent id is the entry for the path in the target tree
 
1897
                if old_dirname == last_source_parent[0]:
 
1898
                    source_parent_id = last_source_parent[1]
 
1899
                else:
 
1900
                    source_parent_entry = state._get_entry(source_index,
 
1901
                                                           path_utf8=old_dirname)
 
1902
                    source_parent_id = source_parent_entry[0][2]
 
1903
                    if source_parent_id == entry[0][2]:
 
1904
                        # This is the root, so the parent is None
 
1905
                        source_parent_id = None
 
1906
                    else:
 
1907
                        last_source_parent[0] = old_dirname
 
1908
                        last_source_parent[1] = source_parent_id
 
1909
                        last_source_parent[2] = source_parent_entry
 
1910
                new_dirname = entry[0][0]
 
1911
                if new_dirname == last_target_parent[0]:
 
1912
                    target_parent_id = last_target_parent[1]
 
1913
                else:
 
1914
                    # TODO: We don't always need to do the lookup, because the
 
1915
                    #       parent entry will be the same as the source entry.
 
1916
                    target_parent_entry = state._get_entry(target_index,
 
1917
                                                           path_utf8=new_dirname)
 
1918
                    target_parent_id = target_parent_entry[0][2]
 
1919
                    if target_parent_id == entry[0][2]:
 
1920
                        # This is the root, so the parent is None
 
1921
                        target_parent_id = None
 
1922
                    else:
 
1923
                        last_target_parent[0] = new_dirname
 
1924
                        last_target_parent[1] = target_parent_id
 
1925
                        last_target_parent[2] = target_parent_entry
 
1926
 
 
1927
                source_exec = source_details[3]
 
1928
                return ((entry[0][2], (old_path, path), content_change,
 
1929
                        (True, True),
 
1930
                        (source_parent_id, target_parent_id),
 
1931
                        (old_basename, entry[0][1]),
 
1932
                        (_minikind_to_kind[source_minikind], target_kind),
 
1933
                        (source_exec, target_exec)),)
 
1934
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1935
                # looks like a new file
 
1936
                if path_info is not None:
 
1937
                    path = pathjoin(entry[0][0], entry[0][1])
 
1938
                    # parent id is the entry for the path in the target tree
 
1939
                    # TODO: these are the same for an entire directory: cache em.
 
1940
                    parent_id = state._get_entry(target_index,
 
1941
                                                 path_utf8=entry[0][0])[0][2]
 
1942
                    if parent_id == entry[0][2]:
 
1943
                        parent_id = None
 
1944
                    if use_filesystem_for_exec:
 
1945
                        # We need S_ISREG here, because we aren't sure if this
 
1946
                        # is a file or not.
 
1947
                        target_exec = bool(
 
1948
                            stat.S_ISREG(path_info[3].st_mode)
 
1949
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1950
                    else:
 
1951
                        target_exec = target_details[3]
 
1952
                    return ((entry[0][2], (None, path), True,
 
1953
                            (False, True),
 
1954
                            (None, parent_id),
 
1955
                            (None, entry[0][1]),
 
1956
                            (None, path_info[2]),
 
1957
                            (None, target_exec)),)
 
1958
                else:
 
1959
                    # but its not on disk: we deliberately treat this as just
 
1960
                    # never-present. (Why ?! - RBC 20070224)
 
1961
                    pass
 
1962
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1963
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1964
                # if its still on disk, *and* theres no other entry at this
 
1965
                # path [we dont know this in this routine at the moment -
 
1966
                # perhaps we should change this - then it would be an unknown.
 
1967
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1968
                # parent id is the entry for the path in the target tree
 
1969
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1970
                if parent_id == entry[0][2]:
 
1971
                    parent_id = None
 
1972
                return ((entry[0][2], (old_path, None), True,
 
1973
                        (True, False),
 
1974
                        (parent_id, None),
 
1975
                        (entry[0][1], None),
 
1976
                        (_minikind_to_kind[source_minikind], None),
 
1977
                        (source_details[3], None)),)
 
1978
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1979
                # a rename; could be a true rename, or a rename inherited from
 
1980
                # a renamed parent. TODO: handle this efficiently. Its not
 
1981
                # common case to rename dirs though, so a correct but slow
 
1982
                # implementation will do.
 
1983
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1984
                    search_specific_files.add(target_details[1])
 
1985
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1986
                # neither of the selected trees contain this file,
 
1987
                # so skip over it. This is not currently directly tested, but
 
1988
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1989
                pass
 
1990
            else:
 
1991
                raise AssertionError("don't know how to compare "
 
1992
                    "source_minikind=%r, target_minikind=%r"
 
1993
                    % (source_minikind, target_minikind))
 
1994
                ## import pdb;pdb.set_trace()
 
1995
            return ()
 
1996
 
 
1997
        while search_specific_files:
 
1998
            # TODO: the pending list should be lexically sorted?  the
 
1999
            # interface doesn't require it.
 
2000
            current_root = search_specific_files.pop()
 
2001
            current_root_unicode = current_root.decode('utf8')
 
2002
            searched_specific_files.add(current_root)
 
2003
            # process the entries for this containing directory: the rest will be
 
2004
            # found by their parents recursively.
 
2005
            root_entries = _entries_for_path(current_root)
 
2006
            root_abspath = self.target.abspath(current_root_unicode)
 
2007
            try:
 
2008
                root_stat = os.lstat(root_abspath)
 
2009
            except OSError, e:
 
2010
                if e.errno == errno.ENOENT:
 
2011
                    # the path does not exist: let _process_entry know that.
 
2012
                    root_dir_info = None
 
2013
                else:
 
2014
                    # some other random error: hand it up.
 
2015
                    raise
 
2016
            else:
 
2017
                root_dir_info = ('', current_root,
 
2018
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2019
                    root_abspath)
 
2020
                if root_dir_info[2] == 'directory':
 
2021
                    if self.target._directory_is_tree_reference(
 
2022
                        current_root.decode('utf8')):
 
2023
                        root_dir_info = root_dir_info[:2] + \
 
2024
                            ('tree-reference',) + root_dir_info[3:]
 
2025
 
 
2026
            if not root_entries and not root_dir_info:
 
2027
                # this specified path is not present at all, skip it.
 
2028
                continue
 
2029
            path_handled = False
 
2030
            for entry in root_entries:
 
2031
                for result in _process_entry(entry, root_dir_info):
 
2032
                    # this check should probably be outside the loop: one
 
2033
                    # 'iterate two trees' api, and then _iter_changes filters
 
2034
                    # unchanged pairs. - RBC 20070226
 
2035
                    path_handled = True
 
2036
                    if (include_unchanged
 
2037
                        or result[2]                    # content change
 
2038
                        or result[3][0] != result[3][1] # versioned status
 
2039
                        or result[4][0] != result[4][1] # parent id
 
2040
                        or result[5][0] != result[5][1] # name
 
2041
                        or result[6][0] != result[6][1] # kind
 
2042
                        or result[7][0] != result[7][1] # executable
 
2043
                        ):
 
2044
                        yield (result[0],
 
2045
                               (utf8_decode_or_none(result[1][0]),
 
2046
                                utf8_decode_or_none(result[1][1])),
 
2047
                               result[2],
 
2048
                               result[3],
 
2049
                               result[4],
 
2050
                               (utf8_decode_or_none(result[5][0]),
 
2051
                                utf8_decode_or_none(result[5][1])),
 
2052
                               result[6],
 
2053
                               result[7],
 
2054
                              )
 
2055
            if want_unversioned and not path_handled and root_dir_info:
 
2056
                new_executable = bool(
 
2057
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2058
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2059
                yield (None,
 
2060
                       (None, current_root_unicode),
 
2061
                       True,
 
2062
                       (False, False),
 
2063
                       (None, None),
 
2064
                       (None, splitpath(current_root_unicode)[-1]),
 
2065
                       (None, root_dir_info[2]),
 
2066
                       (None, new_executable)
 
2067
                      )
 
2068
            initial_key = (current_root, '', '')
 
2069
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2070
            if block_index == 0:
 
2071
                # we have processed the total root already, but because the
 
2072
                # initial key matched it we should skip it here.
 
2073
                block_index +=1
 
2074
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2075
                current_dir_info = None
 
2076
            else:
 
2077
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2078
                try:
 
2079
                    current_dir_info = dir_iterator.next()
 
2080
                except OSError, e:
 
2081
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2082
                    # python 2.5 has e.errno == EINVAL,
 
2083
                    #            and e.winerror == ERROR_DIRECTORY
 
2084
                    e_winerror = getattr(e, 'winerror', None)
 
2085
                    # there may be directories in the inventory even though
 
2086
                    # this path is not a file on disk: so mark it as end of
 
2087
                    # iterator
 
2088
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2089
                        current_dir_info = None
 
2090
                    elif (sys.platform == 'win32'
 
2091
                          and ERROR_DIRECTORY in (e.errno, e_winerror)):
 
2092
                        current_dir_info = None
 
2093
                    else:
 
2094
                        raise
 
2095
                else:
 
2096
                    if current_dir_info[0][0] == '':
 
2097
                        # remove .bzr from iteration
 
2098
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2099
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2100
                        del current_dir_info[1][bzr_index]
 
2101
            # walk until both the directory listing and the versioned metadata
 
2102
            # are exhausted. 
 
2103
            if (block_index < len(state._dirblocks) and
 
2104
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2105
                current_block = state._dirblocks[block_index]
 
2106
            else:
 
2107
                current_block = None
 
2108
            while (current_dir_info is not None or
 
2109
                   current_block is not None):
 
2110
                if (current_dir_info and current_block
 
2111
                    and current_dir_info[0][0] != current_block[0]):
 
2112
                    if current_dir_info[0][0] < current_block[0] :
 
2113
                        # filesystem data refers to paths not covered by the dirblock.
 
2114
                        # this has two possibilities:
 
2115
                        # A) it is versioned but empty, so there is no block for it
 
2116
                        # B) it is not versioned.
 
2117
                        # in either case it was processed by the containing directories walk:
 
2118
                        # if it is root/foo, when we walked root we emitted it,
 
2119
                        # or if we ere given root/foo to walk specifically, we
 
2120
                        # emitted it when checking the walk-root entries
 
2121
                        # advance the iterator and loop - we dont need to emit it.
 
2122
                        try:
 
2123
                            current_dir_info = dir_iterator.next()
 
2124
                        except StopIteration:
 
2125
                            current_dir_info = None
 
2126
                    else:
 
2127
                        # We have a dirblock entry for this location, but there
 
2128
                        # is no filesystem path for this. This is most likely
 
2129
                        # because a directory was removed from the disk.
 
2130
                        # We don't have to report the missing directory,
 
2131
                        # because that should have already been handled, but we
 
2132
                        # need to handle all of the files that are contained
 
2133
                        # within.
 
2134
                        for current_entry in current_block[1]:
 
2135
                            # entry referring to file not present on disk.
 
2136
                            # advance the entry only, after processing.
 
2137
                            for result in _process_entry(current_entry, None):
 
2138
                                # this check should probably be outside the loop: one
 
2139
                                # 'iterate two trees' api, and then _iter_changes filters
 
2140
                                # unchanged pairs. - RBC 20070226
 
2141
                                if (include_unchanged
 
2142
                                    or result[2]                    # content change
 
2143
                                    or result[3][0] != result[3][1] # versioned status
 
2144
                                    or result[4][0] != result[4][1] # parent id
 
2145
                                    or result[5][0] != result[5][1] # name
 
2146
                                    or result[6][0] != result[6][1] # kind
 
2147
                                    or result[7][0] != result[7][1] # executable
 
2148
                                    ):
 
2149
                                    yield (result[0],
 
2150
                                           (utf8_decode_or_none(result[1][0]),
 
2151
                                            utf8_decode_or_none(result[1][1])),
 
2152
                                           result[2],
 
2153
                                           result[3],
 
2154
                                           result[4],
 
2155
                                           (utf8_decode_or_none(result[5][0]),
 
2156
                                            utf8_decode_or_none(result[5][1])),
 
2157
                                           result[6],
 
2158
                                           result[7],
 
2159
                                          )
 
2160
                        block_index +=1
 
2161
                        if (block_index < len(state._dirblocks) and
 
2162
                            osutils.is_inside(current_root,
 
2163
                                              state._dirblocks[block_index][0])):
 
2164
                            current_block = state._dirblocks[block_index]
 
2165
                        else:
 
2166
                            current_block = None
 
2167
                    continue
 
2168
                entry_index = 0
 
2169
                if current_block and entry_index < len(current_block[1]):
 
2170
                    current_entry = current_block[1][entry_index]
 
2171
                else:
 
2172
                    current_entry = None
 
2173
                advance_entry = True
 
2174
                path_index = 0
 
2175
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2176
                    current_path_info = current_dir_info[1][path_index]
 
2177
                    if current_path_info[2] == 'directory':
 
2178
                        if self.target._directory_is_tree_reference(
 
2179
                            current_path_info[0].decode('utf8')):
 
2180
                            current_path_info = current_path_info[:2] + \
 
2181
                                ('tree-reference',) + current_path_info[3:]
 
2182
                else:
 
2183
                    current_path_info = None
 
2184
                advance_path = True
 
2185
                path_handled = False
 
2186
                while (current_entry is not None or
 
2187
                    current_path_info is not None):
 
2188
                    if current_entry is None:
 
2189
                        # the check for path_handled when the path is adnvaced
 
2190
                        # will yield this path if needed.
 
2191
                        pass
 
2192
                    elif current_path_info is None:
 
2193
                        # no path is fine: the per entry code will handle it.
 
2194
                        for result in _process_entry(current_entry, current_path_info):
 
2195
                            # this check should probably be outside the loop: one
 
2196
                            # 'iterate two trees' api, and then _iter_changes filters
 
2197
                            # unchanged pairs. - RBC 20070226
 
2198
                            if (include_unchanged
 
2199
                                or result[2]                    # content change
 
2200
                                or result[3][0] != result[3][1] # versioned status
 
2201
                                or result[4][0] != result[4][1] # parent id
 
2202
                                or result[5][0] != result[5][1] # name
 
2203
                                or result[6][0] != result[6][1] # kind
 
2204
                                or result[7][0] != result[7][1] # executable
 
2205
                                ):
 
2206
                                yield (result[0],
 
2207
                                       (utf8_decode_or_none(result[1][0]),
 
2208
                                        utf8_decode_or_none(result[1][1])),
 
2209
                                       result[2],
 
2210
                                       result[3],
 
2211
                                       result[4],
 
2212
                                       (utf8_decode_or_none(result[5][0]),
 
2213
                                        utf8_decode_or_none(result[5][1])),
 
2214
                                       result[6],
 
2215
                                       result[7],
 
2216
                                      )
 
2217
                    elif current_entry[0][1] != current_path_info[1]:
 
2218
                        if current_path_info[1] < current_entry[0][1]:
 
2219
                            # extra file on disk: pass for now, but only
 
2220
                            # increment the path, not the entry
 
2221
                            advance_entry = False
 
2222
                        else:
 
2223
                            # entry referring to file not present on disk.
 
2224
                            # advance the entry only, after processing.
 
2225
                            for result in _process_entry(current_entry, None):
 
2226
                                # this check should probably be outside the loop: one
 
2227
                                # 'iterate two trees' api, and then _iter_changes filters
 
2228
                                # unchanged pairs. - RBC 20070226
 
2229
                                path_handled = True
 
2230
                                if (include_unchanged
 
2231
                                    or result[2]                    # content change
 
2232
                                    or result[3][0] != result[3][1] # versioned status
 
2233
                                    or result[4][0] != result[4][1] # parent id
 
2234
                                    or result[5][0] != result[5][1] # name
 
2235
                                    or result[6][0] != result[6][1] # kind
 
2236
                                    or result[7][0] != result[7][1] # executable
 
2237
                                    ):
 
2238
                                    yield (result[0],
 
2239
                                           (utf8_decode_or_none(result[1][0]),
 
2240
                                            utf8_decode_or_none(result[1][1])),
 
2241
                                           result[2],
 
2242
                                           result[3],
 
2243
                                           result[4],
 
2244
                                           (utf8_decode_or_none(result[5][0]),
 
2245
                                            utf8_decode_or_none(result[5][1])),
 
2246
                                           result[6],
 
2247
                                           result[7],
 
2248
                                          )
 
2249
                            advance_path = False
 
2250
                    else:
 
2251
                        for result in _process_entry(current_entry, current_path_info):
 
2252
                            # this check should probably be outside the loop: one
 
2253
                            # 'iterate two trees' api, and then _iter_changes filters
 
2254
                            # unchanged pairs. - RBC 20070226
 
2255
                            path_handled = True
 
2256
                            if (include_unchanged
 
2257
                                or result[2]                    # content change
 
2258
                                or result[3][0] != result[3][1] # versioned status
 
2259
                                or result[4][0] != result[4][1] # parent id
 
2260
                                or result[5][0] != result[5][1] # name
 
2261
                                or result[6][0] != result[6][1] # kind
 
2262
                                or result[7][0] != result[7][1] # executable
 
2263
                                ):
 
2264
                                yield (result[0],
 
2265
                                       (utf8_decode_or_none(result[1][0]),
 
2266
                                        utf8_decode_or_none(result[1][1])),
 
2267
                                       result[2],
 
2268
                                       result[3],
 
2269
                                       result[4],
 
2270
                                       (utf8_decode_or_none(result[5][0]),
 
2271
                                        utf8_decode_or_none(result[5][1])),
 
2272
                                       result[6],
 
2273
                                       result[7],
 
2274
                                      )
 
2275
                    if advance_entry and current_entry is not None:
 
2276
                        entry_index += 1
 
2277
                        if entry_index < len(current_block[1]):
 
2278
                            current_entry = current_block[1][entry_index]
 
2279
                        else:
 
2280
                            current_entry = None
 
2281
                    else:
 
2282
                        advance_entry = True # reset the advance flaga
 
2283
                    if advance_path and current_path_info is not None:
 
2284
                        if not path_handled:
 
2285
                            # unversioned in all regards
 
2286
                            if want_unversioned:
 
2287
                                new_executable = bool(
 
2288
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2289
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2290
                                if want_unversioned:
 
2291
                                    yield (None,
 
2292
                                        (None, utf8_decode_or_none(current_path_info[0])),
 
2293
                                        True,
 
2294
                                        (False, False),
 
2295
                                        (None, None),
 
2296
                                        (None, utf8_decode_or_none(current_path_info[1])),
 
2297
                                        (None, current_path_info[2]),
 
2298
                                        (None, new_executable))
 
2299
                            # dont descend into this unversioned path if it is
 
2300
                            # a dir
 
2301
                            if current_path_info[2] in ('directory'):
 
2302
                                del current_dir_info[1][path_index]
 
2303
                                path_index -= 1
 
2304
                        # dont descend the disk iterator into any tree 
 
2305
                        # paths.
 
2306
                        if current_path_info[2] == 'tree-reference':
 
2307
                            del current_dir_info[1][path_index]
 
2308
                            path_index -= 1
 
2309
                        path_index += 1
 
2310
                        if path_index < len(current_dir_info[1]):
 
2311
                            current_path_info = current_dir_info[1][path_index]
 
2312
                            if current_path_info[2] == 'directory':
 
2313
                                if self.target._directory_is_tree_reference(
 
2314
                                    current_path_info[0].decode('utf8')):
 
2315
                                    current_path_info = current_path_info[:2] + \
 
2316
                                        ('tree-reference',) + current_path_info[3:]
 
2317
                        else:
 
2318
                            current_path_info = None
 
2319
                        path_handled = False
 
2320
                    else:
 
2321
                        advance_path = True # reset the advance flagg.
 
2322
                if current_block is not None:
 
2323
                    block_index += 1
 
2324
                    if (block_index < len(state._dirblocks) and
 
2325
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2326
                        current_block = state._dirblocks[block_index]
 
2327
                    else:
 
2328
                        current_block = None
 
2329
                if current_dir_info is not None:
 
2330
                    try:
 
2331
                        current_dir_info = dir_iterator.next()
 
2332
                    except StopIteration:
 
2333
                        current_dir_info = None
 
2334
 
2116
2335
 
2117
2336
    @staticmethod
2118
2337
    def is_compatible(source, target):
2119
2338
        # the target must be a dirstate working tree
2120
 
        if not isinstance(target, DirStateWorkingTree):
 
2339
        if not isinstance(target, WorkingTree4):
2121
2340
            return False
2122
 
        # the source must be a revtree or dirstate rev tree.
 
2341
        # the source must be a revtreee or dirstate rev tree.
2123
2342
        if not isinstance(source,
2124
2343
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2125
2344
            return False
2126
2345
        # the source revid must be in the target dirstate
2127
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2346
        if not (source._revision_id == NULL_REVISION or
2128
2347
            source._revision_id in target.get_parent_ids()):
2129
 
            # TODO: what about ghosts? it may well need to
 
2348
            # TODO: what about ghosts? it may well need to 
2130
2349
            # check for them explicitly.
2131
2350
            return False
2132
2351
        return True
2142
2361
 
2143
2362
    def convert(self, tree):
2144
2363
        # lock the control files not the tree, so that we dont get tree
2145
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2364
        # on-unlock behaviours, and so that noone else diddles with the 
2146
2365
        # tree during upgrade.
2147
2366
        tree._control_files.lock_write()
2148
2367
        try:
2174
2393
 
2175
2394
    def update_format(self, tree):
2176
2395
        """Change the format marker."""
2177
 
        tree._transport.put_bytes('format',
2178
 
            self.target_format.get_format_string(),
2179
 
            mode=tree.bzrdir._get_file_mode())
2180
 
 
2181
 
 
2182
 
class Converter4to5(object):
2183
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2184
 
 
2185
 
    def __init__(self):
2186
 
        self.target_format = WorkingTreeFormat5()
2187
 
 
2188
 
    def convert(self, tree):
2189
 
        # lock the control files not the tree, so that we don't get tree
2190
 
        # on-unlock behaviours, and so that no-one else diddles with the
2191
 
        # tree during upgrade.
2192
 
        tree._control_files.lock_write()
2193
 
        try:
2194
 
            self.update_format(tree)
2195
 
        finally:
2196
 
            tree._control_files.unlock()
2197
 
 
2198
 
    def update_format(self, tree):
2199
 
        """Change the format marker."""
2200
 
        tree._transport.put_bytes('format',
2201
 
            self.target_format.get_format_string(),
2202
 
            mode=tree.bzrdir._get_file_mode())
2203
 
 
2204
 
 
2205
 
class Converter4or5to6(object):
2206
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2207
 
 
2208
 
    def __init__(self):
2209
 
        self.target_format = WorkingTreeFormat6()
2210
 
 
2211
 
    def convert(self, tree):
2212
 
        # lock the control files not the tree, so that we don't get tree
2213
 
        # on-unlock behaviours, and so that no-one else diddles with the
2214
 
        # tree during upgrade.
2215
 
        tree._control_files.lock_write()
2216
 
        try:
2217
 
            self.init_custom_control_files(tree)
2218
 
            self.update_format(tree)
2219
 
        finally:
2220
 
            tree._control_files.unlock()
2221
 
 
2222
 
    def init_custom_control_files(self, tree):
2223
 
        """Initialize custom control files."""
2224
 
        tree._transport.put_bytes('views', '',
2225
 
            mode=tree.bzrdir._get_file_mode())
2226
 
 
2227
 
    def update_format(self, tree):
2228
 
        """Change the format marker."""
2229
 
        tree._transport.put_bytes('format',
2230
 
            self.target_format.get_format_string(),
2231
 
            mode=tree.bzrdir._get_file_mode())
 
2396
        tree._control_files.put_utf8('format',
 
2397
            self.target_format.get_format_string())