/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2007-03-06 12:28:18 UTC
  • mto: (2321.1.1 integration)
  • mto: This revision was merged to the branch mainline in revision 2322.
  • Revision ID: robertc@robertcollins.net-20070306122818-xk0lc3l01ecl6vbc
Get merge_nested finally working: change nested tree iterators to take file_ids, and ensure the right branch is connected to in the merge logic. May not be suitable for shared repositories yet.

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
class WorkingTree4(WorkingTree3):
 
100
    """This is the Format 4 working tree.
 
101
 
 
102
    This differs from WorkingTree3 by:
 
103
     - Having a consolidated internal dirstate, stored in a
 
104
       randomly-accessible sorted file on disk.
 
105
     - Not having a regular inventory attribute.  One can be synthesized 
 
106
       on demand but this is expensive and should be avoided.
 
107
 
 
108
    This is new in bzr 0.15.
 
109
    """
 
110
 
72
111
    def __init__(self, basedir,
73
112
                 branch,
74
113
                 _control_files=None,
83
122
        """
84
123
        self._format = _format
85
124
        self.bzrdir = _bzrdir
 
125
        from bzrlib.trace import note, mutter
 
126
        assert isinstance(basedir, basestring), \
 
127
            "base directory %r is not a string" % basedir
86
128
        basedir = safe_unicode(basedir)
87
129
        mutter("opening working tree %r", basedir)
88
130
        self._branch = branch
 
131
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
132
            "branch %r is not a Branch" % self.branch
89
133
        self.basedir = realpath(basedir)
90
134
        # if branch is at our basedir and is a format 6 or less
91
135
        # assume all other formats have their own control files.
 
136
        assert isinstance(_control_files, LockableFiles), \
 
137
            "_control_files must be a LockableFiles, not %r" % _control_files
92
138
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
139
        self._dirty = None
95
140
        #-------------
96
141
        # during a read or write lock these objects are set, and are
98
143
        self._dirstate = None
99
144
        self._inventory = None
100
145
        #-------------
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
146
 
108
147
    @needs_tree_write_lock
109
148
    def _add(self, files, ids, kinds):
111
150
        state = self.current_dirstate()
112
151
        for f, file_id, kind in zip(files, ids, kinds):
113
152
            f = f.strip('/')
 
153
            assert '//' not in f
 
154
            assert '..' not in f
114
155
            if self.path2id(f):
115
156
                # special case tree root handling.
116
157
                if f == '' and self.path2id(f) == ROOT_ID:
137
178
    @needs_tree_write_lock
138
179
    def add_reference(self, sub_tree):
139
180
        # use standard implementation, which calls back to self._add
140
 
        #
 
181
        # 
141
182
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
183
        # it's just recorded at the moment of commit. 
143
184
        self._add_reference(sub_tree)
144
185
 
145
186
    def break_lock(self):
184
225
            WorkingTree3._comparison_data(self, entry, path)
185
226
        # it looks like a plain directory, but it's really a reference -- see
186
227
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
228
        if kind == 'directory' and self._directory_is_tree_reference(path):
189
229
            kind = 'tree-reference'
190
230
        return kind, executable, stat_value
191
231
 
217
257
            return self._dirstate
218
258
        local_path = self.bzrdir.get_workingtree_transport(None
219
259
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
260
        self._dirstate = dirstate.DirState.on_file(local_path)
222
261
        return self._dirstate
223
262
 
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
 
263
    def _directory_is_tree_reference(self, relpath):
 
264
        # as a special case, if a directory contains control files then 
 
265
        # it's a tree reference, except that the root of the tree is not
 
266
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
267
        # TODO: We could ask all the control formats whether they
 
268
        # recognize this directory, but at the moment there's no cheap api
 
269
        # to do that.  Since we probably can only nest bzr checkouts and
 
270
        # they always use this name it's ok for now.  -- mbp 20060306
 
271
        #
 
272
        # FIXME: There is an unhandled case here of a subdirectory
 
273
        # containing .bzr but not a branch; that will probably blow up
 
274
        # when you try to commit it.  It might happen if there is a
 
275
        # checkout in a subdirectory.  This can be avoided by not adding
 
276
        # it.  mbp 20070306
235
277
 
236
278
    def filter_unversioned_files(self, paths):
237
279
        """Filter out paths that are versioned.
261
303
        self._inventory = None
262
304
        self._dirty = False
263
305
 
264
 
    @needs_tree_write_lock
265
 
    def _gather_kinds(self, files, kinds):
266
 
        """See MutableTree._gather_kinds."""
267
 
        for pos, f in enumerate(files):
268
 
            if kinds[pos] is None:
269
 
                kinds[pos] = self._kind(f)
270
 
 
271
306
    def _generate_inventory(self):
272
307
        """Create and set self.inventory from the dirstate object.
273
 
 
 
308
        
274
309
        This is relatively expensive: we have to walk the entire dirstate.
275
310
        Ideally we would not, and can deprecate this function.
276
311
        """
280
315
        state._read_dirblocks_if_needed()
281
316
        root_key, current_entry = self._get_entry(path='')
282
317
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
318
        assert current_entry[0][0] == 'd' # directory
285
319
        inv = Inventory(root_id=current_id)
286
320
        # Turn some things into local variables
287
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
310
344
                inv_entry = factory[kind](file_id, name_unicode,
311
345
                                          parent_ie.file_id)
312
346
                if kind == 'file':
313
 
                    # This is only needed on win32, where this is the only way
314
 
                    # we know the executable bit.
315
 
                    inv_entry.executable = executable
316
347
                    # not strictly needed: working tree
317
 
                    #inv_entry.text_size = size
318
 
                    #inv_entry.text_sha1 = sha1
 
348
                    #entry.executable = executable
 
349
                    #entry.text_size = size
 
350
                    #entry.text_sha1 = sha1
 
351
                    pass
319
352
                elif kind == 'directory':
320
353
                    # add this entry to the parent map.
321
354
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
355
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
327
356
                    inv_entry.reference_revision = link_or_sha1 or None
328
 
                elif kind != 'symlink':
329
 
                    raise AssertionError("unknown kind %r" % kind)
 
357
                else:
 
358
                    assert 'unknown kind'
330
359
                # 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,))
 
360
                assert file_id not in inv_byid, ('file_id %s already in'
 
361
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
362
                assert name_unicode not in parent_ie.children
337
363
                inv_byid[file_id] = inv_entry
338
364
                parent_ie.children[name_unicode] = inv_entry
339
365
        self._inventory = inv
344
370
        If either file_id or path is supplied, it is used as the key to lookup.
345
371
        If both are supplied, the fastest lookup is used, and an error is
346
372
        raised if they do not both point at the same row.
347
 
 
 
373
        
348
374
        :param file_id: An optional unicode file_id to be looked up.
349
375
        :param path: An optional unicode path to be looked up.
350
376
        :return: The dirstate row tuple for path/file_id, or (None, None)
359
385
    def get_file_sha1(self, file_id, path=None, stat_value=None):
360
386
        # check file id is valid unconditionally.
361
387
        entry = self._get_entry(file_id=file_id, path=path)
362
 
        if entry[0] is None:
363
 
            raise errors.NoSuchId(self, file_id)
 
388
        assert entry[0] is not None, 'what error should this raise'
 
389
        # TODO:
 
390
        # if row stat is valid, use cached sha1, else, get a new sha1.
364
391
        if path is None:
365
392
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
366
393
 
367
394
        file_abspath = self.abspath(path)
368
395
        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)
 
396
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
397
                                          stat_value=stat_value)
379
398
        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
 
399
            return link_or_sha1
390
400
        return None
391
401
 
392
402
    def _get_inventory(self):
393
403
        """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
404
        if self._inventory is not None:
398
405
            return self._inventory
399
406
        self._must_be_locked()
406
413
    @needs_read_lock
407
414
    def get_parent_ids(self):
408
415
        """See Tree.get_parent_ids.
409
 
 
 
416
        
410
417
        This implementation requests the ids list from the dirstate file.
411
418
        """
412
419
        return self.current_dirstate().get_parent_ids()
428
435
 
429
436
    def has_id(self, file_id):
430
437
        state = self.current_dirstate()
 
438
        file_id = osutils.safe_file_id(file_id)
431
439
        row, parents = self._get_entry(file_id=file_id)
432
440
        if row is None:
433
441
            return False
434
442
        return osutils.lexists(pathjoin(
435
443
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
444
 
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
445
    @needs_read_lock
443
446
    def id2path(self, file_id):
444
 
        "Convert a file-id to a path."
 
447
        file_id = osutils.safe_file_id(file_id)
445
448
        state = self.current_dirstate()
446
449
        entry = self._get_entry(file_id=file_id)
447
450
        if entry == (None, None):
449
452
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
453
        return path_utf8.decode('utf8')
451
454
 
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
 
    if not osutils.supports_executable():
459
 
        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
 
            """
464
 
            entry = self._get_entry(file_id=file_id, path=path)
465
 
            if entry == (None, None):
466
 
                return False
467
 
            return entry[1][0][3]
468
 
 
469
 
        _is_executable_from_path_and_stat = \
470
 
            _is_executable_from_path_and_stat_from_basis
471
 
    else:
472
 
        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
 
            if not path:
479
 
                path = self.id2path(file_id)
480
 
            mode = os.lstat(self.abspath(path)).st_mode
481
 
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
 
 
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
455
    @needs_read_lock
494
456
    def __iter__(self):
495
457
        """Iterate through file_ids for this tree.
508
470
        return iter(result)
509
471
 
510
472
    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
473
        for key, tree_details in self.current_dirstate()._iter_entries():
516
474
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
475
                # not relevant to the working tree
519
477
            if not key[1]:
520
478
                # the root is not a reference.
521
479
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
523
 
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
526
 
            except errors.NoSuchFile:
527
 
                # path is missing on disk.
528
 
                continue
529
 
 
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
 
 
 
480
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
481
            if self._kind(path) == 'tree-reference':
 
482
                yield path, key[2]
 
483
 
 
484
    @needs_read_lock
536
485
    def kind(self, file_id):
537
486
        """Return the kind of a file.
538
487
 
539
488
        This is always the actual kind that's on disk, regardless of what it
540
489
        was added as.
541
 
 
542
 
        Note: The caller is expected to take a read-lock before calling this.
543
490
        """
544
491
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
492
        assert relpath != None, \
 
493
            "path for id {%s} is None!" % file_id
548
494
        return self._kind(relpath)
549
495
 
550
496
    def _kind(self, relpath):
551
497
        abspath = self.abspath(relpath)
552
498
        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'
 
499
        if kind == 'directory' and self._directory_is_tree_reference(relpath):
 
500
            kind = 'tree-reference'
558
501
        return kind
559
502
 
560
503
    @needs_read_lock
564
507
        if parent_ids:
565
508
            return parent_ids[0]
566
509
        else:
567
 
            return _mod_revision.NULL_REVISION
 
510
            return None
568
511
 
569
512
    def lock_read(self):
570
513
        """See Branch.lock_read, and WorkingTree.unlock."""
575
518
                state = self.current_dirstate()
576
519
                if not state._lock_token:
577
520
                    state.lock_read()
578
 
                # set our support for tree references from the repository in
579
 
                # use.
580
 
                self._repo_supports_tree_reference = getattr(
581
 
                    self.branch.repository._format, "supports_tree_reference",
582
 
                    False)
583
521
            except:
584
522
                self._control_files.unlock()
585
523
                raise
595
533
                state = self.current_dirstate()
596
534
                if not state._lock_token:
597
535
                    state.lock_write()
598
 
                # set our support for tree references from the repository in
599
 
                # use.
600
 
                self._repo_supports_tree_reference = getattr(
601
 
                    self.branch.repository._format, "supports_tree_reference",
602
 
                    False)
603
536
            except:
604
537
                self._control_files.unlock()
605
538
                raise
623
556
        result = []
624
557
        if not from_paths:
625
558
            return result
 
559
 
626
560
        state = self.current_dirstate()
627
 
        if isinstance(from_paths, basestring):
628
 
            raise ValueError()
 
561
 
 
562
        assert not isinstance(from_paths, basestring)
629
563
        to_dir_utf8 = to_dir.encode('utf8')
630
564
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
631
565
        id_index = state._get_id_index()
653
587
        if self._inventory is not None:
654
588
            update_inventory = True
655
589
            inv = self.inventory
 
590
            to_dir_ie = inv[to_dir_id]
656
591
            to_dir_id = to_entry[0][2]
657
 
            to_dir_ie = inv[to_dir_id]
658
592
        else:
659
593
            update_inventory = False
660
594
 
683
617
            new_entry = to_block[1][added_entry_index]
684
618
            rollbacks.append(lambda:state._make_absent(new_entry))
685
619
 
 
620
        # create rename entries and tuples
686
621
        for from_rel in from_paths:
687
622
            # from_rel is 'pathinroot/foo/bar'
688
623
            from_rel_utf8 = from_rel.encode('utf8')
691
626
            from_entry = self._get_entry(path=from_rel)
692
627
            if from_entry == (None, None):
693
628
                raise errors.BzrMoveFailedError(from_rel,to_dir,
694
 
                    errors.NotVersionedError(path=from_rel))
 
629
                    errors.NotVersionedError(path=str(from_rel)))
695
630
 
696
631
            from_id = from_entry[0][2]
697
632
            to_rel = pathjoin(to_dir, from_tail)
724
659
                if from_missing: # implicitly just update our path mapping
725
660
                    move_file = False
726
661
                elif not after:
727
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
662
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
663
                        extra="(Use --after to update the Bazaar id)")
728
664
 
729
665
            rollbacks = []
730
666
            def rollback_rename():
731
667
                """A single rename has failed, roll it back."""
732
 
                # roll back everything, even if we encounter trouble doing one
733
 
                # of them.
734
 
                #
735
 
                # TODO: at least log the other exceptions rather than just
736
 
                # losing them mbp 20070307
737
668
                exc_info = None
738
669
                for rollback in reversed(rollbacks):
739
670
                    try:
740
671
                        rollback()
741
672
                    except Exception, e:
 
673
                        import pdb;pdb.set_trace()
742
674
                        exc_info = sys.exc_info()
743
675
                if exc_info:
744
676
                    raise exc_info[0], exc_info[1], exc_info[2]
785
717
 
786
718
                if minikind == 'd':
787
719
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
788
 
                        """Recursively update all entries in this dirblock."""
789
 
                        if from_dir == '':
790
 
                            raise AssertionError("renaming root not supported")
 
720
                        """all entries in this block need updating.
 
721
 
 
722
                        TODO: This is pretty ugly, and doesn't support
 
723
                        reverting, but it works.
 
724
                        """
 
725
                        assert from_dir != '', "renaming root not supported"
791
726
                        from_key = (from_dir, '')
792
727
                        from_block_idx, present = \
793
728
                            state._find_block_index_from_key(from_key)
803
738
                        to_block_index = state._ensure_block(
804
739
                            to_block_index, to_entry_index, to_dir_utf8)
805
740
                        to_block = state._dirblocks[to_block_index]
806
 
 
807
 
                        # Grab a copy since move_one may update the list.
808
 
                        for entry in from_block[1][:]:
809
 
                            if not (entry[0][0] == from_dir):
810
 
                                raise AssertionError()
 
741
                        for entry in from_block[1]:
 
742
                            assert entry[0][0] == from_dir
811
743
                            cur_details = entry[1][0]
812
744
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
813
745
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
814
746
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
815
747
                            minikind = cur_details[0]
816
 
                            if minikind in 'ar':
817
 
                                # Deleted children of a renamed directory
818
 
                                # Do not need to be updated.
819
 
                                # Children that have been renamed out of this
820
 
                                # directory should also not be updated
821
 
                                continue
822
748
                            move_one(entry, from_path_utf8=from_path_utf8,
823
749
                                     minikind=minikind,
824
750
                                     executable=cur_details[3],
827
753
                                     size=cur_details[2],
828
754
                                     to_block=to_block,
829
755
                                     to_key=to_key,
830
 
                                     to_path_utf8=to_path_utf8)
 
756
                                     to_path_utf8=to_rel_utf8)
831
757
                            if minikind == 'd':
832
758
                                # We need to move all the children of this
833
759
                                # entry
872
798
        for tree in trees:
873
799
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
874
800
                parents):
875
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
876
 
                    trees, require_versioned)
 
801
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
877
802
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
878
803
        # -- make all paths utf8 --
879
804
        paths_utf8 = set()
939
864
            if not all_versioned:
940
865
                raise errors.PathsNotVersionedError(paths)
941
866
        # -- remove redundancy in supplied paths to prevent over-scanning --
942
 
        search_paths = osutils.minimum_path_selection(paths)
943
 
        # sketch:
 
867
        search_paths = set()
 
868
        for path in paths:
 
869
            other_paths = paths.difference(set([path]))
 
870
            if not osutils.is_inside_any(other_paths, path):
 
871
                # this is a top level path, we must check it.
 
872
                search_paths.add(path)
 
873
        # sketch: 
944
874
        # for all search_indexs in each path at or under each element of
945
875
        # search_paths, if the detail is relocated: add the id, and add the
946
876
        # relocated path as one to search if its not searched already. If the
1002
932
 
1003
933
    def read_working_inventory(self):
1004
934
        """Read the working inventory.
1005
 
 
 
935
        
1006
936
        This is a meaningless operation for dirstate, but we obey it anyhow.
1007
937
        """
1008
938
        return self.inventory
1013
943
 
1014
944
        WorkingTree4 supplies revision_trees for any basis tree.
1015
945
        """
 
946
        revision_id = osutils.safe_revision_id(revision_id)
1016
947
        dirstate = self.current_dirstate()
1017
948
        parent_ids = dirstate.get_parent_ids()
1018
949
        if revision_id not in parent_ids:
1025
956
    @needs_tree_write_lock
1026
957
    def set_last_revision(self, new_revision):
1027
958
        """Change the last revision in the working tree."""
 
959
        new_revision = osutils.safe_revision_id(new_revision)
1028
960
        parents = self.get_parent_ids()
1029
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1030
 
            if len(parents) >= 2:
1031
 
                raise AssertionError(
1032
 
                    "setting the last parent to none with a pending merge is "
1033
 
                    "unsupported.")
 
961
        if new_revision in (NULL_REVISION, None):
 
962
            assert len(parents) < 2, (
 
963
                "setting the last parent to none with a pending merge is "
 
964
                "unsupported.")
1034
965
            self.set_parent_ids([])
1035
966
        else:
1036
967
            self.set_parent_ids([new_revision] + parents[1:],
1039
970
    @needs_tree_write_lock
1040
971
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1041
972
        """Set the parent ids to revision_ids.
1042
 
 
 
973
        
1043
974
        See also set_parent_trees. This api will try to retrieve the tree data
1044
975
        for each element of revision_ids from the trees repository. If you have
1045
976
        tree data already available, it is more efficient to use
1049
980
        :param revision_ids: The revision_ids to set as the parent ids of this
1050
981
            working tree. Any of these may be ghosts.
1051
982
        """
 
983
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1052
984
        trees = []
1053
985
        for revision_id in revision_ids:
1054
986
            try:
1060
992
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1061
993
                revtree = None
1062
994
            trees.append((revision_id, revtree))
 
995
        self.current_dirstate()._validate()
1063
996
        self.set_parent_trees(trees,
1064
997
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
998
        self.current_dirstate()._validate()
1065
999
 
1066
1000
    @needs_tree_write_lock
1067
1001
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1072
1006
            parent tree - i.e. a ghost.
1073
1007
        """
1074
1008
        dirstate = self.current_dirstate()
 
1009
        dirstate._validate()
1075
1010
        if len(parents_list) > 0:
1076
1011
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1077
1012
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1078
1013
        real_trees = []
1079
1014
        ghosts = []
1080
 
 
1081
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1082
 
        graph = self.branch.repository.get_graph()
1083
 
        heads = graph.heads(parent_ids)
1084
 
        accepted_revisions = set()
1085
 
 
1086
1015
        # convert absent trees to the null tree, which we convert back to
1087
1016
        # missing on access.
1088
1017
        for rev_id, tree in parents_list:
1089
 
            if len(accepted_revisions) > 0:
1090
 
                # we always accept the first tree
1091
 
                if rev_id in accepted_revisions or rev_id not in heads:
1092
 
                    # We have already included either this tree, or its
1093
 
                    # descendent, so we skip it.
1094
 
                    continue
1095
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1018
            rev_id = osutils.safe_revision_id(rev_id)
1096
1019
            if tree is not None:
1097
1020
                real_trees.append((rev_id, tree))
1098
1021
            else:
1099
1022
                real_trees.append((rev_id,
1100
 
                    self.branch.repository.revision_tree(
1101
 
                        _mod_revision.NULL_REVISION)))
 
1023
                    self.branch.repository.revision_tree(None)))
1102
1024
                ghosts.append(rev_id)
1103
 
            accepted_revisions.add(rev_id)
 
1025
        dirstate._validate()
1104
1026
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1027
        dirstate._validate()
1105
1028
        self._make_dirty(reset_inventory=False)
 
1029
        dirstate._validate()
1106
1030
 
1107
1031
    def _set_root_id(self, file_id):
1108
1032
        """See WorkingTree.set_root_id."""
1111
1035
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1112
1036
            self._make_dirty(reset_inventory=True)
1113
1037
 
1114
 
    def _sha_from_stat(self, path, stat_result):
1115
 
        """Get a sha digest from the tree's stat cache.
1116
 
 
1117
 
        The default implementation assumes no stat cache is present.
1118
 
 
1119
 
        :param path: The path.
1120
 
        :param stat_result: The stat result being looked up.
1121
 
        """
1122
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1123
 
 
1124
 
    @needs_read_lock
1125
 
    def supports_tree_reference(self):
1126
 
        return self._repo_supports_tree_reference
1127
 
 
1128
1038
    def unlock(self):
1129
1039
        """Unlock in format 4 trees needs to write the entire dirstate."""
1130
 
        # do non-implementation specific cleanup
1131
 
        self._cleanup()
1132
 
 
1133
1040
        if self._control_files._lock_count == 1:
1134
1041
            # eventually we should do signature checking during read locks for
1135
1042
            # dirstate updates.
1166
1073
            return
1167
1074
        state = self.current_dirstate()
1168
1075
        state._read_dirblocks_if_needed()
1169
 
        ids_to_unversion = set(file_ids)
 
1076
        ids_to_unversion = set()
 
1077
        for file_id in file_ids:
 
1078
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1170
1079
        paths_to_unversion = set()
1171
1080
        # sketch:
1172
1081
        # check if the root is to be unversioned, if so, assert for now.
1199
1108
                # just forget the whole block.
1200
1109
                entry_index = 0
1201
1110
                while entry_index < len(block[1]):
1202
 
                    entry = block[1][entry_index]
1203
 
                    if entry[1][0][0] in 'ar':
1204
 
                        # don't remove absent or renamed entries
 
1111
                    # Mark this file id as having been removed
 
1112
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1113
                    if not state._make_absent(block[1][entry_index]):
1205
1114
                        entry_index += 1
1206
 
                    else:
1207
 
                        # Mark this file id as having been removed
1208
 
                        ids_to_unversion.discard(entry[0][2])
1209
 
                        if not state._make_absent(entry):
1210
 
                            # The block has not shrunk.
1211
 
                            entry_index += 1
1212
1115
                # go to the next block. (At the moment we dont delete empty
1213
1116
                # dirblocks)
1214
1117
                block_index += 1
1238
1141
                self._inventory.remove_recursive_id(file_id)
1239
1142
 
1240
1143
    @needs_tree_write_lock
1241
 
    def rename_one(self, from_rel, to_rel, after=False):
1242
 
        """See WorkingTree.rename_one"""
1243
 
        self.flush()
1244
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1245
 
 
1246
 
    @needs_tree_write_lock
1247
 
    def apply_inventory_delta(self, changes):
1248
 
        """See MutableTree.apply_inventory_delta"""
1249
 
        state = self.current_dirstate()
1250
 
        state.update_by_delta(changes)
1251
 
        self._make_dirty(reset_inventory=True)
1252
 
 
1253
 
    def update_basis_by_delta(self, new_revid, delta):
1254
 
        """See MutableTree.update_basis_by_delta."""
1255
 
        if self.last_revision() == new_revid:
1256
 
            raise AssertionError()
1257
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1258
 
 
1259
 
    @needs_read_lock
1260
 
    def _validate(self):
1261
 
        self._dirstate._validate()
1262
 
 
1263
 
    @needs_tree_write_lock
1264
1144
    def _write_inventory(self, inv):
1265
1145
        """Write inventory as the current inventory."""
1266
 
        if self._dirty:
1267
 
            raise AssertionError("attempting to write an inventory when the "
1268
 
                "dirstate is dirty will lose pending changes")
1269
 
        had_inventory = self._inventory is not None
1270
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1271
 
        # working inventory. We do this because self.inventory may be inv, or
1272
 
        # may have been modified, and either case would prevent a clean delta
1273
 
        # being created.
1274
 
        self._inventory = None
1275
 
        # generate a delta,
1276
 
        delta = inv._make_delta(self.inventory)
1277
 
        # and apply it.
1278
 
        self.apply_inventory_delta(delta)
1279
 
        if had_inventory:
 
1146
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1147
        self.current_dirstate().set_state_from_inventory(inv)
 
1148
        self._make_dirty(reset_inventory=False)
 
1149
        if self._inventory is not None:
1280
1150
            self._inventory = inv
1281
1151
        self.flush()
1282
1152
 
1283
1153
 
1284
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1285
 
 
1286
 
    def __init__(self, tree):
1287
 
        self.tree = tree
1288
 
 
1289
 
    def sha1(self, abspath):
1290
 
        """See dirstate.SHA1Provider.sha1()."""
1291
 
        filters = self.tree._content_filter_stack(
1292
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1293
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1294
 
 
1295
 
    def stat_and_sha1(self, abspath):
1296
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1297
 
        filters = self.tree._content_filter_stack(
1298
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1299
 
        file_obj = file(abspath, 'rb', 65000)
1300
 
        try:
1301
 
            statvalue = os.fstat(file_obj.fileno())
1302
 
            if filters:
1303
 
                file_obj = filtered_input_file(file_obj, filters)
1304
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1305
 
        finally:
1306
 
            file_obj.close()
1307
 
        return statvalue, sha1
1308
 
 
1309
 
 
1310
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1311
 
    """Dirstate working tree that supports content filtering.
1312
 
 
1313
 
    The dirstate holds the hash and size of the canonical form of the file, 
1314
 
    and most methods must return that.
1315
 
    """
1316
 
 
1317
 
    def _file_content_summary(self, path, stat_result):
1318
 
        # This is to support the somewhat obsolete path_content_summary method
1319
 
        # with content filtering: see
1320
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1321
 
        #
1322
 
        # If the dirstate cache is up to date and knows the hash and size,
1323
 
        # return that.
1324
 
        # Otherwise if there are no content filters, return the on-disk size
1325
 
        # and leave the hash blank.
1326
 
        # Otherwise, read and filter the on-disk file and use its size and
1327
 
        # hash.
1328
 
        #
1329
 
        # The dirstate doesn't store the size of the canonical form so we
1330
 
        # can't trust it for content-filtered trees.  We just return None.
1331
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1332
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1333
 
        return ('file', None, executable, dirstate_sha1)
1334
 
 
1335
 
 
1336
 
class WorkingTree4(DirStateWorkingTree):
1337
 
    """This is the Format 4 working tree.
1338
 
 
1339
 
    This differs from WorkingTree3 by:
1340
 
     - Having a consolidated internal dirstate, stored in a
1341
 
       randomly-accessible sorted file on disk.
1342
 
     - Not having a regular inventory attribute.  One can be synthesized
1343
 
       on demand but this is expensive and should be avoided.
1344
 
 
1345
 
    This is new in bzr 0.15.
1346
 
    """
1347
 
 
1348
 
 
1349
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1350
 
    """This is the Format 5 working tree.
1351
 
 
1352
 
    This differs from WorkingTree4 by:
1353
 
     - Supporting content filtering.
1354
 
 
1355
 
    This is new in bzr 1.11.
1356
 
    """
1357
 
 
1358
 
 
1359
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1360
 
    """This is the Format 6 working tree.
1361
 
 
1362
 
    This differs from WorkingTree5 by:
1363
 
     - Supporting a current view that may mask the set of files in a tree
1364
 
       impacted by most user operations.
1365
 
 
1366
 
    This is new in bzr 1.14.
1367
 
    """
1368
 
 
1369
 
    def _make_views(self):
1370
 
        return views.PathBasedViews(self)
1371
 
 
1372
 
 
1373
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1374
 
 
1375
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1376
 
                   accelerator_tree=None, hardlink=False):
 
1154
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1155
    """The first consolidated dirstate working tree format.
 
1156
 
 
1157
    This format:
 
1158
        - exists within a metadir controlling .bzr
 
1159
        - includes an explicit version marker for the workingtree control
 
1160
          files, separate from the BzrDir format
 
1161
        - modifies the hash cache format
 
1162
        - is new in bzr TODO FIXME SETBEFOREMERGE
 
1163
        - uses a LockDir to guard access to it.
 
1164
    """
 
1165
 
 
1166
    supports_tree_reference = True
 
1167
 
 
1168
    def get_format_string(self):
 
1169
        """See WorkingTreeFormat.get_format_string()."""
 
1170
        return "Bazaar Working Tree format 4\n"
 
1171
 
 
1172
    def get_format_description(self):
 
1173
        """See WorkingTreeFormat.get_format_description()."""
 
1174
        return "Working tree format 4"
 
1175
 
 
1176
    def initialize(self, a_bzrdir, revision_id=None):
1377
1177
        """See WorkingTreeFormat.initialize().
1378
1178
 
1379
1179
        :param revision_id: allows creating a working tree at a different
1380
1180
        revision than the branch is at.
1381
 
        :param accelerator_tree: A tree which can be used for retrieving file
1382
 
            contents more quickly than the revision tree, i.e. a workingtree.
1383
 
            The revision tree will be used for cases where accelerator_tree's
1384
 
            content is different.
1385
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1386
 
            where possible.
1387
1181
 
1388
 
        These trees get an initial random root id, if their repository supports
1389
 
        rich root data, TREE_ROOT otherwise.
 
1182
        These trees get an initial random root id.
1390
1183
        """
 
1184
        revision_id = osutils.safe_revision_id(revision_id)
1391
1185
        if not isinstance(a_bzrdir.transport, LocalTransport):
1392
1186
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1393
1187
        transport = a_bzrdir.get_workingtree_transport(self)
1394
1188
        control_files = self._open_control_files(a_bzrdir)
1395
1189
        control_files.create_lock()
1396
1190
        control_files.lock_write()
1397
 
        transport.put_bytes('format', self.get_format_string(),
1398
 
            mode=a_bzrdir._get_file_mode())
1399
 
        if from_branch is not None:
1400
 
            branch = from_branch
1401
 
        else:
1402
 
            branch = a_bzrdir.open_branch()
 
1191
        control_files.put_utf8('format', self.get_format_string())
 
1192
        branch = a_bzrdir.open_branch()
1403
1193
        if revision_id is None:
1404
1194
            revision_id = branch.last_revision()
1405
1195
        local_path = transport.local_abspath('dirstate')
1406
1196
        # write out new dirstate (must exist when we create the tree)
1407
1197
        state = dirstate.DirState.initialize(local_path)
1408
1198
        state.unlock()
1409
 
        del state
1410
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1199
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1411
1200
                         branch,
1412
1201
                         _format=self,
1413
1202
                         _bzrdir=a_bzrdir,
1414
1203
                         _control_files=control_files)
1415
1204
        wt._new_tree()
1416
1205
        wt.lock_tree_write()
 
1206
        state._validate()
1417
1207
        try:
1418
 
            self._init_custom_control_files(wt)
1419
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
1420
 
                if branch.repository.supports_rich_root():
1421
 
                    wt._set_root_id(generate_ids.gen_root_id())
1422
 
                else:
1423
 
                    wt._set_root_id(ROOT_ID)
 
1208
            if revision_id in (None, NULL_REVISION):
 
1209
                wt._set_root_id(generate_ids.gen_root_id())
1424
1210
                wt.flush()
1425
 
            basis = None
1426
 
            # frequently, we will get here due to branching.  The accelerator
1427
 
            # tree will be the tree from the branch, so the desired basis
1428
 
            # tree will often be a parent of the accelerator tree.
1429
 
            if accelerator_tree is not None:
1430
 
                try:
1431
 
                    basis = accelerator_tree.revision_tree(revision_id)
1432
 
                except errors.NoSuchRevision:
1433
 
                    pass
1434
 
            if basis is None:
1435
 
                basis = branch.repository.revision_tree(revision_id)
1436
 
            if revision_id == _mod_revision.NULL_REVISION:
1437
 
                parents_list = []
1438
 
            else:
1439
 
                parents_list = [(revision_id, basis)]
 
1211
                wt.current_dirstate()._validate()
 
1212
            wt.set_last_revision(revision_id)
 
1213
            wt.flush()
 
1214
            basis = wt.basis_tree()
1440
1215
            basis.lock_read()
1441
 
            try:
1442
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1216
            # if the basis has a root id we have to use that; otherwise we use
 
1217
            # a new random one
 
1218
            basis_root_id = basis.get_root_id()
 
1219
            if basis_root_id is not None:
 
1220
                wt._set_root_id(basis_root_id)
1443
1221
                wt.flush()
1444
 
                # if the basis has a root id we have to use that; otherwise we
1445
 
                # use a new random one
1446
 
                basis_root_id = basis.get_root_id()
1447
 
                if basis_root_id is not None:
1448
 
                    wt._set_root_id(basis_root_id)
1449
 
                    wt.flush()
1450
 
                if wt.supports_content_filtering():
1451
 
                    # The original tree may not have the same content filters
1452
 
                    # applied so we can't safely build the inventory delta from
1453
 
                    # the source tree.
1454
 
                    delta_from_tree = False
1455
 
                else:
1456
 
                    delta_from_tree = True
1457
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1458
 
                # because wt4.apply_inventory_delta does not mutate the input
1459
 
                # inventory entries.
1460
 
                transform.build_tree(basis, wt, accelerator_tree,
1461
 
                                     hardlink=hardlink,
1462
 
                                     delta_from_tree=delta_from_tree)
1463
 
            finally:
1464
 
                basis.unlock()
 
1222
            transform.build_tree(basis, wt)
 
1223
            basis.unlock()
1465
1224
        finally:
1466
1225
            control_files.unlock()
1467
1226
            wt.unlock()
1468
1227
        return wt
1469
1228
 
1470
 
    def _init_custom_control_files(self, wt):
1471
 
        """Subclasses with custom control files should override this method.
1472
 
 
1473
 
        The working tree and control files are locked for writing when this
1474
 
        method is called.
1475
 
 
1476
 
        :param wt: the WorkingTree object
1477
 
        """
1478
 
 
1479
1229
    def _open(self, a_bzrdir, control_files):
1480
1230
        """Open the tree itself.
1481
1231
 
1482
1232
        :param a_bzrdir: the dir for the tree.
1483
1233
        :param control_files: the control files for the tree.
1484
1234
        """
1485
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1235
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1486
1236
                           branch=a_bzrdir.open_branch(),
1487
1237
                           _format=self,
1488
1238
                           _bzrdir=a_bzrdir,
1489
1239
                           _control_files=control_files)
1490
1240
 
1491
1241
    def __get_matchingbzrdir(self):
1492
 
        return self._get_matchingbzrdir()
1493
 
 
1494
 
    def _get_matchingbzrdir(self):
1495
 
        """Overrideable method to get a bzrdir for testing."""
1496
1242
        # please test against something that will let us do tree references
1497
1243
        return bzrdir.format_registry.make_bzrdir(
1498
1244
            'dirstate-with-subtree')
1500
1246
    _matchingbzrdir = property(__get_matchingbzrdir)
1501
1247
 
1502
1248
 
1503
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1504
 
    """The first consolidated dirstate working tree format.
1505
 
 
1506
 
    This format:
1507
 
        - exists within a metadir controlling .bzr
1508
 
        - includes an explicit version marker for the workingtree control
1509
 
          files, separate from the BzrDir format
1510
 
        - modifies the hash cache format
1511
 
        - is new in bzr 0.15
1512
 
        - uses a LockDir to guard access to it.
1513
 
    """
1514
 
 
1515
 
    upgrade_recommended = False
1516
 
 
1517
 
    _tree_class = WorkingTree4
1518
 
 
1519
 
    def get_format_string(self):
1520
 
        """See WorkingTreeFormat.get_format_string()."""
1521
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1522
 
 
1523
 
    def get_format_description(self):
1524
 
        """See WorkingTreeFormat.get_format_description()."""
1525
 
        return "Working tree format 4"
1526
 
 
1527
 
 
1528
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1529
 
    """WorkingTree format supporting content filtering.
1530
 
    """
1531
 
 
1532
 
    upgrade_recommended = False
1533
 
 
1534
 
    _tree_class = WorkingTree5
1535
 
 
1536
 
    def get_format_string(self):
1537
 
        """See WorkingTreeFormat.get_format_string()."""
1538
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1539
 
 
1540
 
    def get_format_description(self):
1541
 
        """See WorkingTreeFormat.get_format_description()."""
1542
 
        return "Working tree format 5"
1543
 
 
1544
 
    def supports_content_filtering(self):
1545
 
        return True
1546
 
 
1547
 
 
1548
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1549
 
    """WorkingTree format supporting views.
1550
 
    """
1551
 
 
1552
 
    upgrade_recommended = False
1553
 
 
1554
 
    _tree_class = WorkingTree6
1555
 
 
1556
 
    def get_format_string(self):
1557
 
        """See WorkingTreeFormat.get_format_string()."""
1558
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1559
 
 
1560
 
    def get_format_description(self):
1561
 
        """See WorkingTreeFormat.get_format_description()."""
1562
 
        return "Working tree format 6"
1563
 
 
1564
 
    def _init_custom_control_files(self, wt):
1565
 
        """Subclasses with custom control files should override this method."""
1566
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1567
 
 
1568
 
    def supports_content_filtering(self):
1569
 
        return True
1570
 
 
1571
 
    def supports_views(self):
1572
 
        return True
1573
 
 
1574
 
 
1575
1249
class DirStateRevisionTree(Tree):
1576
 
    """A revision tree pulling the inventory from a dirstate.
1577
 
    
1578
 
    Note that this is one of the historical (ie revision) trees cached in the
1579
 
    dirstate for easy access, not the workingtree.
1580
 
    """
 
1250
    """A revision tree pulling the inventory from a dirstate."""
1581
1251
 
1582
1252
    def __init__(self, dirstate, revision_id, repository):
1583
1253
        self._dirstate = dirstate
1584
 
        self._revision_id = revision_id
 
1254
        self._revision_id = osutils.safe_revision_id(revision_id)
1585
1255
        self._repository = repository
1586
1256
        self._inventory = None
1587
1257
        self._locked = 0
1588
1258
        self._dirstate_locked = False
1589
 
        self._repo_supports_tree_reference = getattr(
1590
 
            repository._format, "supports_tree_reference",
1591
 
            False)
1592
1259
 
1593
1260
    def __repr__(self):
1594
1261
        return "<%s of %s in %s>" % \
1595
1262
            (self.__class__.__name__, self._revision_id, self._dirstate)
1596
1263
 
1597
 
    def annotate_iter(self, file_id,
1598
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1264
    def annotate_iter(self, file_id):
1599
1265
        """See Tree.annotate_iter"""
1600
 
        text_key = (file_id, self.inventory[file_id].revision)
1601
 
        annotations = self._repository.texts.annotate(text_key)
1602
 
        return [(key[-1], line) for (key, line) in annotations]
 
1266
        w = self._repository.weave_store.get_weave(file_id,
 
1267
                           self._repository.get_transaction())
 
1268
        return w.annotate_iter(self.inventory[file_id].revision)
1603
1269
 
1604
 
    def _get_ancestors(self, default_revision):
1605
 
        return set(self._repository.get_ancestry(self._revision_id,
1606
 
                                                 topo_sorted=False))
1607
1270
    def _comparison_data(self, entry, path):
1608
1271
        """See Tree._comparison_data."""
1609
1272
        if entry is None:
1626
1289
    def get_root_id(self):
1627
1290
        return self.path2id('')
1628
1291
 
1629
 
    def id2path(self, file_id):
1630
 
        "Convert a file-id to a path."
1631
 
        entry = self._get_entry(file_id=file_id)
1632
 
        if entry == (None, None):
1633
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1634
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1635
 
        return path_utf8.decode('utf8')
1636
 
 
1637
 
    def iter_references(self):
1638
 
        if not self._repo_supports_tree_reference:
1639
 
            # When the repo doesn't support references, we will have nothing to
1640
 
            # return
1641
 
            return iter([])
1642
 
        # Otherwise, fall back to the default implementation
1643
 
        return super(DirStateRevisionTree, self).iter_references()
1644
 
 
1645
1292
    def _get_parent_index(self):
1646
1293
        """Return the index in the dirstate referenced by this tree."""
1647
1294
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1652
1299
        If either file_id or path is supplied, it is used as the key to lookup.
1653
1300
        If both are supplied, the fastest lookup is used, and an error is
1654
1301
        raised if they do not both point at the same row.
1655
 
 
 
1302
        
1656
1303
        :param file_id: An optional unicode file_id to be looked up.
1657
1304
        :param path: An optional unicode path to be looked up.
1658
1305
        :return: The dirstate row tuple for path/file_id, or (None, None)
1659
1306
        """
1660
1307
        if file_id is None and path is None:
1661
1308
            raise errors.BzrError('must supply file_id or path')
 
1309
        file_id = osutils.safe_file_id(file_id)
1662
1310
        if path is not None:
1663
1311
            path = path.encode('utf8')
1664
1312
        parent_index = self._get_parent_index()
1672
1320
 
1673
1321
        This is relatively expensive: we have to walk the entire dirstate.
1674
1322
        """
1675
 
        if not self._locked:
1676
 
            raise AssertionError(
1677
 
                'cannot generate inventory of an unlocked '
1678
 
                'dirstate revision tree')
 
1323
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1324
            'dirstate revision tree'
1679
1325
        # separate call for profiling - makes it clear where the costs are.
1680
1326
        self._dirstate._read_dirblocks_if_needed()
1681
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1682
 
            raise AssertionError(
1683
 
                'parent %s has disappeared from %s' % (
1684
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1327
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1328
            'parent %s has disappeared from %s' % (
 
1329
            self._revision_id, self._dirstate.get_parent_ids())
1685
1330
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1686
1331
        # This is identical now to the WorkingTree _generate_inventory except
1687
1332
        # for the tree index use.
1688
1333
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1689
1334
        current_id = root_key[2]
1690
 
        if current_entry[parent_index][0] != 'd':
1691
 
            raise AssertionError()
 
1335
        assert current_entry[parent_index][0] == 'd'
1692
1336
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1693
1337
        inv.root.revision = current_entry[parent_index][4]
1694
1338
        # Turn some things into local variables
1726
1370
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1727
1371
                elif kind == 'symlink':
1728
1372
                    inv_entry.executable = False
1729
 
                    inv_entry.text_size = None
 
1373
                    inv_entry.text_size = size
1730
1374
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1731
1375
                elif kind == 'tree-reference':
1732
1376
                    inv_entry.reference_revision = fingerprint or None
1734
1378
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1735
1379
                            % entry)
1736
1380
                # These checks cost us around 40ms on a 55k entry tree
1737
 
                if file_id in inv_byid:
1738
 
                    raise AssertionError('file_id %s already in'
1739
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1740
 
                if name_unicode in parent_ie.children:
1741
 
                    raise AssertionError('name %r already in parent'
1742
 
                        % (name_unicode,))
 
1381
                assert file_id not in inv_byid
 
1382
                assert name_unicode not in parent_ie.children
1743
1383
                inv_byid[file_id] = inv_entry
1744
1384
                parent_ie.children[name_unicode] = inv_entry
1745
1385
        self._inventory = inv
1755
1395
            return None
1756
1396
        parent_index = self._get_parent_index()
1757
1397
        last_changed_revision = entry[1][parent_index][4]
1758
 
        try:
1759
 
            rev = self._repository.get_revision(last_changed_revision)
1760
 
        except errors.NoSuchRevision:
1761
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1762
 
        return rev.timestamp
 
1398
        return self._repository.get_revision(last_changed_revision).timestamp
1763
1399
 
1764
1400
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1765
1401
        entry = self._get_entry(file_id=file_id, path=path)
1769
1405
            return parent_details[1]
1770
1406
        return None
1771
1407
 
1772
 
    def get_file(self, file_id, path=None):
 
1408
    def get_file(self, file_id):
1773
1409
        return StringIO(self.get_file_text(file_id))
1774
1410
 
 
1411
    def get_file_lines(self, file_id):
 
1412
        ie = self.inventory[file_id]
 
1413
        return self._repository.weave_store.get_weave(file_id,
 
1414
                self._repository.get_transaction()).get_lines(ie.revision)
 
1415
 
1775
1416
    def get_file_size(self, file_id):
1776
 
        """See Tree.get_file_size"""
1777
1417
        return self.inventory[file_id].text_size
1778
1418
 
1779
 
    def get_file_text(self, file_id, path=None):
1780
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1781
 
        return ''.join(content)
 
1419
    def get_file_text(self, file_id):
 
1420
        return ''.join(self.get_file_lines(file_id))
1782
1421
 
1783
1422
    def get_reference_revision(self, file_id, path=None):
1784
1423
        return self.inventory[file_id].reference_revision
1785
1424
 
1786
 
    def iter_files_bytes(self, desired_files):
1787
 
        """See Tree.iter_files_bytes.
1788
 
 
1789
 
        This version is implemented on top of Repository.iter_files_bytes"""
1790
 
        parent_index = self._get_parent_index()
1791
 
        repo_desired_files = []
1792
 
        for file_id, identifier in desired_files:
1793
 
            entry = self._get_entry(file_id)
1794
 
            if entry == (None, None):
1795
 
                raise errors.NoSuchId(self, file_id)
1796
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1797
 
                                       identifier))
1798
 
        return self._repository.iter_files_bytes(repo_desired_files)
1799
 
 
1800
1425
    def get_symlink_target(self, file_id):
1801
1426
        entry = self._get_entry(file_id=file_id)
1802
1427
        parent_index = self._get_parent_index()
1803
1428
        if entry[1][parent_index][0] != 'l':
1804
1429
            return None
1805
1430
        else:
1806
 
            target = entry[1][parent_index][1]
1807
 
            target = target.decode('utf8')
1808
 
            return target
 
1431
            # At present, none of the tree implementations supports non-ascii
 
1432
            # symlink targets. So we will just assume that the dirstate path is
 
1433
            # correct.
 
1434
            return entry[1][parent_index][1]
1809
1435
 
1810
1436
    def get_revision_id(self):
1811
1437
        """Return the revision id for this tree."""
1829
1455
        return bool(self.path2id(filename))
1830
1456
 
1831
1457
    def kind(self, file_id):
1832
 
        entry = self._get_entry(file_id=file_id)[1]
1833
 
        if entry is None:
1834
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1835
 
        parent_index = self._get_parent_index()
1836
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1837
 
 
1838
 
    def stored_kind(self, file_id):
1839
 
        """See Tree.stored_kind"""
1840
 
        return self.kind(file_id)
1841
 
 
1842
 
    def path_content_summary(self, path):
1843
 
        """See Tree.path_content_summary."""
1844
 
        id = self.inventory.path2id(path)
1845
 
        if id is None:
1846
 
            return ('missing', None, None, None)
1847
 
        entry = self._inventory[id]
1848
 
        kind = entry.kind
1849
 
        if kind == 'file':
1850
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1851
 
        elif kind == 'symlink':
1852
 
            return (kind, None, None, entry.symlink_target)
1853
 
        else:
1854
 
            return (kind, None, None, None)
 
1458
        return self.inventory[file_id].kind
1855
1459
 
1856
1460
    def is_executable(self, file_id, path=None):
1857
1461
        ie = self.inventory[file_id]
1859
1463
            return None
1860
1464
        return ie.executable
1861
1465
 
1862
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1466
    def list_files(self, include_root=False):
1863
1467
        # We use a standard implementation, because DirStateRevisionTree is
1864
1468
        # dealing with one of the parents of the current state
1865
1469
        inv = self._get_inventory()
1866
 
        if from_dir is None:
1867
 
            from_dir_id = None
1868
 
        else:
1869
 
            from_dir_id = inv.path2id(from_dir)
1870
 
            if from_dir_id is None:
1871
 
                # Directory not versioned
1872
 
                return
1873
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1874
 
        if inv.root is not None and not include_root and from_dir is None:
 
1470
        entries = inv.iter_entries()
 
1471
        if self.inventory.root is not None and not include_root:
1875
1472
            entries.next()
1876
1473
        for path, entry in entries:
1877
1474
            yield path, 'V', entry.kind, entry.file_id, entry
1910
1507
                self._dirstate_locked = False
1911
1508
            self._repository.unlock()
1912
1509
 
1913
 
    @needs_read_lock
1914
 
    def supports_tree_reference(self):
1915
 
        return self._repo_supports_tree_reference
1916
 
 
1917
1510
    def walkdirs(self, prefix=""):
1918
 
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1919
 
        # implementation based on an inventory.
1920
 
        # This should be cleaned up to use the much faster Dirstate code
1921
 
        # So for now, we just build up the parent inventory, and extract
1922
 
        # it the same way RevisionTree does.
 
1511
        # TODO: jam 20070215 This is the cheap way by cheating and using the
 
1512
        #       RevisionTree implementation.
 
1513
        #       This should be cleaned up to use the much faster Dirstate code
 
1514
        #       This is a little tricky, though, because the dirstate is
 
1515
        #       indexed by current path, not by parent path.
 
1516
        #       So for now, we just build up the parent inventory, and extract
 
1517
        #       it the same way RevisionTree does.
1923
1518
        _directory = 'directory'
1924
1519
        inv = self._get_inventory()
1925
1520
        top_id = inv.path2id(prefix)
1951
1546
 
1952
1547
class InterDirStateTree(InterTree):
1953
1548
    """Fast path optimiser for changes_from with dirstate trees.
1954
 
 
1955
 
    This is used only when both trees are in the dirstate working file, and
1956
 
    the source is any parent within the dirstate, and the destination is
 
1549
    
 
1550
    This is used only when both trees are in the dirstate working file, and 
 
1551
    the source is any parent within the dirstate, and the destination is 
1957
1552
    the current working tree of the same dirstate.
1958
1553
    """
1959
1554
    # this could be generalized to allow comparisons between any trees in the
1972
1567
        target.set_parent_ids([revid])
1973
1568
        return target.basis_tree(), target
1974
1569
 
1975
 
    @classmethod
1976
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1977
 
        result = klass.make_source_parent_tree(source, target)
1978
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1979
 
        return result
1980
 
 
1981
 
    @classmethod
1982
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
1983
 
                                                  target):
1984
 
        from bzrlib.tests.test__dirstate_helpers import \
1985
 
            compiled_dirstate_helpers_feature
1986
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
1987
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
1988
 
        result = klass.make_source_parent_tree(source, target)
1989
 
        result[1]._iter_changes = ProcessEntryC
1990
 
        return result
1991
 
 
1992
1570
    _matching_from_tree_format = WorkingTreeFormat4()
1993
1571
    _matching_to_tree_format = WorkingTreeFormat4()
1994
 
 
1995
 
    @classmethod
1996
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
1997
 
        # This method shouldn't be called, because we have python and C
1998
 
        # specific flavours.
1999
 
        raise NotImplementedError
2000
 
 
2001
 
    def iter_changes(self, include_unchanged=False,
 
1572
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1573
 
 
1574
    def _iter_changes(self, include_unchanged=False,
2002
1575
                      specific_files=None, pb=None, extra_trees=[],
2003
1576
                      require_versioned=True, want_unversioned=False):
2004
1577
        """Return the changes from source to target.
2005
1578
 
2006
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1579
        :return: An iterator that yields tuples. See InterTree._iter_changes
2007
1580
            for details.
2008
1581
        :param specific_files: An optional list of file paths to restrict the
2009
1582
            comparison to. When mapping filenames to ids, all matches in all
2020
1593
            output. An unversioned file is defined as one with (False, False)
2021
1594
            for the versioned pair.
2022
1595
        """
 
1596
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1597
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1598
        # NB: show_status depends on being able to pass in non-versioned files
 
1599
        # and report them as unknown
2023
1600
        # TODO: handle extra trees in the dirstate.
2024
 
        if (extra_trees or specific_files == []):
 
1601
        # TODO: handle comparisons as an empty tree as a different special
 
1602
        # case? mbp 20070226
 
1603
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2025
1604
            # we can't fast-path these cases (yet)
2026
 
            return super(InterDirStateTree, self).iter_changes(
 
1605
            for f in super(InterDirStateTree, self)._iter_changes(
2027
1606
                include_unchanged, specific_files, pb, extra_trees,
2028
 
                require_versioned, want_unversioned=want_unversioned)
 
1607
                require_versioned, want_unversioned=want_unversioned):
 
1608
                yield f
 
1609
            return
2029
1610
        parent_ids = self.target.get_parent_ids()
2030
 
        if not (self.source._revision_id in parent_ids
2031
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2032
 
            raise AssertionError(
2033
 
                "revision {%s} is not stored in {%s}, but %s "
2034
 
                "can only be used for trees stored in the dirstate"
2035
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1611
        assert (self.source._revision_id in parent_ids), \
 
1612
                "revision {%s} is not stored in {%s}, but %s " \
 
1613
                "can only be used for trees stored in the dirstate" \
 
1614
                % (self.source._revision_id, self.target, self._iter_changes)
2036
1615
        target_index = 0
2037
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1616
        if self.source._revision_id == NULL_REVISION:
2038
1617
            source_index = None
2039
1618
            indices = (target_index,)
2040
1619
        else:
2041
 
            if not (self.source._revision_id in parent_ids):
2042
 
                raise AssertionError(
2043
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2044
 
                    self.source._revision_id, parent_ids))
 
1620
            assert (self.source._revision_id in parent_ids), \
 
1621
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1622
                self.source._revision_id, parent_ids)
2045
1623
            source_index = 1 + parent_ids.index(self.source._revision_id)
2046
 
            indices = (source_index, target_index)
 
1624
            indices = (source_index,target_index)
2047
1625
        # -- make all specific_files utf8 --
2048
1626
        if specific_files:
2049
1627
            specific_files_utf8 = set()
2050
1628
            for path in specific_files:
2051
 
                # Note, if there are many specific files, using cache_utf8
2052
 
                # would be good here.
2053
1629
                specific_files_utf8.add(path.encode('utf8'))
2054
1630
            specific_files = specific_files_utf8
2055
1631
        else:
2056
1632
            specific_files = set([''])
2057
1633
        # -- specific_files is now a utf8 path set --
2058
 
 
2059
1634
        # -- get the state object and prepare it.
2060
1635
        state = self.target.current_dirstate()
2061
1636
        state._read_dirblocks_if_needed()
 
1637
        def _entries_for_path(path):
 
1638
            """Return a list with all the entries that match path for all ids.
 
1639
            """
 
1640
            dirname, basename = os.path.split(path)
 
1641
            key = (dirname, basename, '')
 
1642
            block_index, present = state._find_block_index_from_key(key)
 
1643
            if not present:
 
1644
                # the block which should contain path is absent.
 
1645
                return []
 
1646
            result = []
 
1647
            block = state._dirblocks[block_index][1]
 
1648
            entry_index, _ = state._find_entry_index(key, block)
 
1649
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1650
            while (entry_index < len(block) and
 
1651
                block[entry_index][0][0:2] == key[0:2]):
 
1652
                result.append(block[entry_index])
 
1653
                entry_index += 1
 
1654
            return result
2062
1655
        if require_versioned:
2063
1656
            # -- check all supplied paths are versioned in a search tree. --
2064
 
            not_versioned = []
 
1657
            all_versioned = True
2065
1658
            for path in specific_files:
2066
 
                path_entries = state._entries_for_path(path)
 
1659
                path_entries = _entries_for_path(path)
2067
1660
                if not path_entries:
2068
1661
                    # this specified path is not present at all: error
2069
 
                    not_versioned.append(path)
2070
 
                    continue
 
1662
                    all_versioned = False
 
1663
                    break
2071
1664
                found_versioned = False
2072
1665
                # for each id at this path
2073
1666
                for entry in path_entries:
2080
1673
                if not found_versioned:
2081
1674
                    # none of the indexes was not 'absent' at all ids for this
2082
1675
                    # path.
2083
 
                    not_versioned.append(path)
2084
 
            if len(not_versioned) > 0:
2085
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1676
                    all_versioned = False
 
1677
                    break
 
1678
            if not all_versioned:
 
1679
                raise errors.PathsNotVersionedError(specific_files)
2086
1680
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2087
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1681
        search_specific_files = set()
 
1682
        for path in specific_files:
 
1683
            other_specific_files = specific_files.difference(set([path]))
 
1684
            if not osutils.is_inside_any(other_specific_files, path):
 
1685
                # this is a top level path, we must check it.
 
1686
                search_specific_files.add(path)
 
1687
        # sketch: 
 
1688
        # compare source_index and target_index at or under each element of search_specific_files.
 
1689
        # follow the following comparison table. Note that we only want to do diff operations when
 
1690
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1691
        # for the target.
 
1692
        # cases:
 
1693
        # 
 
1694
        # Source | Target | disk | action
 
1695
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1696
        #        |        |      | diff check on source-target
 
1697
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1698
        #        |        |      | ???
 
1699
        #   r    |  a     |      | add source to search
 
1700
        #   r    |  a     |  a   | 
 
1701
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1702
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1703
        #   a    | fdlt   |      | add new id
 
1704
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1705
        #   a    |  a     |      | not present in either tree, skip
 
1706
        #   a    |  a     |  a   | not present in any tree, skip
 
1707
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1708
        #        |        |      | may not be selected by the users list of paths.
 
1709
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1710
        #        |        |      | may not be selected by the users list of paths.
 
1711
        #  fdlt  | fdlt   |      | content in both: diff them
 
1712
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1713
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1714
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1715
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1716
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1717
        #        |        |      | this id at the other path.
 
1718
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1719
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1720
        #        |        |      | this id at the other path.
 
1721
 
 
1722
        # for all search_indexs in each path at or under each element of
 
1723
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1724
        # relocated path as one to search if its not searched already. If the
 
1725
        # detail is not relocated, add the id.
 
1726
        searched_specific_files = set()
 
1727
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1728
        # Using a list so that we can access the values and change them in
 
1729
        # nested scope. Each one is [path, file_id, entry]
 
1730
        last_source_parent = [None, None, None]
 
1731
        last_target_parent = [None, None, None]
2088
1732
 
2089
1733
        use_filesystem_for_exec = (sys.platform != 'win32')
2090
 
        iter_changes = self.target._iter_changes(include_unchanged,
2091
 
            use_filesystem_for_exec, search_specific_files, state,
2092
 
            source_index, target_index, want_unversioned, self.target)
2093
 
        return iter_changes.iter_changes()
 
1734
 
 
1735
        def _process_entry(entry, path_info):
 
1736
            """Compare an entry and real disk to generate delta information.
 
1737
 
 
1738
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1739
                the path of entry. If None, then the path is considered absent.
 
1740
                (Perhaps we should pass in a concrete entry for this ?)
 
1741
                Basename is returned as a utf8 string because we expect this
 
1742
                tuple will be ignored, and don't want to take the time to
 
1743
                decode.
 
1744
            """
 
1745
            # TODO: when a parent has been renamed, dont emit path renames for children,
 
1746
            ## if path_info[1] == 'sub':
 
1747
            ##     import pdb;pdb.set_trace()
 
1748
            if source_index is None:
 
1749
                source_details = NULL_PARENT_DETAILS
 
1750
            else:
 
1751
                source_details = entry[1][source_index]
 
1752
            target_details = entry[1][target_index]
 
1753
            target_minikind = target_details[0]
 
1754
            if path_info is not None and target_minikind in 'fdlt':
 
1755
                assert target_index == 0
 
1756
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1757
                                                  stat_value=path_info[3])
 
1758
                # The entry may have been modified by update_entry
 
1759
                target_details = entry[1][target_index]
 
1760
                target_minikind = target_details[0]
 
1761
            else:
 
1762
                link_or_sha1 = None
 
1763
            source_minikind = source_details[0]
 
1764
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1765
                # claimed content in both: diff
 
1766
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1767
                #        |        |      | diff check on source-target
 
1768
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1769
                #        |        |      | ???
 
1770
                if source_minikind in 'r':
 
1771
                    # add the source to the search path to find any children it
 
1772
                    # has.  TODO ? : only add if it is a container ?
 
1773
                    if not osutils.is_inside_any(searched_specific_files,
 
1774
                                                 source_details[1]):
 
1775
                        search_specific_files.add(source_details[1])
 
1776
                    # generate the old path; this is needed for stating later
 
1777
                    # as well.
 
1778
                    old_path = source_details[1]
 
1779
                    old_dirname, old_basename = os.path.split(old_path)
 
1780
                    path = pathjoin(entry[0][0], entry[0][1])
 
1781
                    old_entry = state._get_entry(source_index,
 
1782
                                                 path_utf8=old_path)
 
1783
                    # update the source details variable to be the real
 
1784
                    # location.
 
1785
                    source_details = old_entry[1][source_index]
 
1786
                    source_minikind = source_details[0]
 
1787
                else:
 
1788
                    old_dirname = entry[0][0]
 
1789
                    old_basename = entry[0][1]
 
1790
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1791
                if path_info is None:
 
1792
                    # the file is missing on disk, show as removed.
 
1793
                    content_change = True
 
1794
                    target_kind = None
 
1795
                    target_exec = False
 
1796
                else:
 
1797
                    # source and target are both versioned and disk file is present.
 
1798
                    target_kind = path_info[2]
 
1799
                    if target_kind == 'directory':
 
1800
                        if source_minikind != 'd':
 
1801
                            content_change = True
 
1802
                        else:
 
1803
                            # directories have no fingerprint
 
1804
                            content_change = False
 
1805
                        target_exec = False
 
1806
                    elif target_kind == 'file':
 
1807
                        if source_minikind != 'f':
 
1808
                            content_change = True
 
1809
                        else:
 
1810
                            # We could check the size, but we already have the
 
1811
                            # sha1 hash.
 
1812
                            content_change = (link_or_sha1 != source_details[1])
 
1813
                        # Target details is updated at update_entry time
 
1814
                        if use_filesystem_for_exec:
 
1815
                            # We don't need S_ISREG here, because we are sure
 
1816
                            # we are dealing with a file.
 
1817
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1818
                        else:
 
1819
                            target_exec = target_details[3]
 
1820
                    elif target_kind == 'symlink':
 
1821
                        if source_minikind != 'l':
 
1822
                            content_change = True
 
1823
                        else:
 
1824
                            content_change = (link_or_sha1 != source_details[1])
 
1825
                        target_exec = False
 
1826
                    elif target_kind == 'tree-reference':
 
1827
                        if source_minikind != 't':
 
1828
                            content_change = True
 
1829
                        else:
 
1830
                            content_change = False
 
1831
                        target_exec = False
 
1832
                    else:
 
1833
                        raise Exception, "unknown kind %s" % path_info[2]
 
1834
                # parent id is the entry for the path in the target tree
 
1835
                if old_dirname == last_source_parent[0]:
 
1836
                    source_parent_id = last_source_parent[1]
 
1837
                else:
 
1838
                    source_parent_entry = state._get_entry(source_index,
 
1839
                                                           path_utf8=old_dirname)
 
1840
                    source_parent_id = source_parent_entry[0][2]
 
1841
                    if source_parent_id == entry[0][2]:
 
1842
                        # This is the root, so the parent is None
 
1843
                        source_parent_id = None
 
1844
                    else:
 
1845
                        last_source_parent[0] = old_dirname
 
1846
                        last_source_parent[1] = source_parent_id
 
1847
                        last_source_parent[2] = source_parent_entry
 
1848
 
 
1849
                new_dirname = entry[0][0]
 
1850
                if new_dirname == last_target_parent[0]:
 
1851
                    target_parent_id = last_target_parent[1]
 
1852
                else:
 
1853
                    # TODO: We don't always need to do the lookup, because the
 
1854
                    #       parent entry will be the same as the source entry.
 
1855
                    target_parent_entry = state._get_entry(target_index,
 
1856
                                                           path_utf8=new_dirname)
 
1857
                    target_parent_id = target_parent_entry[0][2]
 
1858
                    if target_parent_id == entry[0][2]:
 
1859
                        # This is the root, so the parent is None
 
1860
                        target_parent_id = None
 
1861
                    else:
 
1862
                        last_target_parent[0] = new_dirname
 
1863
                        last_target_parent[1] = target_parent_id
 
1864
                        last_target_parent[2] = target_parent_entry
 
1865
 
 
1866
                source_exec = source_details[3]
 
1867
                return ((entry[0][2], (old_path, path), content_change,
 
1868
                        (True, True),
 
1869
                        (source_parent_id, target_parent_id),
 
1870
                        (old_basename, entry[0][1]),
 
1871
                        (_minikind_to_kind[source_minikind], target_kind),
 
1872
                        (source_exec, target_exec)),)
 
1873
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1874
                # looks like a new file
 
1875
                if path_info is not None:
 
1876
                    path = pathjoin(entry[0][0], entry[0][1])
 
1877
                    # parent id is the entry for the path in the target tree
 
1878
                    # TODO: these are the same for an entire directory: cache em.
 
1879
                    parent_id = state._get_entry(target_index,
 
1880
                                                 path_utf8=entry[0][0])[0][2]
 
1881
                    if parent_id == entry[0][2]:
 
1882
                        parent_id = None
 
1883
                    if use_filesystem_for_exec:
 
1884
                        # We need S_ISREG here, because we aren't sure if this
 
1885
                        # is a file or not.
 
1886
                        target_exec = bool(
 
1887
                            stat.S_ISREG(path_info[3].st_mode)
 
1888
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1889
                    else:
 
1890
                        target_exec = target_details[3]
 
1891
                    return ((entry[0][2], (None, path), True,
 
1892
                            (False, True),
 
1893
                            (None, parent_id),
 
1894
                            (None, entry[0][1]),
 
1895
                            (None, path_info[2]),
 
1896
                            (None, target_exec)),)
 
1897
                else:
 
1898
                    # but its not on disk: we deliberately treat this as just
 
1899
                    # never-present. (Why ?! - RBC 20070224)
 
1900
                    pass
 
1901
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1902
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1903
                # if its still on disk, *and* theres no other entry at this
 
1904
                # path [we dont know this in this routine at the moment -
 
1905
                # perhaps we should change this - then it would be an unknown.
 
1906
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1907
                # parent id is the entry for the path in the target tree
 
1908
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1909
                if parent_id == entry[0][2]:
 
1910
                    parent_id = None
 
1911
                return ((entry[0][2], (old_path, None), True,
 
1912
                        (True, False),
 
1913
                        (parent_id, None),
 
1914
                        (entry[0][1], None),
 
1915
                        (_minikind_to_kind[source_minikind], None),
 
1916
                        (source_details[3], None)),)
 
1917
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1918
                # a rename; could be a true rename, or a rename inherited from
 
1919
                # a renamed parent. TODO: handle this efficiently. Its not
 
1920
                # common case to rename dirs though, so a correct but slow
 
1921
                # implementation will do.
 
1922
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1923
                    search_specific_files.add(target_details[1])
 
1924
            elif source_minikind in 'r' and target_minikind in 'r':
 
1925
                # neither of the selected trees contain this file,
 
1926
                # so skip over it. This is not currently directly tested, but
 
1927
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1928
                pass
 
1929
            else:
 
1930
                raise AssertionError("don't know how to compare "
 
1931
                    "source_minikind=%r, target_minikind=%r"
 
1932
                    % (source_minikind, target_minikind))
 
1933
                ## import pdb;pdb.set_trace()
 
1934
            return ()
 
1935
        while search_specific_files:
 
1936
            # TODO: the pending list should be lexically sorted?
 
1937
            current_root = search_specific_files.pop()
 
1938
            searched_specific_files.add(current_root)
 
1939
            # process the entries for this containing directory: the rest will be
 
1940
            # found by their parents recursively.
 
1941
            root_entries = _entries_for_path(current_root)
 
1942
            root_abspath = self.target.abspath(current_root)
 
1943
            try:
 
1944
                root_stat = os.lstat(root_abspath)
 
1945
            except OSError, e:
 
1946
                if e.errno == errno.ENOENT:
 
1947
                    # the path does not exist: let _process_entry know that.
 
1948
                    root_dir_info = None
 
1949
                else:
 
1950
                    # some other random error: hand it up.
 
1951
                    raise
 
1952
            else:
 
1953
                root_dir_info = ('', current_root,
 
1954
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1955
                    root_abspath)
 
1956
                if root_dir_info[2] == 'directory':
 
1957
                    if self.target._directory_is_tree_reference(
 
1958
                        current_root.decode('utf8')):
 
1959
                        root_dir_info = root_dir_info[:2] + \
 
1960
                            ('tree-reference',) + root_dir_info[3:]
 
1961
 
 
1962
            if not root_entries and not root_dir_info:
 
1963
                # this specified path is not present at all, skip it.
 
1964
                continue
 
1965
            path_handled = False
 
1966
            for entry in root_entries:
 
1967
                for result in _process_entry(entry, root_dir_info):
 
1968
                    # this check should probably be outside the loop: one
 
1969
                    # 'iterate two trees' api, and then _iter_changes filters
 
1970
                    # unchanged pairs. - RBC 20070226
 
1971
                    path_handled = True
 
1972
                    if (include_unchanged
 
1973
                        or result[2]                    # content change
 
1974
                        or result[3][0] != result[3][1] # versioned status
 
1975
                        or result[4][0] != result[4][1] # parent id
 
1976
                        or result[5][0] != result[5][1] # name
 
1977
                        or result[6][0] != result[6][1] # kind
 
1978
                        or result[7][0] != result[7][1] # executable
 
1979
                        ):
 
1980
                        result = (result[0],
 
1981
                            ((utf8_decode(result[1][0])[0]),
 
1982
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
1983
                        yield result
 
1984
            if want_unversioned and not path_handled:
 
1985
                new_executable = bool(
 
1986
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
1987
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
1988
                yield (None, (None, current_root), True, (False, False),
 
1989
                    (None, None),
 
1990
                    (None, splitpath(current_root)[-1]),
 
1991
                    (None, root_dir_info[2]), (None, new_executable))
 
1992
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
1993
            initial_key = (current_root, '', '')
 
1994
            block_index, _ = state._find_block_index_from_key(initial_key)
 
1995
            if block_index == 0:
 
1996
                # we have processed the total root already, but because the
 
1997
                # initial key matched it we should skip it here.
 
1998
                block_index +=1
 
1999
            try:
 
2000
                current_dir_info = dir_iterator.next()
 
2001
            except OSError, e:
 
2002
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
2003
                    # there may be directories in the inventory even though
 
2004
                    # this path is not a file on disk: so mark it as end of
 
2005
                    # iterator
 
2006
                    current_dir_info = None
 
2007
                else:
 
2008
                    raise
 
2009
            else:
 
2010
                if current_dir_info[0][0] == '':
 
2011
                    # remove .bzr from iteration
 
2012
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2013
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2014
                    del current_dir_info[1][bzr_index]
 
2015
            # walk until both the directory listing and the versioned metadata
 
2016
            # are exhausted. TODO: reevaluate this, perhaps we should stop when
 
2017
            # the versioned data runs out.
 
2018
            if (block_index < len(state._dirblocks) and
 
2019
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2020
                current_block = state._dirblocks[block_index]
 
2021
            else:
 
2022
                current_block = None
 
2023
            while (current_dir_info is not None or
 
2024
                   current_block is not None):
 
2025
                if (current_dir_info and current_block
 
2026
                    and current_dir_info[0][0] != current_block[0]):
 
2027
                    if current_dir_info[0][0] < current_block[0] :
 
2028
                        # filesystem data refers to paths not covered by the dirblock.
 
2029
                        # this has two possibilities:
 
2030
                        # A) it is versioned but empty, so there is no block for it
 
2031
                        # B) it is not versioned.
 
2032
                        # in either case it was processed by the containing directories walk:
 
2033
                        # if it is root/foo, when we walked root we emitted it,
 
2034
                        # or if we ere given root/foo to walk specifically, we
 
2035
                        # emitted it when checking the walk-root entries
 
2036
                        # advance the iterator and loop - we dont need to emit it.
 
2037
                        try:
 
2038
                            current_dir_info = dir_iterator.next()
 
2039
                        except StopIteration:
 
2040
                            current_dir_info = None
 
2041
                    else:
 
2042
                        # We have a dirblock entry for this location, but there
 
2043
                        # is no filesystem path for this. This is most likely
 
2044
                        # because a directory was removed from the disk.
 
2045
                        # We don't have to report the missing directory,
 
2046
                        # because that should have already been handled, but we
 
2047
                        # need to handle all of the files that are contained
 
2048
                        # within.
 
2049
                        for current_entry in current_block[1]:
 
2050
                            # entry referring to file not present on disk.
 
2051
                            # advance the entry only, after processing.
 
2052
                            for result in _process_entry(current_entry, None):
 
2053
                                # this check should probably be outside the loop: one
 
2054
                                # 'iterate two trees' api, and then _iter_changes filters
 
2055
                                # unchanged pairs. - RBC 20070226
 
2056
                                if (include_unchanged
 
2057
                                    or result[2]                    # content change
 
2058
                                    or result[3][0] != result[3][1] # versioned status
 
2059
                                    or result[4][0] != result[4][1] # parent id
 
2060
                                    or result[5][0] != result[5][1] # name
 
2061
                                    or result[6][0] != result[6][1] # kind
 
2062
                                    or result[7][0] != result[7][1] # executable
 
2063
                                    ):
 
2064
                                    result = (result[0],
 
2065
                                        ((utf8_decode(result[1][0])[0]),
 
2066
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2067
                                    yield result
 
2068
                        block_index +=1
 
2069
                        if (block_index < len(state._dirblocks) and
 
2070
                            osutils.is_inside(current_root,
 
2071
                                              state._dirblocks[block_index][0])):
 
2072
                            current_block = state._dirblocks[block_index]
 
2073
                        else:
 
2074
                            current_block = None
 
2075
                    continue
 
2076
                entry_index = 0
 
2077
                if current_block and entry_index < len(current_block[1]):
 
2078
                    current_entry = current_block[1][entry_index]
 
2079
                else:
 
2080
                    current_entry = None
 
2081
                advance_entry = True
 
2082
                path_index = 0
 
2083
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2084
                    current_path_info = current_dir_info[1][path_index]
 
2085
                    if current_path_info[2] == 'directory':
 
2086
                        if self.target._directory_is_tree_reference(
 
2087
                            current_path_info[0].decode('utf8')):
 
2088
                            current_path_info = current_path_info[:2] + \
 
2089
                                ('tree-reference',) + current_path_info[3:]
 
2090
                else:
 
2091
                    current_path_info = None
 
2092
                advance_path = True
 
2093
                path_handled = False
 
2094
                while (current_entry is not None or
 
2095
                    current_path_info is not None):
 
2096
                    if current_entry is None:
 
2097
                        # the check for path_handled when the path is adnvaced
 
2098
                        # will yield this path if needed.
 
2099
                        pass
 
2100
                    elif current_path_info is None:
 
2101
                        # no path is fine: the per entry code will handle it.
 
2102
                        for result in _process_entry(current_entry, current_path_info):
 
2103
                            # this check should probably be outside the loop: one
 
2104
                            # 'iterate two trees' api, and then _iter_changes filters
 
2105
                            # unchanged pairs. - RBC 20070226
 
2106
                            if (include_unchanged
 
2107
                                or result[2]                    # content change
 
2108
                                or result[3][0] != result[3][1] # versioned status
 
2109
                                or result[4][0] != result[4][1] # parent id
 
2110
                                or result[5][0] != result[5][1] # name
 
2111
                                or result[6][0] != result[6][1] # kind
 
2112
                                or result[7][0] != result[7][1] # executable
 
2113
                                ):
 
2114
                                result = (result[0],
 
2115
                                    ((utf8_decode(result[1][0])[0]),
 
2116
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2117
                                yield result
 
2118
                    elif current_entry[0][1] != current_path_info[1]:
 
2119
                        if current_path_info[1] < current_entry[0][1]:
 
2120
                            # extra file on disk: pass for now, but only
 
2121
                            # increment the path, not the entry
 
2122
                            # import pdb; pdb.set_trace()
 
2123
                            # print 'unversioned file'
 
2124
                            advance_entry = False
 
2125
                        else:
 
2126
                            # entry referring to file not present on disk.
 
2127
                            # advance the entry only, after processing.
 
2128
                            for result in _process_entry(current_entry, None):
 
2129
                                # this check should probably be outside the loop: one
 
2130
                                # 'iterate two trees' api, and then _iter_changes filters
 
2131
                                # unchanged pairs. - RBC 20070226
 
2132
                                path_handled = True
 
2133
                                if (include_unchanged
 
2134
                                    or result[2]                    # content change
 
2135
                                    or result[3][0] != result[3][1] # versioned status
 
2136
                                    or result[4][0] != result[4][1] # parent id
 
2137
                                    or result[5][0] != result[5][1] # name
 
2138
                                    or result[6][0] != result[6][1] # kind
 
2139
                                    or result[7][0] != result[7][1] # executable
 
2140
                                    ):
 
2141
                                    result = (result[0],
 
2142
                                        ((utf8_decode(result[1][0])[0]),
 
2143
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2144
                                    yield result
 
2145
                            advance_path = False
 
2146
                    else:
 
2147
                        for result in _process_entry(current_entry, current_path_info):
 
2148
                            # this check should probably be outside the loop: one
 
2149
                            # 'iterate two trees' api, and then _iter_changes filters
 
2150
                            # unchanged pairs. - RBC 20070226
 
2151
                            path_handled = True
 
2152
                            if (include_unchanged
 
2153
                                or result[2]                    # content change
 
2154
                                or result[3][0] != result[3][1] # versioned status
 
2155
                                or result[4][0] != result[4][1] # parent id
 
2156
                                or result[5][0] != result[5][1] # name
 
2157
                                or result[6][0] != result[6][1] # kind
 
2158
                                or result[7][0] != result[7][1] # executable
 
2159
                                ):
 
2160
                                result = (result[0],
 
2161
                                    ((utf8_decode(result[1][0])[0]),
 
2162
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2163
                                yield result
 
2164
                    if advance_entry and current_entry is not None:
 
2165
                        entry_index += 1
 
2166
                        if entry_index < len(current_block[1]):
 
2167
                            current_entry = current_block[1][entry_index]
 
2168
                        else:
 
2169
                            current_entry = None
 
2170
                    else:
 
2171
                        advance_entry = True # reset the advance flaga
 
2172
                    if advance_path and current_path_info is not None:
 
2173
                        if not path_handled:
 
2174
                            # unversioned in all regards
 
2175
                            if want_unversioned:
 
2176
                                new_executable = bool(
 
2177
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2178
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2179
                                if want_unversioned:
 
2180
                                    yield (None, (None, current_path_info[0]),
 
2181
                                        True,
 
2182
                                        (False, False),
 
2183
                                        (None, None),
 
2184
                                        (None, current_path_info[1]),
 
2185
                                        (None, current_path_info[2]),
 
2186
                                        (None, new_executable))
 
2187
                            # dont descend into this unversioned path if it is
 
2188
                            # a dir
 
2189
                            if current_path_info[2] in (
 
2190
                                'directory', 'tree-referene'):
 
2191
                                del current_dir_info[1][path_index]
 
2192
                                path_index -= 1
 
2193
                        path_index += 1
 
2194
                        if path_index < len(current_dir_info[1]):
 
2195
                            current_path_info = current_dir_info[1][path_index]
 
2196
                            if current_path_info[2] == 'directory':
 
2197
                                if self.target._directory_is_tree_reference(
 
2198
                                    current_path_info[0].decode('utf8')):
 
2199
                                    current_path_info = current_path_info[:2] + \
 
2200
                                        ('tree-reference',) + current_path_info[3:]
 
2201
                        else:
 
2202
                            current_path_info = None
 
2203
                        path_handled = False
 
2204
                    else:
 
2205
                        advance_path = True # reset the advance flagg.
 
2206
                if current_block is not None:
 
2207
                    block_index += 1
 
2208
                    if (block_index < len(state._dirblocks) and
 
2209
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2210
                        current_block = state._dirblocks[block_index]
 
2211
                    else:
 
2212
                        current_block = None
 
2213
                if current_dir_info is not None:
 
2214
                    try:
 
2215
                        current_dir_info = dir_iterator.next()
 
2216
                    except StopIteration:
 
2217
                        current_dir_info = None
 
2218
 
2094
2219
 
2095
2220
    @staticmethod
2096
2221
    def is_compatible(source, target):
2097
2222
        # the target must be a dirstate working tree
2098
 
        if not isinstance(target, DirStateWorkingTree):
 
2223
        if not isinstance(target, WorkingTree4):
2099
2224
            return False
2100
 
        # the source must be a revtree or dirstate rev tree.
 
2225
        # the source must be a revtreee or dirstate rev tree.
2101
2226
        if not isinstance(source,
2102
2227
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2103
2228
            return False
2104
2229
        # the source revid must be in the target dirstate
2105
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2230
        if not (source._revision_id == NULL_REVISION or
2106
2231
            source._revision_id in target.get_parent_ids()):
2107
 
            # TODO: what about ghosts? it may well need to
 
2232
            # TODO: what about ghosts? it may well need to 
2108
2233
            # check for them explicitly.
2109
2234
            return False
2110
2235
        return True
2120
2245
 
2121
2246
    def convert(self, tree):
2122
2247
        # lock the control files not the tree, so that we dont get tree
2123
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2248
        # on-unlock behaviours, and so that noone else diddles with the 
2124
2249
        # tree during upgrade.
2125
2250
        tree._control_files.lock_write()
2126
2251
        try:
2127
 
            tree.read_working_inventory()
2128
2252
            self.create_dirstate_data(tree)
2129
2253
            self.update_format(tree)
2130
2254
            self.remove_xml_files(tree)
2152
2276
 
2153
2277
    def update_format(self, tree):
2154
2278
        """Change the format marker."""
2155
 
        tree._transport.put_bytes('format',
2156
 
            self.target_format.get_format_string(),
2157
 
            mode=tree.bzrdir._get_file_mode())
2158
 
 
2159
 
 
2160
 
class Converter4to5(object):
2161
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2162
 
 
2163
 
    def __init__(self):
2164
 
        self.target_format = WorkingTreeFormat5()
2165
 
 
2166
 
    def convert(self, tree):
2167
 
        # lock the control files not the tree, so that we don't get tree
2168
 
        # on-unlock behaviours, and so that no-one else diddles with the
2169
 
        # tree during upgrade.
2170
 
        tree._control_files.lock_write()
2171
 
        try:
2172
 
            self.update_format(tree)
2173
 
        finally:
2174
 
            tree._control_files.unlock()
2175
 
 
2176
 
    def update_format(self, tree):
2177
 
        """Change the format marker."""
2178
 
        tree._transport.put_bytes('format',
2179
 
            self.target_format.get_format_string(),
2180
 
            mode=tree.bzrdir._get_file_mode())
2181
 
 
2182
 
 
2183
 
class Converter4or5to6(object):
2184
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2185
 
 
2186
 
    def __init__(self):
2187
 
        self.target_format = WorkingTreeFormat6()
2188
 
 
2189
 
    def convert(self, tree):
2190
 
        # lock the control files not the tree, so that we don't get tree
2191
 
        # on-unlock behaviours, and so that no-one else diddles with the
2192
 
        # tree during upgrade.
2193
 
        tree._control_files.lock_write()
2194
 
        try:
2195
 
            self.init_custom_control_files(tree)
2196
 
            self.update_format(tree)
2197
 
        finally:
2198
 
            tree._control_files.unlock()
2199
 
 
2200
 
    def init_custom_control_files(self, tree):
2201
 
        """Initialize custom control files."""
2202
 
        tree._transport.put_bytes('views', '',
2203
 
            mode=tree.bzrdir._get_file_mode())
2204
 
 
2205
 
    def update_format(self, tree):
2206
 
        """Change the format marker."""
2207
 
        tree._transport.put_bytes('format',
2208
 
            self.target_format.get_format_string(),
2209
 
            mode=tree.bzrdir._get_file_mode())
 
2279
        tree._control_files.put_utf8('format',
 
2280
            self.target_format.get_format_string())