~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

(jelmer) Migrate launchpad plugin configuration to config stacks API.
 (Jelmer Vernooij)

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
1
# Copyright (C) 2007-2011 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
22
22
WorkingTree.open(dir).
23
23
"""
24
24
 
 
25
from __future__ import absolute_import
 
26
 
25
27
from cStringIO import StringIO
26
28
import os
27
29
import sys
28
30
 
29
31
from bzrlib.lazy_import import lazy_import
30
32
lazy_import(globals(), """
31
 
from bisect import bisect_left
32
 
import collections
33
 
from copy import deepcopy
34
33
import errno
35
 
import itertools
36
 
import operator
37
34
import stat
38
 
from time import time
39
 
import warnings
40
35
 
41
 
import bzrlib
42
36
from bzrlib import (
43
37
    bzrdir,
44
38
    cache_utf8,
 
39
    config,
45
40
    conflicts as _mod_conflicts,
46
41
    debug,
47
 
    delta,
48
42
    dirstate,
49
43
    errors,
 
44
    filters as _mod_filters,
50
45
    generate_ids,
51
 
    globbing,
52
 
    hashcache,
53
 
    ignores,
54
 
    merge,
55
46
    osutils,
56
47
    revision as _mod_revision,
57
48
    revisiontree,
58
 
    textui,
59
49
    trace,
60
50
    transform,
61
 
    urlutils,
62
 
    xml5,
63
 
    xml6,
 
51
    views,
64
52
    )
65
 
import bzrlib.branch
66
 
from bzrlib.transport import get_transport
67
 
import bzrlib.ui
68
53
""")
69
54
 
70
 
from bzrlib import symbol_versioning
71
55
from bzrlib.decorators import needs_read_lock, needs_write_lock
72
 
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
73
 
from bzrlib.lockable_files import LockableFiles, TransportLock
 
56
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
57
from bzrlib.lock import LogicalLockResult
 
58
from bzrlib.lockable_files import LockableFiles
74
59
from bzrlib.lockdir import LockDir
75
 
import bzrlib.mutabletree
76
 
from bzrlib.mutabletree import needs_tree_write_lock
 
60
from bzrlib.mutabletree import (
 
61
    MutableTree,
 
62
    needs_tree_write_lock,
 
63
    )
77
64
from bzrlib.osutils import (
78
65
    file_kind,
79
66
    isdir,
80
 
    normpath,
81
67
    pathjoin,
82
 
    rand_chars,
83
68
    realpath,
84
69
    safe_unicode,
85
 
    splitpath,
86
 
    )
87
 
from bzrlib.trace import mutter, note
 
70
    )
 
71
from bzrlib.symbol_versioning import (
 
72
    deprecated_in,
 
73
    deprecated_method,
 
74
    )
88
75
from bzrlib.transport.local import LocalTransport
89
 
from bzrlib.tree import InterTree
90
 
from bzrlib.progress import DummyProgress, ProgressPhase
91
 
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
92
 
from bzrlib.rio import RioReader, rio_file, Stanza
93
 
from bzrlib.symbol_versioning import (deprecated_passed,
94
 
        deprecated_method,
95
 
        deprecated_function,
96
 
        DEPRECATED_PARAMETER,
97
 
        )
98
 
from bzrlib.tree import Tree
99
 
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
100
 
 
101
 
 
102
 
# This is the Windows equivalent of ENOTDIR
103
 
# It is defined in pywin32.winerror, but we don't want a strong dependency for
104
 
# just an error code.
105
 
ERROR_PATH_NOT_FOUND = 3
106
 
ERROR_DIRECTORY = 267
107
 
 
108
 
 
109
 
class WorkingTree4(WorkingTree3):
110
 
    """This is the Format 4 working tree.
111
 
 
112
 
    This differs from WorkingTree3 by:
113
 
     - Having a consolidated internal dirstate, stored in a
114
 
       randomly-accessible sorted file on disk.
115
 
     - Not having a regular inventory attribute.  One can be synthesized 
116
 
       on demand but this is expensive and should be avoided.
117
 
 
118
 
    This is new in bzr 0.15.
119
 
    """
 
76
from bzrlib.tree import (
 
77
    InterTree,
 
78
    InventoryTree,
 
79
    )
 
80
from bzrlib.workingtree import (
 
81
    InventoryWorkingTree,
 
82
    WorkingTree,
 
83
    WorkingTreeFormatMetaDir,
 
84
    )
 
85
 
 
86
 
 
87
class DirStateWorkingTree(InventoryWorkingTree):
120
88
 
121
89
    def __init__(self, basedir,
122
90
                 branch,
132
100
        """
133
101
        self._format = _format
134
102
        self.bzrdir = _bzrdir
135
 
        from bzrlib.trace import note, mutter
136
 
        assert isinstance(basedir, basestring), \
137
 
            "base directory %r is not a string" % basedir
138
103
        basedir = safe_unicode(basedir)
139
 
        mutter("opening working tree %r", basedir)
 
104
        trace.mutter("opening working tree %r", basedir)
140
105
        self._branch = branch
141
 
        assert isinstance(self.branch, bzrlib.branch.Branch), \
142
 
            "branch %r is not a Branch" % self.branch
143
106
        self.basedir = realpath(basedir)
144
107
        # if branch is at our basedir and is a format 6 or less
145
108
        # assume all other formats have their own control files.
146
 
        assert isinstance(_control_files, LockableFiles), \
147
 
            "_control_files must be a LockableFiles, not %r" % _control_files
148
109
        self._control_files = _control_files
 
110
        self._transport = self._control_files._transport
149
111
        self._dirty = None
150
112
        #-------------
151
113
        # during a read or write lock these objects are set, and are
153
115
        self._dirstate = None
154
116
        self._inventory = None
155
117
        #-------------
 
118
        self._setup_directory_is_tree_reference()
 
119
        self._detect_case_handling()
 
120
        self._rules_searcher = None
 
121
        self.views = self._make_views()
 
122
        #--- allow tests to select the dirstate iter_changes implementation
 
123
        self._iter_changes = dirstate._process_entry
156
124
 
157
125
    @needs_tree_write_lock
158
126
    def _add(self, files, ids, kinds):
160
128
        state = self.current_dirstate()
161
129
        for f, file_id, kind in zip(files, ids, kinds):
162
130
            f = f.strip('/')
163
 
            assert '//' not in f
164
 
            assert '..' not in f
165
131
            if self.path2id(f):
166
132
                # special case tree root handling.
167
133
                if f == '' and self.path2id(f) == ROOT_ID:
175
141
            state.add(f, file_id, kind, None, '')
176
142
        self._make_dirty(reset_inventory=True)
177
143
 
 
144
    def _get_check_refs(self):
 
145
        """Return the references needed to perform a check of this tree."""
 
146
        return [('trees', self.last_revision())]
 
147
 
178
148
    def _make_dirty(self, reset_inventory):
179
149
        """Make the tree state dirty.
180
150
 
188
158
    @needs_tree_write_lock
189
159
    def add_reference(self, sub_tree):
190
160
        # use standard implementation, which calls back to self._add
191
 
        # 
 
161
        #
192
162
        # So we don't store the reference_revision in the working dirstate,
193
 
        # it's just recorded at the moment of commit. 
 
163
        # it's just recorded at the moment of commit.
194
164
        self._add_reference(sub_tree)
195
165
 
196
166
    def break_lock(self):
232
202
 
233
203
    def _comparison_data(self, entry, path):
234
204
        kind, executable, stat_value = \
235
 
            WorkingTree3._comparison_data(self, entry, path)
 
205
            WorkingTree._comparison_data(self, entry, path)
236
206
        # it looks like a plain directory, but it's really a reference -- see
237
207
        # also kind()
238
 
        if (self._repo_supports_tree_reference and
239
 
            kind == 'directory' and
240
 
            self._directory_is_tree_reference(path)):
 
208
        if (self._repo_supports_tree_reference and kind == 'directory'
 
209
            and entry is not None and entry.kind == 'tree-reference'):
241
210
            kind = 'tree-reference'
242
211
        return kind, executable, stat_value
243
212
 
245
214
    def commit(self, message=None, revprops=None, *args, **kwargs):
246
215
        # mark the tree as dirty post commit - commit
247
216
        # can change the current versioned list by doing deletes.
248
 
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
217
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
249
218
        self._make_dirty(reset_inventory=True)
250
219
        return result
251
220
 
269
238
            return self._dirstate
270
239
        local_path = self.bzrdir.get_workingtree_transport(None
271
240
            ).local_abspath('dirstate')
272
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
241
        self._dirstate = dirstate.DirState.on_file(local_path,
 
242
            self._sha1_provider(), self._worth_saving_limit())
273
243
        return self._dirstate
274
244
 
275
 
    def _directory_is_tree_reference(self, relpath):
276
 
        # as a special case, if a directory contains control files then 
277
 
        # it's a tree reference, except that the root of the tree is not
278
 
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
279
 
        # TODO: We could ask all the control formats whether they
280
 
        # recognize this directory, but at the moment there's no cheap api
281
 
        # to do that.  Since we probably can only nest bzr checkouts and
282
 
        # they always use this name it's ok for now.  -- mbp 20060306
283
 
        #
284
 
        # FIXME: There is an unhandled case here of a subdirectory
285
 
        # containing .bzr but not a branch; that will probably blow up
286
 
        # when you try to commit it.  It might happen if there is a
287
 
        # checkout in a subdirectory.  This can be avoided by not adding
288
 
        # it.  mbp 20070306
 
245
    def _sha1_provider(self):
 
246
        """A function that returns a SHA1Provider suitable for this tree.
 
247
 
 
248
        :return: None if content filtering is not supported by this tree.
 
249
          Otherwise, a SHA1Provider is returned that sha's the canonical
 
250
          form of files, i.e. after read filters are applied.
 
251
        """
 
252
        if self.supports_content_filtering():
 
253
            return ContentFilterAwareSHA1Provider(self)
 
254
        else:
 
255
            return None
 
256
 
 
257
    def _worth_saving_limit(self):
 
258
        """How many hash changes are ok before we must save the dirstate.
 
259
 
 
260
        :return: an integer. -1 means never save.
 
261
        """
 
262
        conf = self.get_config_stack()
 
263
        return conf.get('bzr.workingtree.worth_saving_limit')
289
264
 
290
265
    def filter_unversioned_files(self, paths):
291
266
        """Filter out paths that are versioned.
324
299
 
325
300
    def _generate_inventory(self):
326
301
        """Create and set self.inventory from the dirstate object.
327
 
        
 
302
 
328
303
        This is relatively expensive: we have to walk the entire dirstate.
329
304
        Ideally we would not, and can deprecate this function.
330
305
        """
334
309
        state._read_dirblocks_if_needed()
335
310
        root_key, current_entry = self._get_entry(path='')
336
311
        current_id = root_key[2]
337
 
        assert current_entry[0][0] == 'd' # directory
 
312
        if not (current_entry[0][0] == 'd'): # directory
 
313
            raise AssertionError(current_entry)
338
314
        inv = Inventory(root_id=current_id)
339
315
        # Turn some things into local variables
340
316
        minikind_to_kind = dirstate.DirState._minikind_to_kind
373
349
                    # add this entry to the parent map.
374
350
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
375
351
                elif kind == 'tree-reference':
376
 
                    assert self._repo_supports_tree_reference, \
377
 
                        "repository of %r " \
378
 
                        "doesn't support tree references " \
379
 
                        "required by entry %r" \
380
 
                        % (self, name)
 
352
                    if not self._repo_supports_tree_reference:
 
353
                        raise errors.UnsupportedOperation(
 
354
                            self._generate_inventory,
 
355
                            self.branch.repository)
381
356
                    inv_entry.reference_revision = link_or_sha1 or None
382
357
                elif kind != 'symlink':
383
358
                    raise AssertionError("unknown kind %r" % kind)
384
359
                # These checks cost us around 40ms on a 55k entry tree
385
 
                assert file_id not in inv_byid, ('file_id %s already in'
386
 
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
387
 
                assert name_unicode not in parent_ie.children
 
360
                if file_id in inv_byid:
 
361
                    raise AssertionError('file_id %s already in'
 
362
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
363
                if name_unicode in parent_ie.children:
 
364
                    raise AssertionError('name %r already in parent'
 
365
                        % (name_unicode,))
388
366
                inv_byid[file_id] = inv_entry
389
367
                parent_ie.children[name_unicode] = inv_entry
390
368
        self._inventory = inv
395
373
        If either file_id or path is supplied, it is used as the key to lookup.
396
374
        If both are supplied, the fastest lookup is used, and an error is
397
375
        raised if they do not both point at the same row.
398
 
        
 
376
 
399
377
        :param file_id: An optional unicode file_id to be looked up.
400
378
        :param path: An optional unicode path to be looked up.
401
379
        :return: The dirstate row tuple for path/file_id, or (None, None)
419
397
        state = self.current_dirstate()
420
398
        if stat_value is None:
421
399
            try:
422
 
                stat_value = os.lstat(file_abspath)
 
400
                stat_value = osutils.lstat(file_abspath)
423
401
            except OSError, e:
424
402
                if e.errno == errno.ENOENT:
425
403
                    return None
426
404
                else:
427
405
                    raise
428
 
        link_or_sha1 = state.update_entry(entry, file_abspath,
429
 
                                          stat_value=stat_value)
 
406
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
407
            stat_value=stat_value)
430
408
        if entry[1][0][0] == 'f':
431
 
            return link_or_sha1
 
409
            if link_or_sha1 is None:
 
410
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
 
411
                try:
 
412
                    sha1 = osutils.sha_file(file_obj)
 
413
                finally:
 
414
                    file_obj.close()
 
415
                self._observed_sha1(file_id, path, (sha1, statvalue))
 
416
                return sha1
 
417
            else:
 
418
                return link_or_sha1
432
419
        return None
433
420
 
434
 
    def _get_inventory(self):
 
421
    def _get_root_inventory(self):
435
422
        """Get the inventory for the tree. This is only valid within a lock."""
436
423
        if 'evil' in debug.debug_flags:
437
424
            trace.mutter_callsite(2,
442
429
        self._generate_inventory()
443
430
        return self._inventory
444
431
 
 
432
    @deprecated_method(deprecated_in((2, 5, 0)))
 
433
    def _get_inventory(self):
 
434
        return self.root_inventory
 
435
 
445
436
    inventory = property(_get_inventory,
446
437
                         doc="Inventory of this Tree")
447
438
 
 
439
    root_inventory = property(_get_root_inventory,
 
440
        "Root inventory of this tree")
 
441
 
448
442
    @needs_read_lock
449
443
    def get_parent_ids(self):
450
444
        """See Tree.get_parent_ids.
451
 
        
 
445
 
452
446
        This implementation requests the ids list from the dirstate file.
453
447
        """
454
448
        return self.current_dirstate().get_parent_ids()
470
464
 
471
465
    def has_id(self, file_id):
472
466
        state = self.current_dirstate()
473
 
        file_id = osutils.safe_file_id(file_id)
474
467
        row, parents = self._get_entry(file_id=file_id)
475
468
        if row is None:
476
469
            return False
477
470
        return osutils.lexists(pathjoin(
478
471
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
479
472
 
 
473
    def has_or_had_id(self, file_id):
 
474
        state = self.current_dirstate()
 
475
        row, parents = self._get_entry(file_id=file_id)
 
476
        return row is not None
 
477
 
480
478
    @needs_read_lock
481
479
    def id2path(self, file_id):
482
480
        "Convert a file-id to a path."
483
 
        file_id = osutils.safe_file_id(file_id)
484
481
        state = self.current_dirstate()
485
482
        entry = self._get_entry(file_id=file_id)
486
483
        if entry == (None, None):
488
485
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
489
486
        return path_utf8.decode('utf8')
490
487
 
491
 
    if not osutils.supports_executable():
492
 
        def is_executable(self, file_id, path=None):
493
 
            """Test if a file is executable or not.
494
 
 
495
 
            Note: The caller is expected to take a read-lock before calling this.
496
 
            """
497
 
            file_id = osutils.safe_file_id(file_id)
 
488
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
489
        entry = self._get_entry(path=path)
 
490
        if entry == (None, None):
 
491
            return False # Missing entries are not executable
 
492
        return entry[1][0][3] # Executable?
 
493
 
 
494
    def is_executable(self, file_id, path=None):
 
495
        """Test if a file is executable or not.
 
496
 
 
497
        Note: The caller is expected to take a read-lock before calling this.
 
498
        """
 
499
        if not self._supports_executable():
498
500
            entry = self._get_entry(file_id=file_id, path=path)
499
501
            if entry == (None, None):
500
502
                return False
501
503
            return entry[1][0][3]
502
 
    else:
503
 
        def is_executable(self, file_id, path=None):
504
 
            """Test if a file is executable or not.
505
 
 
506
 
            Note: The caller is expected to take a read-lock before calling this.
507
 
            """
 
504
        else:
 
505
            self._must_be_locked()
508
506
            if not path:
509
 
                file_id = osutils.safe_file_id(file_id)
510
507
                path = self.id2path(file_id)
511
 
            mode = os.lstat(self.abspath(path)).st_mode
 
508
            mode = osutils.lstat(self.abspath(path)).st_mode
512
509
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
513
510
 
 
511
    def all_file_ids(self):
 
512
        """See Tree.iter_all_file_ids"""
 
513
        self._must_be_locked()
 
514
        result = set()
 
515
        for key, tree_details in self.current_dirstate()._iter_entries():
 
516
            if tree_details[0][0] in ('a', 'r'): # relocated
 
517
                continue
 
518
            result.add(key[2])
 
519
        return result
 
520
 
514
521
    @needs_read_lock
515
522
    def __iter__(self):
516
523
        """Iterate through file_ids for this tree.
529
536
        return iter(result)
530
537
 
531
538
    def iter_references(self):
 
539
        if not self._repo_supports_tree_reference:
 
540
            # When the repo doesn't support references, we will have nothing to
 
541
            # return
 
542
            return
532
543
        for key, tree_details in self.current_dirstate()._iter_entries():
533
544
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
534
545
                # not relevant to the working tree
536
547
            if not key[1]:
537
548
                # the root is not a reference.
538
549
                continue
539
 
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
550
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
540
551
            try:
541
 
                if self._kind(path) == 'tree-reference':
542
 
                    yield path, key[2]
 
552
                if self._kind(relpath) == 'tree-reference':
 
553
                    yield relpath, key[2]
543
554
            except errors.NoSuchFile:
544
555
                # path is missing on disk.
545
556
                continue
546
557
 
 
558
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
 
559
        """See MutableTree._observed_sha1."""
 
560
        state = self.current_dirstate()
 
561
        entry = self._get_entry(file_id=file_id, path=path)
 
562
        state._observed_sha1(entry, sha1, statvalue)
 
563
 
547
564
    def kind(self, file_id):
548
565
        """Return the kind of a file.
549
566
 
553
570
        Note: The caller is expected to take a read-lock before calling this.
554
571
        """
555
572
        relpath = self.id2path(file_id)
556
 
        assert relpath != None, \
557
 
            "path for id {%s} is None!" % file_id
 
573
        if relpath is None:
 
574
            raise AssertionError(
 
575
                "path for id {%s} is None!" % file_id)
558
576
        return self._kind(relpath)
559
577
 
560
578
    def _kind(self, relpath):
561
579
        abspath = self.abspath(relpath)
562
580
        kind = file_kind(abspath)
563
 
        if (self._repo_supports_tree_reference and
564
 
            kind == 'directory' and
565
 
            self._directory_is_tree_reference(relpath)):
566
 
            kind = 'tree-reference'
 
581
        if (self._repo_supports_tree_reference and kind == 'directory'):
 
582
            entry = self._get_entry(path=relpath)
 
583
            if entry[1] is not None:
 
584
                if entry[1][0][0] == 't':
 
585
                    kind = 'tree-reference'
567
586
        return kind
568
587
 
569
588
    @needs_read_lock
576
595
            return _mod_revision.NULL_REVISION
577
596
 
578
597
    def lock_read(self):
579
 
        """See Branch.lock_read, and WorkingTree.unlock."""
 
598
        """See Branch.lock_read, and WorkingTree.unlock.
 
599
 
 
600
        :return: A bzrlib.lock.LogicalLockResult.
 
601
        """
580
602
        self.branch.lock_read()
581
603
        try:
582
604
            self._control_files.lock_read()
595
617
        except:
596
618
            self.branch.unlock()
597
619
            raise
 
620
        return LogicalLockResult(self.unlock)
598
621
 
599
622
    def _lock_self_write(self):
600
623
        """This should be called after the branch is locked."""
615
638
        except:
616
639
            self.branch.unlock()
617
640
            raise
 
641
        return LogicalLockResult(self.unlock)
618
642
 
619
643
    def lock_tree_write(self):
620
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
644
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
 
645
 
 
646
        :return: A bzrlib.lock.LogicalLockResult.
 
647
        """
621
648
        self.branch.lock_read()
622
 
        self._lock_self_write()
 
649
        return self._lock_self_write()
623
650
 
624
651
    def lock_write(self):
625
 
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
652
        """See MutableTree.lock_write, and WorkingTree.unlock.
 
653
 
 
654
        :return: A bzrlib.lock.LogicalLockResult.
 
655
        """
626
656
        self.branch.lock_write()
627
 
        self._lock_self_write()
 
657
        return self._lock_self_write()
628
658
 
629
659
    @needs_tree_write_lock
630
660
    def move(self, from_paths, to_dir, after=False):
632
662
        result = []
633
663
        if not from_paths:
634
664
            return result
635
 
 
636
665
        state = self.current_dirstate()
637
 
 
638
 
        assert not isinstance(from_paths, basestring)
 
666
        if isinstance(from_paths, basestring):
 
667
            raise ValueError()
639
668
        to_dir_utf8 = to_dir.encode('utf8')
640
669
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
641
670
        id_index = state._get_id_index()
662
691
 
663
692
        if self._inventory is not None:
664
693
            update_inventory = True
665
 
            inv = self.inventory
 
694
            inv = self.root_inventory
 
695
            to_dir_id = to_entry[0][2]
666
696
            to_dir_ie = inv[to_dir_id]
667
 
            to_dir_id = to_entry[0][2]
668
697
        else:
669
698
            update_inventory = False
670
699
 
701
730
            from_entry = self._get_entry(path=from_rel)
702
731
            if from_entry == (None, None):
703
732
                raise errors.BzrMoveFailedError(from_rel,to_dir,
704
 
                    errors.NotVersionedError(path=str(from_rel)))
 
733
                    errors.NotVersionedError(path=from_rel))
705
734
 
706
735
            from_id = from_entry[0][2]
707
736
            to_rel = pathjoin(to_dir, from_tail)
734
763
                if from_missing: # implicitly just update our path mapping
735
764
                    move_file = False
736
765
                elif not after:
737
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
738
 
                        extra="(Use --after to update the Bazaar id)")
 
766
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
739
767
 
740
768
            rollbacks = []
741
769
            def rollback_rename():
797
825
                if minikind == 'd':
798
826
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
799
827
                        """Recursively update all entries in this dirblock."""
800
 
                        assert from_dir != '', "renaming root not supported"
 
828
                        if from_dir == '':
 
829
                            raise AssertionError("renaming root not supported")
801
830
                        from_key = (from_dir, '')
802
831
                        from_block_idx, present = \
803
832
                            state._find_block_index_from_key(from_key)
816
845
 
817
846
                        # Grab a copy since move_one may update the list.
818
847
                        for entry in from_block[1][:]:
819
 
                            assert entry[0][0] == from_dir
 
848
                            if not (entry[0][0] == from_dir):
 
849
                                raise AssertionError()
820
850
                            cur_details = entry[1][0]
821
851
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
822
852
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
847
877
                rollback_rename()
848
878
                raise
849
879
            result.append((from_rel, to_rel))
850
 
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
880
            state._mark_modified()
851
881
            self._make_dirty(reset_inventory=False)
852
882
 
853
883
        return result
881
911
        for tree in trees:
882
912
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
883
913
                parents):
884
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
914
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
915
                    trees, require_versioned)
885
916
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
886
917
        # -- make all paths utf8 --
887
918
        paths_utf8 = set()
945
976
                    all_versioned = False
946
977
                    break
947
978
            if not all_versioned:
948
 
                raise errors.PathsNotVersionedError(paths)
 
979
                raise errors.PathsNotVersionedError(
 
980
                    [p.decode('utf-8') for p in paths])
949
981
        # -- remove redundancy in supplied paths to prevent over-scanning --
950
 
        search_paths = set()
951
 
        for path in paths:
952
 
            other_paths = paths.difference(set([path]))
953
 
            if not osutils.is_inside_any(other_paths, path):
954
 
                # this is a top level path, we must check it.
955
 
                search_paths.add(path)
956
 
        # sketch: 
 
982
        search_paths = osutils.minimum_path_selection(paths)
 
983
        # sketch:
957
984
        # for all search_indexs in each path at or under each element of
958
985
        # search_paths, if the detail is relocated: add the id, and add the
959
986
        # relocated path as one to search if its not searched already. If the
1005
1032
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1006
1033
            for dir_name in split_paths:
1007
1034
                if dir_name not in found_dir_names:
1008
 
                    raise errors.PathsNotVersionedError(paths)
 
1035
                    raise errors.PathsNotVersionedError(
 
1036
                        [p.decode('utf-8') for p in paths])
1009
1037
 
1010
1038
        for dir_name_id, trees_info in found.iteritems():
1011
1039
            for index in search_indexes:
1015
1043
 
1016
1044
    def read_working_inventory(self):
1017
1045
        """Read the working inventory.
1018
 
        
 
1046
 
1019
1047
        This is a meaningless operation for dirstate, but we obey it anyhow.
1020
1048
        """
1021
 
        return self.inventory
 
1049
        return self.root_inventory
1022
1050
 
1023
1051
    @needs_read_lock
1024
1052
    def revision_tree(self, revision_id):
1026
1054
 
1027
1055
        WorkingTree4 supplies revision_trees for any basis tree.
1028
1056
        """
1029
 
        revision_id = osutils.safe_revision_id(revision_id)
1030
1057
        dirstate = self.current_dirstate()
1031
1058
        parent_ids = dirstate.get_parent_ids()
1032
1059
        if revision_id not in parent_ids:
1039
1066
    @needs_tree_write_lock
1040
1067
    def set_last_revision(self, new_revision):
1041
1068
        """Change the last revision in the working tree."""
1042
 
        new_revision = osutils.safe_revision_id(new_revision)
1043
1069
        parents = self.get_parent_ids()
1044
 
        if new_revision in (NULL_REVISION, None):
1045
 
            assert len(parents) < 2, (
1046
 
                "setting the last parent to none with a pending merge is "
1047
 
                "unsupported.")
 
1070
        if new_revision in (_mod_revision.NULL_REVISION, None):
 
1071
            if len(parents) >= 2:
 
1072
                raise AssertionError(
 
1073
                    "setting the last parent to none with a pending merge is "
 
1074
                    "unsupported.")
1048
1075
            self.set_parent_ids([])
1049
1076
        else:
1050
1077
            self.set_parent_ids([new_revision] + parents[1:],
1053
1080
    @needs_tree_write_lock
1054
1081
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1055
1082
        """Set the parent ids to revision_ids.
1056
 
        
 
1083
 
1057
1084
        See also set_parent_trees. This api will try to retrieve the tree data
1058
1085
        for each element of revision_ids from the trees repository. If you have
1059
1086
        tree data already available, it is more efficient to use
1063
1090
        :param revision_ids: The revision_ids to set as the parent ids of this
1064
1091
            working tree. Any of these may be ghosts.
1065
1092
        """
1066
 
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1067
1093
        trees = []
1068
1094
        for revision_id in revision_ids:
1069
1095
            try:
1092
1118
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1093
1119
        real_trees = []
1094
1120
        ghosts = []
 
1121
 
 
1122
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1123
        graph = self.branch.repository.get_graph()
 
1124
        heads = graph.heads(parent_ids)
 
1125
        accepted_revisions = set()
 
1126
 
1095
1127
        # convert absent trees to the null tree, which we convert back to
1096
1128
        # missing on access.
1097
1129
        for rev_id, tree in parents_list:
1098
 
            rev_id = osutils.safe_revision_id(rev_id)
 
1130
            if len(accepted_revisions) > 0:
 
1131
                # we always accept the first tree
 
1132
                if rev_id in accepted_revisions or rev_id not in heads:
 
1133
                    # We have already included either this tree, or its
 
1134
                    # descendent, so we skip it.
 
1135
                    continue
1099
1136
            _mod_revision.check_not_reserved_id(rev_id)
1100
1137
            if tree is not None:
1101
1138
                real_trees.append((rev_id, tree))
1102
1139
            else:
1103
1140
                real_trees.append((rev_id,
1104
 
                    self.branch.repository.revision_tree(None)))
 
1141
                    self.branch.repository.revision_tree(
 
1142
                        _mod_revision.NULL_REVISION)))
1105
1143
                ghosts.append(rev_id)
1106
 
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1144
            accepted_revisions.add(rev_id)
 
1145
        updated = False
 
1146
        if (len(real_trees) == 1
 
1147
            and not ghosts
 
1148
            and self.branch.repository._format.fast_deltas
 
1149
            and isinstance(real_trees[0][1],
 
1150
                revisiontree.InventoryRevisionTree)
 
1151
            and self.get_parent_ids()):
 
1152
            rev_id, rev_tree = real_trees[0]
 
1153
            basis_id = self.get_parent_ids()[0]
 
1154
            # There are times when basis_tree won't be in
 
1155
            # self.branch.repository, (switch, for example)
 
1156
            try:
 
1157
                basis_tree = self.branch.repository.revision_tree(basis_id)
 
1158
            except errors.NoSuchRevision:
 
1159
                # Fall back to the set_parent_trees(), since we can't use
 
1160
                # _make_delta if we can't get the RevisionTree
 
1161
                pass
 
1162
            else:
 
1163
                delta = rev_tree.root_inventory._make_delta(
 
1164
                    basis_tree.root_inventory)
 
1165
                dirstate.update_basis_by_delta(delta, rev_id)
 
1166
                updated = True
 
1167
        if not updated:
 
1168
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1107
1169
        self._make_dirty(reset_inventory=False)
1108
1170
 
1109
1171
    def _set_root_id(self, file_id):
1113
1175
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1114
1176
            self._make_dirty(reset_inventory=True)
1115
1177
 
 
1178
    def _sha_from_stat(self, path, stat_result):
 
1179
        """Get a sha digest from the tree's stat cache.
 
1180
 
 
1181
        The default implementation assumes no stat cache is present.
 
1182
 
 
1183
        :param path: The path.
 
1184
        :param stat_result: The stat result being looked up.
 
1185
        """
 
1186
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1187
 
1116
1188
    @needs_read_lock
1117
1189
    def supports_tree_reference(self):
1118
1190
        return self._repo_supports_tree_reference
1119
1191
 
1120
1192
    def unlock(self):
1121
1193
        """Unlock in format 4 trees needs to write the entire dirstate."""
1122
 
        # do non-implementation specific cleanup
1123
 
        self._cleanup()
1124
 
 
1125
1194
        if self._control_files._lock_count == 1:
 
1195
            # do non-implementation specific cleanup
 
1196
            self._cleanup()
 
1197
 
1126
1198
            # eventually we should do signature checking during read locks for
1127
1199
            # dirstate updates.
1128
1200
            if self._control_files._lock_mode == 'w':
1158
1230
            return
1159
1231
        state = self.current_dirstate()
1160
1232
        state._read_dirblocks_if_needed()
1161
 
        ids_to_unversion = set()
1162
 
        for file_id in file_ids:
1163
 
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1233
        ids_to_unversion = set(file_ids)
1164
1234
        paths_to_unversion = set()
1165
1235
        # sketch:
1166
1236
        # check if the root is to be unversioned, if so, assert for now.
1193
1263
                # just forget the whole block.
1194
1264
                entry_index = 0
1195
1265
                while entry_index < len(block[1]):
1196
 
                    # Mark this file id as having been removed
1197
1266
                    entry = block[1][entry_index]
1198
 
                    ids_to_unversion.discard(entry[0][2])
1199
 
                    if (entry[1][0][0] == 'a'
1200
 
                        or not state._make_absent(entry)):
 
1267
                    if entry[1][0][0] in 'ar':
 
1268
                        # don't remove absent or renamed entries
1201
1269
                        entry_index += 1
 
1270
                    else:
 
1271
                        # Mark this file id as having been removed
 
1272
                        ids_to_unversion.discard(entry[0][2])
 
1273
                        if not state._make_absent(entry):
 
1274
                            # The block has not shrunk.
 
1275
                            entry_index += 1
1202
1276
                # go to the next block. (At the moment we dont delete empty
1203
1277
                # dirblocks)
1204
1278
                block_index += 1
1225
1299
        # have to change the legacy inventory too.
1226
1300
        if self._inventory is not None:
1227
1301
            for file_id in file_ids:
1228
 
                self._inventory.remove_recursive_id(file_id)
 
1302
                if self._inventory.has_id(file_id):
 
1303
                    self._inventory.remove_recursive_id(file_id)
 
1304
 
 
1305
    @needs_tree_write_lock
 
1306
    def rename_one(self, from_rel, to_rel, after=False):
 
1307
        """See WorkingTree.rename_one"""
 
1308
        self.flush()
 
1309
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
 
1310
 
 
1311
    @needs_tree_write_lock
 
1312
    def apply_inventory_delta(self, changes):
 
1313
        """See MutableTree.apply_inventory_delta"""
 
1314
        state = self.current_dirstate()
 
1315
        state.update_by_delta(changes)
 
1316
        self._make_dirty(reset_inventory=True)
 
1317
 
 
1318
    def update_basis_by_delta(self, new_revid, delta):
 
1319
        """See MutableTree.update_basis_by_delta."""
 
1320
        if self.last_revision() == new_revid:
 
1321
            raise AssertionError()
 
1322
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1229
1323
 
1230
1324
    @needs_read_lock
1231
1325
    def _validate(self):
1234
1328
    @needs_tree_write_lock
1235
1329
    def _write_inventory(self, inv):
1236
1330
        """Write inventory as the current inventory."""
1237
 
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
1238
 
        self.current_dirstate().set_state_from_inventory(inv)
1239
 
        self._make_dirty(reset_inventory=False)
1240
 
        if self._inventory is not None:
 
1331
        if self._dirty:
 
1332
            raise AssertionError("attempting to write an inventory when the "
 
1333
                "dirstate is dirty will lose pending changes")
 
1334
        had_inventory = self._inventory is not None
 
1335
        # Setting self._inventory = None forces the dirstate to regenerate the
 
1336
        # working inventory. We do this because self.inventory may be inv, or
 
1337
        # may have been modified, and either case would prevent a clean delta
 
1338
        # being created.
 
1339
        self._inventory = None
 
1340
        # generate a delta,
 
1341
        delta = inv._make_delta(self.root_inventory)
 
1342
        # and apply it.
 
1343
        self.apply_inventory_delta(delta)
 
1344
        if had_inventory:
1241
1345
            self._inventory = inv
1242
1346
        self.flush()
1243
1347
 
1244
 
 
1245
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1246
 
    """The first consolidated dirstate working tree format.
1247
 
 
1248
 
    This format:
1249
 
        - exists within a metadir controlling .bzr
1250
 
        - includes an explicit version marker for the workingtree control
1251
 
          files, separate from the BzrDir format
1252
 
        - modifies the hash cache format
1253
 
        - is new in bzr 0.15
1254
 
        - uses a LockDir to guard access to it.
1255
 
    """
1256
 
 
1257
 
    upgrade_recommended = False
1258
 
 
1259
 
    def get_format_string(self):
1260
 
        """See WorkingTreeFormat.get_format_string()."""
1261
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1262
 
 
1263
 
    def get_format_description(self):
1264
 
        """See WorkingTreeFormat.get_format_description()."""
1265
 
        return "Working tree format 4"
1266
 
 
1267
 
    def initialize(self, a_bzrdir, revision_id=None):
 
1348
    @needs_tree_write_lock
 
1349
    def reset_state(self, revision_ids=None):
 
1350
        """Reset the state of the working tree.
 
1351
 
 
1352
        This does a hard-reset to a last-known-good state. This is a way to
 
1353
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
 
1354
        """
 
1355
        if revision_ids is None:
 
1356
            revision_ids = self.get_parent_ids()
 
1357
        if not revision_ids:
 
1358
            base_tree = self.branch.repository.revision_tree(
 
1359
                _mod_revision.NULL_REVISION)
 
1360
            trees = []
 
1361
        else:
 
1362
            trees = zip(revision_ids,
 
1363
                        self.branch.repository.revision_trees(revision_ids))
 
1364
            base_tree = trees[0][1]
 
1365
        state = self.current_dirstate()
 
1366
        # We don't support ghosts yet
 
1367
        state.set_state_from_scratch(base_tree.root_inventory, trees, [])
 
1368
 
 
1369
 
 
1370
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
 
1371
 
 
1372
    def __init__(self, tree):
 
1373
        self.tree = tree
 
1374
 
 
1375
    def sha1(self, abspath):
 
1376
        """See dirstate.SHA1Provider.sha1()."""
 
1377
        filters = self.tree._content_filter_stack(
 
1378
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1379
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
 
1380
 
 
1381
    def stat_and_sha1(self, abspath):
 
1382
        """See dirstate.SHA1Provider.stat_and_sha1()."""
 
1383
        filters = self.tree._content_filter_stack(
 
1384
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1385
        file_obj = file(abspath, 'rb', 65000)
 
1386
        try:
 
1387
            statvalue = os.fstat(file_obj.fileno())
 
1388
            if filters:
 
1389
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
 
1390
            sha1 = osutils.size_sha_file(file_obj)[1]
 
1391
        finally:
 
1392
            file_obj.close()
 
1393
        return statvalue, sha1
 
1394
 
 
1395
 
 
1396
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
 
1397
    """Dirstate working tree that supports content filtering.
 
1398
 
 
1399
    The dirstate holds the hash and size of the canonical form of the file, 
 
1400
    and most methods must return that.
 
1401
    """
 
1402
 
 
1403
    def _file_content_summary(self, path, stat_result):
 
1404
        # This is to support the somewhat obsolete path_content_summary method
 
1405
        # with content filtering: see
 
1406
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
 
1407
        #
 
1408
        # If the dirstate cache is up to date and knows the hash and size,
 
1409
        # return that.
 
1410
        # Otherwise if there are no content filters, return the on-disk size
 
1411
        # and leave the hash blank.
 
1412
        # Otherwise, read and filter the on-disk file and use its size and
 
1413
        # hash.
 
1414
        #
 
1415
        # The dirstate doesn't store the size of the canonical form so we
 
1416
        # can't trust it for content-filtered trees.  We just return None.
 
1417
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
 
1418
        executable = self._is_executable_from_path_and_stat(path, stat_result)
 
1419
        return ('file', None, executable, dirstate_sha1)
 
1420
 
 
1421
 
 
1422
class WorkingTree4(DirStateWorkingTree):
 
1423
    """This is the Format 4 working tree.
 
1424
 
 
1425
    This differs from WorkingTree by:
 
1426
     - Having a consolidated internal dirstate, stored in a
 
1427
       randomly-accessible sorted file on disk.
 
1428
     - Not having a regular inventory attribute.  One can be synthesized
 
1429
       on demand but this is expensive and should be avoided.
 
1430
 
 
1431
    This is new in bzr 0.15.
 
1432
    """
 
1433
 
 
1434
 
 
1435
class WorkingTree5(ContentFilteringDirStateWorkingTree):
 
1436
    """This is the Format 5 working tree.
 
1437
 
 
1438
    This differs from WorkingTree4 by:
 
1439
     - Supporting content filtering.
 
1440
 
 
1441
    This is new in bzr 1.11.
 
1442
    """
 
1443
 
 
1444
 
 
1445
class WorkingTree6(ContentFilteringDirStateWorkingTree):
 
1446
    """This is the Format 6 working tree.
 
1447
 
 
1448
    This differs from WorkingTree5 by:
 
1449
     - Supporting a current view that may mask the set of files in a tree
 
1450
       impacted by most user operations.
 
1451
 
 
1452
    This is new in bzr 1.14.
 
1453
    """
 
1454
 
 
1455
    def _make_views(self):
 
1456
        return views.PathBasedViews(self)
 
1457
 
 
1458
 
 
1459
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
 
1460
 
 
1461
    missing_parent_conflicts = True
 
1462
 
 
1463
    supports_versioned_directories = True
 
1464
 
 
1465
    _lock_class = LockDir
 
1466
    _lock_file_name = 'lock'
 
1467
 
 
1468
    def _open_control_files(self, a_bzrdir):
 
1469
        transport = a_bzrdir.get_workingtree_transport(None)
 
1470
        return LockableFiles(transport, self._lock_file_name,
 
1471
                             self._lock_class)
 
1472
 
 
1473
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1474
                   accelerator_tree=None, hardlink=False):
1268
1475
        """See WorkingTreeFormat.initialize().
1269
1476
 
1270
1477
        :param revision_id: allows creating a working tree at a different
1271
 
        revision than the branch is at.
 
1478
            revision than the branch is at.
 
1479
        :param accelerator_tree: A tree which can be used for retrieving file
 
1480
            contents more quickly than the revision tree, i.e. a workingtree.
 
1481
            The revision tree will be used for cases where accelerator_tree's
 
1482
            content is different.
 
1483
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1484
            where possible.
1272
1485
 
1273
1486
        These trees get an initial random root id, if their repository supports
1274
1487
        rich root data, TREE_ROOT otherwise.
1275
1488
        """
1276
 
        revision_id = osutils.safe_revision_id(revision_id)
1277
1489
        if not isinstance(a_bzrdir.transport, LocalTransport):
1278
1490
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1279
1491
        transport = a_bzrdir.get_workingtree_transport(self)
1280
1492
        control_files = self._open_control_files(a_bzrdir)
1281
1493
        control_files.create_lock()
1282
1494
        control_files.lock_write()
1283
 
        control_files.put_utf8('format', self.get_format_string())
1284
 
        branch = a_bzrdir.open_branch()
 
1495
        transport.put_bytes('format', self.as_string(),
 
1496
            mode=a_bzrdir._get_file_mode())
 
1497
        if from_branch is not None:
 
1498
            branch = from_branch
 
1499
        else:
 
1500
            branch = a_bzrdir.open_branch()
1285
1501
        if revision_id is None:
1286
1502
            revision_id = branch.last_revision()
1287
1503
        local_path = transport.local_abspath('dirstate')
1289
1505
        state = dirstate.DirState.initialize(local_path)
1290
1506
        state.unlock()
1291
1507
        del state
1292
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1508
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1293
1509
                         branch,
1294
1510
                         _format=self,
1295
1511
                         _bzrdir=a_bzrdir,
1297
1513
        wt._new_tree()
1298
1514
        wt.lock_tree_write()
1299
1515
        try:
1300
 
            if revision_id in (None, NULL_REVISION):
 
1516
            self._init_custom_control_files(wt)
 
1517
            if revision_id in (None, _mod_revision.NULL_REVISION):
1301
1518
                if branch.repository.supports_rich_root():
1302
1519
                    wt._set_root_id(generate_ids.gen_root_id())
1303
1520
                else:
1304
1521
                    wt._set_root_id(ROOT_ID)
1305
1522
                wt.flush()
1306
 
            wt.set_last_revision(revision_id)
1307
 
            wt.flush()
1308
 
            basis = wt.basis_tree()
 
1523
            basis = None
 
1524
            # frequently, we will get here due to branching.  The accelerator
 
1525
            # tree will be the tree from the branch, so the desired basis
 
1526
            # tree will often be a parent of the accelerator tree.
 
1527
            if accelerator_tree is not None:
 
1528
                try:
 
1529
                    basis = accelerator_tree.revision_tree(revision_id)
 
1530
                except errors.NoSuchRevision:
 
1531
                    pass
 
1532
            if basis is None:
 
1533
                basis = branch.repository.revision_tree(revision_id)
 
1534
            if revision_id == _mod_revision.NULL_REVISION:
 
1535
                parents_list = []
 
1536
            else:
 
1537
                parents_list = [(revision_id, basis)]
1309
1538
            basis.lock_read()
1310
 
            # if the basis has a root id we have to use that; otherwise we use
1311
 
            # a new random one
1312
 
            basis_root_id = basis.get_root_id()
1313
 
            if basis_root_id is not None:
1314
 
                wt._set_root_id(basis_root_id)
 
1539
            try:
 
1540
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1315
1541
                wt.flush()
1316
 
            transform.build_tree(basis, wt)
1317
 
            basis.unlock()
 
1542
                # if the basis has a root id we have to use that; otherwise we
 
1543
                # use a new random one
 
1544
                basis_root_id = basis.get_root_id()
 
1545
                if basis_root_id is not None:
 
1546
                    wt._set_root_id(basis_root_id)
 
1547
                    wt.flush()
 
1548
                if wt.supports_content_filtering():
 
1549
                    # The original tree may not have the same content filters
 
1550
                    # applied so we can't safely build the inventory delta from
 
1551
                    # the source tree.
 
1552
                    delta_from_tree = False
 
1553
                else:
 
1554
                    delta_from_tree = True
 
1555
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1556
                # because wt4.apply_inventory_delta does not mutate the input
 
1557
                # inventory entries.
 
1558
                transform.build_tree(basis, wt, accelerator_tree,
 
1559
                                     hardlink=hardlink,
 
1560
                                     delta_from_tree=delta_from_tree)
 
1561
                for hook in MutableTree.hooks['post_build_tree']:
 
1562
                    hook(wt)
 
1563
            finally:
 
1564
                basis.unlock()
1318
1565
        finally:
1319
1566
            control_files.unlock()
1320
1567
            wt.unlock()
1321
1568
        return wt
1322
1569
 
 
1570
    def _init_custom_control_files(self, wt):
 
1571
        """Subclasses with custom control files should override this method.
 
1572
 
 
1573
        The working tree and control files are locked for writing when this
 
1574
        method is called.
 
1575
 
 
1576
        :param wt: the WorkingTree object
 
1577
        """
 
1578
 
 
1579
    def open(self, a_bzrdir, _found=False):
 
1580
        """Return the WorkingTree object for a_bzrdir
 
1581
 
 
1582
        _found is a private parameter, do not use it. It is used to indicate
 
1583
               if format probing has already been done.
 
1584
        """
 
1585
        if not _found:
 
1586
            # we are being called directly and must probe.
 
1587
            raise NotImplementedError
 
1588
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1589
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1590
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
 
1591
        return wt
 
1592
 
1323
1593
    def _open(self, a_bzrdir, control_files):
1324
1594
        """Open the tree itself.
1325
1595
 
1326
1596
        :param a_bzrdir: the dir for the tree.
1327
1597
        :param control_files: the control files for the tree.
1328
1598
        """
1329
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1599
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1330
1600
                           branch=a_bzrdir.open_branch(),
1331
1601
                           _format=self,
1332
1602
                           _bzrdir=a_bzrdir,
1333
1603
                           _control_files=control_files)
1334
1604
 
1335
1605
    def __get_matchingbzrdir(self):
 
1606
        return self._get_matchingbzrdir()
 
1607
 
 
1608
    def _get_matchingbzrdir(self):
 
1609
        """Overrideable method to get a bzrdir for testing."""
1336
1610
        # please test against something that will let us do tree references
1337
1611
        return bzrdir.format_registry.make_bzrdir(
1338
 
            'dirstate-with-subtree')
 
1612
            'development-subtree')
1339
1613
 
1340
1614
    _matchingbzrdir = property(__get_matchingbzrdir)
1341
1615
 
1342
1616
 
1343
 
class DirStateRevisionTree(Tree):
1344
 
    """A revision tree pulling the inventory from a dirstate."""
 
1617
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
 
1618
    """The first consolidated dirstate working tree format.
 
1619
 
 
1620
    This format:
 
1621
        - exists within a metadir controlling .bzr
 
1622
        - includes an explicit version marker for the workingtree control
 
1623
          files, separate from the ControlDir format
 
1624
        - modifies the hash cache format
 
1625
        - is new in bzr 0.15
 
1626
        - uses a LockDir to guard access to it.
 
1627
    """
 
1628
 
 
1629
    upgrade_recommended = False
 
1630
 
 
1631
    _tree_class = WorkingTree4
 
1632
 
 
1633
    @classmethod
 
1634
    def get_format_string(cls):
 
1635
        """See WorkingTreeFormat.get_format_string()."""
 
1636
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1637
 
 
1638
    def get_format_description(self):
 
1639
        """See WorkingTreeFormat.get_format_description()."""
 
1640
        return "Working tree format 4"
 
1641
 
 
1642
 
 
1643
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
 
1644
    """WorkingTree format supporting content filtering.
 
1645
    """
 
1646
 
 
1647
    upgrade_recommended = False
 
1648
 
 
1649
    _tree_class = WorkingTree5
 
1650
 
 
1651
    @classmethod
 
1652
    def get_format_string(cls):
 
1653
        """See WorkingTreeFormat.get_format_string()."""
 
1654
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
 
1655
 
 
1656
    def get_format_description(self):
 
1657
        """See WorkingTreeFormat.get_format_description()."""
 
1658
        return "Working tree format 5"
 
1659
 
 
1660
    def supports_content_filtering(self):
 
1661
        return True
 
1662
 
 
1663
 
 
1664
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
 
1665
    """WorkingTree format supporting views.
 
1666
    """
 
1667
 
 
1668
    upgrade_recommended = False
 
1669
 
 
1670
    _tree_class = WorkingTree6
 
1671
 
 
1672
    @classmethod
 
1673
    def get_format_string(cls):
 
1674
        """See WorkingTreeFormat.get_format_string()."""
 
1675
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
 
1676
 
 
1677
    def get_format_description(self):
 
1678
        """See WorkingTreeFormat.get_format_description()."""
 
1679
        return "Working tree format 6"
 
1680
 
 
1681
    def _init_custom_control_files(self, wt):
 
1682
        """Subclasses with custom control files should override this method."""
 
1683
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
 
1684
 
 
1685
    def supports_content_filtering(self):
 
1686
        return True
 
1687
 
 
1688
    def supports_views(self):
 
1689
        return True
 
1690
 
 
1691
 
 
1692
class DirStateRevisionTree(InventoryTree):
 
1693
    """A revision tree pulling the inventory from a dirstate.
 
1694
    
 
1695
    Note that this is one of the historical (ie revision) trees cached in the
 
1696
    dirstate for easy access, not the workingtree.
 
1697
    """
1345
1698
 
1346
1699
    def __init__(self, dirstate, revision_id, repository):
1347
1700
        self._dirstate = dirstate
1348
 
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1701
        self._revision_id = revision_id
1349
1702
        self._repository = repository
1350
1703
        self._inventory = None
1351
1704
        self._locked = 0
1352
1705
        self._dirstate_locked = False
 
1706
        self._repo_supports_tree_reference = getattr(
 
1707
            repository._format, "supports_tree_reference",
 
1708
            False)
1353
1709
 
1354
1710
    def __repr__(self):
1355
1711
        return "<%s of %s in %s>" % \
1358
1714
    def annotate_iter(self, file_id,
1359
1715
                      default_revision=_mod_revision.CURRENT_REVISION):
1360
1716
        """See Tree.annotate_iter"""
1361
 
        w = self._get_weave(file_id)
1362
 
        return w.annotate_iter(self.inventory[file_id].revision)
 
1717
        text_key = (file_id, self.get_file_revision(file_id))
 
1718
        annotations = self._repository.texts.annotate(text_key)
 
1719
        return [(key[-1], line) for (key, line) in annotations]
1363
1720
 
1364
 
    def _get_ancestors(self, default_revision):
1365
 
        return set(self._repository.get_ancestry(self._revision_id,
1366
 
                                                 topo_sorted=False))
1367
1721
    def _comparison_data(self, entry, path):
1368
1722
        """See Tree._comparison_data."""
1369
1723
        if entry is None:
1386
1740
    def get_root_id(self):
1387
1741
        return self.path2id('')
1388
1742
 
 
1743
    def id2path(self, file_id):
 
1744
        "Convert a file-id to a path."
 
1745
        entry = self._get_entry(file_id=file_id)
 
1746
        if entry == (None, None):
 
1747
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1748
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1749
        return path_utf8.decode('utf8')
 
1750
 
 
1751
    def iter_references(self):
 
1752
        if not self._repo_supports_tree_reference:
 
1753
            # When the repo doesn't support references, we will have nothing to
 
1754
            # return
 
1755
            return iter([])
 
1756
        # Otherwise, fall back to the default implementation
 
1757
        return super(DirStateRevisionTree, self).iter_references()
 
1758
 
1389
1759
    def _get_parent_index(self):
1390
1760
        """Return the index in the dirstate referenced by this tree."""
1391
1761
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1396
1766
        If either file_id or path is supplied, it is used as the key to lookup.
1397
1767
        If both are supplied, the fastest lookup is used, and an error is
1398
1768
        raised if they do not both point at the same row.
1399
 
        
 
1769
 
1400
1770
        :param file_id: An optional unicode file_id to be looked up.
1401
1771
        :param path: An optional unicode path to be looked up.
1402
1772
        :return: The dirstate row tuple for path/file_id, or (None, None)
1403
1773
        """
1404
1774
        if file_id is None and path is None:
1405
1775
            raise errors.BzrError('must supply file_id or path')
1406
 
        file_id = osutils.safe_file_id(file_id)
1407
1776
        if path is not None:
1408
1777
            path = path.encode('utf8')
1409
1778
        parent_index = self._get_parent_index()
1417
1786
 
1418
1787
        This is relatively expensive: we have to walk the entire dirstate.
1419
1788
        """
1420
 
        assert self._locked, 'cannot generate inventory of an unlocked '\
1421
 
            'dirstate revision tree'
 
1789
        if not self._locked:
 
1790
            raise AssertionError(
 
1791
                'cannot generate inventory of an unlocked '
 
1792
                'dirstate revision tree')
1422
1793
        # separate call for profiling - makes it clear where the costs are.
1423
1794
        self._dirstate._read_dirblocks_if_needed()
1424
 
        assert self._revision_id in self._dirstate.get_parent_ids(), \
1425
 
            'parent %s has disappeared from %s' % (
1426
 
            self._revision_id, self._dirstate.get_parent_ids())
 
1795
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1796
            raise AssertionError(
 
1797
                'parent %s has disappeared from %s' % (
 
1798
                self._revision_id, self._dirstate.get_parent_ids()))
1427
1799
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1428
1800
        # This is identical now to the WorkingTree _generate_inventory except
1429
1801
        # for the tree index use.
1430
1802
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1431
1803
        current_id = root_key[2]
1432
 
        assert current_entry[parent_index][0] == 'd'
 
1804
        if current_entry[parent_index][0] != 'd':
 
1805
            raise AssertionError()
1433
1806
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1434
1807
        inv.root.revision = current_entry[parent_index][4]
1435
1808
        # Turn some things into local variables
1466
1839
                elif kind == 'directory':
1467
1840
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1468
1841
                elif kind == 'symlink':
1469
 
                    inv_entry.executable = False
1470
 
                    inv_entry.text_size = None
1471
1842
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1472
1843
                elif kind == 'tree-reference':
1473
1844
                    inv_entry.reference_revision = fingerprint or None
1475
1846
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1476
1847
                            % entry)
1477
1848
                # These checks cost us around 40ms on a 55k entry tree
1478
 
                assert file_id not in inv_byid
1479
 
                assert name_unicode not in parent_ie.children
 
1849
                if file_id in inv_byid:
 
1850
                    raise AssertionError('file_id %s already in'
 
1851
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1852
                if name_unicode in parent_ie.children:
 
1853
                    raise AssertionError('name %r already in parent'
 
1854
                        % (name_unicode,))
1480
1855
                inv_byid[file_id] = inv_entry
1481
1856
                parent_ie.children[name_unicode] = inv_entry
1482
1857
        self._inventory = inv
1489
1864
        # Make sure the file exists
1490
1865
        entry = self._get_entry(file_id, path=path)
1491
1866
        if entry == (None, None): # do we raise?
1492
 
            return None
 
1867
            raise errors.NoSuchId(self, file_id)
1493
1868
        parent_index = self._get_parent_index()
1494
1869
        last_changed_revision = entry[1][parent_index][4]
1495
 
        return self._repository.get_revision(last_changed_revision).timestamp
 
1870
        try:
 
1871
            rev = self._repository.get_revision(last_changed_revision)
 
1872
        except errors.NoSuchRevision:
 
1873
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
 
1874
        return rev.timestamp
1496
1875
 
1497
1876
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1498
1877
        entry = self._get_entry(file_id=file_id, path=path)
1502
1881
            return parent_details[1]
1503
1882
        return None
1504
1883
 
1505
 
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
1506
 
    def get_weave(self, file_id):
1507
 
        return self._get_weave(file_id)
1508
 
 
1509
 
    def _get_weave(self, file_id):
1510
 
        return self._repository.weave_store.get_weave(file_id,
1511
 
                self._repository.get_transaction())
 
1884
    @needs_read_lock
 
1885
    def get_file_revision(self, file_id):
 
1886
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1887
        return inv[inv_file_id].revision
1512
1888
 
1513
1889
    def get_file(self, file_id, path=None):
1514
1890
        return StringIO(self.get_file_text(file_id))
1515
1891
 
1516
 
    def get_file_lines(self, file_id):
1517
 
        ie = self.inventory[file_id]
1518
 
        return self._get_weave(file_id).get_lines(ie.revision)
1519
 
 
1520
1892
    def get_file_size(self, file_id):
1521
 
        return self.inventory[file_id].text_size
 
1893
        """See Tree.get_file_size"""
 
1894
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1895
        return inv[inv_file_id].text_size
1522
1896
 
1523
 
    def get_file_text(self, file_id):
1524
 
        return ''.join(self.get_file_lines(file_id))
 
1897
    def get_file_text(self, file_id, path=None):
 
1898
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
 
1899
        return ''.join(content)
1525
1900
 
1526
1901
    def get_reference_revision(self, file_id, path=None):
1527
 
        return self.inventory[file_id].reference_revision
 
1902
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1903
        return inv[inv_file_id].reference_revision
1528
1904
 
1529
1905
    def iter_files_bytes(self, desired_files):
1530
1906
        """See Tree.iter_files_bytes.
1540
1916
                                       identifier))
1541
1917
        return self._repository.iter_files_bytes(repo_desired_files)
1542
1918
 
1543
 
    def get_symlink_target(self, file_id):
 
1919
    def get_symlink_target(self, file_id, path=None):
1544
1920
        entry = self._get_entry(file_id=file_id)
1545
1921
        parent_index = self._get_parent_index()
1546
1922
        if entry[1][parent_index][0] != 'l':
1547
1923
            return None
1548
1924
        else:
1549
 
            # At present, none of the tree implementations supports non-ascii
1550
 
            # symlink targets. So we will just assume that the dirstate path is
1551
 
            # correct.
1552
 
            return entry[1][parent_index][1]
 
1925
            target = entry[1][parent_index][1]
 
1926
            target = target.decode('utf8')
 
1927
            return target
1553
1928
 
1554
1929
    def get_revision_id(self):
1555
1930
        """Return the revision id for this tree."""
1556
1931
        return self._revision_id
1557
1932
 
1558
 
    def _get_inventory(self):
 
1933
    def _get_root_inventory(self):
1559
1934
        if self._inventory is not None:
1560
1935
            return self._inventory
1561
1936
        self._must_be_locked()
1562
1937
        self._generate_inventory()
1563
1938
        return self._inventory
1564
1939
 
 
1940
    root_inventory = property(_get_root_inventory,
 
1941
                         doc="Inventory of this Tree")
 
1942
 
 
1943
    @deprecated_method(deprecated_in((2, 5, 0)))
 
1944
    def _get_inventory(self):
 
1945
        return self.root_inventory
 
1946
 
1565
1947
    inventory = property(_get_inventory,
1566
1948
                         doc="Inventory of this Tree")
1567
1949
 
1573
1955
        return bool(self.path2id(filename))
1574
1956
 
1575
1957
    def kind(self, file_id):
1576
 
        return self.inventory[file_id].kind
 
1958
        entry = self._get_entry(file_id=file_id)[1]
 
1959
        if entry is None:
 
1960
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1961
        parent_index = self._get_parent_index()
 
1962
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1963
 
 
1964
    def stored_kind(self, file_id):
 
1965
        """See Tree.stored_kind"""
 
1966
        return self.kind(file_id)
 
1967
 
 
1968
    def path_content_summary(self, path):
 
1969
        """See Tree.path_content_summary."""
 
1970
        inv, inv_file_id = self._path2inv_file_id(path)
 
1971
        if inv_file_id is None:
 
1972
            return ('missing', None, None, None)
 
1973
        entry = inv[inv_file_id]
 
1974
        kind = entry.kind
 
1975
        if kind == 'file':
 
1976
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1977
        elif kind == 'symlink':
 
1978
            return (kind, None, None, entry.symlink_target)
 
1979
        else:
 
1980
            return (kind, None, None, None)
1577
1981
 
1578
1982
    def is_executable(self, file_id, path=None):
1579
 
        ie = self.inventory[file_id]
 
1983
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1984
        ie = inv[inv_file_id]
1580
1985
        if ie.kind != "file":
1581
 
            return None
 
1986
            return False
1582
1987
        return ie.executable
1583
1988
 
1584
 
    def list_files(self, include_root=False):
 
1989
    def is_locked(self):
 
1990
        return self._locked
 
1991
 
 
1992
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1585
1993
        # We use a standard implementation, because DirStateRevisionTree is
1586
1994
        # dealing with one of the parents of the current state
1587
 
        inv = self._get_inventory()
1588
 
        entries = inv.iter_entries()
1589
 
        if self.inventory.root is not None and not include_root:
 
1995
        if from_dir is None:
 
1996
            inv = self.root_inventory
 
1997
            from_dir_id = None
 
1998
        else:
 
1999
            inv, from_dir_id = self._path2inv_file_id(from_dir)
 
2000
            if from_dir_id is None:
 
2001
                # Directory not versioned
 
2002
                return
 
2003
        # FIXME: Support nested trees
 
2004
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
 
2005
        if inv.root is not None and not include_root and from_dir is None:
1590
2006
            entries.next()
1591
2007
        for path, entry in entries:
1592
2008
            yield path, 'V', entry.kind, entry.file_id, entry
1593
2009
 
1594
2010
    def lock_read(self):
1595
 
        """Lock the tree for a set of operations."""
 
2011
        """Lock the tree for a set of operations.
 
2012
 
 
2013
        :return: A bzrlib.lock.LogicalLockResult.
 
2014
        """
1596
2015
        if not self._locked:
1597
2016
            self._repository.lock_read()
1598
2017
            if self._dirstate._lock_token is None:
1599
2018
                self._dirstate.lock_read()
1600
2019
                self._dirstate_locked = True
1601
2020
        self._locked += 1
 
2021
        return LogicalLockResult(self.unlock)
1602
2022
 
1603
2023
    def _must_be_locked(self):
1604
2024
        if not self._locked:
1625
2045
                self._dirstate_locked = False
1626
2046
            self._repository.unlock()
1627
2047
 
 
2048
    @needs_read_lock
 
2049
    def supports_tree_reference(self):
 
2050
        return self._repo_supports_tree_reference
 
2051
 
1628
2052
    def walkdirs(self, prefix=""):
1629
2053
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1630
 
        # implementation based on an inventory.  
 
2054
        # implementation based on an inventory.
1631
2055
        # This should be cleaned up to use the much faster Dirstate code
1632
2056
        # So for now, we just build up the parent inventory, and extract
1633
2057
        # it the same way RevisionTree does.
1634
2058
        _directory = 'directory'
1635
 
        inv = self._get_inventory()
 
2059
        inv = self._get_root_inventory()
1636
2060
        top_id = inv.path2id(prefix)
1637
2061
        if top_id is None:
1638
2062
            pending = []
1662
2086
 
1663
2087
class InterDirStateTree(InterTree):
1664
2088
    """Fast path optimiser for changes_from with dirstate trees.
1665
 
    
1666
 
    This is used only when both trees are in the dirstate working file, and 
1667
 
    the source is any parent within the dirstate, and the destination is 
 
2089
 
 
2090
    This is used only when both trees are in the dirstate working file, and
 
2091
    the source is any parent within the dirstate, and the destination is
1668
2092
    the current working tree of the same dirstate.
1669
2093
    """
1670
2094
    # this could be generalized to allow comparisons between any trees in the
1679
2103
    def make_source_parent_tree(source, target):
1680
2104
        """Change the source tree into a parent of the target."""
1681
2105
        revid = source.commit('record tree')
1682
 
        target.branch.repository.fetch(source.branch.repository, revid)
 
2106
        target.branch.fetch(source.branch, revid)
1683
2107
        target.set_parent_ids([revid])
1684
2108
        return target.basis_tree(), target
1685
2109
 
 
2110
    @classmethod
 
2111
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
 
2112
        result = klass.make_source_parent_tree(source, target)
 
2113
        result[1]._iter_changes = dirstate.ProcessEntryPython
 
2114
        return result
 
2115
 
 
2116
    @classmethod
 
2117
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
 
2118
                                                  target):
 
2119
        from bzrlib.tests.test__dirstate_helpers import \
 
2120
            compiled_dirstate_helpers_feature
 
2121
        test_case.requireFeature(compiled_dirstate_helpers_feature)
 
2122
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
 
2123
        result = klass.make_source_parent_tree(source, target)
 
2124
        result[1]._iter_changes = ProcessEntryC
 
2125
        return result
 
2126
 
1686
2127
    _matching_from_tree_format = WorkingTreeFormat4()
1687
2128
    _matching_to_tree_format = WorkingTreeFormat4()
1688
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
1689
 
 
1690
 
    def _iter_changes(self, include_unchanged=False,
 
2129
 
 
2130
    @classmethod
 
2131
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
 
2132
        # This method shouldn't be called, because we have python and C
 
2133
        # specific flavours.
 
2134
        raise NotImplementedError
 
2135
 
 
2136
    def iter_changes(self, include_unchanged=False,
1691
2137
                      specific_files=None, pb=None, extra_trees=[],
1692
2138
                      require_versioned=True, want_unversioned=False):
1693
2139
        """Return the changes from source to target.
1694
2140
 
1695
 
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
2141
        :return: An iterator that yields tuples. See InterTree.iter_changes
1696
2142
            for details.
1697
2143
        :param specific_files: An optional list of file paths to restrict the
1698
2144
            comparison to. When mapping filenames to ids, all matches in all
1709
2155
            output. An unversioned file is defined as one with (False, False)
1710
2156
            for the versioned pair.
1711
2157
        """
1712
 
        utf8_decode = cache_utf8._utf8_decode
1713
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1714
 
        cmp_by_dirs = dirstate.cmp_by_dirs
1715
 
        # NB: show_status depends on being able to pass in non-versioned files
1716
 
        # and report them as unknown
1717
2158
        # TODO: handle extra trees in the dirstate.
1718
 
        # TODO: handle comparisons as an empty tree as a different special
1719
 
        # case? mbp 20070226
1720
 
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
2159
        if (extra_trees or specific_files == []):
1721
2160
            # we can't fast-path these cases (yet)
1722
 
            for f in super(InterDirStateTree, self)._iter_changes(
 
2161
            return super(InterDirStateTree, self).iter_changes(
1723
2162
                include_unchanged, specific_files, pb, extra_trees,
1724
 
                require_versioned, want_unversioned=want_unversioned):
1725
 
                yield f
1726
 
            return
 
2163
                require_versioned, want_unversioned=want_unversioned)
1727
2164
        parent_ids = self.target.get_parent_ids()
1728
 
        assert (self.source._revision_id in parent_ids), \
1729
 
                "revision {%s} is not stored in {%s}, but %s " \
1730
 
                "can only be used for trees stored in the dirstate" \
1731
 
                % (self.source._revision_id, self.target, self._iter_changes)
 
2165
        if not (self.source._revision_id in parent_ids
 
2166
                or self.source._revision_id == _mod_revision.NULL_REVISION):
 
2167
            raise AssertionError(
 
2168
                "revision {%s} is not stored in {%s}, but %s "
 
2169
                "can only be used for trees stored in the dirstate"
 
2170
                % (self.source._revision_id, self.target, self.iter_changes))
1732
2171
        target_index = 0
1733
 
        if self.source._revision_id == NULL_REVISION:
 
2172
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1734
2173
            source_index = None
1735
2174
            indices = (target_index,)
1736
2175
        else:
1737
 
            assert (self.source._revision_id in parent_ids), \
1738
 
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
1739
 
                self.source._revision_id, parent_ids)
 
2176
            if not (self.source._revision_id in parent_ids):
 
2177
                raise AssertionError(
 
2178
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
2179
                    self.source._revision_id, parent_ids))
1740
2180
            source_index = 1 + parent_ids.index(self.source._revision_id)
1741
 
            indices = (source_index,target_index)
 
2181
            indices = (source_index, target_index)
1742
2182
        # -- make all specific_files utf8 --
1743
2183
        if specific_files:
1744
2184
            specific_files_utf8 = set()
1745
2185
            for path in specific_files:
 
2186
                # Note, if there are many specific files, using cache_utf8
 
2187
                # would be good here.
1746
2188
                specific_files_utf8.add(path.encode('utf8'))
1747
2189
            specific_files = specific_files_utf8
1748
2190
        else:
1749
2191
            specific_files = set([''])
1750
2192
        # -- specific_files is now a utf8 path set --
 
2193
 
1751
2194
        # -- get the state object and prepare it.
1752
2195
        state = self.target.current_dirstate()
1753
2196
        state._read_dirblocks_if_needed()
1754
 
        def _entries_for_path(path):
1755
 
            """Return a list with all the entries that match path for all ids.
1756
 
            """
1757
 
            dirname, basename = os.path.split(path)
1758
 
            key = (dirname, basename, '')
1759
 
            block_index, present = state._find_block_index_from_key(key)
1760
 
            if not present:
1761
 
                # the block which should contain path is absent.
1762
 
                return []
1763
 
            result = []
1764
 
            block = state._dirblocks[block_index][1]
1765
 
            entry_index, _ = state._find_entry_index(key, block)
1766
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1767
 
            while (entry_index < len(block) and
1768
 
                block[entry_index][0][0:2] == key[0:2]):
1769
 
                result.append(block[entry_index])
1770
 
                entry_index += 1
1771
 
            return result
1772
2197
        if require_versioned:
1773
2198
            # -- check all supplied paths are versioned in a search tree. --
1774
 
            all_versioned = True
 
2199
            not_versioned = []
1775
2200
            for path in specific_files:
1776
 
                path_entries = _entries_for_path(path)
 
2201
                path_entries = state._entries_for_path(path)
1777
2202
                if not path_entries:
1778
2203
                    # this specified path is not present at all: error
1779
 
                    all_versioned = False
1780
 
                    break
 
2204
                    not_versioned.append(path.decode('utf-8'))
 
2205
                    continue
1781
2206
                found_versioned = False
1782
2207
                # for each id at this path
1783
2208
                for entry in path_entries:
1790
2215
                if not found_versioned:
1791
2216
                    # none of the indexes was not 'absent' at all ids for this
1792
2217
                    # path.
1793
 
                    all_versioned = False
1794
 
                    break
1795
 
            if not all_versioned:
1796
 
                raise errors.PathsNotVersionedError(specific_files)
 
2218
                    not_versioned.append(path.decode('utf-8'))
 
2219
            if len(not_versioned) > 0:
 
2220
                raise errors.PathsNotVersionedError(not_versioned)
1797
2221
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1798
 
        search_specific_files = set()
1799
 
        for path in specific_files:
1800
 
            other_specific_files = specific_files.difference(set([path]))
1801
 
            if not osutils.is_inside_any(other_specific_files, path):
1802
 
                # this is a top level path, we must check it.
1803
 
                search_specific_files.add(path)
1804
 
        # sketch: 
1805
 
        # compare source_index and target_index at or under each element of search_specific_files.
1806
 
        # follow the following comparison table. Note that we only want to do diff operations when
1807
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1808
 
        # for the target.
1809
 
        # cases:
1810
 
        # 
1811
 
        # Source | Target | disk | action
1812
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1813
 
        #        |        |      | diff check on source-target
1814
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1815
 
        #        |        |      | ???
1816
 
        #   r    |  a     |      | add source to search
1817
 
        #   r    |  a     |  a   | 
1818
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1819
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1820
 
        #   a    | fdlt   |      | add new id
1821
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1822
 
        #   a    |  a     |      | not present in either tree, skip
1823
 
        #   a    |  a     |  a   | not present in any tree, skip
1824
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1825
 
        #        |        |      | may not be selected by the users list of paths.
1826
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1827
 
        #        |        |      | may not be selected by the users list of paths.
1828
 
        #  fdlt  | fdlt   |      | content in both: diff them
1829
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1830
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1831
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1832
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1833
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1834
 
        #        |        |      | this id at the other path.
1835
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1836
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1837
 
        #        |        |      | this id at the other path.
1838
 
 
1839
 
        # for all search_indexs in each path at or under each element of
1840
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1841
 
        # relocated path as one to search if its not searched already. If the
1842
 
        # detail is not relocated, add the id.
1843
 
        searched_specific_files = set()
1844
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1845
 
        # Using a list so that we can access the values and change them in
1846
 
        # nested scope. Each one is [path, file_id, entry]
1847
 
        last_source_parent = [None, None]
1848
 
        last_target_parent = [None, None]
 
2222
        search_specific_files = osutils.minimum_path_selection(specific_files)
1849
2223
 
1850
2224
        use_filesystem_for_exec = (sys.platform != 'win32')
1851
 
 
1852
 
        # Just a sentry, so that _process_entry can say that this
1853
 
        # record is handled, but isn't interesting to process (unchanged)
1854
 
        uninteresting = object()
1855
 
 
1856
 
 
1857
 
        old_dirname_to_file_id = {}
1858
 
        new_dirname_to_file_id = {}
1859
 
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1860
 
        #       keeping a cache of directories that we have seen.
1861
 
 
1862
 
        def _process_entry(entry, path_info):
1863
 
            """Compare an entry and real disk to generate delta information.
1864
 
 
1865
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1866
 
                the path of entry. If None, then the path is considered absent.
1867
 
                (Perhaps we should pass in a concrete entry for this ?)
1868
 
                Basename is returned as a utf8 string because we expect this
1869
 
                tuple will be ignored, and don't want to take the time to
1870
 
                decode.
1871
 
            :return: None if these don't match
1872
 
                     A tuple of information about the change, or
1873
 
                     the object 'uninteresting' if these match, but are
1874
 
                     basically identical.
1875
 
            """
1876
 
            if source_index is None:
1877
 
                source_details = NULL_PARENT_DETAILS
1878
 
            else:
1879
 
                source_details = entry[1][source_index]
1880
 
            target_details = entry[1][target_index]
1881
 
            target_minikind = target_details[0]
1882
 
            if path_info is not None and target_minikind in 'fdlt':
1883
 
                assert target_index == 0
1884
 
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
1885
 
                                                  stat_value=path_info[3])
1886
 
                # The entry may have been modified by update_entry
1887
 
                target_details = entry[1][target_index]
1888
 
                target_minikind = target_details[0]
1889
 
            else:
1890
 
                link_or_sha1 = None
1891
 
            file_id = entry[0][2]
1892
 
            source_minikind = source_details[0]
1893
 
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
1894
 
                # claimed content in both: diff
1895
 
                #   r    | fdlt   |      | add source to search, add id path move and perform
1896
 
                #        |        |      | diff check on source-target
1897
 
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
1898
 
                #        |        |      | ???
1899
 
                if source_minikind in 'r':
1900
 
                    # add the source to the search path to find any children it
1901
 
                    # has.  TODO ? : only add if it is a container ?
1902
 
                    if not osutils.is_inside_any(searched_specific_files,
1903
 
                                                 source_details[1]):
1904
 
                        search_specific_files.add(source_details[1])
1905
 
                    # generate the old path; this is needed for stating later
1906
 
                    # as well.
1907
 
                    old_path = source_details[1]
1908
 
                    old_dirname, old_basename = os.path.split(old_path)
1909
 
                    path = pathjoin(entry[0][0], entry[0][1])
1910
 
                    old_entry = state._get_entry(source_index,
1911
 
                                                 path_utf8=old_path)
1912
 
                    # update the source details variable to be the real
1913
 
                    # location.
1914
 
                    source_details = old_entry[1][source_index]
1915
 
                    source_minikind = source_details[0]
1916
 
                else:
1917
 
                    old_dirname = entry[0][0]
1918
 
                    old_basename = entry[0][1]
1919
 
                    old_path = path = None
1920
 
                if path_info is None:
1921
 
                    # the file is missing on disk, show as removed.
1922
 
                    content_change = True
1923
 
                    target_kind = None
1924
 
                    target_exec = False
1925
 
                else:
1926
 
                    # source and target are both versioned and disk file is present.
1927
 
                    target_kind = path_info[2]
1928
 
                    if target_kind == 'directory':
1929
 
                        if path is None:
1930
 
                            old_path = path = pathjoin(old_dirname, old_basename)
1931
 
                        new_dirname_to_file_id[path] = file_id
1932
 
                        if source_minikind != 'd':
1933
 
                            content_change = True
1934
 
                        else:
1935
 
                            # directories have no fingerprint
1936
 
                            content_change = False
1937
 
                        target_exec = False
1938
 
                    elif target_kind == 'file':
1939
 
                        if source_minikind != 'f':
1940
 
                            content_change = True
1941
 
                        else:
1942
 
                            # We could check the size, but we already have the
1943
 
                            # sha1 hash.
1944
 
                            content_change = (link_or_sha1 != source_details[1])
1945
 
                        # Target details is updated at update_entry time
1946
 
                        if use_filesystem_for_exec:
1947
 
                            # We don't need S_ISREG here, because we are sure
1948
 
                            # we are dealing with a file.
1949
 
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
1950
 
                        else:
1951
 
                            target_exec = target_details[3]
1952
 
                    elif target_kind == 'symlink':
1953
 
                        if source_minikind != 'l':
1954
 
                            content_change = True
1955
 
                        else:
1956
 
                            content_change = (link_or_sha1 != source_details[1])
1957
 
                        target_exec = False
1958
 
                    elif target_kind == 'tree-reference':
1959
 
                        if source_minikind != 't':
1960
 
                            content_change = True
1961
 
                        else:
1962
 
                            content_change = False
1963
 
                        target_exec = False
1964
 
                    else:
1965
 
                        raise Exception, "unknown kind %s" % path_info[2]
1966
 
                if source_minikind == 'd':
1967
 
                    if path is None:
1968
 
                        old_path = path = pathjoin(old_dirname, old_basename)
1969
 
                    old_dirname_to_file_id[old_path] = file_id
1970
 
                # parent id is the entry for the path in the target tree
1971
 
                if old_dirname == last_source_parent[0]:
1972
 
                    source_parent_id = last_source_parent[1]
1973
 
                else:
1974
 
                    try:
1975
 
                        source_parent_id = old_dirname_to_file_id[old_dirname]
1976
 
                    except KeyError:
1977
 
                        source_parent_entry = state._get_entry(source_index,
1978
 
                                                               path_utf8=old_dirname)
1979
 
                        source_parent_id = source_parent_entry[0][2]
1980
 
                    if source_parent_id == entry[0][2]:
1981
 
                        # This is the root, so the parent is None
1982
 
                        source_parent_id = None
1983
 
                    else:
1984
 
                        last_source_parent[0] = old_dirname
1985
 
                        last_source_parent[1] = source_parent_id
1986
 
                new_dirname = entry[0][0]
1987
 
                if new_dirname == last_target_parent[0]:
1988
 
                    target_parent_id = last_target_parent[1]
1989
 
                else:
1990
 
                    try:
1991
 
                        target_parent_id = new_dirname_to_file_id[new_dirname]
1992
 
                    except KeyError:
1993
 
                        # TODO: We don't always need to do the lookup, because the
1994
 
                        #       parent entry will be the same as the source entry.
1995
 
                        target_parent_entry = state._get_entry(target_index,
1996
 
                                                               path_utf8=new_dirname)
1997
 
                        assert target_parent_entry != (None, None), (
1998
 
                            "Could not find target parent in wt: %s\nparent of: %s"
1999
 
                            % (new_dirname, entry))
2000
 
                        target_parent_id = target_parent_entry[0][2]
2001
 
                    if target_parent_id == entry[0][2]:
2002
 
                        # This is the root, so the parent is None
2003
 
                        target_parent_id = None
2004
 
                    else:
2005
 
                        last_target_parent[0] = new_dirname
2006
 
                        last_target_parent[1] = target_parent_id
2007
 
 
2008
 
                source_exec = source_details[3]
2009
 
                if (include_unchanged
2010
 
                    or content_change
2011
 
                    or source_parent_id != target_parent_id
2012
 
                    or old_basename != entry[0][1]
2013
 
                    or source_exec != target_exec
2014
 
                    ):
2015
 
                    if old_path is None:
2016
 
                        old_path = path = pathjoin(old_dirname, old_basename)
2017
 
                        old_path_u = utf8_decode(old_path)[0]
2018
 
                        path_u = old_path_u
2019
 
                    else:
2020
 
                        old_path_u = utf8_decode(old_path)[0]
2021
 
                        if old_path == path:
2022
 
                            path_u = old_path_u
2023
 
                        else:
2024
 
                            path_u = utf8_decode(path)[0]
2025
 
                    source_kind = _minikind_to_kind[source_minikind]
2026
 
                    return (entry[0][2],
2027
 
                           (old_path_u, path_u),
2028
 
                           content_change,
2029
 
                           (True, True),
2030
 
                           (source_parent_id, target_parent_id),
2031
 
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
2032
 
                           (source_kind, target_kind),
2033
 
                           (source_exec, target_exec))
2034
 
                else:
2035
 
                    return uninteresting
2036
 
            elif source_minikind in 'a' and target_minikind in 'fdlt':
2037
 
                # looks like a new file
2038
 
                if path_info is not None:
2039
 
                    path = pathjoin(entry[0][0], entry[0][1])
2040
 
                    # parent id is the entry for the path in the target tree
2041
 
                    # TODO: these are the same for an entire directory: cache em.
2042
 
                    parent_id = state._get_entry(target_index,
2043
 
                                                 path_utf8=entry[0][0])[0][2]
2044
 
                    if parent_id == entry[0][2]:
2045
 
                        parent_id = None
2046
 
                    if use_filesystem_for_exec:
2047
 
                        # We need S_ISREG here, because we aren't sure if this
2048
 
                        # is a file or not.
2049
 
                        target_exec = bool(
2050
 
                            stat.S_ISREG(path_info[3].st_mode)
2051
 
                            and stat.S_IEXEC & path_info[3].st_mode)
2052
 
                    else:
2053
 
                        target_exec = target_details[3]
2054
 
                    return (entry[0][2],
2055
 
                           (None, utf8_decode(path)[0]),
2056
 
                           True,
2057
 
                           (False, True),
2058
 
                           (None, parent_id),
2059
 
                           (None, utf8_decode(entry[0][1])[0]),
2060
 
                           (None, path_info[2]),
2061
 
                           (None, target_exec))
2062
 
                else:
2063
 
                    # but its not on disk: we deliberately treat this as just
2064
 
                    # never-present. (Why ?! - RBC 20070224)
2065
 
                    pass
2066
 
            elif source_minikind in 'fdlt' and target_minikind in 'a':
2067
 
                # unversioned, possibly, or possibly not deleted: we dont care.
2068
 
                # if its still on disk, *and* theres no other entry at this
2069
 
                # path [we dont know this in this routine at the moment -
2070
 
                # perhaps we should change this - then it would be an unknown.
2071
 
                old_path = pathjoin(entry[0][0], entry[0][1])
2072
 
                # parent id is the entry for the path in the target tree
2073
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
2074
 
                if parent_id == entry[0][2]:
2075
 
                    parent_id = None
2076
 
                return (entry[0][2],
2077
 
                       (utf8_decode(old_path)[0], None),
2078
 
                       True,
2079
 
                       (True, False),
2080
 
                       (parent_id, None),
2081
 
                       (utf8_decode(entry[0][1])[0], None),
2082
 
                       (_minikind_to_kind[source_minikind], None),
2083
 
                       (source_details[3], None))
2084
 
            elif source_minikind in 'fdlt' and target_minikind in 'r':
2085
 
                # a rename; could be a true rename, or a rename inherited from
2086
 
                # a renamed parent. TODO: handle this efficiently. Its not
2087
 
                # common case to rename dirs though, so a correct but slow
2088
 
                # implementation will do.
2089
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
2090
 
                    search_specific_files.add(target_details[1])
2091
 
            elif source_minikind in 'ra' and target_minikind in 'ra':
2092
 
                # neither of the selected trees contain this file,
2093
 
                # so skip over it. This is not currently directly tested, but
2094
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
2095
 
                pass
2096
 
            else:
2097
 
                raise AssertionError("don't know how to compare "
2098
 
                    "source_minikind=%r, target_minikind=%r"
2099
 
                    % (source_minikind, target_minikind))
2100
 
                ## import pdb;pdb.set_trace()
2101
 
            return None
2102
 
 
2103
 
        while search_specific_files:
2104
 
            # TODO: the pending list should be lexically sorted?  the
2105
 
            # interface doesn't require it.
2106
 
            current_root = search_specific_files.pop()
2107
 
            current_root_unicode = current_root.decode('utf8')
2108
 
            searched_specific_files.add(current_root)
2109
 
            # process the entries for this containing directory: the rest will be
2110
 
            # found by their parents recursively.
2111
 
            root_entries = _entries_for_path(current_root)
2112
 
            root_abspath = self.target.abspath(current_root_unicode)
2113
 
            try:
2114
 
                root_stat = os.lstat(root_abspath)
2115
 
            except OSError, e:
2116
 
                if e.errno == errno.ENOENT:
2117
 
                    # the path does not exist: let _process_entry know that.
2118
 
                    root_dir_info = None
2119
 
                else:
2120
 
                    # some other random error: hand it up.
2121
 
                    raise
2122
 
            else:
2123
 
                root_dir_info = ('', current_root,
2124
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
2125
 
                    root_abspath)
2126
 
                if root_dir_info[2] == 'directory':
2127
 
                    if self.target._directory_is_tree_reference(
2128
 
                        current_root.decode('utf8')):
2129
 
                        root_dir_info = root_dir_info[:2] + \
2130
 
                            ('tree-reference',) + root_dir_info[3:]
2131
 
 
2132
 
            if not root_entries and not root_dir_info:
2133
 
                # this specified path is not present at all, skip it.
2134
 
                continue
2135
 
            path_handled = False
2136
 
            for entry in root_entries:
2137
 
                result = _process_entry(entry, root_dir_info)
2138
 
                if result is not None:
2139
 
                    path_handled = True
2140
 
                    if result is not uninteresting:
2141
 
                        yield result
2142
 
            if want_unversioned and not path_handled and root_dir_info:
2143
 
                new_executable = bool(
2144
 
                    stat.S_ISREG(root_dir_info[3].st_mode)
2145
 
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
2146
 
                yield (None,
2147
 
                       (None, current_root_unicode),
2148
 
                       True,
2149
 
                       (False, False),
2150
 
                       (None, None),
2151
 
                       (None, splitpath(current_root_unicode)[-1]),
2152
 
                       (None, root_dir_info[2]),
2153
 
                       (None, new_executable)
2154
 
                      )
2155
 
            initial_key = (current_root, '', '')
2156
 
            block_index, _ = state._find_block_index_from_key(initial_key)
2157
 
            if block_index == 0:
2158
 
                # we have processed the total root already, but because the
2159
 
                # initial key matched it we should skip it here.
2160
 
                block_index +=1
2161
 
            if root_dir_info and root_dir_info[2] == 'tree-reference':
2162
 
                current_dir_info = None
2163
 
            else:
2164
 
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
2165
 
                try:
2166
 
                    current_dir_info = dir_iterator.next()
2167
 
                except OSError, e:
2168
 
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
2169
 
                    # python 2.5 has e.errno == EINVAL,
2170
 
                    #            and e.winerror == ERROR_DIRECTORY
2171
 
                    e_winerror = getattr(e, 'winerror', None)
2172
 
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
2173
 
                    # there may be directories in the inventory even though
2174
 
                    # this path is not a file on disk: so mark it as end of
2175
 
                    # iterator
2176
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
2177
 
                        current_dir_info = None
2178
 
                    elif (sys.platform == 'win32'
2179
 
                          and (e.errno in win_errors
2180
 
                               or e_winerror in win_errors)):
2181
 
                        current_dir_info = None
2182
 
                    else:
2183
 
                        raise
2184
 
                else:
2185
 
                    if current_dir_info[0][0] == '':
2186
 
                        # remove .bzr from iteration
2187
 
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
2188
 
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
2189
 
                        del current_dir_info[1][bzr_index]
2190
 
            # walk until both the directory listing and the versioned metadata
2191
 
            # are exhausted. 
2192
 
            if (block_index < len(state._dirblocks) and
2193
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2194
 
                current_block = state._dirblocks[block_index]
2195
 
            else:
2196
 
                current_block = None
2197
 
            while (current_dir_info is not None or
2198
 
                   current_block is not None):
2199
 
                if (current_dir_info and current_block
2200
 
                    and current_dir_info[0][0] != current_block[0]):
2201
 
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
2202
 
                        # filesystem data refers to paths not covered by the dirblock.
2203
 
                        # this has two possibilities:
2204
 
                        # A) it is versioned but empty, so there is no block for it
2205
 
                        # B) it is not versioned.
2206
 
 
2207
 
                        # if (A) then we need to recurse into it to check for
2208
 
                        # new unknown files or directories.
2209
 
                        # if (B) then we should ignore it, because we don't
2210
 
                        # recurse into unknown directories.
2211
 
                        path_index = 0
2212
 
                        while path_index < len(current_dir_info[1]):
2213
 
                                current_path_info = current_dir_info[1][path_index]
2214
 
                                if want_unversioned:
2215
 
                                    if current_path_info[2] == 'directory':
2216
 
                                        if self.target._directory_is_tree_reference(
2217
 
                                            current_path_info[0].decode('utf8')):
2218
 
                                            current_path_info = current_path_info[:2] + \
2219
 
                                                ('tree-reference',) + current_path_info[3:]
2220
 
                                    new_executable = bool(
2221
 
                                        stat.S_ISREG(current_path_info[3].st_mode)
2222
 
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
2223
 
                                    yield (None,
2224
 
                                        (None, utf8_decode(current_path_info[0])[0]),
2225
 
                                        True,
2226
 
                                        (False, False),
2227
 
                                        (None, None),
2228
 
                                        (None, utf8_decode(current_path_info[1])[0]),
2229
 
                                        (None, current_path_info[2]),
2230
 
                                        (None, new_executable))
2231
 
                                # dont descend into this unversioned path if it is
2232
 
                                # a dir
2233
 
                                if current_path_info[2] in ('directory',
2234
 
                                                            'tree-reference'):
2235
 
                                    del current_dir_info[1][path_index]
2236
 
                                    path_index -= 1
2237
 
                                path_index += 1
2238
 
 
2239
 
                        # This dir info has been handled, go to the next
2240
 
                        try:
2241
 
                            current_dir_info = dir_iterator.next()
2242
 
                        except StopIteration:
2243
 
                            current_dir_info = None
2244
 
                    else:
2245
 
                        # We have a dirblock entry for this location, but there
2246
 
                        # is no filesystem path for this. This is most likely
2247
 
                        # because a directory was removed from the disk.
2248
 
                        # We don't have to report the missing directory,
2249
 
                        # because that should have already been handled, but we
2250
 
                        # need to handle all of the files that are contained
2251
 
                        # within.
2252
 
                        for current_entry in current_block[1]:
2253
 
                            # entry referring to file not present on disk.
2254
 
                            # advance the entry only, after processing.
2255
 
                            result = _process_entry(current_entry, None)
2256
 
                            if result is not None:
2257
 
                                if result is not uninteresting:
2258
 
                                    yield result
2259
 
                        block_index +=1
2260
 
                        if (block_index < len(state._dirblocks) and
2261
 
                            osutils.is_inside(current_root,
2262
 
                                              state._dirblocks[block_index][0])):
2263
 
                            current_block = state._dirblocks[block_index]
2264
 
                        else:
2265
 
                            current_block = None
2266
 
                    continue
2267
 
                entry_index = 0
2268
 
                if current_block and entry_index < len(current_block[1]):
2269
 
                    current_entry = current_block[1][entry_index]
2270
 
                else:
2271
 
                    current_entry = None
2272
 
                advance_entry = True
2273
 
                path_index = 0
2274
 
                if current_dir_info and path_index < len(current_dir_info[1]):
2275
 
                    current_path_info = current_dir_info[1][path_index]
2276
 
                    if current_path_info[2] == 'directory':
2277
 
                        if self.target._directory_is_tree_reference(
2278
 
                            current_path_info[0].decode('utf8')):
2279
 
                            current_path_info = current_path_info[:2] + \
2280
 
                                ('tree-reference',) + current_path_info[3:]
2281
 
                else:
2282
 
                    current_path_info = None
2283
 
                advance_path = True
2284
 
                path_handled = False
2285
 
                while (current_entry is not None or
2286
 
                    current_path_info is not None):
2287
 
                    if current_entry is None:
2288
 
                        # the check for path_handled when the path is adnvaced
2289
 
                        # will yield this path if needed.
2290
 
                        pass
2291
 
                    elif current_path_info is None:
2292
 
                        # no path is fine: the per entry code will handle it.
2293
 
                        result = _process_entry(current_entry, current_path_info)
2294
 
                        if result is not None:
2295
 
                            if result is not uninteresting:
2296
 
                                yield result
2297
 
                    elif (current_entry[0][1] != current_path_info[1]
2298
 
                          or current_entry[1][target_index][0] in 'ar'):
2299
 
                        # The current path on disk doesn't match the dirblock
2300
 
                        # record. Either the dirblock is marked as absent, or
2301
 
                        # the file on disk is not present at all in the
2302
 
                        # dirblock. Either way, report about the dirblock
2303
 
                        # entry, and let other code handle the filesystem one.
2304
 
 
2305
 
                        # Compare the basename for these files to determine
2306
 
                        # which comes first
2307
 
                        if current_path_info[1] < current_entry[0][1]:
2308
 
                            # extra file on disk: pass for now, but only
2309
 
                            # increment the path, not the entry
2310
 
                            advance_entry = False
2311
 
                        else:
2312
 
                            # entry referring to file not present on disk.
2313
 
                            # advance the entry only, after processing.
2314
 
                            result = _process_entry(current_entry, None)
2315
 
                            if result is not None:
2316
 
                                if result is not uninteresting:
2317
 
                                    yield result
2318
 
                            advance_path = False
2319
 
                    else:
2320
 
                        result = _process_entry(current_entry, current_path_info)
2321
 
                        if result is not None:
2322
 
                            path_handled = True
2323
 
                            if result is not uninteresting:
2324
 
                                yield result
2325
 
                    if advance_entry and current_entry is not None:
2326
 
                        entry_index += 1
2327
 
                        if entry_index < len(current_block[1]):
2328
 
                            current_entry = current_block[1][entry_index]
2329
 
                        else:
2330
 
                            current_entry = None
2331
 
                    else:
2332
 
                        advance_entry = True # reset the advance flaga
2333
 
                    if advance_path and current_path_info is not None:
2334
 
                        if not path_handled:
2335
 
                            # unversioned in all regards
2336
 
                            if want_unversioned:
2337
 
                                new_executable = bool(
2338
 
                                    stat.S_ISREG(current_path_info[3].st_mode)
2339
 
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
2340
 
                                yield (None,
2341
 
                                    (None, utf8_decode(current_path_info[0])[0]),
2342
 
                                    True,
2343
 
                                    (False, False),
2344
 
                                    (None, None),
2345
 
                                    (None, utf8_decode(current_path_info[1])[0]),
2346
 
                                    (None, current_path_info[2]),
2347
 
                                    (None, new_executable))
2348
 
                            # dont descend into this unversioned path if it is
2349
 
                            # a dir
2350
 
                            if current_path_info[2] in ('directory'):
2351
 
                                del current_dir_info[1][path_index]
2352
 
                                path_index -= 1
2353
 
                        # dont descend the disk iterator into any tree 
2354
 
                        # paths.
2355
 
                        if current_path_info[2] == 'tree-reference':
2356
 
                            del current_dir_info[1][path_index]
2357
 
                            path_index -= 1
2358
 
                        path_index += 1
2359
 
                        if path_index < len(current_dir_info[1]):
2360
 
                            current_path_info = current_dir_info[1][path_index]
2361
 
                            if current_path_info[2] == 'directory':
2362
 
                                if self.target._directory_is_tree_reference(
2363
 
                                    current_path_info[0].decode('utf8')):
2364
 
                                    current_path_info = current_path_info[:2] + \
2365
 
                                        ('tree-reference',) + current_path_info[3:]
2366
 
                        else:
2367
 
                            current_path_info = None
2368
 
                        path_handled = False
2369
 
                    else:
2370
 
                        advance_path = True # reset the advance flagg.
2371
 
                if current_block is not None:
2372
 
                    block_index += 1
2373
 
                    if (block_index < len(state._dirblocks) and
2374
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2375
 
                        current_block = state._dirblocks[block_index]
2376
 
                    else:
2377
 
                        current_block = None
2378
 
                if current_dir_info is not None:
2379
 
                    try:
2380
 
                        current_dir_info = dir_iterator.next()
2381
 
                    except StopIteration:
2382
 
                        current_dir_info = None
2383
 
 
 
2225
        iter_changes = self.target._iter_changes(include_unchanged,
 
2226
            use_filesystem_for_exec, search_specific_files, state,
 
2227
            source_index, target_index, want_unversioned, self.target)
 
2228
        return iter_changes.iter_changes()
2384
2229
 
2385
2230
    @staticmethod
2386
2231
    def is_compatible(source, target):
2387
2232
        # the target must be a dirstate working tree
2388
 
        if not isinstance(target, WorkingTree4):
 
2233
        if not isinstance(target, DirStateWorkingTree):
2389
2234
            return False
2390
 
        # the source must be a revtreee or dirstate rev tree.
 
2235
        # the source must be a revtree or dirstate rev tree.
2391
2236
        if not isinstance(source,
2392
2237
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2393
2238
            return False
2394
2239
        # the source revid must be in the target dirstate
2395
 
        if not (source._revision_id == NULL_REVISION or
 
2240
        if not (source._revision_id == _mod_revision.NULL_REVISION or
2396
2241
            source._revision_id in target.get_parent_ids()):
2397
 
            # TODO: what about ghosts? it may well need to 
 
2242
            # TODO: what about ghosts? it may well need to
2398
2243
            # check for them explicitly.
2399
2244
            return False
2400
2245
        return True
2410
2255
 
2411
2256
    def convert(self, tree):
2412
2257
        # lock the control files not the tree, so that we dont get tree
2413
 
        # on-unlock behaviours, and so that noone else diddles with the 
 
2258
        # on-unlock behaviours, and so that noone else diddles with the
2414
2259
        # tree during upgrade.
2415
2260
        tree._control_files.lock_write()
2416
2261
        try:
2442
2287
 
2443
2288
    def update_format(self, tree):
2444
2289
        """Change the format marker."""
2445
 
        tree._control_files.put_utf8('format',
2446
 
            self.target_format.get_format_string())
 
2290
        tree._transport.put_bytes('format',
 
2291
            self.target_format.as_string(),
 
2292
            mode=tree.bzrdir._get_file_mode())
 
2293
 
 
2294
 
 
2295
class Converter4to5(object):
 
2296
    """Perform an in-place upgrade of format 4 to format 5 trees."""
 
2297
 
 
2298
    def __init__(self):
 
2299
        self.target_format = WorkingTreeFormat5()
 
2300
 
 
2301
    def convert(self, tree):
 
2302
        # lock the control files not the tree, so that we don't get tree
 
2303
        # on-unlock behaviours, and so that no-one else diddles with the
 
2304
        # tree during upgrade.
 
2305
        tree._control_files.lock_write()
 
2306
        try:
 
2307
            self.update_format(tree)
 
2308
        finally:
 
2309
            tree._control_files.unlock()
 
2310
 
 
2311
    def update_format(self, tree):
 
2312
        """Change the format marker."""
 
2313
        tree._transport.put_bytes('format',
 
2314
            self.target_format.as_string(),
 
2315
            mode=tree.bzrdir._get_file_mode())
 
2316
 
 
2317
 
 
2318
class Converter4or5to6(object):
 
2319
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
 
2320
 
 
2321
    def __init__(self):
 
2322
        self.target_format = WorkingTreeFormat6()
 
2323
 
 
2324
    def convert(self, tree):
 
2325
        # lock the control files not the tree, so that we don't get tree
 
2326
        # on-unlock behaviours, and so that no-one else diddles with the
 
2327
        # tree during upgrade.
 
2328
        tree._control_files.lock_write()
 
2329
        try:
 
2330
            self.init_custom_control_files(tree)
 
2331
            self.update_format(tree)
 
2332
        finally:
 
2333
            tree._control_files.unlock()
 
2334
 
 
2335
    def init_custom_control_files(self, tree):
 
2336
        """Initialize custom control files."""
 
2337
        tree._transport.put_bytes('views', '',
 
2338
            mode=tree.bzrdir._get_file_mode())
 
2339
 
 
2340
    def update_format(self, tree):
 
2341
        """Change the format marker."""
 
2342
        tree._transport.put_bytes('format',
 
2343
            self.target_format.as_string(),
 
2344
            mode=tree.bzrdir._get_file_mode())