~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Patch Queue Manager
  • Date: 2014-10-06 16:32:42 UTC
  • mfrom: (6597.2.4 split-diff-tests)
  • Revision ID: pqm@pqm.ubuntu.com-20141006163242-c2cll01cwc24grkk
(vila) Split some tests to be able to get finer grained failures (Vincent
 Ladeuil)

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
1
# Copyright (C) 2007-2012 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
 
    delta,
 
41
    controldir,
 
42
    debug,
47
43
    dirstate,
48
44
    errors,
 
45
    filters as _mod_filters,
49
46
    generate_ids,
50
 
    globbing,
51
 
    hashcache,
52
 
    ignores,
53
 
    merge,
54
47
    osutils,
55
48
    revision as _mod_revision,
56
49
    revisiontree,
57
 
    textui,
 
50
    trace,
58
51
    transform,
59
 
    urlutils,
60
 
    xml5,
61
 
    xml6,
 
52
    views,
62
53
    )
63
 
import bzrlib.branch
64
 
from bzrlib.transport import get_transport
65
 
import bzrlib.ui
66
54
""")
67
55
 
68
 
from bzrlib import symbol_versioning
69
56
from bzrlib.decorators import needs_read_lock, needs_write_lock
70
 
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
71
 
from bzrlib.lockable_files import LockableFiles, TransportLock
 
57
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
58
from bzrlib.lock import LogicalLockResult
 
59
from bzrlib.lockable_files import LockableFiles
72
60
from bzrlib.lockdir import LockDir
73
 
import bzrlib.mutabletree
74
 
from bzrlib.mutabletree import needs_tree_write_lock
 
61
from bzrlib.mutabletree import (
 
62
    MutableTree,
 
63
    needs_tree_write_lock,
 
64
    )
75
65
from bzrlib.osutils import (
76
66
    file_kind,
77
67
    isdir,
78
 
    normpath,
79
68
    pathjoin,
80
 
    rand_chars,
81
69
    realpath,
82
70
    safe_unicode,
83
 
    splitpath,
84
 
    )
85
 
from bzrlib.trace import mutter, note
 
71
    )
 
72
from bzrlib.symbol_versioning import (
 
73
    deprecated_in,
 
74
    deprecated_method,
 
75
    )
86
76
from bzrlib.transport.local import LocalTransport
87
 
from bzrlib.tree import InterTree
88
 
from bzrlib.progress import DummyProgress, ProgressPhase
89
 
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
90
 
from bzrlib.rio import RioReader, rio_file, Stanza
91
 
from bzrlib.symbol_versioning import (deprecated_passed,
92
 
        deprecated_method,
93
 
        deprecated_function,
94
 
        DEPRECATED_PARAMETER,
95
 
        )
96
 
from bzrlib.tree import Tree
97
 
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
98
 
 
99
 
 
100
 
# This is the Windows equivalent of ENOTDIR
101
 
# It is defined in pywin32.winerror, but we don't want a strong dependency for
102
 
# just an error code.
103
 
ERROR_PATH_NOT_FOUND = 3
104
 
ERROR_DIRECTORY = 267
105
 
 
106
 
 
107
 
class WorkingTree4(WorkingTree3):
108
 
    """This is the Format 4 working tree.
109
 
 
110
 
    This differs from WorkingTree3 by:
111
 
     - Having a consolidated internal dirstate, stored in a
112
 
       randomly-accessible sorted file on disk.
113
 
     - Not having a regular inventory attribute.  One can be synthesized 
114
 
       on demand but this is expensive and should be avoided.
115
 
 
116
 
    This is new in bzr 0.15.
117
 
    """
 
77
from bzrlib.tree import (
 
78
    InterTree,
 
79
    InventoryTree,
 
80
    )
 
81
from bzrlib.workingtree import (
 
82
    InventoryWorkingTree,
 
83
    WorkingTree,
 
84
    WorkingTreeFormatMetaDir,
 
85
    )
 
86
 
 
87
 
 
88
class DirStateWorkingTree(InventoryWorkingTree):
118
89
 
119
90
    def __init__(self, basedir,
120
91
                 branch,
130
101
        """
131
102
        self._format = _format
132
103
        self.bzrdir = _bzrdir
133
 
        from bzrlib.trace import note, mutter
134
 
        assert isinstance(basedir, basestring), \
135
 
            "base directory %r is not a string" % basedir
136
104
        basedir = safe_unicode(basedir)
137
 
        mutter("opening working tree %r", basedir)
 
105
        trace.mutter("opening working tree %r", basedir)
138
106
        self._branch = branch
139
 
        assert isinstance(self.branch, bzrlib.branch.Branch), \
140
 
            "branch %r is not a Branch" % self.branch
141
107
        self.basedir = realpath(basedir)
142
108
        # if branch is at our basedir and is a format 6 or less
143
109
        # assume all other formats have their own control files.
144
 
        assert isinstance(_control_files, LockableFiles), \
145
 
            "_control_files must be a LockableFiles, not %r" % _control_files
146
110
        self._control_files = _control_files
 
111
        self._transport = self._control_files._transport
147
112
        self._dirty = None
148
113
        #-------------
149
114
        # during a read or write lock these objects are set, and are
151
116
        self._dirstate = None
152
117
        self._inventory = None
153
118
        #-------------
 
119
        self._setup_directory_is_tree_reference()
 
120
        self._detect_case_handling()
 
121
        self._rules_searcher = None
 
122
        self.views = self._make_views()
 
123
        #--- allow tests to select the dirstate iter_changes implementation
 
124
        self._iter_changes = dirstate._process_entry
154
125
 
155
126
    @needs_tree_write_lock
156
127
    def _add(self, files, ids, kinds):
158
129
        state = self.current_dirstate()
159
130
        for f, file_id, kind in zip(files, ids, kinds):
160
131
            f = f.strip('/')
161
 
            assert '//' not in f
162
 
            assert '..' not in f
163
132
            if self.path2id(f):
164
133
                # special case tree root handling.
165
134
                if f == '' and self.path2id(f) == ROOT_ID:
173
142
            state.add(f, file_id, kind, None, '')
174
143
        self._make_dirty(reset_inventory=True)
175
144
 
 
145
    def _get_check_refs(self):
 
146
        """Return the references needed to perform a check of this tree."""
 
147
        return [('trees', self.last_revision())]
 
148
 
176
149
    def _make_dirty(self, reset_inventory):
177
150
        """Make the tree state dirty.
178
151
 
186
159
    @needs_tree_write_lock
187
160
    def add_reference(self, sub_tree):
188
161
        # use standard implementation, which calls back to self._add
189
 
        # 
 
162
        #
190
163
        # So we don't store the reference_revision in the working dirstate,
191
 
        # it's just recorded at the moment of commit. 
 
164
        # it's just recorded at the moment of commit.
192
165
        self._add_reference(sub_tree)
193
166
 
194
167
    def break_lock(self):
230
203
 
231
204
    def _comparison_data(self, entry, path):
232
205
        kind, executable, stat_value = \
233
 
            WorkingTree3._comparison_data(self, entry, path)
 
206
            WorkingTree._comparison_data(self, entry, path)
234
207
        # it looks like a plain directory, but it's really a reference -- see
235
208
        # also kind()
236
 
        if (self._repo_supports_tree_reference and
237
 
            kind == 'directory' and
238
 
            self._directory_is_tree_reference(path)):
 
209
        if (self._repo_supports_tree_reference and kind == 'directory'
 
210
            and entry is not None and entry.kind == 'tree-reference'):
239
211
            kind = 'tree-reference'
240
212
        return kind, executable, stat_value
241
213
 
243
215
    def commit(self, message=None, revprops=None, *args, **kwargs):
244
216
        # mark the tree as dirty post commit - commit
245
217
        # can change the current versioned list by doing deletes.
246
 
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
218
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
247
219
        self._make_dirty(reset_inventory=True)
248
220
        return result
249
221
 
267
239
            return self._dirstate
268
240
        local_path = self.bzrdir.get_workingtree_transport(None
269
241
            ).local_abspath('dirstate')
270
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
242
        self._dirstate = dirstate.DirState.on_file(local_path,
 
243
            self._sha1_provider(), self._worth_saving_limit())
271
244
        return self._dirstate
272
245
 
273
 
    def _directory_is_tree_reference(self, relpath):
274
 
        # as a special case, if a directory contains control files then 
275
 
        # it's a tree reference, except that the root of the tree is not
276
 
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
277
 
        # TODO: We could ask all the control formats whether they
278
 
        # recognize this directory, but at the moment there's no cheap api
279
 
        # to do that.  Since we probably can only nest bzr checkouts and
280
 
        # they always use this name it's ok for now.  -- mbp 20060306
281
 
        #
282
 
        # FIXME: There is an unhandled case here of a subdirectory
283
 
        # containing .bzr but not a branch; that will probably blow up
284
 
        # when you try to commit it.  It might happen if there is a
285
 
        # checkout in a subdirectory.  This can be avoided by not adding
286
 
        # it.  mbp 20070306
 
246
    def _sha1_provider(self):
 
247
        """A function that returns a SHA1Provider suitable for this tree.
 
248
 
 
249
        :return: None if content filtering is not supported by this tree.
 
250
          Otherwise, a SHA1Provider is returned that sha's the canonical
 
251
          form of files, i.e. after read filters are applied.
 
252
        """
 
253
        if self.supports_content_filtering():
 
254
            return ContentFilterAwareSHA1Provider(self)
 
255
        else:
 
256
            return None
 
257
 
 
258
    def _worth_saving_limit(self):
 
259
        """How many hash changes are ok before we must save the dirstate.
 
260
 
 
261
        :return: an integer. -1 means never save.
 
262
        """
 
263
        conf = self.get_config_stack()
 
264
        return conf.get('bzr.workingtree.worth_saving_limit')
287
265
 
288
266
    def filter_unversioned_files(self, paths):
289
267
        """Filter out paths that are versioned.
322
300
 
323
301
    def _generate_inventory(self):
324
302
        """Create and set self.inventory from the dirstate object.
325
 
        
 
303
 
326
304
        This is relatively expensive: we have to walk the entire dirstate.
327
305
        Ideally we would not, and can deprecate this function.
328
306
        """
332
310
        state._read_dirblocks_if_needed()
333
311
        root_key, current_entry = self._get_entry(path='')
334
312
        current_id = root_key[2]
335
 
        assert current_entry[0][0] == 'd' # directory
 
313
        if not (current_entry[0][0] == 'd'): # directory
 
314
            raise AssertionError(current_entry)
336
315
        inv = Inventory(root_id=current_id)
337
316
        # Turn some things into local variables
338
317
        minikind_to_kind = dirstate.DirState._minikind_to_kind
371
350
                    # add this entry to the parent map.
372
351
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
373
352
                elif kind == 'tree-reference':
374
 
                    assert self._repo_supports_tree_reference, \
375
 
                        "repository of %r " \
376
 
                        "doesn't support tree references " \
377
 
                        "required by entry %r" \
378
 
                        % (self, name)
 
353
                    if not self._repo_supports_tree_reference:
 
354
                        raise errors.UnsupportedOperation(
 
355
                            self._generate_inventory,
 
356
                            self.branch.repository)
379
357
                    inv_entry.reference_revision = link_or_sha1 or None
380
358
                elif kind != 'symlink':
381
359
                    raise AssertionError("unknown kind %r" % kind)
382
360
                # These checks cost us around 40ms on a 55k entry tree
383
 
                assert file_id not in inv_byid, ('file_id %s already in'
384
 
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
385
 
                assert name_unicode not in parent_ie.children
 
361
                if file_id in inv_byid:
 
362
                    raise AssertionError('file_id %s already in'
 
363
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
364
                if name_unicode in parent_ie.children:
 
365
                    raise AssertionError('name %r already in parent'
 
366
                        % (name_unicode,))
386
367
                inv_byid[file_id] = inv_entry
387
368
                parent_ie.children[name_unicode] = inv_entry
388
369
        self._inventory = inv
393
374
        If either file_id or path is supplied, it is used as the key to lookup.
394
375
        If both are supplied, the fastest lookup is used, and an error is
395
376
        raised if they do not both point at the same row.
396
 
        
 
377
 
397
378
        :param file_id: An optional unicode file_id to be looked up.
398
379
        :param path: An optional unicode path to be looked up.
399
380
        :return: The dirstate row tuple for path/file_id, or (None, None)
417
398
        state = self.current_dirstate()
418
399
        if stat_value is None:
419
400
            try:
420
 
                stat_value = os.lstat(file_abspath)
 
401
                stat_value = osutils.lstat(file_abspath)
421
402
            except OSError, e:
422
403
                if e.errno == errno.ENOENT:
423
404
                    return None
424
405
                else:
425
406
                    raise
426
 
        link_or_sha1 = state.update_entry(entry, file_abspath,
427
 
                                          stat_value=stat_value)
 
407
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
408
            stat_value=stat_value)
428
409
        if entry[1][0][0] == 'f':
429
 
            return link_or_sha1
 
410
            if link_or_sha1 is None:
 
411
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
 
412
                try:
 
413
                    sha1 = osutils.sha_file(file_obj)
 
414
                finally:
 
415
                    file_obj.close()
 
416
                self._observed_sha1(file_id, path, (sha1, statvalue))
 
417
                return sha1
 
418
            else:
 
419
                return link_or_sha1
430
420
        return None
431
421
 
432
 
    def _get_inventory(self):
 
422
    def _get_root_inventory(self):
433
423
        """Get the inventory for the tree. This is only valid within a lock."""
 
424
        if 'evil' in debug.debug_flags:
 
425
            trace.mutter_callsite(2,
 
426
                "accessing .inventory forces a size of tree translation.")
434
427
        if self._inventory is not None:
435
428
            return self._inventory
436
429
        self._must_be_locked()
437
430
        self._generate_inventory()
438
431
        return self._inventory
439
432
 
 
433
    @deprecated_method(deprecated_in((2, 5, 0)))
 
434
    def _get_inventory(self):
 
435
        return self.root_inventory
 
436
 
440
437
    inventory = property(_get_inventory,
441
438
                         doc="Inventory of this Tree")
442
439
 
 
440
    root_inventory = property(_get_root_inventory,
 
441
        "Root inventory of this tree")
 
442
 
443
443
    @needs_read_lock
444
444
    def get_parent_ids(self):
445
445
        """See Tree.get_parent_ids.
446
 
        
 
446
 
447
447
        This implementation requests the ids list from the dirstate file.
448
448
        """
449
449
        return self.current_dirstate().get_parent_ids()
465
465
 
466
466
    def has_id(self, file_id):
467
467
        state = self.current_dirstate()
468
 
        file_id = osutils.safe_file_id(file_id)
469
468
        row, parents = self._get_entry(file_id=file_id)
470
469
        if row is None:
471
470
            return False
472
471
        return osutils.lexists(pathjoin(
473
472
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
474
473
 
 
474
    def has_or_had_id(self, file_id):
 
475
        state = self.current_dirstate()
 
476
        row, parents = self._get_entry(file_id=file_id)
 
477
        return row is not None
 
478
 
475
479
    @needs_read_lock
476
480
    def id2path(self, file_id):
477
481
        "Convert a file-id to a path."
478
 
        file_id = osutils.safe_file_id(file_id)
479
482
        state = self.current_dirstate()
480
483
        entry = self._get_entry(file_id=file_id)
481
484
        if entry == (None, None):
483
486
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
484
487
        return path_utf8.decode('utf8')
485
488
 
486
 
    if not osutils.supports_executable():
487
 
        def is_executable(self, file_id, path=None):
488
 
            """Test if a file is executable or not.
489
 
 
490
 
            Note: The caller is expected to take a read-lock before calling this.
491
 
            """
492
 
            file_id = osutils.safe_file_id(file_id)
 
489
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
490
        entry = self._get_entry(path=path)
 
491
        if entry == (None, None):
 
492
            return False # Missing entries are not executable
 
493
        return entry[1][0][3] # Executable?
 
494
 
 
495
    def is_executable(self, file_id, path=None):
 
496
        """Test if a file is executable or not.
 
497
 
 
498
        Note: The caller is expected to take a read-lock before calling this.
 
499
        """
 
500
        if not self._supports_executable():
493
501
            entry = self._get_entry(file_id=file_id, path=path)
494
502
            if entry == (None, None):
495
503
                return False
496
504
            return entry[1][0][3]
497
 
    else:
498
 
        def is_executable(self, file_id, path=None):
499
 
            """Test if a file is executable or not.
500
 
 
501
 
            Note: The caller is expected to take a read-lock before calling this.
502
 
            """
 
505
        else:
 
506
            self._must_be_locked()
503
507
            if not path:
504
 
                file_id = osutils.safe_file_id(file_id)
505
508
                path = self.id2path(file_id)
506
 
            mode = os.lstat(self.abspath(path)).st_mode
 
509
            mode = osutils.lstat(self.abspath(path)).st_mode
507
510
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
508
511
 
 
512
    def all_file_ids(self):
 
513
        """See Tree.iter_all_file_ids"""
 
514
        self._must_be_locked()
 
515
        result = set()
 
516
        for key, tree_details in self.current_dirstate()._iter_entries():
 
517
            if tree_details[0][0] in ('a', 'r'): # relocated
 
518
                continue
 
519
            result.add(key[2])
 
520
        return result
 
521
 
509
522
    @needs_read_lock
510
523
    def __iter__(self):
511
524
        """Iterate through file_ids for this tree.
524
537
        return iter(result)
525
538
 
526
539
    def iter_references(self):
 
540
        if not self._repo_supports_tree_reference:
 
541
            # When the repo doesn't support references, we will have nothing to
 
542
            # return
 
543
            return
527
544
        for key, tree_details in self.current_dirstate()._iter_entries():
528
545
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
529
546
                # not relevant to the working tree
531
548
            if not key[1]:
532
549
                # the root is not a reference.
533
550
                continue
534
 
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
551
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
535
552
            try:
536
 
                if self._kind(path) == 'tree-reference':
537
 
                    yield path, key[2]
 
553
                if self._kind(relpath) == 'tree-reference':
 
554
                    yield relpath, key[2]
538
555
            except errors.NoSuchFile:
539
556
                # path is missing on disk.
540
557
                continue
541
558
 
 
559
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
 
560
        """See MutableTree._observed_sha1."""
 
561
        state = self.current_dirstate()
 
562
        entry = self._get_entry(file_id=file_id, path=path)
 
563
        state._observed_sha1(entry, sha1, statvalue)
 
564
 
542
565
    def kind(self, file_id):
543
566
        """Return the kind of a file.
544
567
 
548
571
        Note: The caller is expected to take a read-lock before calling this.
549
572
        """
550
573
        relpath = self.id2path(file_id)
551
 
        assert relpath != None, \
552
 
            "path for id {%s} is None!" % file_id
 
574
        if relpath is None:
 
575
            raise AssertionError(
 
576
                "path for id {%s} is None!" % file_id)
553
577
        return self._kind(relpath)
554
578
 
555
579
    def _kind(self, relpath):
556
580
        abspath = self.abspath(relpath)
557
581
        kind = file_kind(abspath)
558
 
        if (self._repo_supports_tree_reference and
559
 
            kind == 'directory' and
560
 
            self._directory_is_tree_reference(relpath)):
561
 
            kind = 'tree-reference'
 
582
        if (self._repo_supports_tree_reference and kind == 'directory'):
 
583
            entry = self._get_entry(path=relpath)
 
584
            if entry[1] is not None:
 
585
                if entry[1][0][0] == 't':
 
586
                    kind = 'tree-reference'
562
587
        return kind
563
588
 
564
589
    @needs_read_lock
568
593
        if parent_ids:
569
594
            return parent_ids[0]
570
595
        else:
571
 
            return None
 
596
            return _mod_revision.NULL_REVISION
572
597
 
573
598
    def lock_read(self):
574
 
        """See Branch.lock_read, and WorkingTree.unlock."""
 
599
        """See Branch.lock_read, and WorkingTree.unlock.
 
600
 
 
601
        :return: A bzrlib.lock.LogicalLockResult.
 
602
        """
575
603
        self.branch.lock_read()
576
604
        try:
577
605
            self._control_files.lock_read()
590
618
        except:
591
619
            self.branch.unlock()
592
620
            raise
 
621
        return LogicalLockResult(self.unlock)
593
622
 
594
623
    def _lock_self_write(self):
595
624
        """This should be called after the branch is locked."""
610
639
        except:
611
640
            self.branch.unlock()
612
641
            raise
 
642
        return LogicalLockResult(self.unlock)
613
643
 
614
644
    def lock_tree_write(self):
615
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
645
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
 
646
 
 
647
        :return: A bzrlib.lock.LogicalLockResult.
 
648
        """
616
649
        self.branch.lock_read()
617
 
        self._lock_self_write()
 
650
        return self._lock_self_write()
618
651
 
619
652
    def lock_write(self):
620
 
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
653
        """See MutableTree.lock_write, and WorkingTree.unlock.
 
654
 
 
655
        :return: A bzrlib.lock.LogicalLockResult.
 
656
        """
621
657
        self.branch.lock_write()
622
 
        self._lock_self_write()
 
658
        return self._lock_self_write()
623
659
 
624
660
    @needs_tree_write_lock
625
661
    def move(self, from_paths, to_dir, after=False):
627
663
        result = []
628
664
        if not from_paths:
629
665
            return result
630
 
 
631
666
        state = self.current_dirstate()
632
 
 
633
 
        assert not isinstance(from_paths, basestring)
 
667
        if isinstance(from_paths, basestring):
 
668
            raise ValueError()
634
669
        to_dir_utf8 = to_dir.encode('utf8')
635
670
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
636
671
        id_index = state._get_id_index()
657
692
 
658
693
        if self._inventory is not None:
659
694
            update_inventory = True
660
 
            inv = self.inventory
 
695
            inv = self.root_inventory
 
696
            to_dir_id = to_entry[0][2]
661
697
            to_dir_ie = inv[to_dir_id]
662
 
            to_dir_id = to_entry[0][2]
663
698
        else:
664
699
            update_inventory = False
665
700
 
696
731
            from_entry = self._get_entry(path=from_rel)
697
732
            if from_entry == (None, None):
698
733
                raise errors.BzrMoveFailedError(from_rel,to_dir,
699
 
                    errors.NotVersionedError(path=str(from_rel)))
 
734
                    errors.NotVersionedError(path=from_rel))
700
735
 
701
736
            from_id = from_entry[0][2]
702
737
            to_rel = pathjoin(to_dir, from_tail)
729
764
                if from_missing: # implicitly just update our path mapping
730
765
                    move_file = False
731
766
                elif not after:
732
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
733
 
                        extra="(Use --after to update the Bazaar id)")
 
767
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
734
768
 
735
769
            rollbacks = []
736
770
            def rollback_rename():
792
826
                if minikind == 'd':
793
827
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
794
828
                        """Recursively update all entries in this dirblock."""
795
 
                        assert from_dir != '', "renaming root not supported"
 
829
                        if from_dir == '':
 
830
                            raise AssertionError("renaming root not supported")
796
831
                        from_key = (from_dir, '')
797
832
                        from_block_idx, present = \
798
833
                            state._find_block_index_from_key(from_key)
811
846
 
812
847
                        # Grab a copy since move_one may update the list.
813
848
                        for entry in from_block[1][:]:
814
 
                            assert entry[0][0] == from_dir
 
849
                            if not (entry[0][0] == from_dir):
 
850
                                raise AssertionError()
815
851
                            cur_details = entry[1][0]
816
852
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
817
853
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
842
878
                rollback_rename()
843
879
                raise
844
880
            result.append((from_rel, to_rel))
845
 
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
881
            state._mark_modified()
846
882
            self._make_dirty(reset_inventory=False)
847
883
 
848
884
        return result
858
894
    @needs_read_lock
859
895
    def path2id(self, path):
860
896
        """Return the id for path in this tree."""
 
897
        if isinstance(path, list):
 
898
            if path == []:
 
899
                path = [""]
 
900
            path = osutils.pathjoin(*path)
861
901
        path = path.strip('/')
862
902
        entry = self._get_entry(path=path)
863
903
        if entry == (None, None):
876
916
        for tree in trees:
877
917
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
878
918
                parents):
879
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
919
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
920
                    trees, require_versioned)
880
921
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
881
922
        # -- make all paths utf8 --
882
923
        paths_utf8 = set()
940
981
                    all_versioned = False
941
982
                    break
942
983
            if not all_versioned:
943
 
                raise errors.PathsNotVersionedError(paths)
 
984
                raise errors.PathsNotVersionedError(
 
985
                    [p.decode('utf-8') for p in paths])
944
986
        # -- remove redundancy in supplied paths to prevent over-scanning --
945
 
        search_paths = set()
946
 
        for path in paths:
947
 
            other_paths = paths.difference(set([path]))
948
 
            if not osutils.is_inside_any(other_paths, path):
949
 
                # this is a top level path, we must check it.
950
 
                search_paths.add(path)
951
 
        # sketch: 
 
987
        search_paths = osutils.minimum_path_selection(paths)
 
988
        # sketch:
952
989
        # for all search_indexs in each path at or under each element of
953
990
        # search_paths, if the detail is relocated: add the id, and add the
954
991
        # relocated path as one to search if its not searched already. If the
1000
1037
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1001
1038
            for dir_name in split_paths:
1002
1039
                if dir_name not in found_dir_names:
1003
 
                    raise errors.PathsNotVersionedError(paths)
 
1040
                    raise errors.PathsNotVersionedError(
 
1041
                        [p.decode('utf-8') for p in paths])
1004
1042
 
1005
1043
        for dir_name_id, trees_info in found.iteritems():
1006
1044
            for index in search_indexes:
1010
1048
 
1011
1049
    def read_working_inventory(self):
1012
1050
        """Read the working inventory.
1013
 
        
 
1051
 
1014
1052
        This is a meaningless operation for dirstate, but we obey it anyhow.
1015
1053
        """
1016
 
        return self.inventory
 
1054
        return self.root_inventory
1017
1055
 
1018
1056
    @needs_read_lock
1019
1057
    def revision_tree(self, revision_id):
1021
1059
 
1022
1060
        WorkingTree4 supplies revision_trees for any basis tree.
1023
1061
        """
1024
 
        revision_id = osutils.safe_revision_id(revision_id)
1025
1062
        dirstate = self.current_dirstate()
1026
1063
        parent_ids = dirstate.get_parent_ids()
1027
1064
        if revision_id not in parent_ids:
1034
1071
    @needs_tree_write_lock
1035
1072
    def set_last_revision(self, new_revision):
1036
1073
        """Change the last revision in the working tree."""
1037
 
        new_revision = osutils.safe_revision_id(new_revision)
1038
1074
        parents = self.get_parent_ids()
1039
 
        if new_revision in (NULL_REVISION, None):
1040
 
            assert len(parents) < 2, (
1041
 
                "setting the last parent to none with a pending merge is "
1042
 
                "unsupported.")
 
1075
        if new_revision in (_mod_revision.NULL_REVISION, None):
 
1076
            if len(parents) >= 2:
 
1077
                raise AssertionError(
 
1078
                    "setting the last parent to none with a pending merge is "
 
1079
                    "unsupported.")
1043
1080
            self.set_parent_ids([])
1044
1081
        else:
1045
1082
            self.set_parent_ids([new_revision] + parents[1:],
1048
1085
    @needs_tree_write_lock
1049
1086
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1050
1087
        """Set the parent ids to revision_ids.
1051
 
        
 
1088
 
1052
1089
        See also set_parent_trees. This api will try to retrieve the tree data
1053
1090
        for each element of revision_ids from the trees repository. If you have
1054
1091
        tree data already available, it is more efficient to use
1058
1095
        :param revision_ids: The revision_ids to set as the parent ids of this
1059
1096
            working tree. Any of these may be ghosts.
1060
1097
        """
1061
 
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1062
1098
        trees = []
1063
1099
        for revision_id in revision_ids:
1064
1100
            try:
1087
1123
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1088
1124
        real_trees = []
1089
1125
        ghosts = []
 
1126
 
 
1127
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1128
        graph = self.branch.repository.get_graph()
 
1129
        heads = graph.heads(parent_ids)
 
1130
        accepted_revisions = set()
 
1131
 
1090
1132
        # convert absent trees to the null tree, which we convert back to
1091
1133
        # missing on access.
1092
1134
        for rev_id, tree in parents_list:
1093
 
            rev_id = osutils.safe_revision_id(rev_id)
 
1135
            if len(accepted_revisions) > 0:
 
1136
                # we always accept the first tree
 
1137
                if rev_id in accepted_revisions or rev_id not in heads:
 
1138
                    # We have already included either this tree, or its
 
1139
                    # descendent, so we skip it.
 
1140
                    continue
1094
1141
            _mod_revision.check_not_reserved_id(rev_id)
1095
1142
            if tree is not None:
1096
1143
                real_trees.append((rev_id, tree))
1097
1144
            else:
1098
1145
                real_trees.append((rev_id,
1099
 
                    self.branch.repository.revision_tree(None)))
 
1146
                    self.branch.repository.revision_tree(
 
1147
                        _mod_revision.NULL_REVISION)))
1100
1148
                ghosts.append(rev_id)
1101
 
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1149
            accepted_revisions.add(rev_id)
 
1150
        updated = False
 
1151
        if (len(real_trees) == 1
 
1152
            and not ghosts
 
1153
            and self.branch.repository._format.fast_deltas
 
1154
            and isinstance(real_trees[0][1],
 
1155
                revisiontree.InventoryRevisionTree)
 
1156
            and self.get_parent_ids()):
 
1157
            rev_id, rev_tree = real_trees[0]
 
1158
            basis_id = self.get_parent_ids()[0]
 
1159
            # There are times when basis_tree won't be in
 
1160
            # self.branch.repository, (switch, for example)
 
1161
            try:
 
1162
                basis_tree = self.branch.repository.revision_tree(basis_id)
 
1163
            except errors.NoSuchRevision:
 
1164
                # Fall back to the set_parent_trees(), since we can't use
 
1165
                # _make_delta if we can't get the RevisionTree
 
1166
                pass
 
1167
            else:
 
1168
                delta = rev_tree.root_inventory._make_delta(
 
1169
                    basis_tree.root_inventory)
 
1170
                dirstate.update_basis_by_delta(delta, rev_id)
 
1171
                updated = True
 
1172
        if not updated:
 
1173
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1102
1174
        self._make_dirty(reset_inventory=False)
1103
1175
 
1104
1176
    def _set_root_id(self, file_id):
1108
1180
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1109
1181
            self._make_dirty(reset_inventory=True)
1110
1182
 
 
1183
    def _sha_from_stat(self, path, stat_result):
 
1184
        """Get a sha digest from the tree's stat cache.
 
1185
 
 
1186
        The default implementation assumes no stat cache is present.
 
1187
 
 
1188
        :param path: The path.
 
1189
        :param stat_result: The stat result being looked up.
 
1190
        """
 
1191
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1192
 
1111
1193
    @needs_read_lock
1112
1194
    def supports_tree_reference(self):
1113
1195
        return self._repo_supports_tree_reference
1115
1197
    def unlock(self):
1116
1198
        """Unlock in format 4 trees needs to write the entire dirstate."""
1117
1199
        if self._control_files._lock_count == 1:
 
1200
            # do non-implementation specific cleanup
 
1201
            self._cleanup()
 
1202
 
1118
1203
            # eventually we should do signature checking during read locks for
1119
1204
            # dirstate updates.
1120
1205
            if self._control_files._lock_mode == 'w':
1150
1235
            return
1151
1236
        state = self.current_dirstate()
1152
1237
        state._read_dirblocks_if_needed()
1153
 
        ids_to_unversion = set()
1154
 
        for file_id in file_ids:
1155
 
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1238
        ids_to_unversion = set(file_ids)
1156
1239
        paths_to_unversion = set()
1157
1240
        # sketch:
1158
1241
        # check if the root is to be unversioned, if so, assert for now.
1185
1268
                # just forget the whole block.
1186
1269
                entry_index = 0
1187
1270
                while entry_index < len(block[1]):
1188
 
                    # Mark this file id as having been removed
1189
1271
                    entry = block[1][entry_index]
1190
 
                    ids_to_unversion.discard(entry[0][2])
1191
 
                    if (entry[1][0][0] == 'a'
1192
 
                        or not state._make_absent(entry)):
 
1272
                    if entry[1][0][0] in 'ar':
 
1273
                        # don't remove absent or renamed entries
1193
1274
                        entry_index += 1
 
1275
                    else:
 
1276
                        # Mark this file id as having been removed
 
1277
                        ids_to_unversion.discard(entry[0][2])
 
1278
                        if not state._make_absent(entry):
 
1279
                            # The block has not shrunk.
 
1280
                            entry_index += 1
1194
1281
                # go to the next block. (At the moment we dont delete empty
1195
1282
                # dirblocks)
1196
1283
                block_index += 1
1217
1304
        # have to change the legacy inventory too.
1218
1305
        if self._inventory is not None:
1219
1306
            for file_id in file_ids:
1220
 
                self._inventory.remove_recursive_id(file_id)
 
1307
                if self._inventory.has_id(file_id):
 
1308
                    self._inventory.remove_recursive_id(file_id)
 
1309
 
 
1310
    @needs_tree_write_lock
 
1311
    def rename_one(self, from_rel, to_rel, after=False):
 
1312
        """See WorkingTree.rename_one"""
 
1313
        self.flush()
 
1314
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
 
1315
 
 
1316
    @needs_tree_write_lock
 
1317
    def apply_inventory_delta(self, changes):
 
1318
        """See MutableTree.apply_inventory_delta"""
 
1319
        state = self.current_dirstate()
 
1320
        state.update_by_delta(changes)
 
1321
        self._make_dirty(reset_inventory=True)
 
1322
 
 
1323
    def update_basis_by_delta(self, new_revid, delta):
 
1324
        """See MutableTree.update_basis_by_delta."""
 
1325
        if self.last_revision() == new_revid:
 
1326
            raise AssertionError()
 
1327
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1221
1328
 
1222
1329
    @needs_read_lock
1223
1330
    def _validate(self):
1226
1333
    @needs_tree_write_lock
1227
1334
    def _write_inventory(self, inv):
1228
1335
        """Write inventory as the current inventory."""
1229
 
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
1230
 
        self.current_dirstate().set_state_from_inventory(inv)
1231
 
        self._make_dirty(reset_inventory=False)
1232
 
        if self._inventory is not None:
 
1336
        if self._dirty:
 
1337
            raise AssertionError("attempting to write an inventory when the "
 
1338
                "dirstate is dirty will lose pending changes")
 
1339
        had_inventory = self._inventory is not None
 
1340
        # Setting self._inventory = None forces the dirstate to regenerate the
 
1341
        # working inventory. We do this because self.inventory may be inv, or
 
1342
        # may have been modified, and either case would prevent a clean delta
 
1343
        # being created.
 
1344
        self._inventory = None
 
1345
        # generate a delta,
 
1346
        delta = inv._make_delta(self.root_inventory)
 
1347
        # and apply it.
 
1348
        self.apply_inventory_delta(delta)
 
1349
        if had_inventory:
1233
1350
            self._inventory = inv
1234
1351
        self.flush()
1235
1352
 
1236
 
 
1237
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1238
 
    """The first consolidated dirstate working tree format.
1239
 
 
1240
 
    This format:
1241
 
        - exists within a metadir controlling .bzr
1242
 
        - includes an explicit version marker for the workingtree control
1243
 
          files, separate from the BzrDir format
1244
 
        - modifies the hash cache format
1245
 
        - is new in bzr 0.15
1246
 
        - uses a LockDir to guard access to it.
1247
 
    """
1248
 
 
1249
 
    upgrade_recommended = False
1250
 
 
1251
 
    def get_format_string(self):
1252
 
        """See WorkingTreeFormat.get_format_string()."""
1253
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1254
 
 
1255
 
    def get_format_description(self):
1256
 
        """See WorkingTreeFormat.get_format_description()."""
1257
 
        return "Working tree format 4"
1258
 
 
1259
 
    def initialize(self, a_bzrdir, revision_id=None):
 
1353
    @needs_tree_write_lock
 
1354
    def reset_state(self, revision_ids=None):
 
1355
        """Reset the state of the working tree.
 
1356
 
 
1357
        This does a hard-reset to a last-known-good state. This is a way to
 
1358
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
 
1359
        """
 
1360
        if revision_ids is None:
 
1361
            revision_ids = self.get_parent_ids()
 
1362
        if not revision_ids:
 
1363
            base_tree = self.branch.repository.revision_tree(
 
1364
                _mod_revision.NULL_REVISION)
 
1365
            trees = []
 
1366
        else:
 
1367
            trees = zip(revision_ids,
 
1368
                        self.branch.repository.revision_trees(revision_ids))
 
1369
            base_tree = trees[0][1]
 
1370
        state = self.current_dirstate()
 
1371
        # We don't support ghosts yet
 
1372
        state.set_state_from_scratch(base_tree.root_inventory, trees, [])
 
1373
 
 
1374
 
 
1375
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
 
1376
 
 
1377
    def __init__(self, tree):
 
1378
        self.tree = tree
 
1379
 
 
1380
    def sha1(self, abspath):
 
1381
        """See dirstate.SHA1Provider.sha1()."""
 
1382
        filters = self.tree._content_filter_stack(
 
1383
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1384
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
 
1385
 
 
1386
    def stat_and_sha1(self, abspath):
 
1387
        """See dirstate.SHA1Provider.stat_and_sha1()."""
 
1388
        filters = self.tree._content_filter_stack(
 
1389
            self.tree.relpath(osutils.safe_unicode(abspath)))
 
1390
        file_obj = file(abspath, 'rb', 65000)
 
1391
        try:
 
1392
            statvalue = os.fstat(file_obj.fileno())
 
1393
            if filters:
 
1394
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
 
1395
            sha1 = osutils.size_sha_file(file_obj)[1]
 
1396
        finally:
 
1397
            file_obj.close()
 
1398
        return statvalue, sha1
 
1399
 
 
1400
 
 
1401
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
 
1402
    """Dirstate working tree that supports content filtering.
 
1403
 
 
1404
    The dirstate holds the hash and size of the canonical form of the file, 
 
1405
    and most methods must return that.
 
1406
    """
 
1407
 
 
1408
    def _file_content_summary(self, path, stat_result):
 
1409
        # This is to support the somewhat obsolete path_content_summary method
 
1410
        # with content filtering: see
 
1411
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
 
1412
        #
 
1413
        # If the dirstate cache is up to date and knows the hash and size,
 
1414
        # return that.
 
1415
        # Otherwise if there are no content filters, return the on-disk size
 
1416
        # and leave the hash blank.
 
1417
        # Otherwise, read and filter the on-disk file and use its size and
 
1418
        # hash.
 
1419
        #
 
1420
        # The dirstate doesn't store the size of the canonical form so we
 
1421
        # can't trust it for content-filtered trees.  We just return None.
 
1422
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
 
1423
        executable = self._is_executable_from_path_and_stat(path, stat_result)
 
1424
        return ('file', None, executable, dirstate_sha1)
 
1425
 
 
1426
 
 
1427
class WorkingTree4(DirStateWorkingTree):
 
1428
    """This is the Format 4 working tree.
 
1429
 
 
1430
    This differs from WorkingTree by:
 
1431
     - Having a consolidated internal dirstate, stored in a
 
1432
       randomly-accessible sorted file on disk.
 
1433
     - Not having a regular inventory attribute.  One can be synthesized
 
1434
       on demand but this is expensive and should be avoided.
 
1435
 
 
1436
    This is new in bzr 0.15.
 
1437
    """
 
1438
 
 
1439
 
 
1440
class WorkingTree5(ContentFilteringDirStateWorkingTree):
 
1441
    """This is the Format 5 working tree.
 
1442
 
 
1443
    This differs from WorkingTree4 by:
 
1444
     - Supporting content filtering.
 
1445
 
 
1446
    This is new in bzr 1.11.
 
1447
    """
 
1448
 
 
1449
 
 
1450
class WorkingTree6(ContentFilteringDirStateWorkingTree):
 
1451
    """This is the Format 6 working tree.
 
1452
 
 
1453
    This differs from WorkingTree5 by:
 
1454
     - Supporting a current view that may mask the set of files in a tree
 
1455
       impacted by most user operations.
 
1456
 
 
1457
    This is new in bzr 1.14.
 
1458
    """
 
1459
 
 
1460
    def _make_views(self):
 
1461
        return views.PathBasedViews(self)
 
1462
 
 
1463
 
 
1464
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
 
1465
 
 
1466
    missing_parent_conflicts = True
 
1467
 
 
1468
    supports_versioned_directories = True
 
1469
 
 
1470
    _lock_class = LockDir
 
1471
    _lock_file_name = 'lock'
 
1472
 
 
1473
    def _open_control_files(self, a_bzrdir):
 
1474
        transport = a_bzrdir.get_workingtree_transport(None)
 
1475
        return LockableFiles(transport, self._lock_file_name,
 
1476
                             self._lock_class)
 
1477
 
 
1478
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1479
                   accelerator_tree=None, hardlink=False):
1260
1480
        """See WorkingTreeFormat.initialize().
1261
1481
 
1262
1482
        :param revision_id: allows creating a working tree at a different
1263
 
        revision than the branch is at.
 
1483
            revision than the branch is at.
 
1484
        :param accelerator_tree: A tree which can be used for retrieving file
 
1485
            contents more quickly than the revision tree, i.e. a workingtree.
 
1486
            The revision tree will be used for cases where accelerator_tree's
 
1487
            content is different.
 
1488
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1489
            where possible.
1264
1490
 
1265
1491
        These trees get an initial random root id, if their repository supports
1266
1492
        rich root data, TREE_ROOT otherwise.
1267
1493
        """
1268
 
        revision_id = osutils.safe_revision_id(revision_id)
1269
1494
        if not isinstance(a_bzrdir.transport, LocalTransport):
1270
1495
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1271
1496
        transport = a_bzrdir.get_workingtree_transport(self)
1272
1497
        control_files = self._open_control_files(a_bzrdir)
1273
1498
        control_files.create_lock()
1274
1499
        control_files.lock_write()
1275
 
        control_files.put_utf8('format', self.get_format_string())
1276
 
        branch = a_bzrdir.open_branch()
 
1500
        transport.put_bytes('format', self.as_string(),
 
1501
            mode=a_bzrdir._get_file_mode())
 
1502
        if from_branch is not None:
 
1503
            branch = from_branch
 
1504
        else:
 
1505
            branch = a_bzrdir.open_branch()
1277
1506
        if revision_id is None:
1278
1507
            revision_id = branch.last_revision()
1279
1508
        local_path = transport.local_abspath('dirstate')
1281
1510
        state = dirstate.DirState.initialize(local_path)
1282
1511
        state.unlock()
1283
1512
        del state
1284
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1513
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1285
1514
                         branch,
1286
1515
                         _format=self,
1287
1516
                         _bzrdir=a_bzrdir,
1289
1518
        wt._new_tree()
1290
1519
        wt.lock_tree_write()
1291
1520
        try:
1292
 
            if revision_id in (None, NULL_REVISION):
 
1521
            self._init_custom_control_files(wt)
 
1522
            if revision_id in (None, _mod_revision.NULL_REVISION):
1293
1523
                if branch.repository.supports_rich_root():
1294
1524
                    wt._set_root_id(generate_ids.gen_root_id())
1295
1525
                else:
1296
1526
                    wt._set_root_id(ROOT_ID)
1297
1527
                wt.flush()
1298
 
            wt.set_last_revision(revision_id)
1299
 
            wt.flush()
1300
 
            basis = wt.basis_tree()
 
1528
            basis = None
 
1529
            # frequently, we will get here due to branching.  The accelerator
 
1530
            # tree will be the tree from the branch, so the desired basis
 
1531
            # tree will often be a parent of the accelerator tree.
 
1532
            if accelerator_tree is not None:
 
1533
                try:
 
1534
                    basis = accelerator_tree.revision_tree(revision_id)
 
1535
                except errors.NoSuchRevision:
 
1536
                    pass
 
1537
            if basis is None:
 
1538
                basis = branch.repository.revision_tree(revision_id)
 
1539
            if revision_id == _mod_revision.NULL_REVISION:
 
1540
                parents_list = []
 
1541
            else:
 
1542
                parents_list = [(revision_id, basis)]
1301
1543
            basis.lock_read()
1302
 
            # if the basis has a root id we have to use that; otherwise we use
1303
 
            # a new random one
1304
 
            basis_root_id = basis.get_root_id()
1305
 
            if basis_root_id is not None:
1306
 
                wt._set_root_id(basis_root_id)
 
1544
            try:
 
1545
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1307
1546
                wt.flush()
1308
 
            transform.build_tree(basis, wt)
1309
 
            basis.unlock()
 
1547
                # if the basis has a root id we have to use that; otherwise we
 
1548
                # use a new random one
 
1549
                basis_root_id = basis.get_root_id()
 
1550
                if basis_root_id is not None:
 
1551
                    wt._set_root_id(basis_root_id)
 
1552
                    wt.flush()
 
1553
                if wt.supports_content_filtering():
 
1554
                    # The original tree may not have the same content filters
 
1555
                    # applied so we can't safely build the inventory delta from
 
1556
                    # the source tree.
 
1557
                    delta_from_tree = False
 
1558
                else:
 
1559
                    delta_from_tree = True
 
1560
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1561
                # because wt4.apply_inventory_delta does not mutate the input
 
1562
                # inventory entries.
 
1563
                transform.build_tree(basis, wt, accelerator_tree,
 
1564
                                     hardlink=hardlink,
 
1565
                                     delta_from_tree=delta_from_tree)
 
1566
                for hook in MutableTree.hooks['post_build_tree']:
 
1567
                    hook(wt)
 
1568
            finally:
 
1569
                basis.unlock()
1310
1570
        finally:
1311
1571
            control_files.unlock()
1312
1572
            wt.unlock()
1313
1573
        return wt
1314
1574
 
 
1575
    def _init_custom_control_files(self, wt):
 
1576
        """Subclasses with custom control files should override this method.
 
1577
 
 
1578
        The working tree and control files are locked for writing when this
 
1579
        method is called.
 
1580
 
 
1581
        :param wt: the WorkingTree object
 
1582
        """
 
1583
 
 
1584
    def open(self, a_bzrdir, _found=False):
 
1585
        """Return the WorkingTree object for a_bzrdir
 
1586
 
 
1587
        _found is a private parameter, do not use it. It is used to indicate
 
1588
               if format probing has already been done.
 
1589
        """
 
1590
        if not _found:
 
1591
            # we are being called directly and must probe.
 
1592
            raise NotImplementedError
 
1593
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1594
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1595
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
 
1596
        return wt
 
1597
 
1315
1598
    def _open(self, a_bzrdir, control_files):
1316
1599
        """Open the tree itself.
1317
1600
 
1318
1601
        :param a_bzrdir: the dir for the tree.
1319
1602
        :param control_files: the control files for the tree.
1320
1603
        """
1321
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1604
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1322
1605
                           branch=a_bzrdir.open_branch(),
1323
1606
                           _format=self,
1324
1607
                           _bzrdir=a_bzrdir,
1325
1608
                           _control_files=control_files)
1326
1609
 
1327
1610
    def __get_matchingbzrdir(self):
 
1611
        return self._get_matchingbzrdir()
 
1612
 
 
1613
    def _get_matchingbzrdir(self):
 
1614
        """Overrideable method to get a bzrdir for testing."""
1328
1615
        # please test against something that will let us do tree references
1329
 
        return bzrdir.format_registry.make_bzrdir(
1330
 
            'dirstate-with-subtree')
 
1616
        return controldir.format_registry.make_bzrdir(
 
1617
            'development-subtree')
1331
1618
 
1332
1619
    _matchingbzrdir = property(__get_matchingbzrdir)
1333
1620
 
1334
1621
 
1335
 
class DirStateRevisionTree(Tree):
1336
 
    """A revision tree pulling the inventory from a dirstate."""
 
1622
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
 
1623
    """The first consolidated dirstate working tree format.
 
1624
 
 
1625
    This format:
 
1626
        - exists within a metadir controlling .bzr
 
1627
        - includes an explicit version marker for the workingtree control
 
1628
          files, separate from the ControlDir format
 
1629
        - modifies the hash cache format
 
1630
        - is new in bzr 0.15
 
1631
        - uses a LockDir to guard access to it.
 
1632
    """
 
1633
 
 
1634
    upgrade_recommended = False
 
1635
 
 
1636
    _tree_class = WorkingTree4
 
1637
 
 
1638
    @classmethod
 
1639
    def get_format_string(cls):
 
1640
        """See WorkingTreeFormat.get_format_string()."""
 
1641
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1642
 
 
1643
    def get_format_description(self):
 
1644
        """See WorkingTreeFormat.get_format_description()."""
 
1645
        return "Working tree format 4"
 
1646
 
 
1647
 
 
1648
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
 
1649
    """WorkingTree format supporting content filtering.
 
1650
    """
 
1651
 
 
1652
    upgrade_recommended = False
 
1653
 
 
1654
    _tree_class = WorkingTree5
 
1655
 
 
1656
    @classmethod
 
1657
    def get_format_string(cls):
 
1658
        """See WorkingTreeFormat.get_format_string()."""
 
1659
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
 
1660
 
 
1661
    def get_format_description(self):
 
1662
        """See WorkingTreeFormat.get_format_description()."""
 
1663
        return "Working tree format 5"
 
1664
 
 
1665
    def supports_content_filtering(self):
 
1666
        return True
 
1667
 
 
1668
 
 
1669
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
 
1670
    """WorkingTree format supporting views.
 
1671
    """
 
1672
 
 
1673
    upgrade_recommended = False
 
1674
 
 
1675
    _tree_class = WorkingTree6
 
1676
 
 
1677
    @classmethod
 
1678
    def get_format_string(cls):
 
1679
        """See WorkingTreeFormat.get_format_string()."""
 
1680
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
 
1681
 
 
1682
    def get_format_description(self):
 
1683
        """See WorkingTreeFormat.get_format_description()."""
 
1684
        return "Working tree format 6"
 
1685
 
 
1686
    def _init_custom_control_files(self, wt):
 
1687
        """Subclasses with custom control files should override this method."""
 
1688
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
 
1689
 
 
1690
    def supports_content_filtering(self):
 
1691
        return True
 
1692
 
 
1693
    def supports_views(self):
 
1694
        return True
 
1695
 
 
1696
    def _get_matchingbzrdir(self):
 
1697
        """Overrideable method to get a bzrdir for testing."""
 
1698
        # We use 'development-subtree' instead of '2a', because we have a
 
1699
        # few tests that want to test tree references
 
1700
        return bzrdir.format_registry.make_bzrdir('development-subtree')
 
1701
 
 
1702
 
 
1703
class DirStateRevisionTree(InventoryTree):
 
1704
    """A revision tree pulling the inventory from a dirstate.
 
1705
    
 
1706
    Note that this is one of the historical (ie revision) trees cached in the
 
1707
    dirstate for easy access, not the workingtree.
 
1708
    """
1337
1709
 
1338
1710
    def __init__(self, dirstate, revision_id, repository):
1339
1711
        self._dirstate = dirstate
1340
 
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1712
        self._revision_id = revision_id
1341
1713
        self._repository = repository
1342
1714
        self._inventory = None
1343
1715
        self._locked = 0
1344
1716
        self._dirstate_locked = False
 
1717
        self._repo_supports_tree_reference = getattr(
 
1718
            repository._format, "supports_tree_reference",
 
1719
            False)
1345
1720
 
1346
1721
    def __repr__(self):
1347
1722
        return "<%s of %s in %s>" % \
1350
1725
    def annotate_iter(self, file_id,
1351
1726
                      default_revision=_mod_revision.CURRENT_REVISION):
1352
1727
        """See Tree.annotate_iter"""
1353
 
        w = self._get_weave(file_id)
1354
 
        return w.annotate_iter(self.inventory[file_id].revision)
 
1728
        text_key = (file_id, self.get_file_revision(file_id))
 
1729
        annotations = self._repository.texts.annotate(text_key)
 
1730
        return [(key[-1], line) for (key, line) in annotations]
1355
1731
 
1356
 
    def _get_ancestors(self, default_revision):
1357
 
        return set(self._repository.get_ancestry(self._revision_id,
1358
 
                                                 topo_sorted=False))
1359
1732
    def _comparison_data(self, entry, path):
1360
1733
        """See Tree._comparison_data."""
1361
1734
        if entry is None:
1378
1751
    def get_root_id(self):
1379
1752
        return self.path2id('')
1380
1753
 
 
1754
    def id2path(self, file_id):
 
1755
        "Convert a file-id to a path."
 
1756
        entry = self._get_entry(file_id=file_id)
 
1757
        if entry == (None, None):
 
1758
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1759
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1760
        return path_utf8.decode('utf8')
 
1761
 
 
1762
    def iter_references(self):
 
1763
        if not self._repo_supports_tree_reference:
 
1764
            # When the repo doesn't support references, we will have nothing to
 
1765
            # return
 
1766
            return iter([])
 
1767
        # Otherwise, fall back to the default implementation
 
1768
        return super(DirStateRevisionTree, self).iter_references()
 
1769
 
1381
1770
    def _get_parent_index(self):
1382
1771
        """Return the index in the dirstate referenced by this tree."""
1383
1772
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1388
1777
        If either file_id or path is supplied, it is used as the key to lookup.
1389
1778
        If both are supplied, the fastest lookup is used, and an error is
1390
1779
        raised if they do not both point at the same row.
1391
 
        
 
1780
 
1392
1781
        :param file_id: An optional unicode file_id to be looked up.
1393
1782
        :param path: An optional unicode path to be looked up.
1394
1783
        :return: The dirstate row tuple for path/file_id, or (None, None)
1395
1784
        """
1396
1785
        if file_id is None and path is None:
1397
1786
            raise errors.BzrError('must supply file_id or path')
1398
 
        file_id = osutils.safe_file_id(file_id)
1399
1787
        if path is not None:
1400
1788
            path = path.encode('utf8')
1401
1789
        parent_index = self._get_parent_index()
1402
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1790
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
 
1791
            path_utf8=path)
1403
1792
 
1404
1793
    def _generate_inventory(self):
1405
1794
        """Create and set self.inventory from the dirstate object.
1409
1798
 
1410
1799
        This is relatively expensive: we have to walk the entire dirstate.
1411
1800
        """
1412
 
        assert self._locked, 'cannot generate inventory of an unlocked '\
1413
 
            'dirstate revision tree'
 
1801
        if not self._locked:
 
1802
            raise AssertionError(
 
1803
                'cannot generate inventory of an unlocked '
 
1804
                'dirstate revision tree')
1414
1805
        # separate call for profiling - makes it clear where the costs are.
1415
1806
        self._dirstate._read_dirblocks_if_needed()
1416
 
        assert self._revision_id in self._dirstate.get_parent_ids(), \
1417
 
            'parent %s has disappeared from %s' % (
1418
 
            self._revision_id, self._dirstate.get_parent_ids())
 
1807
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1808
            raise AssertionError(
 
1809
                'parent %s has disappeared from %s' % (
 
1810
                self._revision_id, self._dirstate.get_parent_ids()))
1419
1811
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1420
1812
        # This is identical now to the WorkingTree _generate_inventory except
1421
1813
        # for the tree index use.
1422
1814
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1423
1815
        current_id = root_key[2]
1424
 
        assert current_entry[parent_index][0] == 'd'
 
1816
        if current_entry[parent_index][0] != 'd':
 
1817
            raise AssertionError()
1425
1818
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1426
1819
        inv.root.revision = current_entry[parent_index][4]
1427
1820
        # Turn some things into local variables
1458
1851
                elif kind == 'directory':
1459
1852
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1460
1853
                elif kind == 'symlink':
1461
 
                    inv_entry.executable = False
1462
 
                    inv_entry.text_size = None
1463
1854
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1464
1855
                elif kind == 'tree-reference':
1465
1856
                    inv_entry.reference_revision = fingerprint or None
1467
1858
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1468
1859
                            % entry)
1469
1860
                # These checks cost us around 40ms on a 55k entry tree
1470
 
                assert file_id not in inv_byid
1471
 
                assert name_unicode not in parent_ie.children
 
1861
                if file_id in inv_byid:
 
1862
                    raise AssertionError('file_id %s already in'
 
1863
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1864
                if name_unicode in parent_ie.children:
 
1865
                    raise AssertionError('name %r already in parent'
 
1866
                        % (name_unicode,))
1472
1867
                inv_byid[file_id] = inv_entry
1473
1868
                parent_ie.children[name_unicode] = inv_entry
1474
1869
        self._inventory = inv
1481
1876
        # Make sure the file exists
1482
1877
        entry = self._get_entry(file_id, path=path)
1483
1878
        if entry == (None, None): # do we raise?
1484
 
            return None
 
1879
            raise errors.NoSuchId(self, file_id)
1485
1880
        parent_index = self._get_parent_index()
1486
1881
        last_changed_revision = entry[1][parent_index][4]
1487
 
        return self._repository.get_revision(last_changed_revision).timestamp
 
1882
        try:
 
1883
            rev = self._repository.get_revision(last_changed_revision)
 
1884
        except errors.NoSuchRevision:
 
1885
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
 
1886
        return rev.timestamp
1488
1887
 
1489
1888
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1490
1889
        entry = self._get_entry(file_id=file_id, path=path)
1494
1893
            return parent_details[1]
1495
1894
        return None
1496
1895
 
1497
 
    @symbol_versioning.deprecated_method(symbol_versioning.zero_nineteen)
1498
 
    def get_weave(self, file_id):
1499
 
        return self._get_weave(file_id)
1500
 
 
1501
 
    def _get_weave(self, file_id):
1502
 
        return self._repository.weave_store.get_weave(file_id,
1503
 
                self._repository.get_transaction())
1504
 
 
1505
 
    def get_file(self, file_id):
 
1896
    @needs_read_lock
 
1897
    def get_file_revision(self, file_id):
 
1898
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1899
        return inv[inv_file_id].revision
 
1900
 
 
1901
    def get_file(self, file_id, path=None):
1506
1902
        return StringIO(self.get_file_text(file_id))
1507
1903
 
1508
 
    def get_file_lines(self, file_id):
1509
 
        ie = self.inventory[file_id]
1510
 
        return self._get_weave(file_id).get_lines(ie.revision)
1511
 
 
1512
1904
    def get_file_size(self, file_id):
1513
 
        return self.inventory[file_id].text_size
 
1905
        """See Tree.get_file_size"""
 
1906
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1907
        return inv[inv_file_id].text_size
1514
1908
 
1515
 
    def get_file_text(self, file_id):
1516
 
        return ''.join(self.get_file_lines(file_id))
 
1909
    def get_file_text(self, file_id, path=None):
 
1910
        content = None
 
1911
        for _, content_iter in self.iter_files_bytes([(file_id, None)]):
 
1912
            if content is not None:
 
1913
                raise AssertionError('iter_files_bytes returned'
 
1914
                    ' too many entries')
 
1915
            # For each entry returned by iter_files_bytes, we must consume the
 
1916
            # content_iter before we step the files iterator.
 
1917
            content = ''.join(content_iter)
 
1918
        if content is None:
 
1919
            raise AssertionError('iter_files_bytes did not return'
 
1920
                ' the requested data')
 
1921
        return content
1517
1922
 
1518
1923
    def get_reference_revision(self, file_id, path=None):
1519
 
        return self.inventory[file_id].reference_revision
1520
 
 
1521
 
    def get_symlink_target(self, file_id):
 
1924
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1925
        return inv[inv_file_id].reference_revision
 
1926
 
 
1927
    def iter_files_bytes(self, desired_files):
 
1928
        """See Tree.iter_files_bytes.
 
1929
 
 
1930
        This version is implemented on top of Repository.iter_files_bytes"""
 
1931
        parent_index = self._get_parent_index()
 
1932
        repo_desired_files = []
 
1933
        for file_id, identifier in desired_files:
 
1934
            entry = self._get_entry(file_id)
 
1935
            if entry == (None, None):
 
1936
                raise errors.NoSuchId(self, file_id)
 
1937
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1938
                                       identifier))
 
1939
        return self._repository.iter_files_bytes(repo_desired_files)
 
1940
 
 
1941
    def get_symlink_target(self, file_id, path=None):
1522
1942
        entry = self._get_entry(file_id=file_id)
1523
1943
        parent_index = self._get_parent_index()
1524
1944
        if entry[1][parent_index][0] != 'l':
1525
1945
            return None
1526
1946
        else:
1527
 
            # At present, none of the tree implementations supports non-ascii
1528
 
            # symlink targets. So we will just assume that the dirstate path is
1529
 
            # correct.
1530
 
            return entry[1][parent_index][1]
 
1947
            target = entry[1][parent_index][1]
 
1948
            target = target.decode('utf8')
 
1949
            return target
1531
1950
 
1532
1951
    def get_revision_id(self):
1533
1952
        """Return the revision id for this tree."""
1534
1953
        return self._revision_id
1535
1954
 
1536
 
    def _get_inventory(self):
 
1955
    def _get_root_inventory(self):
1537
1956
        if self._inventory is not None:
1538
1957
            return self._inventory
1539
1958
        self._must_be_locked()
1540
1959
        self._generate_inventory()
1541
1960
        return self._inventory
1542
1961
 
 
1962
    root_inventory = property(_get_root_inventory,
 
1963
                         doc="Inventory of this Tree")
 
1964
 
 
1965
    @deprecated_method(deprecated_in((2, 5, 0)))
 
1966
    def _get_inventory(self):
 
1967
        return self.root_inventory
 
1968
 
1543
1969
    inventory = property(_get_inventory,
1544
1970
                         doc="Inventory of this Tree")
1545
1971
 
1551
1977
        return bool(self.path2id(filename))
1552
1978
 
1553
1979
    def kind(self, file_id):
1554
 
        return self.inventory[file_id].kind
 
1980
        entry = self._get_entry(file_id=file_id)[1]
 
1981
        if entry is None:
 
1982
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1983
        parent_index = self._get_parent_index()
 
1984
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1985
 
 
1986
    def stored_kind(self, file_id):
 
1987
        """See Tree.stored_kind"""
 
1988
        return self.kind(file_id)
 
1989
 
 
1990
    def path_content_summary(self, path):
 
1991
        """See Tree.path_content_summary."""
 
1992
        inv, inv_file_id = self._path2inv_file_id(path)
 
1993
        if inv_file_id is None:
 
1994
            return ('missing', None, None, None)
 
1995
        entry = inv[inv_file_id]
 
1996
        kind = entry.kind
 
1997
        if kind == 'file':
 
1998
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1999
        elif kind == 'symlink':
 
2000
            return (kind, None, None, entry.symlink_target)
 
2001
        else:
 
2002
            return (kind, None, None, None)
1555
2003
 
1556
2004
    def is_executable(self, file_id, path=None):
1557
 
        ie = self.inventory[file_id]
 
2005
        inv, inv_file_id = self._unpack_file_id(file_id)
 
2006
        ie = inv[inv_file_id]
1558
2007
        if ie.kind != "file":
1559
 
            return None
 
2008
            return False
1560
2009
        return ie.executable
1561
2010
 
1562
 
    def list_files(self, include_root=False):
 
2011
    def is_locked(self):
 
2012
        return self._locked
 
2013
 
 
2014
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1563
2015
        # We use a standard implementation, because DirStateRevisionTree is
1564
2016
        # dealing with one of the parents of the current state
1565
 
        inv = self._get_inventory()
1566
 
        entries = inv.iter_entries()
1567
 
        if self.inventory.root is not None and not include_root:
 
2017
        if from_dir is None:
 
2018
            inv = self.root_inventory
 
2019
            from_dir_id = None
 
2020
        else:
 
2021
            inv, from_dir_id = self._path2inv_file_id(from_dir)
 
2022
            if from_dir_id is None:
 
2023
                # Directory not versioned
 
2024
                return
 
2025
        # FIXME: Support nested trees
 
2026
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
 
2027
        if inv.root is not None and not include_root and from_dir is None:
1568
2028
            entries.next()
1569
2029
        for path, entry in entries:
1570
2030
            yield path, 'V', entry.kind, entry.file_id, entry
1571
2031
 
1572
2032
    def lock_read(self):
1573
 
        """Lock the tree for a set of operations."""
 
2033
        """Lock the tree for a set of operations.
 
2034
 
 
2035
        :return: A bzrlib.lock.LogicalLockResult.
 
2036
        """
1574
2037
        if not self._locked:
1575
2038
            self._repository.lock_read()
1576
2039
            if self._dirstate._lock_token is None:
1577
2040
                self._dirstate.lock_read()
1578
2041
                self._dirstate_locked = True
1579
2042
        self._locked += 1
 
2043
        return LogicalLockResult(self.unlock)
1580
2044
 
1581
2045
    def _must_be_locked(self):
1582
2046
        if not self._locked:
1586
2050
    def path2id(self, path):
1587
2051
        """Return the id for path in this tree."""
1588
2052
        # lookup by path: faster than splitting and walking the ivnentory.
 
2053
        if isinstance(path, list):
 
2054
            if path == []:
 
2055
                path = [""]
 
2056
            path = osutils.pathjoin(*path)
1589
2057
        entry = self._get_entry(path=path)
1590
2058
        if entry == (None, None):
1591
2059
            return None
1603
2071
                self._dirstate_locked = False
1604
2072
            self._repository.unlock()
1605
2073
 
 
2074
    @needs_read_lock
 
2075
    def supports_tree_reference(self):
 
2076
        return self._repo_supports_tree_reference
 
2077
 
1606
2078
    def walkdirs(self, prefix=""):
1607
2079
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1608
 
        # implementation based on an inventory.  
 
2080
        # implementation based on an inventory.
1609
2081
        # This should be cleaned up to use the much faster Dirstate code
1610
2082
        # So for now, we just build up the parent inventory, and extract
1611
2083
        # it the same way RevisionTree does.
1612
2084
        _directory = 'directory'
1613
 
        inv = self._get_inventory()
 
2085
        inv = self._get_root_inventory()
1614
2086
        top_id = inv.path2id(prefix)
1615
2087
        if top_id is None:
1616
2088
            pending = []
1640
2112
 
1641
2113
class InterDirStateTree(InterTree):
1642
2114
    """Fast path optimiser for changes_from with dirstate trees.
1643
 
    
1644
 
    This is used only when both trees are in the dirstate working file, and 
1645
 
    the source is any parent within the dirstate, and the destination is 
 
2115
 
 
2116
    This is used only when both trees are in the dirstate working file, and
 
2117
    the source is any parent within the dirstate, and the destination is
1646
2118
    the current working tree of the same dirstate.
1647
2119
    """
1648
2120
    # this could be generalized to allow comparisons between any trees in the
1657
2129
    def make_source_parent_tree(source, target):
1658
2130
        """Change the source tree into a parent of the target."""
1659
2131
        revid = source.commit('record tree')
1660
 
        target.branch.repository.fetch(source.branch.repository, revid)
 
2132
        target.branch.fetch(source.branch, revid)
1661
2133
        target.set_parent_ids([revid])
1662
2134
        return target.basis_tree(), target
1663
2135
 
 
2136
    @classmethod
 
2137
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
 
2138
        result = klass.make_source_parent_tree(source, target)
 
2139
        result[1]._iter_changes = dirstate.ProcessEntryPython
 
2140
        return result
 
2141
 
 
2142
    @classmethod
 
2143
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
 
2144
                                                  target):
 
2145
        from bzrlib.tests.test__dirstate_helpers import \
 
2146
            compiled_dirstate_helpers_feature
 
2147
        test_case.requireFeature(compiled_dirstate_helpers_feature)
 
2148
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
 
2149
        result = klass.make_source_parent_tree(source, target)
 
2150
        result[1]._iter_changes = ProcessEntryC
 
2151
        return result
 
2152
 
1664
2153
    _matching_from_tree_format = WorkingTreeFormat4()
1665
2154
    _matching_to_tree_format = WorkingTreeFormat4()
1666
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
1667
 
 
1668
 
    def _iter_changes(self, include_unchanged=False,
 
2155
 
 
2156
    @classmethod
 
2157
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
 
2158
        # This method shouldn't be called, because we have python and C
 
2159
        # specific flavours.
 
2160
        raise NotImplementedError
 
2161
 
 
2162
    def iter_changes(self, include_unchanged=False,
1669
2163
                      specific_files=None, pb=None, extra_trees=[],
1670
2164
                      require_versioned=True, want_unversioned=False):
1671
2165
        """Return the changes from source to target.
1672
2166
 
1673
 
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
2167
        :return: An iterator that yields tuples. See InterTree.iter_changes
1674
2168
            for details.
1675
2169
        :param specific_files: An optional list of file paths to restrict the
1676
2170
            comparison to. When mapping filenames to ids, all matches in all
1687
2181
            output. An unversioned file is defined as one with (False, False)
1688
2182
            for the versioned pair.
1689
2183
        """
1690
 
        utf8_decode = cache_utf8._utf8_decode
1691
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1692
 
        cmp_by_dirs = dirstate.cmp_by_dirs
1693
 
        # NB: show_status depends on being able to pass in non-versioned files
1694
 
        # and report them as unknown
1695
2184
        # TODO: handle extra trees in the dirstate.
1696
 
        # TODO: handle comparisons as an empty tree as a different special
1697
 
        # case? mbp 20070226
1698
 
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
2185
        if (extra_trees or specific_files == []):
1699
2186
            # we can't fast-path these cases (yet)
1700
 
            for f in super(InterDirStateTree, self)._iter_changes(
 
2187
            return super(InterDirStateTree, self).iter_changes(
1701
2188
                include_unchanged, specific_files, pb, extra_trees,
1702
 
                require_versioned, want_unversioned=want_unversioned):
1703
 
                yield f
1704
 
            return
 
2189
                require_versioned, want_unversioned=want_unversioned)
1705
2190
        parent_ids = self.target.get_parent_ids()
1706
 
        assert (self.source._revision_id in parent_ids), \
1707
 
                "revision {%s} is not stored in {%s}, but %s " \
1708
 
                "can only be used for trees stored in the dirstate" \
1709
 
                % (self.source._revision_id, self.target, self._iter_changes)
 
2191
        if not (self.source._revision_id in parent_ids
 
2192
                or self.source._revision_id == _mod_revision.NULL_REVISION):
 
2193
            raise AssertionError(
 
2194
                "revision {%s} is not stored in {%s}, but %s "
 
2195
                "can only be used for trees stored in the dirstate"
 
2196
                % (self.source._revision_id, self.target, self.iter_changes))
1710
2197
        target_index = 0
1711
 
        if self.source._revision_id == NULL_REVISION:
 
2198
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1712
2199
            source_index = None
1713
2200
            indices = (target_index,)
1714
2201
        else:
1715
 
            assert (self.source._revision_id in parent_ids), \
1716
 
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
1717
 
                self.source._revision_id, parent_ids)
 
2202
            if not (self.source._revision_id in parent_ids):
 
2203
                raise AssertionError(
 
2204
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
2205
                    self.source._revision_id, parent_ids))
1718
2206
            source_index = 1 + parent_ids.index(self.source._revision_id)
1719
 
            indices = (source_index,target_index)
 
2207
            indices = (source_index, target_index)
1720
2208
        # -- make all specific_files utf8 --
1721
2209
        if specific_files:
1722
2210
            specific_files_utf8 = set()
1723
2211
            for path in specific_files:
 
2212
                # Note, if there are many specific files, using cache_utf8
 
2213
                # would be good here.
1724
2214
                specific_files_utf8.add(path.encode('utf8'))
1725
2215
            specific_files = specific_files_utf8
1726
2216
        else:
1727
2217
            specific_files = set([''])
1728
2218
        # -- specific_files is now a utf8 path set --
 
2219
 
1729
2220
        # -- get the state object and prepare it.
1730
2221
        state = self.target.current_dirstate()
1731
2222
        state._read_dirblocks_if_needed()
1732
 
        def _entries_for_path(path):
1733
 
            """Return a list with all the entries that match path for all ids.
1734
 
            """
1735
 
            dirname, basename = os.path.split(path)
1736
 
            key = (dirname, basename, '')
1737
 
            block_index, present = state._find_block_index_from_key(key)
1738
 
            if not present:
1739
 
                # the block which should contain path is absent.
1740
 
                return []
1741
 
            result = []
1742
 
            block = state._dirblocks[block_index][1]
1743
 
            entry_index, _ = state._find_entry_index(key, block)
1744
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1745
 
            while (entry_index < len(block) and
1746
 
                block[entry_index][0][0:2] == key[0:2]):
1747
 
                result.append(block[entry_index])
1748
 
                entry_index += 1
1749
 
            return result
1750
2223
        if require_versioned:
1751
2224
            # -- check all supplied paths are versioned in a search tree. --
1752
 
            all_versioned = True
 
2225
            not_versioned = []
1753
2226
            for path in specific_files:
1754
 
                path_entries = _entries_for_path(path)
 
2227
                path_entries = state._entries_for_path(path)
1755
2228
                if not path_entries:
1756
2229
                    # this specified path is not present at all: error
1757
 
                    all_versioned = False
1758
 
                    break
 
2230
                    not_versioned.append(path.decode('utf-8'))
 
2231
                    continue
1759
2232
                found_versioned = False
1760
2233
                # for each id at this path
1761
2234
                for entry in path_entries:
1768
2241
                if not found_versioned:
1769
2242
                    # none of the indexes was not 'absent' at all ids for this
1770
2243
                    # path.
1771
 
                    all_versioned = False
1772
 
                    break
1773
 
            if not all_versioned:
1774
 
                raise errors.PathsNotVersionedError(specific_files)
 
2244
                    not_versioned.append(path.decode('utf-8'))
 
2245
            if len(not_versioned) > 0:
 
2246
                raise errors.PathsNotVersionedError(not_versioned)
1775
2247
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1776
 
        search_specific_files = set()
1777
 
        for path in specific_files:
1778
 
            other_specific_files = specific_files.difference(set([path]))
1779
 
            if not osutils.is_inside_any(other_specific_files, path):
1780
 
                # this is a top level path, we must check it.
1781
 
                search_specific_files.add(path)
1782
 
        # sketch: 
1783
 
        # compare source_index and target_index at or under each element of search_specific_files.
1784
 
        # follow the following comparison table. Note that we only want to do diff operations when
1785
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1786
 
        # for the target.
1787
 
        # cases:
1788
 
        # 
1789
 
        # Source | Target | disk | action
1790
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1791
 
        #        |        |      | diff check on source-target
1792
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1793
 
        #        |        |      | ???
1794
 
        #   r    |  a     |      | add source to search
1795
 
        #   r    |  a     |  a   | 
1796
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1797
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1798
 
        #   a    | fdlt   |      | add new id
1799
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1800
 
        #   a    |  a     |      | not present in either tree, skip
1801
 
        #   a    |  a     |  a   | not present in any tree, skip
1802
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1803
 
        #        |        |      | may not be selected by the users list of paths.
1804
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1805
 
        #        |        |      | may not be selected by the users list of paths.
1806
 
        #  fdlt  | fdlt   |      | content in both: diff them
1807
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1808
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1809
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1810
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1811
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1812
 
        #        |        |      | this id at the other path.
1813
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1814
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1815
 
        #        |        |      | this id at the other path.
1816
 
 
1817
 
        # for all search_indexs in each path at or under each element of
1818
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1819
 
        # relocated path as one to search if its not searched already. If the
1820
 
        # detail is not relocated, add the id.
1821
 
        searched_specific_files = set()
1822
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1823
 
        # Using a list so that we can access the values and change them in
1824
 
        # nested scope. Each one is [path, file_id, entry]
1825
 
        last_source_parent = [None, None]
1826
 
        last_target_parent = [None, None]
 
2248
        search_specific_files = osutils.minimum_path_selection(specific_files)
1827
2249
 
1828
2250
        use_filesystem_for_exec = (sys.platform != 'win32')
1829
 
 
1830
 
        # Just a sentry, so that _process_entry can say that this
1831
 
        # record is handled, but isn't interesting to process (unchanged)
1832
 
        uninteresting = object()
1833
 
 
1834
 
 
1835
 
        old_dirname_to_file_id = {}
1836
 
        new_dirname_to_file_id = {}
1837
 
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1838
 
        #       keeping a cache of directories that we have seen.
1839
 
 
1840
 
        def _process_entry(entry, path_info):
1841
 
            """Compare an entry and real disk to generate delta information.
1842
 
 
1843
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1844
 
                the path of entry. If None, then the path is considered absent.
1845
 
                (Perhaps we should pass in a concrete entry for this ?)
1846
 
                Basename is returned as a utf8 string because we expect this
1847
 
                tuple will be ignored, and don't want to take the time to
1848
 
                decode.
1849
 
            :return: None if these don't match
1850
 
                     A tuple of information about the change, or
1851
 
                     the object 'uninteresting' if these match, but are
1852
 
                     basically identical.
1853
 
            """
1854
 
            if source_index is None:
1855
 
                source_details = NULL_PARENT_DETAILS
1856
 
            else:
1857
 
                source_details = entry[1][source_index]
1858
 
            target_details = entry[1][target_index]
1859
 
            target_minikind = target_details[0]
1860
 
            if path_info is not None and target_minikind in 'fdlt':
1861
 
                assert target_index == 0
1862
 
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
1863
 
                                                  stat_value=path_info[3])
1864
 
                # The entry may have been modified by update_entry
1865
 
                target_details = entry[1][target_index]
1866
 
                target_minikind = target_details[0]
1867
 
            else:
1868
 
                link_or_sha1 = None
1869
 
            file_id = entry[0][2]
1870
 
            source_minikind = source_details[0]
1871
 
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
1872
 
                # claimed content in both: diff
1873
 
                #   r    | fdlt   |      | add source to search, add id path move and perform
1874
 
                #        |        |      | diff check on source-target
1875
 
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
1876
 
                #        |        |      | ???
1877
 
                if source_minikind in 'r':
1878
 
                    # add the source to the search path to find any children it
1879
 
                    # has.  TODO ? : only add if it is a container ?
1880
 
                    if not osutils.is_inside_any(searched_specific_files,
1881
 
                                                 source_details[1]):
1882
 
                        search_specific_files.add(source_details[1])
1883
 
                    # generate the old path; this is needed for stating later
1884
 
                    # as well.
1885
 
                    old_path = source_details[1]
1886
 
                    old_dirname, old_basename = os.path.split(old_path)
1887
 
                    path = pathjoin(entry[0][0], entry[0][1])
1888
 
                    old_entry = state._get_entry(source_index,
1889
 
                                                 path_utf8=old_path)
1890
 
                    # update the source details variable to be the real
1891
 
                    # location.
1892
 
                    source_details = old_entry[1][source_index]
1893
 
                    source_minikind = source_details[0]
1894
 
                else:
1895
 
                    old_dirname = entry[0][0]
1896
 
                    old_basename = entry[0][1]
1897
 
                    old_path = path = None
1898
 
                if path_info is None:
1899
 
                    # the file is missing on disk, show as removed.
1900
 
                    content_change = True
1901
 
                    target_kind = None
1902
 
                    target_exec = False
1903
 
                else:
1904
 
                    # source and target are both versioned and disk file is present.
1905
 
                    target_kind = path_info[2]
1906
 
                    if target_kind == 'directory':
1907
 
                        if path is None:
1908
 
                            old_path = path = pathjoin(old_dirname, old_basename)
1909
 
                        new_dirname_to_file_id[path] = file_id
1910
 
                        if source_minikind != 'd':
1911
 
                            content_change = True
1912
 
                        else:
1913
 
                            # directories have no fingerprint
1914
 
                            content_change = False
1915
 
                        target_exec = False
1916
 
                    elif target_kind == 'file':
1917
 
                        if source_minikind != 'f':
1918
 
                            content_change = True
1919
 
                        else:
1920
 
                            # We could check the size, but we already have the
1921
 
                            # sha1 hash.
1922
 
                            content_change = (link_or_sha1 != source_details[1])
1923
 
                        # Target details is updated at update_entry time
1924
 
                        if use_filesystem_for_exec:
1925
 
                            # We don't need S_ISREG here, because we are sure
1926
 
                            # we are dealing with a file.
1927
 
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
1928
 
                        else:
1929
 
                            target_exec = target_details[3]
1930
 
                    elif target_kind == 'symlink':
1931
 
                        if source_minikind != 'l':
1932
 
                            content_change = True
1933
 
                        else:
1934
 
                            content_change = (link_or_sha1 != source_details[1])
1935
 
                        target_exec = False
1936
 
                    elif target_kind == 'tree-reference':
1937
 
                        if source_minikind != 't':
1938
 
                            content_change = True
1939
 
                        else:
1940
 
                            content_change = False
1941
 
                        target_exec = False
1942
 
                    else:
1943
 
                        raise Exception, "unknown kind %s" % path_info[2]
1944
 
                if source_minikind == 'd':
1945
 
                    if path is None:
1946
 
                        old_path = path = pathjoin(old_dirname, old_basename)
1947
 
                    old_dirname_to_file_id[old_path] = file_id
1948
 
                # parent id is the entry for the path in the target tree
1949
 
                if old_dirname == last_source_parent[0]:
1950
 
                    source_parent_id = last_source_parent[1]
1951
 
                else:
1952
 
                    try:
1953
 
                        source_parent_id = old_dirname_to_file_id[old_dirname]
1954
 
                    except KeyError:
1955
 
                        source_parent_entry = state._get_entry(source_index,
1956
 
                                                               path_utf8=old_dirname)
1957
 
                        source_parent_id = source_parent_entry[0][2]
1958
 
                    if source_parent_id == entry[0][2]:
1959
 
                        # This is the root, so the parent is None
1960
 
                        source_parent_id = None
1961
 
                    else:
1962
 
                        last_source_parent[0] = old_dirname
1963
 
                        last_source_parent[1] = source_parent_id
1964
 
                new_dirname = entry[0][0]
1965
 
                if new_dirname == last_target_parent[0]:
1966
 
                    target_parent_id = last_target_parent[1]
1967
 
                else:
1968
 
                    try:
1969
 
                        target_parent_id = new_dirname_to_file_id[new_dirname]
1970
 
                    except KeyError:
1971
 
                        # TODO: We don't always need to do the lookup, because the
1972
 
                        #       parent entry will be the same as the source entry.
1973
 
                        target_parent_entry = state._get_entry(target_index,
1974
 
                                                               path_utf8=new_dirname)
1975
 
                        assert target_parent_entry != (None, None), (
1976
 
                            "Could not find target parent in wt: %s\nparent of: %s"
1977
 
                            % (new_dirname, entry))
1978
 
                        target_parent_id = target_parent_entry[0][2]
1979
 
                    if target_parent_id == entry[0][2]:
1980
 
                        # This is the root, so the parent is None
1981
 
                        target_parent_id = None
1982
 
                    else:
1983
 
                        last_target_parent[0] = new_dirname
1984
 
                        last_target_parent[1] = target_parent_id
1985
 
 
1986
 
                source_exec = source_details[3]
1987
 
                if (include_unchanged
1988
 
                    or content_change
1989
 
                    or source_parent_id != target_parent_id
1990
 
                    or old_basename != entry[0][1]
1991
 
                    or source_exec != target_exec
1992
 
                    ):
1993
 
                    if old_path is None:
1994
 
                        old_path = path = pathjoin(old_dirname, old_basename)
1995
 
                        old_path_u = utf8_decode(old_path)[0]
1996
 
                        path_u = old_path_u
1997
 
                    else:
1998
 
                        old_path_u = utf8_decode(old_path)[0]
1999
 
                        if old_path == path:
2000
 
                            path_u = old_path_u
2001
 
                        else:
2002
 
                            path_u = utf8_decode(path)[0]
2003
 
                    source_kind = _minikind_to_kind[source_minikind]
2004
 
                    return (entry[0][2],
2005
 
                           (old_path_u, path_u),
2006
 
                           content_change,
2007
 
                           (True, True),
2008
 
                           (source_parent_id, target_parent_id),
2009
 
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
2010
 
                           (source_kind, target_kind),
2011
 
                           (source_exec, target_exec))
2012
 
                else:
2013
 
                    return uninteresting
2014
 
            elif source_minikind in 'a' and target_minikind in 'fdlt':
2015
 
                # looks like a new file
2016
 
                if path_info is not None:
2017
 
                    path = pathjoin(entry[0][0], entry[0][1])
2018
 
                    # parent id is the entry for the path in the target tree
2019
 
                    # TODO: these are the same for an entire directory: cache em.
2020
 
                    parent_id = state._get_entry(target_index,
2021
 
                                                 path_utf8=entry[0][0])[0][2]
2022
 
                    if parent_id == entry[0][2]:
2023
 
                        parent_id = None
2024
 
                    if use_filesystem_for_exec:
2025
 
                        # We need S_ISREG here, because we aren't sure if this
2026
 
                        # is a file or not.
2027
 
                        target_exec = bool(
2028
 
                            stat.S_ISREG(path_info[3].st_mode)
2029
 
                            and stat.S_IEXEC & path_info[3].st_mode)
2030
 
                    else:
2031
 
                        target_exec = target_details[3]
2032
 
                    return (entry[0][2],
2033
 
                           (None, utf8_decode(path)[0]),
2034
 
                           True,
2035
 
                           (False, True),
2036
 
                           (None, parent_id),
2037
 
                           (None, utf8_decode(entry[0][1])[0]),
2038
 
                           (None, path_info[2]),
2039
 
                           (None, target_exec))
2040
 
                else:
2041
 
                    # but its not on disk: we deliberately treat this as just
2042
 
                    # never-present. (Why ?! - RBC 20070224)
2043
 
                    pass
2044
 
            elif source_minikind in 'fdlt' and target_minikind in 'a':
2045
 
                # unversioned, possibly, or possibly not deleted: we dont care.
2046
 
                # if its still on disk, *and* theres no other entry at this
2047
 
                # path [we dont know this in this routine at the moment -
2048
 
                # perhaps we should change this - then it would be an unknown.
2049
 
                old_path = pathjoin(entry[0][0], entry[0][1])
2050
 
                # parent id is the entry for the path in the target tree
2051
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
2052
 
                if parent_id == entry[0][2]:
2053
 
                    parent_id = None
2054
 
                return (entry[0][2],
2055
 
                       (utf8_decode(old_path)[0], None),
2056
 
                       True,
2057
 
                       (True, False),
2058
 
                       (parent_id, None),
2059
 
                       (utf8_decode(entry[0][1])[0], None),
2060
 
                       (_minikind_to_kind[source_minikind], None),
2061
 
                       (source_details[3], None))
2062
 
            elif source_minikind in 'fdlt' and target_minikind in 'r':
2063
 
                # a rename; could be a true rename, or a rename inherited from
2064
 
                # a renamed parent. TODO: handle this efficiently. Its not
2065
 
                # common case to rename dirs though, so a correct but slow
2066
 
                # implementation will do.
2067
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
2068
 
                    search_specific_files.add(target_details[1])
2069
 
            elif source_minikind in 'ra' and target_minikind in 'ra':
2070
 
                # neither of the selected trees contain this file,
2071
 
                # so skip over it. This is not currently directly tested, but
2072
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
2073
 
                pass
2074
 
            else:
2075
 
                raise AssertionError("don't know how to compare "
2076
 
                    "source_minikind=%r, target_minikind=%r"
2077
 
                    % (source_minikind, target_minikind))
2078
 
                ## import pdb;pdb.set_trace()
2079
 
            return None
2080
 
 
2081
 
        while search_specific_files:
2082
 
            # TODO: the pending list should be lexically sorted?  the
2083
 
            # interface doesn't require it.
2084
 
            current_root = search_specific_files.pop()
2085
 
            current_root_unicode = current_root.decode('utf8')
2086
 
            searched_specific_files.add(current_root)
2087
 
            # process the entries for this containing directory: the rest will be
2088
 
            # found by their parents recursively.
2089
 
            root_entries = _entries_for_path(current_root)
2090
 
            root_abspath = self.target.abspath(current_root_unicode)
2091
 
            try:
2092
 
                root_stat = os.lstat(root_abspath)
2093
 
            except OSError, e:
2094
 
                if e.errno == errno.ENOENT:
2095
 
                    # the path does not exist: let _process_entry know that.
2096
 
                    root_dir_info = None
2097
 
                else:
2098
 
                    # some other random error: hand it up.
2099
 
                    raise
2100
 
            else:
2101
 
                root_dir_info = ('', current_root,
2102
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
2103
 
                    root_abspath)
2104
 
                if root_dir_info[2] == 'directory':
2105
 
                    if self.target._directory_is_tree_reference(
2106
 
                        current_root.decode('utf8')):
2107
 
                        root_dir_info = root_dir_info[:2] + \
2108
 
                            ('tree-reference',) + root_dir_info[3:]
2109
 
 
2110
 
            if not root_entries and not root_dir_info:
2111
 
                # this specified path is not present at all, skip it.
2112
 
                continue
2113
 
            path_handled = False
2114
 
            for entry in root_entries:
2115
 
                result = _process_entry(entry, root_dir_info)
2116
 
                if result is not None:
2117
 
                    path_handled = True
2118
 
                    if result is not uninteresting:
2119
 
                        yield result
2120
 
            if want_unversioned and not path_handled and root_dir_info:
2121
 
                new_executable = bool(
2122
 
                    stat.S_ISREG(root_dir_info[3].st_mode)
2123
 
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
2124
 
                yield (None,
2125
 
                       (None, current_root_unicode),
2126
 
                       True,
2127
 
                       (False, False),
2128
 
                       (None, None),
2129
 
                       (None, splitpath(current_root_unicode)[-1]),
2130
 
                       (None, root_dir_info[2]),
2131
 
                       (None, new_executable)
2132
 
                      )
2133
 
            initial_key = (current_root, '', '')
2134
 
            block_index, _ = state._find_block_index_from_key(initial_key)
2135
 
            if block_index == 0:
2136
 
                # we have processed the total root already, but because the
2137
 
                # initial key matched it we should skip it here.
2138
 
                block_index +=1
2139
 
            if root_dir_info and root_dir_info[2] == 'tree-reference':
2140
 
                current_dir_info = None
2141
 
            else:
2142
 
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
2143
 
                try:
2144
 
                    current_dir_info = dir_iterator.next()
2145
 
                except OSError, e:
2146
 
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
2147
 
                    # python 2.5 has e.errno == EINVAL,
2148
 
                    #            and e.winerror == ERROR_DIRECTORY
2149
 
                    e_winerror = getattr(e, 'winerror', None)
2150
 
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
2151
 
                    # there may be directories in the inventory even though
2152
 
                    # this path is not a file on disk: so mark it as end of
2153
 
                    # iterator
2154
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
2155
 
                        current_dir_info = None
2156
 
                    elif (sys.platform == 'win32'
2157
 
                          and (e.errno in win_errors
2158
 
                               or e_winerror in win_errors)):
2159
 
                        current_dir_info = None
2160
 
                    else:
2161
 
                        raise
2162
 
                else:
2163
 
                    if current_dir_info[0][0] == '':
2164
 
                        # remove .bzr from iteration
2165
 
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
2166
 
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
2167
 
                        del current_dir_info[1][bzr_index]
2168
 
            # walk until both the directory listing and the versioned metadata
2169
 
            # are exhausted. 
2170
 
            if (block_index < len(state._dirblocks) and
2171
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2172
 
                current_block = state._dirblocks[block_index]
2173
 
            else:
2174
 
                current_block = None
2175
 
            while (current_dir_info is not None or
2176
 
                   current_block is not None):
2177
 
                if (current_dir_info and current_block
2178
 
                    and current_dir_info[0][0] != current_block[0]):
2179
 
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
2180
 
                        # filesystem data refers to paths not covered by the dirblock.
2181
 
                        # this has two possibilities:
2182
 
                        # A) it is versioned but empty, so there is no block for it
2183
 
                        # B) it is not versioned.
2184
 
 
2185
 
                        # if (A) then we need to recurse into it to check for
2186
 
                        # new unknown files or directories.
2187
 
                        # if (B) then we should ignore it, because we don't
2188
 
                        # recurse into unknown directories.
2189
 
                        path_index = 0
2190
 
                        while path_index < len(current_dir_info[1]):
2191
 
                                current_path_info = current_dir_info[1][path_index]
2192
 
                                if want_unversioned:
2193
 
                                    if current_path_info[2] == 'directory':
2194
 
                                        if self.target._directory_is_tree_reference(
2195
 
                                            current_path_info[0].decode('utf8')):
2196
 
                                            current_path_info = current_path_info[:2] + \
2197
 
                                                ('tree-reference',) + current_path_info[3:]
2198
 
                                    new_executable = bool(
2199
 
                                        stat.S_ISREG(current_path_info[3].st_mode)
2200
 
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
2201
 
                                    yield (None,
2202
 
                                        (None, utf8_decode(current_path_info[0])[0]),
2203
 
                                        True,
2204
 
                                        (False, False),
2205
 
                                        (None, None),
2206
 
                                        (None, utf8_decode(current_path_info[1])[0]),
2207
 
                                        (None, current_path_info[2]),
2208
 
                                        (None, new_executable))
2209
 
                                # dont descend into this unversioned path if it is
2210
 
                                # a dir
2211
 
                                if current_path_info[2] in ('directory',
2212
 
                                                            'tree-reference'):
2213
 
                                    del current_dir_info[1][path_index]
2214
 
                                    path_index -= 1
2215
 
                                path_index += 1
2216
 
 
2217
 
                        # This dir info has been handled, go to the next
2218
 
                        try:
2219
 
                            current_dir_info = dir_iterator.next()
2220
 
                        except StopIteration:
2221
 
                            current_dir_info = None
2222
 
                    else:
2223
 
                        # We have a dirblock entry for this location, but there
2224
 
                        # is no filesystem path for this. This is most likely
2225
 
                        # because a directory was removed from the disk.
2226
 
                        # We don't have to report the missing directory,
2227
 
                        # because that should have already been handled, but we
2228
 
                        # need to handle all of the files that are contained
2229
 
                        # within.
2230
 
                        for current_entry in current_block[1]:
2231
 
                            # entry referring to file not present on disk.
2232
 
                            # advance the entry only, after processing.
2233
 
                            result = _process_entry(current_entry, None)
2234
 
                            if result is not None:
2235
 
                                if result is not uninteresting:
2236
 
                                    yield result
2237
 
                        block_index +=1
2238
 
                        if (block_index < len(state._dirblocks) and
2239
 
                            osutils.is_inside(current_root,
2240
 
                                              state._dirblocks[block_index][0])):
2241
 
                            current_block = state._dirblocks[block_index]
2242
 
                        else:
2243
 
                            current_block = None
2244
 
                    continue
2245
 
                entry_index = 0
2246
 
                if current_block and entry_index < len(current_block[1]):
2247
 
                    current_entry = current_block[1][entry_index]
2248
 
                else:
2249
 
                    current_entry = None
2250
 
                advance_entry = True
2251
 
                path_index = 0
2252
 
                if current_dir_info and path_index < len(current_dir_info[1]):
2253
 
                    current_path_info = current_dir_info[1][path_index]
2254
 
                    if current_path_info[2] == 'directory':
2255
 
                        if self.target._directory_is_tree_reference(
2256
 
                            current_path_info[0].decode('utf8')):
2257
 
                            current_path_info = current_path_info[:2] + \
2258
 
                                ('tree-reference',) + current_path_info[3:]
2259
 
                else:
2260
 
                    current_path_info = None
2261
 
                advance_path = True
2262
 
                path_handled = False
2263
 
                while (current_entry is not None or
2264
 
                    current_path_info is not None):
2265
 
                    if current_entry is None:
2266
 
                        # the check for path_handled when the path is adnvaced
2267
 
                        # will yield this path if needed.
2268
 
                        pass
2269
 
                    elif current_path_info is None:
2270
 
                        # no path is fine: the per entry code will handle it.
2271
 
                        result = _process_entry(current_entry, current_path_info)
2272
 
                        if result is not None:
2273
 
                            if result is not uninteresting:
2274
 
                                yield result
2275
 
                    elif (current_entry[0][1] != current_path_info[1]
2276
 
                          or current_entry[1][target_index][0] in 'ar'):
2277
 
                        # The current path on disk doesn't match the dirblock
2278
 
                        # record. Either the dirblock is marked as absent, or
2279
 
                        # the file on disk is not present at all in the
2280
 
                        # dirblock. Either way, report about the dirblock
2281
 
                        # entry, and let other code handle the filesystem one.
2282
 
 
2283
 
                        # Compare the basename for these files to determine
2284
 
                        # which comes first
2285
 
                        if current_path_info[1] < current_entry[0][1]:
2286
 
                            # extra file on disk: pass for now, but only
2287
 
                            # increment the path, not the entry
2288
 
                            advance_entry = False
2289
 
                        else:
2290
 
                            # entry referring to file not present on disk.
2291
 
                            # advance the entry only, after processing.
2292
 
                            result = _process_entry(current_entry, None)
2293
 
                            if result is not None:
2294
 
                                if result is not uninteresting:
2295
 
                                    yield result
2296
 
                            advance_path = False
2297
 
                    else:
2298
 
                        result = _process_entry(current_entry, current_path_info)
2299
 
                        if result is not None:
2300
 
                            path_handled = True
2301
 
                            if result is not uninteresting:
2302
 
                                yield result
2303
 
                    if advance_entry and current_entry is not None:
2304
 
                        entry_index += 1
2305
 
                        if entry_index < len(current_block[1]):
2306
 
                            current_entry = current_block[1][entry_index]
2307
 
                        else:
2308
 
                            current_entry = None
2309
 
                    else:
2310
 
                        advance_entry = True # reset the advance flaga
2311
 
                    if advance_path and current_path_info is not None:
2312
 
                        if not path_handled:
2313
 
                            # unversioned in all regards
2314
 
                            if want_unversioned:
2315
 
                                new_executable = bool(
2316
 
                                    stat.S_ISREG(current_path_info[3].st_mode)
2317
 
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
2318
 
                                yield (None,
2319
 
                                    (None, utf8_decode(current_path_info[0])[0]),
2320
 
                                    True,
2321
 
                                    (False, False),
2322
 
                                    (None, None),
2323
 
                                    (None, utf8_decode(current_path_info[1])[0]),
2324
 
                                    (None, current_path_info[2]),
2325
 
                                    (None, new_executable))
2326
 
                            # dont descend into this unversioned path if it is
2327
 
                            # a dir
2328
 
                            if current_path_info[2] in ('directory'):
2329
 
                                del current_dir_info[1][path_index]
2330
 
                                path_index -= 1
2331
 
                        # dont descend the disk iterator into any tree 
2332
 
                        # paths.
2333
 
                        if current_path_info[2] == 'tree-reference':
2334
 
                            del current_dir_info[1][path_index]
2335
 
                            path_index -= 1
2336
 
                        path_index += 1
2337
 
                        if path_index < len(current_dir_info[1]):
2338
 
                            current_path_info = current_dir_info[1][path_index]
2339
 
                            if current_path_info[2] == 'directory':
2340
 
                                if self.target._directory_is_tree_reference(
2341
 
                                    current_path_info[0].decode('utf8')):
2342
 
                                    current_path_info = current_path_info[:2] + \
2343
 
                                        ('tree-reference',) + current_path_info[3:]
2344
 
                        else:
2345
 
                            current_path_info = None
2346
 
                        path_handled = False
2347
 
                    else:
2348
 
                        advance_path = True # reset the advance flagg.
2349
 
                if current_block is not None:
2350
 
                    block_index += 1
2351
 
                    if (block_index < len(state._dirblocks) and
2352
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2353
 
                        current_block = state._dirblocks[block_index]
2354
 
                    else:
2355
 
                        current_block = None
2356
 
                if current_dir_info is not None:
2357
 
                    try:
2358
 
                        current_dir_info = dir_iterator.next()
2359
 
                    except StopIteration:
2360
 
                        current_dir_info = None
2361
 
 
 
2251
        iter_changes = self.target._iter_changes(include_unchanged,
 
2252
            use_filesystem_for_exec, search_specific_files, state,
 
2253
            source_index, target_index, want_unversioned, self.target)
 
2254
        return iter_changes.iter_changes()
2362
2255
 
2363
2256
    @staticmethod
2364
2257
    def is_compatible(source, target):
2365
2258
        # the target must be a dirstate working tree
2366
 
        if not isinstance(target, WorkingTree4):
 
2259
        if not isinstance(target, DirStateWorkingTree):
2367
2260
            return False
2368
 
        # the source must be a revtreee or dirstate rev tree.
 
2261
        # the source must be a revtree or dirstate rev tree.
2369
2262
        if not isinstance(source,
2370
2263
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2371
2264
            return False
2372
2265
        # the source revid must be in the target dirstate
2373
 
        if not (source._revision_id == NULL_REVISION or
 
2266
        if not (source._revision_id == _mod_revision.NULL_REVISION or
2374
2267
            source._revision_id in target.get_parent_ids()):
2375
 
            # TODO: what about ghosts? it may well need to 
 
2268
            # TODO: what about ghosts? it may well need to
2376
2269
            # check for them explicitly.
2377
2270
            return False
2378
2271
        return True
2388
2281
 
2389
2282
    def convert(self, tree):
2390
2283
        # lock the control files not the tree, so that we dont get tree
2391
 
        # on-unlock behaviours, and so that noone else diddles with the 
 
2284
        # on-unlock behaviours, and so that noone else diddles with the
2392
2285
        # tree during upgrade.
2393
2286
        tree._control_files.lock_write()
2394
2287
        try:
2420
2313
 
2421
2314
    def update_format(self, tree):
2422
2315
        """Change the format marker."""
2423
 
        tree._control_files.put_utf8('format',
2424
 
            self.target_format.get_format_string())
 
2316
        tree._transport.put_bytes('format',
 
2317
            self.target_format.as_string(),
 
2318
            mode=tree.bzrdir._get_file_mode())
 
2319
 
 
2320
 
 
2321
class Converter4to5(object):
 
2322
    """Perform an in-place upgrade of format 4 to format 5 trees."""
 
2323
 
 
2324
    def __init__(self):
 
2325
        self.target_format = WorkingTreeFormat5()
 
2326
 
 
2327
    def convert(self, tree):
 
2328
        # lock the control files not the tree, so that we don't get tree
 
2329
        # on-unlock behaviours, and so that no-one else diddles with the
 
2330
        # tree during upgrade.
 
2331
        tree._control_files.lock_write()
 
2332
        try:
 
2333
            self.update_format(tree)
 
2334
        finally:
 
2335
            tree._control_files.unlock()
 
2336
 
 
2337
    def update_format(self, tree):
 
2338
        """Change the format marker."""
 
2339
        tree._transport.put_bytes('format',
 
2340
            self.target_format.as_string(),
 
2341
            mode=tree.bzrdir._get_file_mode())
 
2342
 
 
2343
 
 
2344
class Converter4or5to6(object):
 
2345
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
 
2346
 
 
2347
    def __init__(self):
 
2348
        self.target_format = WorkingTreeFormat6()
 
2349
 
 
2350
    def convert(self, tree):
 
2351
        # lock the control files not the tree, so that we don't get tree
 
2352
        # on-unlock behaviours, and so that no-one else diddles with the
 
2353
        # tree during upgrade.
 
2354
        tree._control_files.lock_write()
 
2355
        try:
 
2356
            self.init_custom_control_files(tree)
 
2357
            self.update_format(tree)
 
2358
        finally:
 
2359
            tree._control_files.unlock()
 
2360
 
 
2361
    def init_custom_control_files(self, tree):
 
2362
        """Initialize custom control files."""
 
2363
        tree._transport.put_bytes('views', '',
 
2364
            mode=tree.bzrdir._get_file_mode())
 
2365
 
 
2366
    def update_format(self, tree):
 
2367
        """Change the format marker."""
 
2368
        tree._transport.put_bytes('format',
 
2369
            self.target_format.as_string(),
 
2370
            mode=tree.bzrdir._get_file_mode())