/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

Merge from bzr.dev

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
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
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
 
47
    dirstate,
 
48
    errors,
 
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    transform,
 
59
    urlutils,
 
60
    xml5,
 
61
    xml6,
 
62
    )
 
63
import bzrlib.branch
 
64
from bzrlib.transport import get_transport
 
65
import bzrlib.ui
 
66
""")
 
67
 
 
68
from bzrlib import symbol_versioning
 
69
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
 
72
from bzrlib.lockdir import LockDir
 
73
import bzrlib.mutabletree
 
74
from bzrlib.mutabletree import needs_tree_write_lock
 
75
from bzrlib.osutils import (
 
76
    file_kind,
 
77
    isdir,
 
78
    normpath,
 
79
    pathjoin,
 
80
    rand_chars,
 
81
    realpath,
 
82
    safe_unicode,
 
83
    splitpath,
 
84
    )
 
85
from bzrlib.trace import mutter, note
 
86
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
    """
 
118
 
 
119
    def __init__(self, basedir,
 
120
                 branch,
 
121
                 _control_files=None,
 
122
                 _format=None,
 
123
                 _bzrdir=None):
 
124
        """Construct a WorkingTree for basedir.
 
125
 
 
126
        If the branch is not supplied, it is opened automatically.
 
127
        If the branch is supplied, it must be the branch for this basedir.
 
128
        (branch.base is not cross checked, because for remote branches that
 
129
        would be meaningless).
 
130
        """
 
131
        self._format = _format
 
132
        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
        basedir = safe_unicode(basedir)
 
137
        mutter("opening working tree %r", basedir)
 
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
 
141
        self.basedir = realpath(basedir)
 
142
        # if branch is at our basedir and is a format 6 or less
 
143
        # 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
        self._control_files = _control_files
 
147
        self._dirty = None
 
148
        #-------------
 
149
        # during a read or write lock these objects are set, and are
 
150
        # None the rest of the time.
 
151
        self._dirstate = None
 
152
        self._inventory = None
 
153
        #-------------
 
154
 
 
155
    @needs_tree_write_lock
 
156
    def _add(self, files, ids, kinds):
 
157
        """See MutableTree._add."""
 
158
        state = self.current_dirstate()
 
159
        for f, file_id, kind in zip(files, ids, kinds):
 
160
            f = f.strip('/')
 
161
            assert '//' not in f
 
162
            assert '..' not in f
 
163
            if self.path2id(f):
 
164
                # special case tree root handling.
 
165
                if f == '' and self.path2id(f) == ROOT_ID:
 
166
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
167
                continue
 
168
            if file_id is None:
 
169
                file_id = generate_ids.gen_file_id(f)
 
170
            # deliberately add the file with no cached stat or sha1
 
171
            # - on the first access it will be gathered, and we can
 
172
            # always change this once tests are all passing.
 
173
            state.add(f, file_id, kind, None, '')
 
174
        self._make_dirty(reset_inventory=True)
 
175
 
 
176
    def _make_dirty(self, reset_inventory):
 
177
        """Make the tree state dirty.
 
178
 
 
179
        :param reset_inventory: True if the cached inventory should be removed
 
180
            (presuming there is one).
 
181
        """
 
182
        self._dirty = True
 
183
        if reset_inventory and self._inventory is not None:
 
184
            self._inventory = None
 
185
 
 
186
    @needs_tree_write_lock
 
187
    def add_reference(self, sub_tree):
 
188
        # use standard implementation, which calls back to self._add
 
189
        # 
 
190
        # So we don't store the reference_revision in the working dirstate,
 
191
        # it's just recorded at the moment of commit. 
 
192
        self._add_reference(sub_tree)
 
193
 
 
194
    def break_lock(self):
 
195
        """Break a lock if one is present from another instance.
 
196
 
 
197
        Uses the ui factory to ask for confirmation if the lock may be from
 
198
        an active process.
 
199
 
 
200
        This will probe the repository for its lock as well.
 
201
        """
 
202
        # if the dirstate is locked by an active process, reject the break lock
 
203
        # call.
 
204
        try:
 
205
            if self._dirstate is None:
 
206
                clear = True
 
207
            else:
 
208
                clear = False
 
209
            state = self._current_dirstate()
 
210
            if state._lock_token is not None:
 
211
                # we already have it locked. sheese, cant break our own lock.
 
212
                raise errors.LockActive(self.basedir)
 
213
            else:
 
214
                try:
 
215
                    # try for a write lock - need permission to get one anyhow
 
216
                    # to break locks.
 
217
                    state.lock_write()
 
218
                except errors.LockContention:
 
219
                    # oslocks fail when a process is still live: fail.
 
220
                    # TODO: get the locked lockdir info and give to the user to
 
221
                    # assist in debugging.
 
222
                    raise errors.LockActive(self.basedir)
 
223
                else:
 
224
                    state.unlock()
 
225
        finally:
 
226
            if clear:
 
227
                self._dirstate = None
 
228
        self._control_files.break_lock()
 
229
        self.branch.break_lock()
 
230
 
 
231
    def _comparison_data(self, entry, path):
 
232
        kind, executable, stat_value = \
 
233
            WorkingTree3._comparison_data(self, entry, path)
 
234
        # it looks like a plain directory, but it's really a reference -- see
 
235
        # also kind()
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
 
239
            kind = 'tree-reference'
 
240
        return kind, executable, stat_value
 
241
 
 
242
    @needs_write_lock
 
243
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
244
        # mark the tree as dirty post commit - commit
 
245
        # can change the current versioned list by doing deletes.
 
246
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
247
        self._make_dirty(reset_inventory=True)
 
248
        return result
 
249
 
 
250
    def current_dirstate(self):
 
251
        """Return the current dirstate object.
 
252
 
 
253
        This is not part of the tree interface and only exposed for ease of
 
254
        testing.
 
255
 
 
256
        :raises errors.NotWriteLocked: when not in a lock.
 
257
        """
 
258
        self._must_be_locked()
 
259
        return self._current_dirstate()
 
260
 
 
261
    def _current_dirstate(self):
 
262
        """Internal function that does not check lock status.
 
263
 
 
264
        This is needed for break_lock which also needs the dirstate.
 
265
        """
 
266
        if self._dirstate is not None:
 
267
            return self._dirstate
 
268
        local_path = self.bzrdir.get_workingtree_transport(None
 
269
            ).local_abspath('dirstate')
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
 
271
        return self._dirstate
 
272
 
 
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
 
287
 
 
288
    def filter_unversioned_files(self, paths):
 
289
        """Filter out paths that are versioned.
 
290
 
 
291
        :return: set of paths.
 
292
        """
 
293
        # TODO: make a generic multi-bisect routine roughly that should list
 
294
        # the paths, then process one half at a time recursively, and feed the
 
295
        # results of each bisect in further still
 
296
        paths = sorted(paths)
 
297
        result = set()
 
298
        state = self.current_dirstate()
 
299
        # TODO we want a paths_to_dirblocks helper I think
 
300
        for path in paths:
 
301
            dirname, basename = os.path.split(path.encode('utf8'))
 
302
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
303
                dirname, basename, 0)
 
304
            if not path_is_versioned:
 
305
                result.add(path)
 
306
        return result
 
307
 
 
308
    def flush(self):
 
309
        """Write all cached data to disk."""
 
310
        if self._control_files._lock_mode != 'w':
 
311
            raise errors.NotWriteLocked(self)
 
312
        self.current_dirstate().save()
 
313
        self._inventory = None
 
314
        self._dirty = False
 
315
 
 
316
    @needs_tree_write_lock
 
317
    def _gather_kinds(self, files, kinds):
 
318
        """See MutableTree._gather_kinds."""
 
319
        for pos, f in enumerate(files):
 
320
            if kinds[pos] is None:
 
321
                kinds[pos] = self._kind(f)
 
322
 
 
323
    def _generate_inventory(self):
 
324
        """Create and set self.inventory from the dirstate object.
 
325
        
 
326
        This is relatively expensive: we have to walk the entire dirstate.
 
327
        Ideally we would not, and can deprecate this function.
 
328
        """
 
329
        #: uncomment to trap on inventory requests.
 
330
        # import pdb;pdb.set_trace()
 
331
        state = self.current_dirstate()
 
332
        state._read_dirblocks_if_needed()
 
333
        root_key, current_entry = self._get_entry(path='')
 
334
        current_id = root_key[2]
 
335
        assert current_entry[0][0] == 'd' # directory
 
336
        inv = Inventory(root_id=current_id)
 
337
        # Turn some things into local variables
 
338
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
339
        factory = entry_factory
 
340
        utf8_decode = cache_utf8._utf8_decode
 
341
        inv_byid = inv._byid
 
342
        # we could do this straight out of the dirstate; it might be fast
 
343
        # and should be profiled - RBC 20070216
 
344
        parent_ies = {'' : inv.root}
 
345
        for block in state._dirblocks[1:]: # skip the root
 
346
            dirname = block[0]
 
347
            try:
 
348
                parent_ie = parent_ies[dirname]
 
349
            except KeyError:
 
350
                # all the paths in this block are not versioned in this tree
 
351
                continue
 
352
            for key, entry in block[1]:
 
353
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
354
                if minikind in ('a', 'r'): # absent, relocated
 
355
                    # a parent tree only entry
 
356
                    continue
 
357
                name = key[1]
 
358
                name_unicode = utf8_decode(name)[0]
 
359
                file_id = key[2]
 
360
                kind = minikind_to_kind[minikind]
 
361
                inv_entry = factory[kind](file_id, name_unicode,
 
362
                                          parent_ie.file_id)
 
363
                if kind == 'file':
 
364
                    # This is only needed on win32, where this is the only way
 
365
                    # we know the executable bit.
 
366
                    inv_entry.executable = executable
 
367
                    # not strictly needed: working tree
 
368
                    #inv_entry.text_size = size
 
369
                    #inv_entry.text_sha1 = sha1
 
370
                elif kind == 'directory':
 
371
                    # add this entry to the parent map.
 
372
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
373
                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)
 
379
                    inv_entry.reference_revision = link_or_sha1 or None
 
380
                elif kind != 'symlink':
 
381
                    raise AssertionError("unknown kind %r" % kind)
 
382
                # 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
 
386
                inv_byid[file_id] = inv_entry
 
387
                parent_ie.children[name_unicode] = inv_entry
 
388
        self._inventory = inv
 
389
 
 
390
    def _get_entry(self, file_id=None, path=None):
 
391
        """Get the dirstate row for file_id or path.
 
392
 
 
393
        If either file_id or path is supplied, it is used as the key to lookup.
 
394
        If both are supplied, the fastest lookup is used, and an error is
 
395
        raised if they do not both point at the same row.
 
396
        
 
397
        :param file_id: An optional unicode file_id to be looked up.
 
398
        :param path: An optional unicode path to be looked up.
 
399
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
400
        """
 
401
        if file_id is None and path is None:
 
402
            raise errors.BzrError('must supply file_id or path')
 
403
        state = self.current_dirstate()
 
404
        if path is not None:
 
405
            path = path.encode('utf8')
 
406
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
407
 
 
408
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
409
        # check file id is valid unconditionally.
 
410
        entry = self._get_entry(file_id=file_id, path=path)
 
411
        if entry[0] is None:
 
412
            raise errors.NoSuchId(self, file_id)
 
413
        if path is None:
 
414
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
415
 
 
416
        file_abspath = self.abspath(path)
 
417
        state = self.current_dirstate()
 
418
        if stat_value is None:
 
419
            try:
 
420
                stat_value = os.lstat(file_abspath)
 
421
            except OSError, e:
 
422
                if e.errno == errno.ENOENT:
 
423
                    return None
 
424
                else:
 
425
                    raise
 
426
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
427
                                          stat_value=stat_value)
 
428
        if entry[1][0][0] == 'f':
 
429
            return link_or_sha1
 
430
        return None
 
431
 
 
432
    def _get_inventory(self):
 
433
        """Get the inventory for the tree. This is only valid within a lock."""
 
434
        if self._inventory is not None:
 
435
            return self._inventory
 
436
        self._must_be_locked()
 
437
        self._generate_inventory()
 
438
        return self._inventory
 
439
 
 
440
    inventory = property(_get_inventory,
 
441
                         doc="Inventory of this Tree")
 
442
 
 
443
    @needs_read_lock
 
444
    def get_parent_ids(self):
 
445
        """See Tree.get_parent_ids.
 
446
        
 
447
        This implementation requests the ids list from the dirstate file.
 
448
        """
 
449
        return self.current_dirstate().get_parent_ids()
 
450
 
 
451
    def get_reference_revision(self, file_id, path=None):
 
452
        # referenced tree's revision is whatever's currently there
 
453
        return self.get_nested_tree(file_id, path).last_revision()
 
454
 
 
455
    def get_nested_tree(self, file_id, path=None):
 
456
        if path is None:
 
457
            path = self.id2path(file_id)
 
458
        # else: check file_id is at path?
 
459
        return WorkingTree.open(self.abspath(path))
 
460
 
 
461
    @needs_read_lock
 
462
    def get_root_id(self):
 
463
        """Return the id of this trees root"""
 
464
        return self._get_entry(path='')[0][2]
 
465
 
 
466
    def has_id(self, file_id):
 
467
        state = self.current_dirstate()
 
468
        file_id = osutils.safe_file_id(file_id)
 
469
        row, parents = self._get_entry(file_id=file_id)
 
470
        if row is None:
 
471
            return False
 
472
        return osutils.lexists(pathjoin(
 
473
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
474
 
 
475
    @needs_read_lock
 
476
    def id2path(self, file_id):
 
477
        "Convert a file-id to a path."
 
478
        file_id = osutils.safe_file_id(file_id)
 
479
        state = self.current_dirstate()
 
480
        entry = self._get_entry(file_id=file_id)
 
481
        if entry == (None, None):
 
482
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
483
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
484
        return path_utf8.decode('utf8')
 
485
 
 
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)
 
493
            entry = self._get_entry(file_id=file_id, path=path)
 
494
            if entry == (None, None):
 
495
                return False
 
496
            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
            """
 
503
            if not path:
 
504
                file_id = osutils.safe_file_id(file_id)
 
505
                path = self.id2path(file_id)
 
506
            mode = os.lstat(self.abspath(path)).st_mode
 
507
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
508
 
 
509
    @needs_read_lock
 
510
    def __iter__(self):
 
511
        """Iterate through file_ids for this tree.
 
512
 
 
513
        file_ids are in a WorkingTree if they are in the working inventory
 
514
        and the working file exists.
 
515
        """
 
516
        result = []
 
517
        for key, tree_details in self.current_dirstate()._iter_entries():
 
518
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
519
                # not relevant to the working tree
 
520
                continue
 
521
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
522
            if osutils.lexists(path):
 
523
                result.append(key[2])
 
524
        return iter(result)
 
525
 
 
526
    def iter_references(self):
 
527
        for key, tree_details in self.current_dirstate()._iter_entries():
 
528
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
529
                # not relevant to the working tree
 
530
                continue
 
531
            if not key[1]:
 
532
                # the root is not a reference.
 
533
                continue
 
534
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
535
            try:
 
536
                if self._kind(path) == 'tree-reference':
 
537
                    yield path, key[2]
 
538
            except errors.NoSuchFile:
 
539
                # path is missing on disk.
 
540
                continue
 
541
 
 
542
    def kind(self, file_id):
 
543
        """Return the kind of a file.
 
544
 
 
545
        This is always the actual kind that's on disk, regardless of what it
 
546
        was added as.
 
547
 
 
548
        Note: The caller is expected to take a read-lock before calling this.
 
549
        """
 
550
        relpath = self.id2path(file_id)
 
551
        assert relpath != None, \
 
552
            "path for id {%s} is None!" % file_id
 
553
        return self._kind(relpath)
 
554
 
 
555
    def _kind(self, relpath):
 
556
        abspath = self.abspath(relpath)
 
557
        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'
 
562
        return kind
 
563
 
 
564
    @needs_read_lock
 
565
    def _last_revision(self):
 
566
        """See Mutable.last_revision."""
 
567
        parent_ids = self.current_dirstate().get_parent_ids()
 
568
        if parent_ids:
 
569
            return parent_ids[0]
 
570
        else:
 
571
            return None
 
572
 
 
573
    def lock_read(self):
 
574
        """See Branch.lock_read, and WorkingTree.unlock."""
 
575
        self.branch.lock_read()
 
576
        try:
 
577
            self._control_files.lock_read()
 
578
            try:
 
579
                state = self.current_dirstate()
 
580
                if not state._lock_token:
 
581
                    state.lock_read()
 
582
                # set our support for tree references from the repository in
 
583
                # use.
 
584
                self._repo_supports_tree_reference = getattr(
 
585
                    self.branch.repository._format, "supports_tree_reference",
 
586
                    False)
 
587
            except:
 
588
                self._control_files.unlock()
 
589
                raise
 
590
        except:
 
591
            self.branch.unlock()
 
592
            raise
 
593
 
 
594
    def _lock_self_write(self):
 
595
        """This should be called after the branch is locked."""
 
596
        try:
 
597
            self._control_files.lock_write()
 
598
            try:
 
599
                state = self.current_dirstate()
 
600
                if not state._lock_token:
 
601
                    state.lock_write()
 
602
                # set our support for tree references from the repository in
 
603
                # use.
 
604
                self._repo_supports_tree_reference = getattr(
 
605
                    self.branch.repository._format, "supports_tree_reference",
 
606
                    False)
 
607
            except:
 
608
                self._control_files.unlock()
 
609
                raise
 
610
        except:
 
611
            self.branch.unlock()
 
612
            raise
 
613
 
 
614
    def lock_tree_write(self):
 
615
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
616
        self.branch.lock_read()
 
617
        self._lock_self_write()
 
618
 
 
619
    def lock_write(self):
 
620
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
621
        self.branch.lock_write()
 
622
        self._lock_self_write()
 
623
 
 
624
    @needs_tree_write_lock
 
625
    def move(self, from_paths, to_dir, after=False):
 
626
        """See WorkingTree.move()."""
 
627
        result = []
 
628
        if not from_paths:
 
629
            return result
 
630
 
 
631
        state = self.current_dirstate()
 
632
 
 
633
        assert not isinstance(from_paths, basestring)
 
634
        to_dir_utf8 = to_dir.encode('utf8')
 
635
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
636
        id_index = state._get_id_index()
 
637
        # check destination directory
 
638
        # get the details for it
 
639
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
640
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
641
        if not entry_present:
 
642
            raise errors.BzrMoveFailedError('', to_dir,
 
643
                errors.NotVersionedError(to_dir))
 
644
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
645
        # get a handle on the block itself.
 
646
        to_block_index = state._ensure_block(
 
647
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
648
        to_block = state._dirblocks[to_block_index]
 
649
        to_abs = self.abspath(to_dir)
 
650
        if not isdir(to_abs):
 
651
            raise errors.BzrMoveFailedError('',to_dir,
 
652
                errors.NotADirectory(to_abs))
 
653
 
 
654
        if to_entry[1][0][0] != 'd':
 
655
            raise errors.BzrMoveFailedError('',to_dir,
 
656
                errors.NotADirectory(to_abs))
 
657
 
 
658
        if self._inventory is not None:
 
659
            update_inventory = True
 
660
            inv = self.inventory
 
661
            to_dir_ie = inv[to_dir_id]
 
662
            to_dir_id = to_entry[0][2]
 
663
        else:
 
664
            update_inventory = False
 
665
 
 
666
        rollbacks = []
 
667
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
668
                     fingerprint, packed_stat, size,
 
669
                     to_block, to_key, to_path_utf8):
 
670
            state._make_absent(old_entry)
 
671
            from_key = old_entry[0]
 
672
            rollbacks.append(
 
673
                lambda:state.update_minimal(from_key,
 
674
                    minikind,
 
675
                    executable=executable,
 
676
                    fingerprint=fingerprint,
 
677
                    packed_stat=packed_stat,
 
678
                    size=size,
 
679
                    path_utf8=from_path_utf8))
 
680
            state.update_minimal(to_key,
 
681
                    minikind,
 
682
                    executable=executable,
 
683
                    fingerprint=fingerprint,
 
684
                    packed_stat=packed_stat,
 
685
                    size=size,
 
686
                    path_utf8=to_path_utf8)
 
687
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
688
            new_entry = to_block[1][added_entry_index]
 
689
            rollbacks.append(lambda:state._make_absent(new_entry))
 
690
 
 
691
        for from_rel in from_paths:
 
692
            # from_rel is 'pathinroot/foo/bar'
 
693
            from_rel_utf8 = from_rel.encode('utf8')
 
694
            from_dirname, from_tail = osutils.split(from_rel)
 
695
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
696
            from_entry = self._get_entry(path=from_rel)
 
697
            if from_entry == (None, None):
 
698
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
699
                    errors.NotVersionedError(path=str(from_rel)))
 
700
 
 
701
            from_id = from_entry[0][2]
 
702
            to_rel = pathjoin(to_dir, from_tail)
 
703
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
704
            item_to_entry = self._get_entry(path=to_rel)
 
705
            if item_to_entry != (None, None):
 
706
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
707
                    "Target is already versioned.")
 
708
 
 
709
            if from_rel == to_rel:
 
710
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
711
                    "Source and target are identical.")
 
712
 
 
713
            from_missing = not self.has_filename(from_rel)
 
714
            to_missing = not self.has_filename(to_rel)
 
715
            if after:
 
716
                move_file = False
 
717
            else:
 
718
                move_file = True
 
719
            if to_missing:
 
720
                if not move_file:
 
721
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
722
                        errors.NoSuchFile(path=to_rel,
 
723
                        extra="New file has not been created yet"))
 
724
                elif from_missing:
 
725
                    # neither path exists
 
726
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
727
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
728
            else:
 
729
                if from_missing: # implicitly just update our path mapping
 
730
                    move_file = False
 
731
                elif not after:
 
732
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
733
                        extra="(Use --after to update the Bazaar id)")
 
734
 
 
735
            rollbacks = []
 
736
            def rollback_rename():
 
737
                """A single rename has failed, roll it back."""
 
738
                # roll back everything, even if we encounter trouble doing one
 
739
                # of them.
 
740
                #
 
741
                # TODO: at least log the other exceptions rather than just
 
742
                # losing them mbp 20070307
 
743
                exc_info = None
 
744
                for rollback in reversed(rollbacks):
 
745
                    try:
 
746
                        rollback()
 
747
                    except Exception, e:
 
748
                        exc_info = sys.exc_info()
 
749
                if exc_info:
 
750
                    raise exc_info[0], exc_info[1], exc_info[2]
 
751
 
 
752
            # perform the disk move first - its the most likely failure point.
 
753
            if move_file:
 
754
                from_rel_abs = self.abspath(from_rel)
 
755
                to_rel_abs = self.abspath(to_rel)
 
756
                try:
 
757
                    osutils.rename(from_rel_abs, to_rel_abs)
 
758
                except OSError, e:
 
759
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
760
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
761
            try:
 
762
                # perform the rename in the inventory next if needed: its easy
 
763
                # to rollback
 
764
                if update_inventory:
 
765
                    # rename the entry
 
766
                    from_entry = inv[from_id]
 
767
                    current_parent = from_entry.parent_id
 
768
                    inv.rename(from_id, to_dir_id, from_tail)
 
769
                    rollbacks.append(
 
770
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
771
                # finally do the rename in the dirstate, which is a little
 
772
                # tricky to rollback, but least likely to need it.
 
773
                old_block_index, old_entry_index, dir_present, file_present = \
 
774
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
775
                old_block = state._dirblocks[old_block_index][1]
 
776
                old_entry = old_block[old_entry_index]
 
777
                from_key, old_entry_details = old_entry
 
778
                cur_details = old_entry_details[0]
 
779
                # remove the old row
 
780
                to_key = ((to_block[0],) + from_key[1:3])
 
781
                minikind = cur_details[0]
 
782
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
783
                         minikind=minikind,
 
784
                         executable=cur_details[3],
 
785
                         fingerprint=cur_details[1],
 
786
                         packed_stat=cur_details[4],
 
787
                         size=cur_details[2],
 
788
                         to_block=to_block,
 
789
                         to_key=to_key,
 
790
                         to_path_utf8=to_rel_utf8)
 
791
 
 
792
                if minikind == 'd':
 
793
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
794
                        """Recursively update all entries in this dirblock."""
 
795
                        assert from_dir != '', "renaming root not supported"
 
796
                        from_key = (from_dir, '')
 
797
                        from_block_idx, present = \
 
798
                            state._find_block_index_from_key(from_key)
 
799
                        if not present:
 
800
                            # This is the old record, if it isn't present, then
 
801
                            # there is theoretically nothing to update.
 
802
                            # (Unless it isn't present because of lazy loading,
 
803
                            # but we don't do that yet)
 
804
                            return
 
805
                        from_block = state._dirblocks[from_block_idx]
 
806
                        to_block_index, to_entry_index, _, _ = \
 
807
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
808
                        to_block_index = state._ensure_block(
 
809
                            to_block_index, to_entry_index, to_dir_utf8)
 
810
                        to_block = state._dirblocks[to_block_index]
 
811
 
 
812
                        # Grab a copy since move_one may update the list.
 
813
                        for entry in from_block[1][:]:
 
814
                            assert entry[0][0] == from_dir
 
815
                            cur_details = entry[1][0]
 
816
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
817
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
818
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
819
                            minikind = cur_details[0]
 
820
                            if minikind in 'ar':
 
821
                                # Deleted children of a renamed directory
 
822
                                # Do not need to be updated.
 
823
                                # Children that have been renamed out of this
 
824
                                # directory should also not be updated
 
825
                                continue
 
826
                            move_one(entry, from_path_utf8=from_path_utf8,
 
827
                                     minikind=minikind,
 
828
                                     executable=cur_details[3],
 
829
                                     fingerprint=cur_details[1],
 
830
                                     packed_stat=cur_details[4],
 
831
                                     size=cur_details[2],
 
832
                                     to_block=to_block,
 
833
                                     to_key=to_key,
 
834
                                     to_path_utf8=to_path_utf8)
 
835
                            if minikind == 'd':
 
836
                                # We need to move all the children of this
 
837
                                # entry
 
838
                                update_dirblock(from_path_utf8, to_key,
 
839
                                                to_path_utf8)
 
840
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
841
            except:
 
842
                rollback_rename()
 
843
                raise
 
844
            result.append((from_rel, to_rel))
 
845
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
846
            self._make_dirty(reset_inventory=False)
 
847
 
 
848
        return result
 
849
 
 
850
    def _must_be_locked(self):
 
851
        if not self._control_files._lock_count:
 
852
            raise errors.ObjectNotLocked(self)
 
853
 
 
854
    def _new_tree(self):
 
855
        """Initialize the state in this tree to be a new tree."""
 
856
        self._dirty = True
 
857
 
 
858
    @needs_read_lock
 
859
    def path2id(self, path):
 
860
        """Return the id for path in this tree."""
 
861
        path = path.strip('/')
 
862
        entry = self._get_entry(path=path)
 
863
        if entry == (None, None):
 
864
            return None
 
865
        return entry[0][2]
 
866
 
 
867
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
868
        """See Tree.paths2ids().
 
869
 
 
870
        This specialisation fast-paths the case where all the trees are in the
 
871
        dirstate.
 
872
        """
 
873
        if paths is None:
 
874
            return None
 
875
        parents = self.get_parent_ids()
 
876
        for tree in trees:
 
877
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
878
                parents):
 
879
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
880
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
881
        # -- make all paths utf8 --
 
882
        paths_utf8 = set()
 
883
        for path in paths:
 
884
            paths_utf8.add(path.encode('utf8'))
 
885
        paths = paths_utf8
 
886
        # -- paths is now a utf8 path set --
 
887
        # -- get the state object and prepare it.
 
888
        state = self.current_dirstate()
 
889
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
890
            and '' not in paths):
 
891
            paths2ids = self._paths2ids_using_bisect
 
892
        else:
 
893
            paths2ids = self._paths2ids_in_memory
 
894
        return paths2ids(paths, search_indexes,
 
895
                         require_versioned=require_versioned)
 
896
 
 
897
    def _paths2ids_in_memory(self, paths, search_indexes,
 
898
                             require_versioned=True):
 
899
        state = self.current_dirstate()
 
900
        state._read_dirblocks_if_needed()
 
901
        def _entries_for_path(path):
 
902
            """Return a list with all the entries that match path for all ids.
 
903
            """
 
904
            dirname, basename = os.path.split(path)
 
905
            key = (dirname, basename, '')
 
906
            block_index, present = state._find_block_index_from_key(key)
 
907
            if not present:
 
908
                # the block which should contain path is absent.
 
909
                return []
 
910
            result = []
 
911
            block = state._dirblocks[block_index][1]
 
912
            entry_index, _ = state._find_entry_index(key, block)
 
913
            # we may need to look at multiple entries at this path: walk while the paths match.
 
914
            while (entry_index < len(block) and
 
915
                block[entry_index][0][0:2] == key[0:2]):
 
916
                result.append(block[entry_index])
 
917
                entry_index += 1
 
918
            return result
 
919
        if require_versioned:
 
920
            # -- check all supplied paths are versioned in a search tree. --
 
921
            all_versioned = True
 
922
            for path in paths:
 
923
                path_entries = _entries_for_path(path)
 
924
                if not path_entries:
 
925
                    # this specified path is not present at all: error
 
926
                    all_versioned = False
 
927
                    break
 
928
                found_versioned = False
 
929
                # for each id at this path
 
930
                for entry in path_entries:
 
931
                    # for each tree.
 
932
                    for index in search_indexes:
 
933
                        if entry[1][index][0] != 'a': # absent
 
934
                            found_versioned = True
 
935
                            # all good: found a versioned cell
 
936
                            break
 
937
                if not found_versioned:
 
938
                    # none of the indexes was not 'absent' at all ids for this
 
939
                    # path.
 
940
                    all_versioned = False
 
941
                    break
 
942
            if not all_versioned:
 
943
                raise errors.PathsNotVersionedError(paths)
 
944
        # -- 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: 
 
952
        # for all search_indexs in each path at or under each element of
 
953
        # search_paths, if the detail is relocated: add the id, and add the
 
954
        # relocated path as one to search if its not searched already. If the
 
955
        # detail is not relocated, add the id.
 
956
        searched_paths = set()
 
957
        found_ids = set()
 
958
        def _process_entry(entry):
 
959
            """Look at search_indexes within entry.
 
960
 
 
961
            If a specific tree's details are relocated, add the relocation
 
962
            target to search_paths if not searched already. If it is absent, do
 
963
            nothing. Otherwise add the id to found_ids.
 
964
            """
 
965
            for index in search_indexes:
 
966
                if entry[1][index][0] == 'r': # relocated
 
967
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
968
                        search_paths.add(entry[1][index][1])
 
969
                elif entry[1][index][0] != 'a': # absent
 
970
                    found_ids.add(entry[0][2])
 
971
        while search_paths:
 
972
            current_root = search_paths.pop()
 
973
            searched_paths.add(current_root)
 
974
            # process the entries for this containing directory: the rest will be
 
975
            # found by their parents recursively.
 
976
            root_entries = _entries_for_path(current_root)
 
977
            if not root_entries:
 
978
                # this specified path is not present at all, skip it.
 
979
                continue
 
980
            for entry in root_entries:
 
981
                _process_entry(entry)
 
982
            initial_key = (current_root, '', '')
 
983
            block_index, _ = state._find_block_index_from_key(initial_key)
 
984
            while (block_index < len(state._dirblocks) and
 
985
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
986
                for entry in state._dirblocks[block_index][1]:
 
987
                    _process_entry(entry)
 
988
                block_index += 1
 
989
        return found_ids
 
990
 
 
991
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
992
                                require_versioned=True):
 
993
        state = self.current_dirstate()
 
994
        found_ids = set()
 
995
 
 
996
        split_paths = sorted(osutils.split(p) for p in paths)
 
997
        found = state._bisect_recursive(split_paths)
 
998
 
 
999
        if require_versioned:
 
1000
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
1001
            for dir_name in split_paths:
 
1002
                if dir_name not in found_dir_names:
 
1003
                    raise errors.PathsNotVersionedError(paths)
 
1004
 
 
1005
        for dir_name_id, trees_info in found.iteritems():
 
1006
            for index in search_indexes:
 
1007
                if trees_info[index][0] not in ('r', 'a'):
 
1008
                    found_ids.add(dir_name_id[2])
 
1009
        return found_ids
 
1010
 
 
1011
    def read_working_inventory(self):
 
1012
        """Read the working inventory.
 
1013
        
 
1014
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1015
        """
 
1016
        return self.inventory
 
1017
 
 
1018
    @needs_read_lock
 
1019
    def revision_tree(self, revision_id):
 
1020
        """See Tree.revision_tree.
 
1021
 
 
1022
        WorkingTree4 supplies revision_trees for any basis tree.
 
1023
        """
 
1024
        revision_id = osutils.safe_revision_id(revision_id)
 
1025
        dirstate = self.current_dirstate()
 
1026
        parent_ids = dirstate.get_parent_ids()
 
1027
        if revision_id not in parent_ids:
 
1028
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1029
        if revision_id in dirstate.get_ghosts():
 
1030
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1031
        return DirStateRevisionTree(dirstate, revision_id,
 
1032
            self.branch.repository)
 
1033
 
 
1034
    @needs_tree_write_lock
 
1035
    def set_last_revision(self, new_revision):
 
1036
        """Change the last revision in the working tree."""
 
1037
        new_revision = osutils.safe_revision_id(new_revision)
 
1038
        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.")
 
1043
            self.set_parent_ids([])
 
1044
        else:
 
1045
            self.set_parent_ids([new_revision] + parents[1:],
 
1046
                allow_leftmost_as_ghost=True)
 
1047
 
 
1048
    @needs_tree_write_lock
 
1049
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1050
        """Set the parent ids to revision_ids.
 
1051
        
 
1052
        See also set_parent_trees. This api will try to retrieve the tree data
 
1053
        for each element of revision_ids from the trees repository. If you have
 
1054
        tree data already available, it is more efficient to use
 
1055
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1056
        an easier API to use.
 
1057
 
 
1058
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1059
            working tree. Any of these may be ghosts.
 
1060
        """
 
1061
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1062
        trees = []
 
1063
        for revision_id in revision_ids:
 
1064
            try:
 
1065
                revtree = self.branch.repository.revision_tree(revision_id)
 
1066
                # TODO: jam 20070213 KnitVersionedFile raises
 
1067
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1068
                #       given revision_id is not present. Should Repository be
 
1069
                #       catching it and re-raising NoSuchRevision?
 
1070
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1071
                revtree = None
 
1072
            trees.append((revision_id, revtree))
 
1073
        self.set_parent_trees(trees,
 
1074
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1075
 
 
1076
    @needs_tree_write_lock
 
1077
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1078
        """Set the parents of the working tree.
 
1079
 
 
1080
        :param parents_list: A list of (revision_id, tree) tuples.
 
1081
            If tree is None, then that element is treated as an unreachable
 
1082
            parent tree - i.e. a ghost.
 
1083
        """
 
1084
        dirstate = self.current_dirstate()
 
1085
        if len(parents_list) > 0:
 
1086
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1087
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1088
        real_trees = []
 
1089
        ghosts = []
 
1090
        # convert absent trees to the null tree, which we convert back to
 
1091
        # missing on access.
 
1092
        for rev_id, tree in parents_list:
 
1093
            rev_id = osutils.safe_revision_id(rev_id)
 
1094
            _mod_revision.check_not_reserved_id(rev_id)
 
1095
            if tree is not None:
 
1096
                real_trees.append((rev_id, tree))
 
1097
            else:
 
1098
                real_trees.append((rev_id,
 
1099
                    self.branch.repository.revision_tree(None)))
 
1100
                ghosts.append(rev_id)
 
1101
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1102
        self._make_dirty(reset_inventory=False)
 
1103
 
 
1104
    def _set_root_id(self, file_id):
 
1105
        """See WorkingTree.set_root_id."""
 
1106
        state = self.current_dirstate()
 
1107
        state.set_path_id('', file_id)
 
1108
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1109
            self._make_dirty(reset_inventory=True)
 
1110
 
 
1111
    @needs_read_lock
 
1112
    def supports_tree_reference(self):
 
1113
        return self._repo_supports_tree_reference
 
1114
 
 
1115
    def unlock(self):
 
1116
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1117
        # do non-implementation specific cleanup
 
1118
        self._cleanup()
 
1119
 
 
1120
        if self._control_files._lock_count == 1:
 
1121
            # eventually we should do signature checking during read locks for
 
1122
            # dirstate updates.
 
1123
            if self._control_files._lock_mode == 'w':
 
1124
                if self._dirty:
 
1125
                    self.flush()
 
1126
            if self._dirstate is not None:
 
1127
                # This is a no-op if there are no modifications.
 
1128
                self._dirstate.save()
 
1129
                self._dirstate.unlock()
 
1130
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1131
            #       point. Instead, it could check if the header has been
 
1132
            #       modified when it is locked, and if not, it can hang on to
 
1133
            #       the data it has in memory.
 
1134
            self._dirstate = None
 
1135
            self._inventory = None
 
1136
        # reverse order of locking.
 
1137
        try:
 
1138
            return self._control_files.unlock()
 
1139
        finally:
 
1140
            self.branch.unlock()
 
1141
 
 
1142
    @needs_tree_write_lock
 
1143
    def unversion(self, file_ids):
 
1144
        """Remove the file ids in file_ids from the current versioned set.
 
1145
 
 
1146
        When a file_id is unversioned, all of its children are automatically
 
1147
        unversioned.
 
1148
 
 
1149
        :param file_ids: The file ids to stop versioning.
 
1150
        :raises: NoSuchId if any fileid is not currently versioned.
 
1151
        """
 
1152
        if not file_ids:
 
1153
            return
 
1154
        state = self.current_dirstate()
 
1155
        state._read_dirblocks_if_needed()
 
1156
        ids_to_unversion = set()
 
1157
        for file_id in file_ids:
 
1158
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1159
        paths_to_unversion = set()
 
1160
        # sketch:
 
1161
        # check if the root is to be unversioned, if so, assert for now.
 
1162
        # walk the state marking unversioned things as absent.
 
1163
        # if there are any un-unversioned ids at the end, raise
 
1164
        for key, details in state._dirblocks[0][1]:
 
1165
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1166
                key[2] in ids_to_unversion):
 
1167
                # I haven't written the code to unversion / yet - it should be
 
1168
                # supported.
 
1169
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1170
        block_index = 0
 
1171
        while block_index < len(state._dirblocks):
 
1172
            # process one directory at a time.
 
1173
            block = state._dirblocks[block_index]
 
1174
            # first check: is the path one to remove - it or its children
 
1175
            delete_block = False
 
1176
            for path in paths_to_unversion:
 
1177
                if (block[0].startswith(path) and
 
1178
                    (len(block[0]) == len(path) or
 
1179
                     block[0][len(path)] == '/')):
 
1180
                    # this entire block should be deleted - its the block for a
 
1181
                    # path to unversion; or the child of one
 
1182
                    delete_block = True
 
1183
                    break
 
1184
            # TODO: trim paths_to_unversion as we pass by paths
 
1185
            if delete_block:
 
1186
                # this block is to be deleted: process it.
 
1187
                # TODO: we can special case the no-parents case and
 
1188
                # just forget the whole block.
 
1189
                entry_index = 0
 
1190
                while entry_index < len(block[1]):
 
1191
                    # Mark this file id as having been removed
 
1192
                    entry = block[1][entry_index]
 
1193
                    ids_to_unversion.discard(entry[0][2])
 
1194
                    if (entry[1][0][0] == 'a'
 
1195
                        or not state._make_absent(entry)):
 
1196
                        entry_index += 1
 
1197
                # go to the next block. (At the moment we dont delete empty
 
1198
                # dirblocks)
 
1199
                block_index += 1
 
1200
                continue
 
1201
            entry_index = 0
 
1202
            while entry_index < len(block[1]):
 
1203
                entry = block[1][entry_index]
 
1204
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1205
                    # ^ some parent row.
 
1206
                    entry[0][2] not in ids_to_unversion):
 
1207
                    # ^ not an id to unversion
 
1208
                    entry_index += 1
 
1209
                    continue
 
1210
                if entry[1][0][0] == 'd':
 
1211
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1212
                if not state._make_absent(entry):
 
1213
                    entry_index += 1
 
1214
                # we have unversioned this id
 
1215
                ids_to_unversion.remove(entry[0][2])
 
1216
            block_index += 1
 
1217
        if ids_to_unversion:
 
1218
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1219
        self._make_dirty(reset_inventory=False)
 
1220
        # have to change the legacy inventory too.
 
1221
        if self._inventory is not None:
 
1222
            for file_id in file_ids:
 
1223
                self._inventory.remove_recursive_id(file_id)
 
1224
 
 
1225
    @needs_read_lock
 
1226
    def _validate(self):
 
1227
        self._dirstate._validate()
 
1228
 
 
1229
    @needs_tree_write_lock
 
1230
    def _write_inventory(self, inv):
 
1231
        """Write inventory as the current inventory."""
 
1232
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1233
        self.current_dirstate().set_state_from_inventory(inv)
 
1234
        self._make_dirty(reset_inventory=False)
 
1235
        if self._inventory is not None:
 
1236
            self._inventory = inv
 
1237
        self.flush()
 
1238
 
 
1239
 
 
1240
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1241
    """The first consolidated dirstate working tree format.
 
1242
 
 
1243
    This format:
 
1244
        - exists within a metadir controlling .bzr
 
1245
        - includes an explicit version marker for the workingtree control
 
1246
          files, separate from the BzrDir format
 
1247
        - modifies the hash cache format
 
1248
        - is new in bzr 0.15
 
1249
        - uses a LockDir to guard access to it.
 
1250
    """
 
1251
 
 
1252
    upgrade_recommended = False
 
1253
 
 
1254
    def get_format_string(self):
 
1255
        """See WorkingTreeFormat.get_format_string()."""
 
1256
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1257
 
 
1258
    def get_format_description(self):
 
1259
        """See WorkingTreeFormat.get_format_description()."""
 
1260
        return "Working tree format 4"
 
1261
 
 
1262
    def initialize(self, a_bzrdir, revision_id=None):
 
1263
        """See WorkingTreeFormat.initialize().
 
1264
 
 
1265
        :param revision_id: allows creating a working tree at a different
 
1266
        revision than the branch is at.
 
1267
 
 
1268
        These trees get an initial random root id, if their repository supports
 
1269
        rich root data, TREE_ROOT otherwise.
 
1270
        """
 
1271
        revision_id = osutils.safe_revision_id(revision_id)
 
1272
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1273
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1274
        transport = a_bzrdir.get_workingtree_transport(self)
 
1275
        control_files = self._open_control_files(a_bzrdir)
 
1276
        control_files.create_lock()
 
1277
        control_files.lock_write()
 
1278
        control_files.put_utf8('format', self.get_format_string())
 
1279
        branch = a_bzrdir.open_branch()
 
1280
        if revision_id is None:
 
1281
            revision_id = branch.last_revision()
 
1282
        local_path = transport.local_abspath('dirstate')
 
1283
        # write out new dirstate (must exist when we create the tree)
 
1284
        state = dirstate.DirState.initialize(local_path)
 
1285
        state.unlock()
 
1286
        del state
 
1287
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1288
                         branch,
 
1289
                         _format=self,
 
1290
                         _bzrdir=a_bzrdir,
 
1291
                         _control_files=control_files)
 
1292
        wt._new_tree()
 
1293
        wt.lock_tree_write()
 
1294
        try:
 
1295
            if revision_id in (None, NULL_REVISION):
 
1296
                if branch.repository.supports_rich_root():
 
1297
                    wt._set_root_id(generate_ids.gen_root_id())
 
1298
                else:
 
1299
                    wt._set_root_id(ROOT_ID)
 
1300
                wt.flush()
 
1301
            wt.set_last_revision(revision_id)
 
1302
            wt.flush()
 
1303
            basis = wt.basis_tree()
 
1304
            basis.lock_read()
 
1305
            # if the basis has a root id we have to use that; otherwise we use
 
1306
            # a new random one
 
1307
            basis_root_id = basis.get_root_id()
 
1308
            if basis_root_id is not None:
 
1309
                wt._set_root_id(basis_root_id)
 
1310
                wt.flush()
 
1311
            transform.build_tree(basis, wt)
 
1312
            basis.unlock()
 
1313
        finally:
 
1314
            control_files.unlock()
 
1315
            wt.unlock()
 
1316
        return wt
 
1317
 
 
1318
    def _open(self, a_bzrdir, control_files):
 
1319
        """Open the tree itself.
 
1320
 
 
1321
        :param a_bzrdir: the dir for the tree.
 
1322
        :param control_files: the control files for the tree.
 
1323
        """
 
1324
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1325
                           branch=a_bzrdir.open_branch(),
 
1326
                           _format=self,
 
1327
                           _bzrdir=a_bzrdir,
 
1328
                           _control_files=control_files)
 
1329
 
 
1330
    def __get_matchingbzrdir(self):
 
1331
        # please test against something that will let us do tree references
 
1332
        return bzrdir.format_registry.make_bzrdir(
 
1333
            'dirstate-with-subtree')
 
1334
 
 
1335
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1336
 
 
1337
 
 
1338
class DirStateRevisionTree(Tree):
 
1339
    """A revision tree pulling the inventory from a dirstate."""
 
1340
 
 
1341
    def __init__(self, dirstate, revision_id, repository):
 
1342
        self._dirstate = dirstate
 
1343
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1344
        self._repository = repository
 
1345
        self._inventory = None
 
1346
        self._locked = 0
 
1347
        self._dirstate_locked = False
 
1348
 
 
1349
    def __repr__(self):
 
1350
        return "<%s of %s in %s>" % \
 
1351
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1352
 
 
1353
    def annotate_iter(self, file_id,
 
1354
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1355
        """See Tree.annotate_iter"""
 
1356
        w = self._get_weave(file_id)
 
1357
        return w.annotate_iter(self.inventory[file_id].revision)
 
1358
 
 
1359
    def _get_ancestors(self, default_revision):
 
1360
        return set(self._repository.get_ancestry(self._revision_id,
 
1361
                                                 topo_sorted=False))
 
1362
    def _comparison_data(self, entry, path):
 
1363
        """See Tree._comparison_data."""
 
1364
        if entry is None:
 
1365
            return None, False, None
 
1366
        # trust the entry as RevisionTree does, but this may not be
 
1367
        # sensible: the entry might not have come from us?
 
1368
        return entry.kind, entry.executable, None
 
1369
 
 
1370
    def _file_size(self, entry, stat_value):
 
1371
        return entry.text_size
 
1372
 
 
1373
    def filter_unversioned_files(self, paths):
 
1374
        """Filter out paths that are not versioned.
 
1375
 
 
1376
        :return: set of paths.
 
1377
        """
 
1378
        pred = self.has_filename
 
1379
        return set((p for p in paths if not pred(p)))
 
1380
 
 
1381
    def get_root_id(self):
 
1382
        return self.path2id('')
 
1383
 
 
1384
    def _get_parent_index(self):
 
1385
        """Return the index in the dirstate referenced by this tree."""
 
1386
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1387
 
 
1388
    def _get_entry(self, file_id=None, path=None):
 
1389
        """Get the dirstate row for file_id or path.
 
1390
 
 
1391
        If either file_id or path is supplied, it is used as the key to lookup.
 
1392
        If both are supplied, the fastest lookup is used, and an error is
 
1393
        raised if they do not both point at the same row.
 
1394
        
 
1395
        :param file_id: An optional unicode file_id to be looked up.
 
1396
        :param path: An optional unicode path to be looked up.
 
1397
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1398
        """
 
1399
        if file_id is None and path is None:
 
1400
            raise errors.BzrError('must supply file_id or path')
 
1401
        file_id = osutils.safe_file_id(file_id)
 
1402
        if path is not None:
 
1403
            path = path.encode('utf8')
 
1404
        parent_index = self._get_parent_index()
 
1405
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1406
 
 
1407
    def _generate_inventory(self):
 
1408
        """Create and set self.inventory from the dirstate object.
 
1409
 
 
1410
        (So this is only called the first time the inventory is requested for
 
1411
        this tree; it then remains in memory until it's out of date.)
 
1412
 
 
1413
        This is relatively expensive: we have to walk the entire dirstate.
 
1414
        """
 
1415
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1416
            'dirstate revision tree'
 
1417
        # separate call for profiling - makes it clear where the costs are.
 
1418
        self._dirstate._read_dirblocks_if_needed()
 
1419
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1420
            'parent %s has disappeared from %s' % (
 
1421
            self._revision_id, self._dirstate.get_parent_ids())
 
1422
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1423
        # This is identical now to the WorkingTree _generate_inventory except
 
1424
        # for the tree index use.
 
1425
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1426
        current_id = root_key[2]
 
1427
        assert current_entry[parent_index][0] == 'd'
 
1428
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1429
        inv.root.revision = current_entry[parent_index][4]
 
1430
        # Turn some things into local variables
 
1431
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1432
        factory = entry_factory
 
1433
        utf8_decode = cache_utf8._utf8_decode
 
1434
        inv_byid = inv._byid
 
1435
        # we could do this straight out of the dirstate; it might be fast
 
1436
        # and should be profiled - RBC 20070216
 
1437
        parent_ies = {'' : inv.root}
 
1438
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1439
            dirname = block[0]
 
1440
            try:
 
1441
                parent_ie = parent_ies[dirname]
 
1442
            except KeyError:
 
1443
                # all the paths in this block are not versioned in this tree
 
1444
                continue
 
1445
            for key, entry in block[1]:
 
1446
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1447
                if minikind in ('a', 'r'): # absent, relocated
 
1448
                    # not this tree
 
1449
                    continue
 
1450
                name = key[1]
 
1451
                name_unicode = utf8_decode(name)[0]
 
1452
                file_id = key[2]
 
1453
                kind = minikind_to_kind[minikind]
 
1454
                inv_entry = factory[kind](file_id, name_unicode,
 
1455
                                          parent_ie.file_id)
 
1456
                inv_entry.revision = revid
 
1457
                if kind == 'file':
 
1458
                    inv_entry.executable = executable
 
1459
                    inv_entry.text_size = size
 
1460
                    inv_entry.text_sha1 = fingerprint
 
1461
                elif kind == 'directory':
 
1462
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1463
                elif kind == 'symlink':
 
1464
                    inv_entry.executable = False
 
1465
                    inv_entry.text_size = None
 
1466
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1467
                elif kind == 'tree-reference':
 
1468
                    inv_entry.reference_revision = fingerprint or None
 
1469
                else:
 
1470
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1471
                            % entry)
 
1472
                # These checks cost us around 40ms on a 55k entry tree
 
1473
                assert file_id not in inv_byid
 
1474
                assert name_unicode not in parent_ie.children
 
1475
                inv_byid[file_id] = inv_entry
 
1476
                parent_ie.children[name_unicode] = inv_entry
 
1477
        self._inventory = inv
 
1478
 
 
1479
    def get_file_mtime(self, file_id, path=None):
 
1480
        """Return the modification time for this record.
 
1481
 
 
1482
        We return the timestamp of the last-changed revision.
 
1483
        """
 
1484
        # Make sure the file exists
 
1485
        entry = self._get_entry(file_id, path=path)
 
1486
        if entry == (None, None): # do we raise?
 
1487
            return None
 
1488
        parent_index = self._get_parent_index()
 
1489
        last_changed_revision = entry[1][parent_index][4]
 
1490
        return self._repository.get_revision(last_changed_revision).timestamp
 
1491
 
 
1492
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1493
        entry = self._get_entry(file_id=file_id, path=path)
 
1494
        parent_index = self._get_parent_index()
 
1495
        parent_details = entry[1][parent_index]
 
1496
        if parent_details[0] == 'f':
 
1497
            return parent_details[1]
 
1498
        return None
 
1499
 
 
1500
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1501
    def get_weave(self, file_id):
 
1502
        return self._get_weave(file_id)
 
1503
 
 
1504
    def _get_weave(self, file_id):
 
1505
        return self._repository.weave_store.get_weave(file_id,
 
1506
                self._repository.get_transaction())
 
1507
 
 
1508
    def get_file(self, file_id):
 
1509
        return StringIO(self.get_file_text(file_id))
 
1510
 
 
1511
    def get_file_lines(self, file_id):
 
1512
        ie = self.inventory[file_id]
 
1513
        return self._get_weave(file_id).get_lines(ie.revision)
 
1514
 
 
1515
    def get_file_size(self, file_id):
 
1516
        return self.inventory[file_id].text_size
 
1517
 
 
1518
    def get_file_text(self, file_id):
 
1519
        return ''.join(self.get_file_lines(file_id))
 
1520
 
 
1521
    def get_reference_revision(self, file_id, path=None):
 
1522
        return self.inventory[file_id].reference_revision
 
1523
 
 
1524
    def get_symlink_target(self, file_id):
 
1525
        entry = self._get_entry(file_id=file_id)
 
1526
        parent_index = self._get_parent_index()
 
1527
        if entry[1][parent_index][0] != 'l':
 
1528
            return None
 
1529
        else:
 
1530
            # At present, none of the tree implementations supports non-ascii
 
1531
            # symlink targets. So we will just assume that the dirstate path is
 
1532
            # correct.
 
1533
            return entry[1][parent_index][1]
 
1534
 
 
1535
    def get_revision_id(self):
 
1536
        """Return the revision id for this tree."""
 
1537
        return self._revision_id
 
1538
 
 
1539
    def _get_inventory(self):
 
1540
        if self._inventory is not None:
 
1541
            return self._inventory
 
1542
        self._must_be_locked()
 
1543
        self._generate_inventory()
 
1544
        return self._inventory
 
1545
 
 
1546
    inventory = property(_get_inventory,
 
1547
                         doc="Inventory of this Tree")
 
1548
 
 
1549
    def get_parent_ids(self):
 
1550
        """The parents of a tree in the dirstate are not cached."""
 
1551
        return self._repository.get_revision(self._revision_id).parent_ids
 
1552
 
 
1553
    def has_filename(self, filename):
 
1554
        return bool(self.path2id(filename))
 
1555
 
 
1556
    def kind(self, file_id):
 
1557
        return self.inventory[file_id].kind
 
1558
 
 
1559
    def is_executable(self, file_id, path=None):
 
1560
        ie = self.inventory[file_id]
 
1561
        if ie.kind != "file":
 
1562
            return None
 
1563
        return ie.executable
 
1564
 
 
1565
    def list_files(self, include_root=False):
 
1566
        # We use a standard implementation, because DirStateRevisionTree is
 
1567
        # dealing with one of the parents of the current state
 
1568
        inv = self._get_inventory()
 
1569
        entries = inv.iter_entries()
 
1570
        if self.inventory.root is not None and not include_root:
 
1571
            entries.next()
 
1572
        for path, entry in entries:
 
1573
            yield path, 'V', entry.kind, entry.file_id, entry
 
1574
 
 
1575
    def lock_read(self):
 
1576
        """Lock the tree for a set of operations."""
 
1577
        if not self._locked:
 
1578
            self._repository.lock_read()
 
1579
            if self._dirstate._lock_token is None:
 
1580
                self._dirstate.lock_read()
 
1581
                self._dirstate_locked = True
 
1582
        self._locked += 1
 
1583
 
 
1584
    def _must_be_locked(self):
 
1585
        if not self._locked:
 
1586
            raise errors.ObjectNotLocked(self)
 
1587
 
 
1588
    @needs_read_lock
 
1589
    def path2id(self, path):
 
1590
        """Return the id for path in this tree."""
 
1591
        # lookup by path: faster than splitting and walking the ivnentory.
 
1592
        entry = self._get_entry(path=path)
 
1593
        if entry == (None, None):
 
1594
            return None
 
1595
        return entry[0][2]
 
1596
 
 
1597
    def unlock(self):
 
1598
        """Unlock, freeing any cache memory used during the lock."""
 
1599
        # outside of a lock, the inventory is suspect: release it.
 
1600
        self._locked -=1
 
1601
        if not self._locked:
 
1602
            self._inventory = None
 
1603
            self._locked = 0
 
1604
            if self._dirstate_locked:
 
1605
                self._dirstate.unlock()
 
1606
                self._dirstate_locked = False
 
1607
            self._repository.unlock()
 
1608
 
 
1609
    def walkdirs(self, prefix=""):
 
1610
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1611
        # implementation based on an inventory.  
 
1612
        # This should be cleaned up to use the much faster Dirstate code
 
1613
        # So for now, we just build up the parent inventory, and extract
 
1614
        # it the same way RevisionTree does.
 
1615
        _directory = 'directory'
 
1616
        inv = self._get_inventory()
 
1617
        top_id = inv.path2id(prefix)
 
1618
        if top_id is None:
 
1619
            pending = []
 
1620
        else:
 
1621
            pending = [(prefix, top_id)]
 
1622
        while pending:
 
1623
            dirblock = []
 
1624
            relpath, file_id = pending.pop()
 
1625
            # 0 - relpath, 1- file-id
 
1626
            if relpath:
 
1627
                relroot = relpath + '/'
 
1628
            else:
 
1629
                relroot = ""
 
1630
            # FIXME: stash the node in pending
 
1631
            entry = inv[file_id]
 
1632
            for name, child in entry.sorted_children():
 
1633
                toppath = relroot + name
 
1634
                dirblock.append((toppath, name, child.kind, None,
 
1635
                    child.file_id, child.kind
 
1636
                    ))
 
1637
            yield (relpath, entry.file_id), dirblock
 
1638
            # push the user specified dirs from dirblock
 
1639
            for dir in reversed(dirblock):
 
1640
                if dir[2] == _directory:
 
1641
                    pending.append((dir[0], dir[4]))
 
1642
 
 
1643
 
 
1644
class InterDirStateTree(InterTree):
 
1645
    """Fast path optimiser for changes_from with dirstate trees.
 
1646
    
 
1647
    This is used only when both trees are in the dirstate working file, and 
 
1648
    the source is any parent within the dirstate, and the destination is 
 
1649
    the current working tree of the same dirstate.
 
1650
    """
 
1651
    # this could be generalized to allow comparisons between any trees in the
 
1652
    # dirstate, and possibly between trees stored in different dirstates.
 
1653
 
 
1654
    def __init__(self, source, target):
 
1655
        super(InterDirStateTree, self).__init__(source, target)
 
1656
        if not InterDirStateTree.is_compatible(source, target):
 
1657
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1658
 
 
1659
    @staticmethod
 
1660
    def make_source_parent_tree(source, target):
 
1661
        """Change the source tree into a parent of the target."""
 
1662
        revid = source.commit('record tree')
 
1663
        target.branch.repository.fetch(source.branch.repository, revid)
 
1664
        target.set_parent_ids([revid])
 
1665
        return target.basis_tree(), target
 
1666
 
 
1667
    _matching_from_tree_format = WorkingTreeFormat4()
 
1668
    _matching_to_tree_format = WorkingTreeFormat4()
 
1669
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1670
 
 
1671
    def _iter_changes(self, include_unchanged=False,
 
1672
                      specific_files=None, pb=None, extra_trees=[],
 
1673
                      require_versioned=True, want_unversioned=False):
 
1674
        """Return the changes from source to target.
 
1675
 
 
1676
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1677
            for details.
 
1678
        :param specific_files: An optional list of file paths to restrict the
 
1679
            comparison to. When mapping filenames to ids, all matches in all
 
1680
            trees (including optional extra_trees) are used, and all children of
 
1681
            matched directories are included.
 
1682
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1683
            unchanged entries in the result.
 
1684
        :param extra_trees: An optional list of additional trees to use when
 
1685
            mapping the contents of specific_files (paths) to file_ids.
 
1686
        :param require_versioned: If True, all files in specific_files must be
 
1687
            versioned in one of source, target, extra_trees or
 
1688
            PathsNotVersionedError is raised.
 
1689
        :param want_unversioned: Should unversioned files be returned in the
 
1690
            output. An unversioned file is defined as one with (False, False)
 
1691
            for the versioned pair.
 
1692
        """
 
1693
        utf8_decode = cache_utf8._utf8_decode
 
1694
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1695
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1696
        # NB: show_status depends on being able to pass in non-versioned files
 
1697
        # and report them as unknown
 
1698
        # TODO: handle extra trees in the dirstate.
 
1699
        # TODO: handle comparisons as an empty tree as a different special
 
1700
        # case? mbp 20070226
 
1701
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1702
            # we can't fast-path these cases (yet)
 
1703
            for f in super(InterDirStateTree, self)._iter_changes(
 
1704
                include_unchanged, specific_files, pb, extra_trees,
 
1705
                require_versioned, want_unversioned=want_unversioned):
 
1706
                yield f
 
1707
            return
 
1708
        parent_ids = self.target.get_parent_ids()
 
1709
        assert (self.source._revision_id in parent_ids), \
 
1710
                "revision {%s} is not stored in {%s}, but %s " \
 
1711
                "can only be used for trees stored in the dirstate" \
 
1712
                % (self.source._revision_id, self.target, self._iter_changes)
 
1713
        target_index = 0
 
1714
        if self.source._revision_id == NULL_REVISION:
 
1715
            source_index = None
 
1716
            indices = (target_index,)
 
1717
        else:
 
1718
            assert (self.source._revision_id in parent_ids), \
 
1719
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1720
                self.source._revision_id, parent_ids)
 
1721
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1722
            indices = (source_index,target_index)
 
1723
        # -- make all specific_files utf8 --
 
1724
        if specific_files:
 
1725
            specific_files_utf8 = set()
 
1726
            for path in specific_files:
 
1727
                specific_files_utf8.add(path.encode('utf8'))
 
1728
            specific_files = specific_files_utf8
 
1729
        else:
 
1730
            specific_files = set([''])
 
1731
        # -- specific_files is now a utf8 path set --
 
1732
        # -- get the state object and prepare it.
 
1733
        state = self.target.current_dirstate()
 
1734
        state._read_dirblocks_if_needed()
 
1735
        def _entries_for_path(path):
 
1736
            """Return a list with all the entries that match path for all ids.
 
1737
            """
 
1738
            dirname, basename = os.path.split(path)
 
1739
            key = (dirname, basename, '')
 
1740
            block_index, present = state._find_block_index_from_key(key)
 
1741
            if not present:
 
1742
                # the block which should contain path is absent.
 
1743
                return []
 
1744
            result = []
 
1745
            block = state._dirblocks[block_index][1]
 
1746
            entry_index, _ = state._find_entry_index(key, block)
 
1747
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1748
            while (entry_index < len(block) and
 
1749
                block[entry_index][0][0:2] == key[0:2]):
 
1750
                result.append(block[entry_index])
 
1751
                entry_index += 1
 
1752
            return result
 
1753
        if require_versioned:
 
1754
            # -- check all supplied paths are versioned in a search tree. --
 
1755
            all_versioned = True
 
1756
            for path in specific_files:
 
1757
                path_entries = _entries_for_path(path)
 
1758
                if not path_entries:
 
1759
                    # this specified path is not present at all: error
 
1760
                    all_versioned = False
 
1761
                    break
 
1762
                found_versioned = False
 
1763
                # for each id at this path
 
1764
                for entry in path_entries:
 
1765
                    # for each tree.
 
1766
                    for index in indices:
 
1767
                        if entry[1][index][0] != 'a': # absent
 
1768
                            found_versioned = True
 
1769
                            # all good: found a versioned cell
 
1770
                            break
 
1771
                if not found_versioned:
 
1772
                    # none of the indexes was not 'absent' at all ids for this
 
1773
                    # path.
 
1774
                    all_versioned = False
 
1775
                    break
 
1776
            if not all_versioned:
 
1777
                raise errors.PathsNotVersionedError(specific_files)
 
1778
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1779
        search_specific_files = set()
 
1780
        for path in specific_files:
 
1781
            other_specific_files = specific_files.difference(set([path]))
 
1782
            if not osutils.is_inside_any(other_specific_files, path):
 
1783
                # this is a top level path, we must check it.
 
1784
                search_specific_files.add(path)
 
1785
        # sketch: 
 
1786
        # compare source_index and target_index at or under each element of search_specific_files.
 
1787
        # follow the following comparison table. Note that we only want to do diff operations when
 
1788
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1789
        # for the target.
 
1790
        # cases:
 
1791
        # 
 
1792
        # Source | Target | disk | action
 
1793
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1794
        #        |        |      | diff check on source-target
 
1795
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1796
        #        |        |      | ???
 
1797
        #   r    |  a     |      | add source to search
 
1798
        #   r    |  a     |  a   | 
 
1799
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1800
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1801
        #   a    | fdlt   |      | add new id
 
1802
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1803
        #   a    |  a     |      | not present in either tree, skip
 
1804
        #   a    |  a     |  a   | not present in any tree, skip
 
1805
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1806
        #        |        |      | may not be selected by the users list of paths.
 
1807
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1808
        #        |        |      | may not be selected by the users list of paths.
 
1809
        #  fdlt  | fdlt   |      | content in both: diff them
 
1810
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1811
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1812
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1813
        #  fdlt  |  r     |      | 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
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1817
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1818
        #        |        |      | this id at the other path.
 
1819
 
 
1820
        # for all search_indexs in each path at or under each element of
 
1821
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1822
        # relocated path as one to search if its not searched already. If the
 
1823
        # detail is not relocated, add the id.
 
1824
        searched_specific_files = set()
 
1825
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1826
        # Using a list so that we can access the values and change them in
 
1827
        # nested scope. Each one is [path, file_id, entry]
 
1828
        last_source_parent = [None, None]
 
1829
        last_target_parent = [None, None]
 
1830
 
 
1831
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1832
 
 
1833
        # Just a sentry, so that _process_entry can say that this
 
1834
        # record is handled, but isn't interesting to process (unchanged)
 
1835
        uninteresting = object()
 
1836
 
 
1837
 
 
1838
        old_dirname_to_file_id = {}
 
1839
        new_dirname_to_file_id = {}
 
1840
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1841
        #       keeping a cache of directories that we have seen.
 
1842
 
 
1843
        def _process_entry(entry, path_info):
 
1844
            """Compare an entry and real disk to generate delta information.
 
1845
 
 
1846
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1847
                the path of entry. If None, then the path is considered absent.
 
1848
                (Perhaps we should pass in a concrete entry for this ?)
 
1849
                Basename is returned as a utf8 string because we expect this
 
1850
                tuple will be ignored, and don't want to take the time to
 
1851
                decode.
 
1852
            :return: None if these don't match
 
1853
                     A tuple of information about the change, or
 
1854
                     the object 'uninteresting' if these match, but are
 
1855
                     basically identical.
 
1856
            """
 
1857
            if source_index is None:
 
1858
                source_details = NULL_PARENT_DETAILS
 
1859
            else:
 
1860
                source_details = entry[1][source_index]
 
1861
            target_details = entry[1][target_index]
 
1862
            target_minikind = target_details[0]
 
1863
            if path_info is not None and target_minikind in 'fdlt':
 
1864
                assert target_index == 0
 
1865
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1866
                                                  stat_value=path_info[3])
 
1867
                # The entry may have been modified by update_entry
 
1868
                target_details = entry[1][target_index]
 
1869
                target_minikind = target_details[0]
 
1870
            else:
 
1871
                link_or_sha1 = None
 
1872
            file_id = entry[0][2]
 
1873
            source_minikind = source_details[0]
 
1874
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1875
                # claimed content in both: diff
 
1876
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1877
                #        |        |      | diff check on source-target
 
1878
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1879
                #        |        |      | ???
 
1880
                if source_minikind in 'r':
 
1881
                    # add the source to the search path to find any children it
 
1882
                    # has.  TODO ? : only add if it is a container ?
 
1883
                    if not osutils.is_inside_any(searched_specific_files,
 
1884
                                                 source_details[1]):
 
1885
                        search_specific_files.add(source_details[1])
 
1886
                    # generate the old path; this is needed for stating later
 
1887
                    # as well.
 
1888
                    old_path = source_details[1]
 
1889
                    old_dirname, old_basename = os.path.split(old_path)
 
1890
                    path = pathjoin(entry[0][0], entry[0][1])
 
1891
                    old_entry = state._get_entry(source_index,
 
1892
                                                 path_utf8=old_path)
 
1893
                    # update the source details variable to be the real
 
1894
                    # location.
 
1895
                    source_details = old_entry[1][source_index]
 
1896
                    source_minikind = source_details[0]
 
1897
                else:
 
1898
                    old_dirname = entry[0][0]
 
1899
                    old_basename = entry[0][1]
 
1900
                    old_path = path = None
 
1901
                if path_info is None:
 
1902
                    # the file is missing on disk, show as removed.
 
1903
                    content_change = True
 
1904
                    target_kind = None
 
1905
                    target_exec = False
 
1906
                else:
 
1907
                    # source and target are both versioned and disk file is present.
 
1908
                    target_kind = path_info[2]
 
1909
                    if target_kind == 'directory':
 
1910
                        if path is None:
 
1911
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1912
                        new_dirname_to_file_id[path] = file_id
 
1913
                        if source_minikind != 'd':
 
1914
                            content_change = True
 
1915
                        else:
 
1916
                            # directories have no fingerprint
 
1917
                            content_change = False
 
1918
                        target_exec = False
 
1919
                    elif target_kind == 'file':
 
1920
                        if source_minikind != 'f':
 
1921
                            content_change = True
 
1922
                        else:
 
1923
                            # We could check the size, but we already have the
 
1924
                            # sha1 hash.
 
1925
                            content_change = (link_or_sha1 != source_details[1])
 
1926
                        # Target details is updated at update_entry time
 
1927
                        if use_filesystem_for_exec:
 
1928
                            # We don't need S_ISREG here, because we are sure
 
1929
                            # we are dealing with a file.
 
1930
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1931
                        else:
 
1932
                            target_exec = target_details[3]
 
1933
                    elif target_kind == 'symlink':
 
1934
                        if source_minikind != 'l':
 
1935
                            content_change = True
 
1936
                        else:
 
1937
                            content_change = (link_or_sha1 != source_details[1])
 
1938
                        target_exec = False
 
1939
                    elif target_kind == 'tree-reference':
 
1940
                        if source_minikind != 't':
 
1941
                            content_change = True
 
1942
                        else:
 
1943
                            content_change = False
 
1944
                        target_exec = False
 
1945
                    else:
 
1946
                        raise Exception, "unknown kind %s" % path_info[2]
 
1947
                if source_minikind == 'd':
 
1948
                    if path is None:
 
1949
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1950
                    old_dirname_to_file_id[old_path] = file_id
 
1951
                # parent id is the entry for the path in the target tree
 
1952
                if old_dirname == last_source_parent[0]:
 
1953
                    source_parent_id = last_source_parent[1]
 
1954
                else:
 
1955
                    try:
 
1956
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1957
                    except KeyError:
 
1958
                        source_parent_entry = state._get_entry(source_index,
 
1959
                                                               path_utf8=old_dirname)
 
1960
                        source_parent_id = source_parent_entry[0][2]
 
1961
                    if source_parent_id == entry[0][2]:
 
1962
                        # This is the root, so the parent is None
 
1963
                        source_parent_id = None
 
1964
                    else:
 
1965
                        last_source_parent[0] = old_dirname
 
1966
                        last_source_parent[1] = source_parent_id
 
1967
                new_dirname = entry[0][0]
 
1968
                if new_dirname == last_target_parent[0]:
 
1969
                    target_parent_id = last_target_parent[1]
 
1970
                else:
 
1971
                    try:
 
1972
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1973
                    except KeyError:
 
1974
                        # TODO: We don't always need to do the lookup, because the
 
1975
                        #       parent entry will be the same as the source entry.
 
1976
                        target_parent_entry = state._get_entry(target_index,
 
1977
                                                               path_utf8=new_dirname)
 
1978
                        assert target_parent_entry != (None, None), (
 
1979
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1980
                            % (new_dirname, entry))
 
1981
                        target_parent_id = target_parent_entry[0][2]
 
1982
                    if target_parent_id == entry[0][2]:
 
1983
                        # This is the root, so the parent is None
 
1984
                        target_parent_id = None
 
1985
                    else:
 
1986
                        last_target_parent[0] = new_dirname
 
1987
                        last_target_parent[1] = target_parent_id
 
1988
 
 
1989
                source_exec = source_details[3]
 
1990
                if (include_unchanged
 
1991
                    or content_change
 
1992
                    or source_parent_id != target_parent_id
 
1993
                    or old_basename != entry[0][1]
 
1994
                    or source_exec != target_exec
 
1995
                    ):
 
1996
                    if old_path is None:
 
1997
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1998
                        old_path_u = utf8_decode(old_path)[0]
 
1999
                        path_u = old_path_u
 
2000
                    else:
 
2001
                        old_path_u = utf8_decode(old_path)[0]
 
2002
                        if old_path == path:
 
2003
                            path_u = old_path_u
 
2004
                        else:
 
2005
                            path_u = utf8_decode(path)[0]
 
2006
                    source_kind = _minikind_to_kind[source_minikind]
 
2007
                    return (entry[0][2],
 
2008
                           (old_path_u, path_u),
 
2009
                           content_change,
 
2010
                           (True, True),
 
2011
                           (source_parent_id, target_parent_id),
 
2012
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2013
                           (source_kind, target_kind),
 
2014
                           (source_exec, target_exec))
 
2015
                else:
 
2016
                    return uninteresting
 
2017
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2018
                # looks like a new file
 
2019
                if path_info is not None:
 
2020
                    path = pathjoin(entry[0][0], entry[0][1])
 
2021
                    # parent id is the entry for the path in the target tree
 
2022
                    # TODO: these are the same for an entire directory: cache em.
 
2023
                    parent_id = state._get_entry(target_index,
 
2024
                                                 path_utf8=entry[0][0])[0][2]
 
2025
                    if parent_id == entry[0][2]:
 
2026
                        parent_id = None
 
2027
                    if use_filesystem_for_exec:
 
2028
                        # We need S_ISREG here, because we aren't sure if this
 
2029
                        # is a file or not.
 
2030
                        target_exec = bool(
 
2031
                            stat.S_ISREG(path_info[3].st_mode)
 
2032
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2033
                    else:
 
2034
                        target_exec = target_details[3]
 
2035
                    return (entry[0][2],
 
2036
                           (None, utf8_decode(path)[0]),
 
2037
                           True,
 
2038
                           (False, True),
 
2039
                           (None, parent_id),
 
2040
                           (None, utf8_decode(entry[0][1])[0]),
 
2041
                           (None, path_info[2]),
 
2042
                           (None, target_exec))
 
2043
                else:
 
2044
                    # but its not on disk: we deliberately treat this as just
 
2045
                    # never-present. (Why ?! - RBC 20070224)
 
2046
                    pass
 
2047
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2048
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2049
                # if its still on disk, *and* theres no other entry at this
 
2050
                # path [we dont know this in this routine at the moment -
 
2051
                # perhaps we should change this - then it would be an unknown.
 
2052
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2053
                # parent id is the entry for the path in the target tree
 
2054
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2055
                if parent_id == entry[0][2]:
 
2056
                    parent_id = None
 
2057
                return (entry[0][2],
 
2058
                       (utf8_decode(old_path)[0], None),
 
2059
                       True,
 
2060
                       (True, False),
 
2061
                       (parent_id, None),
 
2062
                       (utf8_decode(entry[0][1])[0], None),
 
2063
                       (_minikind_to_kind[source_minikind], None),
 
2064
                       (source_details[3], None))
 
2065
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2066
                # a rename; could be a true rename, or a rename inherited from
 
2067
                # a renamed parent. TODO: handle this efficiently. Its not
 
2068
                # common case to rename dirs though, so a correct but slow
 
2069
                # implementation will do.
 
2070
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2071
                    search_specific_files.add(target_details[1])
 
2072
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2073
                # neither of the selected trees contain this file,
 
2074
                # so skip over it. This is not currently directly tested, but
 
2075
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2076
                pass
 
2077
            else:
 
2078
                raise AssertionError("don't know how to compare "
 
2079
                    "source_minikind=%r, target_minikind=%r"
 
2080
                    % (source_minikind, target_minikind))
 
2081
                ## import pdb;pdb.set_trace()
 
2082
            return None
 
2083
 
 
2084
        while search_specific_files:
 
2085
            # TODO: the pending list should be lexically sorted?  the
 
2086
            # interface doesn't require it.
 
2087
            current_root = search_specific_files.pop()
 
2088
            current_root_unicode = current_root.decode('utf8')
 
2089
            searched_specific_files.add(current_root)
 
2090
            # process the entries for this containing directory: the rest will be
 
2091
            # found by their parents recursively.
 
2092
            root_entries = _entries_for_path(current_root)
 
2093
            root_abspath = self.target.abspath(current_root_unicode)
 
2094
            try:
 
2095
                root_stat = os.lstat(root_abspath)
 
2096
            except OSError, e:
 
2097
                if e.errno == errno.ENOENT:
 
2098
                    # the path does not exist: let _process_entry know that.
 
2099
                    root_dir_info = None
 
2100
                else:
 
2101
                    # some other random error: hand it up.
 
2102
                    raise
 
2103
            else:
 
2104
                root_dir_info = ('', current_root,
 
2105
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2106
                    root_abspath)
 
2107
                if root_dir_info[2] == 'directory':
 
2108
                    if self.target._directory_is_tree_reference(
 
2109
                        current_root.decode('utf8')):
 
2110
                        root_dir_info = root_dir_info[:2] + \
 
2111
                            ('tree-reference',) + root_dir_info[3:]
 
2112
 
 
2113
            if not root_entries and not root_dir_info:
 
2114
                # this specified path is not present at all, skip it.
 
2115
                continue
 
2116
            path_handled = False
 
2117
            for entry in root_entries:
 
2118
                result = _process_entry(entry, root_dir_info)
 
2119
                if result is not None:
 
2120
                    path_handled = True
 
2121
                    if result is not uninteresting:
 
2122
                        yield result
 
2123
            if want_unversioned and not path_handled and root_dir_info:
 
2124
                new_executable = bool(
 
2125
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2126
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2127
                yield (None,
 
2128
                       (None, current_root_unicode),
 
2129
                       True,
 
2130
                       (False, False),
 
2131
                       (None, None),
 
2132
                       (None, splitpath(current_root_unicode)[-1]),
 
2133
                       (None, root_dir_info[2]),
 
2134
                       (None, new_executable)
 
2135
                      )
 
2136
            initial_key = (current_root, '', '')
 
2137
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2138
            if block_index == 0:
 
2139
                # we have processed the total root already, but because the
 
2140
                # initial key matched it we should skip it here.
 
2141
                block_index +=1
 
2142
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2143
                current_dir_info = None
 
2144
            else:
 
2145
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2146
                try:
 
2147
                    current_dir_info = dir_iterator.next()
 
2148
                except OSError, e:
 
2149
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2150
                    # python 2.5 has e.errno == EINVAL,
 
2151
                    #            and e.winerror == ERROR_DIRECTORY
 
2152
                    e_winerror = getattr(e, 'winerror', None)
 
2153
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2154
                    # there may be directories in the inventory even though
 
2155
                    # this path is not a file on disk: so mark it as end of
 
2156
                    # iterator
 
2157
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2158
                        current_dir_info = None
 
2159
                    elif (sys.platform == 'win32'
 
2160
                          and (e.errno in win_errors
 
2161
                               or e_winerror in win_errors)):
 
2162
                        current_dir_info = None
 
2163
                    else:
 
2164
                        raise
 
2165
                else:
 
2166
                    if current_dir_info[0][0] == '':
 
2167
                        # remove .bzr from iteration
 
2168
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2169
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2170
                        del current_dir_info[1][bzr_index]
 
2171
            # walk until both the directory listing and the versioned metadata
 
2172
            # are exhausted. 
 
2173
            if (block_index < len(state._dirblocks) and
 
2174
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2175
                current_block = state._dirblocks[block_index]
 
2176
            else:
 
2177
                current_block = None
 
2178
            while (current_dir_info is not None or
 
2179
                   current_block is not None):
 
2180
                if (current_dir_info and current_block
 
2181
                    and current_dir_info[0][0] != current_block[0]):
 
2182
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2183
                        # filesystem data refers to paths not covered by the dirblock.
 
2184
                        # this has two possibilities:
 
2185
                        # A) it is versioned but empty, so there is no block for it
 
2186
                        # B) it is not versioned.
 
2187
 
 
2188
                        # if (A) then we need to recurse into it to check for
 
2189
                        # new unknown files or directories.
 
2190
                        # if (B) then we should ignore it, because we don't
 
2191
                        # recurse into unknown directories.
 
2192
                        path_index = 0
 
2193
                        while path_index < len(current_dir_info[1]):
 
2194
                                current_path_info = current_dir_info[1][path_index]
 
2195
                                if want_unversioned:
 
2196
                                    if current_path_info[2] == 'directory':
 
2197
                                        if self.target._directory_is_tree_reference(
 
2198
                                            current_path_info[0].decode('utf8')):
 
2199
                                            current_path_info = current_path_info[:2] + \
 
2200
                                                ('tree-reference',) + current_path_info[3:]
 
2201
                                    new_executable = bool(
 
2202
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2203
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2204
                                    yield (None,
 
2205
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2206
                                        True,
 
2207
                                        (False, False),
 
2208
                                        (None, None),
 
2209
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2210
                                        (None, current_path_info[2]),
 
2211
                                        (None, new_executable))
 
2212
                                # dont descend into this unversioned path if it is
 
2213
                                # a dir
 
2214
                                if current_path_info[2] in ('directory',
 
2215
                                                            'tree-reference'):
 
2216
                                    del current_dir_info[1][path_index]
 
2217
                                    path_index -= 1
 
2218
                                path_index += 1
 
2219
 
 
2220
                        # This dir info has been handled, go to the next
 
2221
                        try:
 
2222
                            current_dir_info = dir_iterator.next()
 
2223
                        except StopIteration:
 
2224
                            current_dir_info = None
 
2225
                    else:
 
2226
                        # We have a dirblock entry for this location, but there
 
2227
                        # is no filesystem path for this. This is most likely
 
2228
                        # because a directory was removed from the disk.
 
2229
                        # We don't have to report the missing directory,
 
2230
                        # because that should have already been handled, but we
 
2231
                        # need to handle all of the files that are contained
 
2232
                        # within.
 
2233
                        for current_entry in current_block[1]:
 
2234
                            # entry referring to file not present on disk.
 
2235
                            # advance the entry only, after processing.
 
2236
                            result = _process_entry(current_entry, None)
 
2237
                            if result is not None:
 
2238
                                if result is not uninteresting:
 
2239
                                    yield result
 
2240
                        block_index +=1
 
2241
                        if (block_index < len(state._dirblocks) and
 
2242
                            osutils.is_inside(current_root,
 
2243
                                              state._dirblocks[block_index][0])):
 
2244
                            current_block = state._dirblocks[block_index]
 
2245
                        else:
 
2246
                            current_block = None
 
2247
                    continue
 
2248
                entry_index = 0
 
2249
                if current_block and entry_index < len(current_block[1]):
 
2250
                    current_entry = current_block[1][entry_index]
 
2251
                else:
 
2252
                    current_entry = None
 
2253
                advance_entry = True
 
2254
                path_index = 0
 
2255
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2256
                    current_path_info = current_dir_info[1][path_index]
 
2257
                    if current_path_info[2] == 'directory':
 
2258
                        if self.target._directory_is_tree_reference(
 
2259
                            current_path_info[0].decode('utf8')):
 
2260
                            current_path_info = current_path_info[:2] + \
 
2261
                                ('tree-reference',) + current_path_info[3:]
 
2262
                else:
 
2263
                    current_path_info = None
 
2264
                advance_path = True
 
2265
                path_handled = False
 
2266
                while (current_entry is not None or
 
2267
                    current_path_info is not None):
 
2268
                    if current_entry is None:
 
2269
                        # the check for path_handled when the path is adnvaced
 
2270
                        # will yield this path if needed.
 
2271
                        pass
 
2272
                    elif current_path_info is None:
 
2273
                        # no path is fine: the per entry code will handle it.
 
2274
                        result = _process_entry(current_entry, current_path_info)
 
2275
                        if result is not None:
 
2276
                            if result is not uninteresting:
 
2277
                                yield result
 
2278
                    elif (current_entry[0][1] != current_path_info[1]
 
2279
                          or current_entry[1][target_index][0] in 'ar'):
 
2280
                        # The current path on disk doesn't match the dirblock
 
2281
                        # record. Either the dirblock is marked as absent, or
 
2282
                        # the file on disk is not present at all in the
 
2283
                        # dirblock. Either way, report about the dirblock
 
2284
                        # entry, and let other code handle the filesystem one.
 
2285
 
 
2286
                        # Compare the basename for these files to determine
 
2287
                        # which comes first
 
2288
                        if current_path_info[1] < current_entry[0][1]:
 
2289
                            # extra file on disk: pass for now, but only
 
2290
                            # increment the path, not the entry
 
2291
                            advance_entry = False
 
2292
                        else:
 
2293
                            # entry referring to file not present on disk.
 
2294
                            # advance the entry only, after processing.
 
2295
                            result = _process_entry(current_entry, None)
 
2296
                            if result is not None:
 
2297
                                if result is not uninteresting:
 
2298
                                    yield result
 
2299
                            advance_path = False
 
2300
                    else:
 
2301
                        result = _process_entry(current_entry, current_path_info)
 
2302
                        if result is not None:
 
2303
                            path_handled = True
 
2304
                            if result is not uninteresting:
 
2305
                                yield result
 
2306
                    if advance_entry and current_entry is not None:
 
2307
                        entry_index += 1
 
2308
                        if entry_index < len(current_block[1]):
 
2309
                            current_entry = current_block[1][entry_index]
 
2310
                        else:
 
2311
                            current_entry = None
 
2312
                    else:
 
2313
                        advance_entry = True # reset the advance flaga
 
2314
                    if advance_path and current_path_info is not None:
 
2315
                        if not path_handled:
 
2316
                            # unversioned in all regards
 
2317
                            if want_unversioned:
 
2318
                                new_executable = bool(
 
2319
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2320
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2321
                                yield (None,
 
2322
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2323
                                    True,
 
2324
                                    (False, False),
 
2325
                                    (None, None),
 
2326
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2327
                                    (None, current_path_info[2]),
 
2328
                                    (None, new_executable))
 
2329
                            # dont descend into this unversioned path if it is
 
2330
                            # a dir
 
2331
                            if current_path_info[2] in ('directory'):
 
2332
                                del current_dir_info[1][path_index]
 
2333
                                path_index -= 1
 
2334
                        # dont descend the disk iterator into any tree 
 
2335
                        # paths.
 
2336
                        if current_path_info[2] == 'tree-reference':
 
2337
                            del current_dir_info[1][path_index]
 
2338
                            path_index -= 1
 
2339
                        path_index += 1
 
2340
                        if path_index < len(current_dir_info[1]):
 
2341
                            current_path_info = current_dir_info[1][path_index]
 
2342
                            if current_path_info[2] == 'directory':
 
2343
                                if self.target._directory_is_tree_reference(
 
2344
                                    current_path_info[0].decode('utf8')):
 
2345
                                    current_path_info = current_path_info[:2] + \
 
2346
                                        ('tree-reference',) + current_path_info[3:]
 
2347
                        else:
 
2348
                            current_path_info = None
 
2349
                        path_handled = False
 
2350
                    else:
 
2351
                        advance_path = True # reset the advance flagg.
 
2352
                if current_block is not None:
 
2353
                    block_index += 1
 
2354
                    if (block_index < len(state._dirblocks) and
 
2355
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2356
                        current_block = state._dirblocks[block_index]
 
2357
                    else:
 
2358
                        current_block = None
 
2359
                if current_dir_info is not None:
 
2360
                    try:
 
2361
                        current_dir_info = dir_iterator.next()
 
2362
                    except StopIteration:
 
2363
                        current_dir_info = None
 
2364
 
 
2365
 
 
2366
    @staticmethod
 
2367
    def is_compatible(source, target):
 
2368
        # the target must be a dirstate working tree
 
2369
        if not isinstance(target, WorkingTree4):
 
2370
            return False
 
2371
        # the source must be a revtreee or dirstate rev tree.
 
2372
        if not isinstance(source,
 
2373
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2374
            return False
 
2375
        # the source revid must be in the target dirstate
 
2376
        if not (source._revision_id == NULL_REVISION or
 
2377
            source._revision_id in target.get_parent_ids()):
 
2378
            # TODO: what about ghosts? it may well need to 
 
2379
            # check for them explicitly.
 
2380
            return False
 
2381
        return True
 
2382
 
 
2383
InterTree.register_optimiser(InterDirStateTree)
 
2384
 
 
2385
 
 
2386
class Converter3to4(object):
 
2387
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2388
 
 
2389
    def __init__(self):
 
2390
        self.target_format = WorkingTreeFormat4()
 
2391
 
 
2392
    def convert(self, tree):
 
2393
        # lock the control files not the tree, so that we dont get tree
 
2394
        # on-unlock behaviours, and so that noone else diddles with the 
 
2395
        # tree during upgrade.
 
2396
        tree._control_files.lock_write()
 
2397
        try:
 
2398
            tree.read_working_inventory()
 
2399
            self.create_dirstate_data(tree)
 
2400
            self.update_format(tree)
 
2401
            self.remove_xml_files(tree)
 
2402
        finally:
 
2403
            tree._control_files.unlock()
 
2404
 
 
2405
    def create_dirstate_data(self, tree):
 
2406
        """Create the dirstate based data for tree."""
 
2407
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2408
            ).local_abspath('dirstate')
 
2409
        state = dirstate.DirState.from_tree(tree, local_path)
 
2410
        state.save()
 
2411
        state.unlock()
 
2412
 
 
2413
    def remove_xml_files(self, tree):
 
2414
        """Remove the oldformat 3 data."""
 
2415
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2416
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2417
            'pending-merges', 'stat-cache']:
 
2418
            try:
 
2419
                transport.delete(path)
 
2420
            except errors.NoSuchFile:
 
2421
                # some files are optional - just deal.
 
2422
                pass
 
2423
 
 
2424
    def update_format(self, tree):
 
2425
        """Change the format marker."""
 
2426
        tree._control_files.put_utf8('format',
 
2427
            self.target_format.get_format_string())