~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/repofmt/groupcompress_repo.py

  • Committer: Martin Pool
  • Date: 2009-06-10 03:31:01 UTC
  • mfrom: (4070.12.3 249908-doc-generate)
  • mto: This revision was merged to the branch mainline in revision 4464.
  • Revision ID: mbp@sourcefrog.net-20090610033101-ce26xd79i9cbmpsd
Merge other outstanding work on 249908

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2008-2011 Canonical Ltd
 
1
# Copyright (C) 2008, 2009 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
26
26
    errors,
27
27
    index as _mod_index,
28
28
    inventory,
 
29
    knit,
29
30
    osutils,
30
31
    pack,
 
32
    remote,
 
33
    repository,
31
34
    revision as _mod_revision,
32
35
    trace,
33
36
    ui,
34
 
    versionedfile,
35
37
    )
36
38
from bzrlib.btree_index import (
37
39
    BTreeGraphIndex,
38
40
    BTreeBuilder,
39
41
    )
40
 
from bzrlib.decorators import needs_write_lock
 
42
from bzrlib.index import GraphIndex, GraphIndexBuilder
41
43
from bzrlib.groupcompress import (
42
44
    _GCGraphIndex,
43
45
    GroupCompressVersionedFiles,
44
46
    )
45
47
from bzrlib.repofmt.pack_repo import (
46
 
    _DirectPackAccess,
47
48
    Pack,
48
49
    NewPack,
49
 
    PackRepository,
 
50
    KnitPackRepository,
50
51
    PackRootCommitBuilder,
51
52
    RepositoryPackCollection,
52
53
    RepositoryFormatPack,
53
54
    ResumedPack,
54
55
    Packer,
55
56
    )
56
 
from bzrlib.vf_repository import (
57
 
    StreamSource,
58
 
    )
59
 
from bzrlib.static_tuple import StaticTuple
60
57
 
61
58
 
62
59
class GCPack(NewPack):
158
155
        self._writer.begin()
159
156
        # what state is the pack in? (open, finished, aborted)
160
157
        self._state = 'open'
161
 
        # no name until we finish writing the content
162
 
        self.name = None
163
158
 
164
159
    def _check_references(self):
165
160
        """Make sure our external references are present.
222
217
            p_id_roots_set = set()
223
218
            stream = source_vf.get_record_stream(keys, 'groupcompress', True)
224
219
            for idx, record in enumerate(stream):
225
 
                # Inventories should always be with revisions; assume success.
226
220
                bytes = record.get_bytes_as('fulltext')
227
221
                chk_inv = inventory.CHKInventory.deserialise(None, bytes,
228
222
                                                             record.key)
266
260
        remaining_keys = set(keys)
267
261
        counter = [0]
268
262
        if self._gather_text_refs:
 
263
            bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
269
264
            self._text_refs = set()
270
265
        def _get_referenced_stream(root_keys, parse_leaf_nodes=False):
271
266
            cur_keys = root_keys
292
287
                    # Store is None, because we know we have a LeafNode, and we
293
288
                    # just want its entries
294
289
                    for file_id, bytes in node.iteritems(None):
295
 
                        self._text_refs.add(chk_map._bytes_to_text_key(bytes))
 
290
                        name_utf8, file_id, revision_id = bytes_to_info(bytes)
 
291
                        self._text_refs.add((file_id, revision_id))
296
292
                def next_stream():
297
293
                    stream = source_vf.get_record_stream(cur_keys,
298
294
                                                         'as-requested', True)
299
295
                    for record in stream:
300
 
                        if record.storage_kind == 'absent':
301
 
                            # An absent CHK record: we assume that the missing
302
 
                            # record is in a different pack - e.g. a page not
303
 
                            # altered by the commit we're packing.
304
 
                            continue
305
296
                        bytes = record.get_bytes_as('fulltext')
306
297
                        # We don't care about search_key_func for this code,
307
298
                        # because we only care about external references.
354
345
        """Build a VersionedFiles instance on top of this group of packs."""
355
346
        index_name = index_name + '_index'
356
347
        index_to_pack = {}
357
 
        access = _DirectPackAccess(index_to_pack,
358
 
                                   reload_func=self._reload_func)
 
348
        access = knit._DirectPackAccess(index_to_pack)
359
349
        if for_write:
360
350
            # Use new_pack
361
351
            if self.new_pack is None:
415
405
 
416
406
    def _copy_inventory_texts(self):
417
407
        source_vf, target_vf = self._build_vfs('inventory', True, True)
418
 
        # It is not sufficient to just use self.revision_keys, as stacked
419
 
        # repositories can have more inventories than they have revisions.
420
 
        # One alternative would be to do something with
421
 
        # get_parent_map(self.revision_keys), but that shouldn't be any faster
422
 
        # than this.
423
 
        inventory_keys = source_vf.keys()
424
 
        missing_inventories = set(self.revision_keys).difference(inventory_keys)
425
 
        if missing_inventories:
426
 
            # Go back to the original repo, to see if these are really missing
427
 
            # https://bugs.launchpad.net/bzr/+bug/437003
428
 
            # If we are packing a subset of the repo, it is fine to just have
429
 
            # the data in another Pack file, which is not included in this pack
430
 
            # operation.
431
 
            inv_index = self._pack_collection.repo.inventories._index
432
 
            pmap = inv_index.get_parent_map(missing_inventories)
433
 
            really_missing = missing_inventories.difference(pmap)
434
 
            if really_missing:
435
 
                missing_inventories = sorted(really_missing)
436
 
                raise ValueError('We are missing inventories for revisions: %s'
437
 
                    % (missing_inventories,))
438
 
        self._copy_stream(source_vf, target_vf, inventory_keys,
 
408
        self._copy_stream(source_vf, target_vf, self.revision_keys,
439
409
                          'inventories', self._get_filtered_inv_stream, 2)
440
410
 
441
 
    def _get_chk_vfs_for_copy(self):
442
 
        return self._build_vfs('chk', False, False)
443
 
 
444
411
    def _copy_chk_texts(self):
445
 
        source_vf, target_vf = self._get_chk_vfs_for_copy()
 
412
        source_vf, target_vf = self._build_vfs('chk', False, False)
446
413
        # TODO: This is technically spurious... if it is a performance issue,
447
414
        #       remove it
448
415
        total_keys = source_vf.keys()
471
438
        #      is grabbing too many keys...
472
439
        text_keys = source_vf.keys()
473
440
        self._copy_stream(source_vf, target_vf, text_keys,
474
 
                          'texts', self._get_progress_stream, 4)
 
441
                          'text', self._get_progress_stream, 4)
475
442
 
476
443
    def _copy_signature_texts(self):
477
444
        source_vf, target_vf = self._build_vfs('signature', False, False)
494
461
        if not self._use_pack(self.new_pack):
495
462
            self.new_pack.abort()
496
463
            return None
497
 
        self.new_pack.finish_content()
498
 
        if len(self.packs) == 1:
499
 
            old_pack = self.packs[0]
500
 
            if old_pack.name == self.new_pack._hash.hexdigest():
501
 
                # The single old pack was already optimally packed.
502
 
                trace.mutter('single pack %s was already optimally packed',
503
 
                    old_pack.name)
504
 
                self.new_pack.abort()
505
 
                return None
506
464
        self.pb.update('finishing repack', 6, 7)
507
465
        self.new_pack.finish()
508
466
        self._pack_collection.allocate(self.new_pack)
594
552
        return new_pack.data_inserted() and self._data_changed
595
553
 
596
554
 
597
 
class GCCHKCanonicalizingPacker(GCCHKPacker):
598
 
    """A packer that ensures inventories have canonical-form CHK maps.
599
 
    
600
 
    Ideally this would be part of reconcile, but it's very slow and rarely
601
 
    needed.  (It repairs repositories affected by
602
 
    https://bugs.launchpad.net/bzr/+bug/522637).
603
 
    """
604
 
 
605
 
    def __init__(self, *args, **kwargs):
606
 
        super(GCCHKCanonicalizingPacker, self).__init__(*args, **kwargs)
607
 
        self._data_changed = False
608
 
 
609
 
    def _exhaust_stream(self, source_vf, keys, message, vf_to_stream, pb_offset):
610
 
        """Create and exhaust a stream, but don't insert it.
611
 
 
612
 
        This is useful to get the side-effects of generating a stream.
613
 
        """
614
 
        self.pb.update('scanning %s' % (message,), pb_offset)
615
 
        child_pb = ui.ui_factory.nested_progress_bar()
616
 
        try:
617
 
            list(vf_to_stream(source_vf, keys, message, child_pb))
618
 
        finally:
619
 
            child_pb.finished()
620
 
 
621
 
    def _copy_inventory_texts(self):
622
 
        source_vf, target_vf = self._build_vfs('inventory', True, True)
623
 
        source_chk_vf, target_chk_vf = self._get_chk_vfs_for_copy()
624
 
        inventory_keys = source_vf.keys()
625
 
        # First, copy the existing CHKs on the assumption that most of them
626
 
        # will be correct.  This will save us from having to reinsert (and
627
 
        # recompress) these records later at the cost of perhaps preserving a
628
 
        # few unused CHKs. 
629
 
        # (Iterate but don't insert _get_filtered_inv_stream to populate the
630
 
        # variables needed by GCCHKPacker._copy_chk_texts.)
631
 
        self._exhaust_stream(source_vf, inventory_keys, 'inventories',
632
 
                self._get_filtered_inv_stream, 2)
633
 
        GCCHKPacker._copy_chk_texts(self)
634
 
        # Now copy and fix the inventories, and any regenerated CHKs.
635
 
        def chk_canonicalizing_inv_stream(source_vf, keys, message, pb=None):
636
 
            return self._get_filtered_canonicalizing_inv_stream(
637
 
                source_vf, keys, message, pb, source_chk_vf, target_chk_vf)
638
 
        self._copy_stream(source_vf, target_vf, inventory_keys,
639
 
                          'inventories', chk_canonicalizing_inv_stream, 4)
640
 
 
641
 
    def _copy_chk_texts(self):
642
 
        # No-op; in this class this happens during _copy_inventory_texts.
643
 
        pass
644
 
 
645
 
    def _get_filtered_canonicalizing_inv_stream(self, source_vf, keys, message,
646
 
            pb=None, source_chk_vf=None, target_chk_vf=None):
647
 
        """Filter the texts of inventories, regenerating CHKs to make sure they
648
 
        are canonical.
649
 
        """
650
 
        total_keys = len(keys)
651
 
        target_chk_vf = versionedfile.NoDupeAddLinesDecorator(target_chk_vf)
652
 
        def _filtered_inv_stream():
653
 
            stream = source_vf.get_record_stream(keys, 'groupcompress', True)
654
 
            search_key_name = None
655
 
            for idx, record in enumerate(stream):
656
 
                # Inventories should always be with revisions; assume success.
657
 
                bytes = record.get_bytes_as('fulltext')
658
 
                chk_inv = inventory.CHKInventory.deserialise(
659
 
                    source_chk_vf, bytes, record.key)
660
 
                if pb is not None:
661
 
                    pb.update('inv', idx, total_keys)
662
 
                chk_inv.id_to_entry._ensure_root()
663
 
                if search_key_name is None:
664
 
                    # Find the name corresponding to the search_key_func
665
 
                    search_key_reg = chk_map.search_key_registry
666
 
                    for search_key_name, func in search_key_reg.iteritems():
667
 
                        if func == chk_inv.id_to_entry._search_key_func:
668
 
                            break
669
 
                canonical_inv = inventory.CHKInventory.from_inventory(
670
 
                    target_chk_vf, chk_inv,
671
 
                    maximum_size=chk_inv.id_to_entry._root_node._maximum_size,
672
 
                    search_key_name=search_key_name)
673
 
                if chk_inv.id_to_entry.key() != canonical_inv.id_to_entry.key():
674
 
                    trace.mutter(
675
 
                        'Non-canonical CHK map for id_to_entry of inv: %s '
676
 
                        '(root is %s, should be %s)' % (chk_inv.revision_id,
677
 
                        chk_inv.id_to_entry.key()[0],
678
 
                        canonical_inv.id_to_entry.key()[0]))
679
 
                    self._data_changed = True
680
 
                p_id_map = chk_inv.parent_id_basename_to_file_id
681
 
                p_id_map._ensure_root()
682
 
                canon_p_id_map = canonical_inv.parent_id_basename_to_file_id
683
 
                if p_id_map.key() != canon_p_id_map.key():
684
 
                    trace.mutter(
685
 
                        'Non-canonical CHK map for parent_id_to_basename of '
686
 
                        'inv: %s (root is %s, should be %s)'
687
 
                        % (chk_inv.revision_id, p_id_map.key()[0],
688
 
                           canon_p_id_map.key()[0]))
689
 
                    self._data_changed = True
690
 
                yield versionedfile.ChunkedContentFactory(record.key,
691
 
                        record.parents, record.sha1,
692
 
                        canonical_inv.to_lines())
693
 
            # We have finished processing all of the inventory records, we
694
 
            # don't need these sets anymore
695
 
        return _filtered_inv_stream()
696
 
 
697
 
    def _use_pack(self, new_pack):
698
 
        """Override _use_pack to check for reconcile having changed content."""
699
 
        return new_pack.data_inserted() and self._data_changed
700
 
 
701
 
 
702
555
class GCRepositoryPackCollection(RepositoryPackCollection):
703
556
 
704
557
    pack_factory = GCPack
705
558
    resumed_pack_factory = ResumedGCPack
706
 
    normal_packer_class = GCCHKPacker
707
 
    optimising_packer_class = GCCHKPacker
708
 
 
709
 
    def _check_new_inventories(self):
710
 
        """Detect missing inventories or chk root entries for the new revisions
711
 
        in this write group.
712
 
 
713
 
        :returns: list of strs, summarising any problems found.  If the list is
714
 
            empty no problems were found.
 
559
 
 
560
    def _already_packed(self):
 
561
        """Is the collection already packed?"""
 
562
        # Always repack GC repositories for now
 
563
        return False
 
564
 
 
565
    def _execute_pack_operations(self, pack_operations,
 
566
                                 _packer_class=GCCHKPacker,
 
567
                                 reload_func=None):
 
568
        """Execute a series of pack operations.
 
569
 
 
570
        :param pack_operations: A list of [revision_count, packs_to_combine].
 
571
        :param _packer_class: The class of packer to use (default: Packer).
 
572
        :return: None.
715
573
        """
716
 
        # Ensure that all revisions added in this write group have:
717
 
        #   - corresponding inventories,
718
 
        #   - chk root entries for those inventories,
719
 
        #   - and any present parent inventories have their chk root
720
 
        #     entries too.
721
 
        # And all this should be independent of any fallback repository.
722
 
        problems = []
723
 
        key_deps = self.repo.revisions._index._key_dependencies
724
 
        new_revisions_keys = key_deps.get_new_keys()
725
 
        no_fallback_inv_index = self.repo.inventories._index
726
 
        no_fallback_chk_bytes_index = self.repo.chk_bytes._index
727
 
        no_fallback_texts_index = self.repo.texts._index
728
 
        inv_parent_map = no_fallback_inv_index.get_parent_map(
729
 
            new_revisions_keys)
730
 
        # Are any inventories for corresponding to the new revisions missing?
731
 
        corresponding_invs = set(inv_parent_map)
732
 
        missing_corresponding = set(new_revisions_keys)
733
 
        missing_corresponding.difference_update(corresponding_invs)
734
 
        if missing_corresponding:
735
 
            problems.append("inventories missing for revisions %s" %
736
 
                (sorted(missing_corresponding),))
737
 
            return problems
738
 
        # Are any chk root entries missing for any inventories?  This includes
739
 
        # any present parent inventories, which may be used when calculating
740
 
        # deltas for streaming.
741
 
        all_inv_keys = set(corresponding_invs)
742
 
        for parent_inv_keys in inv_parent_map.itervalues():
743
 
            all_inv_keys.update(parent_inv_keys)
744
 
        # Filter out ghost parents.
745
 
        all_inv_keys.intersection_update(
746
 
            no_fallback_inv_index.get_parent_map(all_inv_keys))
747
 
        parent_invs_only_keys = all_inv_keys.symmetric_difference(
748
 
            corresponding_invs)
749
 
        all_missing = set()
750
 
        inv_ids = [key[-1] for key in all_inv_keys]
751
 
        parent_invs_only_ids = [key[-1] for key in parent_invs_only_keys]
752
 
        root_key_info = _build_interesting_key_sets(
753
 
            self.repo, inv_ids, parent_invs_only_ids)
754
 
        expected_chk_roots = root_key_info.all_keys()
755
 
        present_chk_roots = no_fallback_chk_bytes_index.get_parent_map(
756
 
            expected_chk_roots)
757
 
        missing_chk_roots = expected_chk_roots.difference(present_chk_roots)
758
 
        if missing_chk_roots:
759
 
            problems.append("missing referenced chk root keys: %s"
760
 
                % (sorted(missing_chk_roots),))
761
 
            # Don't bother checking any further.
762
 
            return problems
763
 
        # Find all interesting chk_bytes records, and make sure they are
764
 
        # present, as well as the text keys they reference.
765
 
        chk_bytes_no_fallbacks = self.repo.chk_bytes.without_fallbacks()
766
 
        chk_bytes_no_fallbacks._search_key_func = \
767
 
            self.repo.chk_bytes._search_key_func
768
 
        chk_diff = chk_map.iter_interesting_nodes(
769
 
            chk_bytes_no_fallbacks, root_key_info.interesting_root_keys,
770
 
            root_key_info.uninteresting_root_keys)
771
 
        text_keys = set()
772
 
        try:
773
 
            for record in _filter_text_keys(chk_diff, text_keys,
774
 
                                            chk_map._bytes_to_text_key):
775
 
                pass
776
 
        except errors.NoSuchRevision, e:
777
 
            # XXX: It would be nice if we could give a more precise error here.
778
 
            problems.append("missing chk node(s) for id_to_entry maps")
779
 
        chk_diff = chk_map.iter_interesting_nodes(
780
 
            chk_bytes_no_fallbacks, root_key_info.interesting_pid_root_keys,
781
 
            root_key_info.uninteresting_pid_root_keys)
782
 
        try:
783
 
            for interesting_rec, interesting_map in chk_diff:
784
 
                pass
785
 
        except errors.NoSuchRevision, e:
786
 
            problems.append(
787
 
                "missing chk node(s) for parent_id_basename_to_file_id maps")
788
 
        present_text_keys = no_fallback_texts_index.get_parent_map(text_keys)
789
 
        missing_text_keys = text_keys.difference(present_text_keys)
790
 
        if missing_text_keys:
791
 
            problems.append("missing text keys: %r"
792
 
                % (sorted(missing_text_keys),))
793
 
        return problems
794
 
 
795
 
 
796
 
class CHKInventoryRepository(PackRepository):
797
 
    """subclass of PackRepository that uses CHK based inventories."""
 
574
        # XXX: Copied across from RepositoryPackCollection simply because we
 
575
        #      want to override the _packer_class ... :(
 
576
        for revision_count, packs in pack_operations:
 
577
            # we may have no-ops from the setup logic
 
578
            if len(packs) == 0:
 
579
                continue
 
580
            packer = GCCHKPacker(self, packs, '.autopack',
 
581
                                 reload_func=reload_func)
 
582
            try:
 
583
                packer.pack()
 
584
            except errors.RetryWithNewPacks:
 
585
                # An exception is propagating out of this context, make sure
 
586
                # this packer has cleaned up. Packer() doesn't set its new_pack
 
587
                # state into the RepositoryPackCollection object, so we only
 
588
                # have access to it directly here.
 
589
                if packer.new_pack is not None:
 
590
                    packer.new_pack.abort()
 
591
                raise
 
592
            for pack in packs:
 
593
                self._remove_pack_from_memory(pack)
 
594
        # record the newly available packs and stop advertising the old
 
595
        # packs
 
596
        self._save_pack_names(clear_obsolete_packs=True)
 
597
        # Move the old packs out of the way now they are no longer referenced.
 
598
        for revision_count, packs in pack_operations:
 
599
            self._obsolete_packs(packs)
 
600
 
 
601
 
 
602
class CHKInventoryRepository(KnitPackRepository):
 
603
    """subclass of KnitPackRepository that uses CHK based inventories."""
798
604
 
799
605
    def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
800
606
        _serializer):
801
607
        """Overridden to change pack collection class."""
802
 
        super(CHKInventoryRepository, self).__init__(_format, a_bzrdir,
803
 
            control_files, _commit_builder_class, _serializer)
 
608
        KnitPackRepository.__init__(self, _format, a_bzrdir, control_files,
 
609
            _commit_builder_class, _serializer)
 
610
        # and now replace everything it did :)
804
611
        index_transport = self._transport.clone('indices')
805
612
        self._pack_collection = GCRepositoryPackCollection(self,
806
613
            self._transport, index_transport,
813
620
        self.inventories = GroupCompressVersionedFiles(
814
621
            _GCGraphIndex(self._pack_collection.inventory_index.combined_index,
815
622
                add_callback=self._pack_collection.inventory_index.add_callback,
816
 
                parents=True, is_locked=self.is_locked,
817
 
                inconsistency_fatal=False),
 
623
                parents=True, is_locked=self.is_locked),
818
624
            access=self._pack_collection.inventory_index.data_access)
819
625
        self.revisions = GroupCompressVersionedFiles(
820
626
            _GCGraphIndex(self._pack_collection.revision_index.combined_index,
821
627
                add_callback=self._pack_collection.revision_index.add_callback,
822
628
                parents=True, is_locked=self.is_locked,
823
 
                track_external_parent_refs=True, track_new_keys=True),
 
629
                track_external_parent_refs=True),
824
630
            access=self._pack_collection.revision_index.data_access,
825
631
            delta=False)
826
632
        self.signatures = GroupCompressVersionedFiles(
827
633
            _GCGraphIndex(self._pack_collection.signature_index.combined_index,
828
634
                add_callback=self._pack_collection.signature_index.add_callback,
829
 
                parents=False, is_locked=self.is_locked,
830
 
                inconsistency_fatal=False),
 
635
                parents=False, is_locked=self.is_locked),
831
636
            access=self._pack_collection.signature_index.data_access,
832
637
            delta=False)
833
638
        self.texts = GroupCompressVersionedFiles(
834
639
            _GCGraphIndex(self._pack_collection.text_index.combined_index,
835
640
                add_callback=self._pack_collection.text_index.add_callback,
836
 
                parents=True, is_locked=self.is_locked,
837
 
                inconsistency_fatal=False),
 
641
                parents=True, is_locked=self.is_locked),
838
642
            access=self._pack_collection.text_index.data_access)
839
643
        # No parents, individual CHK pages don't have specific ancestry
840
644
        self.chk_bytes = GroupCompressVersionedFiles(
841
645
            _GCGraphIndex(self._pack_collection.chk_index.combined_index,
842
646
                add_callback=self._pack_collection.chk_index.add_callback,
843
 
                parents=False, is_locked=self.is_locked,
844
 
                inconsistency_fatal=False),
 
647
                parents=False, is_locked=self.is_locked),
845
648
            access=self._pack_collection.chk_index.data_access)
846
 
        search_key_name = self._format._serializer.search_key_name
847
 
        search_key_func = chk_map.search_key_registry.get(search_key_name)
848
 
        self.chk_bytes._search_key_func = search_key_func
849
649
        # True when the repository object is 'write locked' (as opposed to the
850
650
        # physical lock only taken out around changes to the pack-names list.)
851
651
        # Another way to represent this would be a decorator around the control
874
674
        return self._inventory_add_lines(revision_id, parents,
875
675
            inv_lines, check_content=False)
876
676
 
877
 
    def _create_inv_from_null(self, delta, revision_id):
878
 
        """This will mutate new_inv directly.
879
 
 
880
 
        This is a simplified form of create_by_apply_delta which knows that all
881
 
        the old values must be None, so everything is a create.
882
 
        """
883
 
        serializer = self._format._serializer
884
 
        new_inv = inventory.CHKInventory(serializer.search_key_name)
885
 
        new_inv.revision_id = revision_id
886
 
        entry_to_bytes = new_inv._entry_to_bytes
887
 
        id_to_entry_dict = {}
888
 
        parent_id_basename_dict = {}
889
 
        for old_path, new_path, file_id, entry in delta:
890
 
            if old_path is not None:
891
 
                raise ValueError('Invalid delta, somebody tried to delete %r'
892
 
                                 ' from the NULL_REVISION'
893
 
                                 % ((old_path, file_id),))
894
 
            if new_path is None:
895
 
                raise ValueError('Invalid delta, delta from NULL_REVISION has'
896
 
                                 ' no new_path %r' % (file_id,))
897
 
            if new_path == '':
898
 
                new_inv.root_id = file_id
899
 
                parent_id_basename_key = StaticTuple('', '').intern()
900
 
            else:
901
 
                utf8_entry_name = entry.name.encode('utf-8')
902
 
                parent_id_basename_key = StaticTuple(entry.parent_id,
903
 
                                                     utf8_entry_name).intern()
904
 
            new_value = entry_to_bytes(entry)
905
 
            # Populate Caches?
906
 
            # new_inv._path_to_fileid_cache[new_path] = file_id
907
 
            key = StaticTuple(file_id).intern()
908
 
            id_to_entry_dict[key] = new_value
909
 
            parent_id_basename_dict[parent_id_basename_key] = file_id
910
 
 
911
 
        new_inv._populate_from_dicts(self.chk_bytes, id_to_entry_dict,
912
 
            parent_id_basename_dict, maximum_size=serializer.maximum_size)
913
 
        return new_inv
914
 
 
915
677
    def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
916
678
                               parents, basis_inv=None, propagate_caches=False):
917
679
        """Add a new inventory expressed as a delta against another revision.
937
699
            repository format specific) of the serialized inventory, and the
938
700
            resulting inventory.
939
701
        """
 
702
        if basis_revision_id == _mod_revision.NULL_REVISION:
 
703
            return KnitPackRepository.add_inventory_by_delta(self,
 
704
                basis_revision_id, delta, new_revision_id, parents)
940
705
        if not self.is_in_write_group():
941
706
            raise AssertionError("%r not in write group" % (self,))
942
707
        _mod_revision.check_not_reserved_id(new_revision_id)
943
 
        basis_tree = None
944
 
        if basis_inv is None:
945
 
            if basis_revision_id == _mod_revision.NULL_REVISION:
946
 
                new_inv = self._create_inv_from_null(delta, new_revision_id)
947
 
                if new_inv.root_id is None:
948
 
                    raise errors.RootMissing()
949
 
                inv_lines = new_inv.to_lines()
950
 
                return self._inventory_add_lines(new_revision_id, parents,
951
 
                    inv_lines, check_content=False), new_inv
952
 
            else:
953
 
                basis_tree = self.revision_tree(basis_revision_id)
954
 
                basis_tree.lock_read()
 
708
        basis_tree = self.revision_tree(basis_revision_id)
 
709
        basis_tree.lock_read()
 
710
        try:
 
711
            if basis_inv is None:
955
712
                basis_inv = basis_tree.inventory
956
 
        try:
957
713
            result = basis_inv.create_by_apply_delta(delta, new_revision_id,
958
714
                propagate_caches=propagate_caches)
959
715
            inv_lines = result.to_lines()
960
716
            return self._inventory_add_lines(new_revision_id, parents,
961
717
                inv_lines, check_content=False), result
962
718
        finally:
963
 
            if basis_tree is not None:
964
 
                basis_tree.unlock()
965
 
 
966
 
    def _deserialise_inventory(self, revision_id, bytes):
967
 
        return inventory.CHKInventory.deserialise(self.chk_bytes, bytes,
968
 
            (revision_id,))
969
 
 
970
 
    def _iter_inventories(self, revision_ids, ordering):
 
719
            basis_tree.unlock()
 
720
 
 
721
    def _iter_inventories(self, revision_ids):
971
722
        """Iterate over many inventory objects."""
972
 
        if ordering is None:
973
 
            ordering = 'unordered'
974
723
        keys = [(revision_id,) for revision_id in revision_ids]
975
 
        stream = self.inventories.get_record_stream(keys, ordering, True)
 
724
        stream = self.inventories.get_record_stream(keys, 'unordered', True)
976
725
        texts = {}
977
726
        for record in stream:
978
727
            if record.storage_kind != 'absent':
982
731
        for key in keys:
983
732
            yield inventory.CHKInventory.deserialise(self.chk_bytes, texts[key], key)
984
733
 
985
 
    def _iter_inventory_xmls(self, revision_ids, ordering):
986
 
        # Without a native 'xml' inventory, this method doesn't make sense.
987
 
        # However older working trees, and older bundles want it - so we supply
988
 
        # it allowing _get_inventory_xml to work. Bundles currently use the
989
 
        # serializer directly; this also isn't ideal, but there isn't an xml
990
 
        # iteration interface offered at all for repositories. We could make
991
 
        # _iter_inventory_xmls be part of the contract, even if kept private.
992
 
        inv_to_str = self._serializer.write_inventory_to_string
993
 
        for inv in self.iter_inventories(revision_ids, ordering=ordering):
994
 
            yield inv_to_str(inv), inv.revision_id
995
 
 
996
 
    def _find_present_inventory_keys(self, revision_keys):
997
 
        parent_map = self.inventories.get_parent_map(revision_keys)
998
 
        present_inventory_keys = set(k for k in parent_map)
999
 
        return present_inventory_keys
 
734
    def _iter_inventory_xmls(self, revision_ids):
 
735
        # Without a native 'xml' inventory, this method doesn't make sense, so
 
736
        # make it raise to trap naughty direct users.
 
737
        raise NotImplementedError(self._iter_inventory_xmls)
 
738
 
 
739
    def _find_parent_ids_of_revisions(self, revision_ids):
 
740
        # TODO: we probably want to make this a helper that other code can get
 
741
        #       at
 
742
        parent_map = self.get_parent_map(revision_ids)
 
743
        parents = set()
 
744
        map(parents.update, parent_map.itervalues())
 
745
        parents.difference_update(revision_ids)
 
746
        parents.discard(_mod_revision.NULL_REVISION)
 
747
        return parents
 
748
 
 
749
    def _find_present_inventory_ids(self, revision_ids):
 
750
        keys = [(r,) for r in revision_ids]
 
751
        parent_map = self.inventories.get_parent_map(keys)
 
752
        present_inventory_ids = set(k[-1] for k in parent_map)
 
753
        return present_inventory_ids
1000
754
 
1001
755
    def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
1002
756
        """Find the file ids and versions affected by revisions.
1013
767
        file_id_revisions = {}
1014
768
        pb = ui.ui_factory.nested_progress_bar()
1015
769
        try:
1016
 
            revision_keys = [(r,) for r in revision_ids]
1017
 
            parent_keys = self._find_parent_keys_of_revisions(revision_keys)
1018
 
            # TODO: instead of using _find_present_inventory_keys, change the
1019
 
            #       code paths to allow missing inventories to be tolerated.
1020
 
            #       However, we only want to tolerate missing parent
1021
 
            #       inventories, not missing inventories for revision_ids
1022
 
            present_parent_inv_keys = self._find_present_inventory_keys(
1023
 
                                        parent_keys)
1024
 
            present_parent_inv_ids = set(
1025
 
                [k[-1] for k in present_parent_inv_keys])
1026
 
            inventories_to_read = set(revision_ids)
1027
 
            inventories_to_read.update(present_parent_inv_ids)
1028
 
            root_key_info = _build_interesting_key_sets(
1029
 
                self, inventories_to_read, present_parent_inv_ids)
1030
 
            interesting_root_keys = root_key_info.interesting_root_keys
1031
 
            uninteresting_root_keys = root_key_info.uninteresting_root_keys
 
770
            parent_ids = self._find_parent_ids_of_revisions(revision_ids)
 
771
            present_parent_inv_ids = self._find_present_inventory_ids(parent_ids)
 
772
            uninteresting_root_keys = set()
 
773
            interesting_root_keys = set()
 
774
            inventories_to_read = set(present_parent_inv_ids)
 
775
            inventories_to_read.update(revision_ids)
 
776
            for inv in self.iter_inventories(inventories_to_read):
 
777
                entry_chk_root_key = inv.id_to_entry.key()
 
778
                if inv.revision_id in present_parent_inv_ids:
 
779
                    uninteresting_root_keys.add(entry_chk_root_key)
 
780
                else:
 
781
                    interesting_root_keys.add(entry_chk_root_key)
 
782
 
1032
783
            chk_bytes = self.chk_bytes
1033
784
            for record, items in chk_map.iter_interesting_nodes(chk_bytes,
1034
785
                        interesting_root_keys, uninteresting_root_keys,
1035
786
                        pb=pb):
1036
787
                for name, bytes in items:
1037
788
                    (name_utf8, file_id, revision_id) = bytes_to_info(bytes)
1038
 
                    # TODO: consider interning file_id, revision_id here, or
1039
 
                    #       pushing that intern() into bytes_to_info()
1040
 
                    # TODO: rich_root should always be True here, for all
1041
 
                    #       repositories that support chk_bytes
1042
789
                    if not rich_root and name_utf8 == '':
1043
790
                        continue
1044
791
                    try:
1081
828
        finally:
1082
829
            pb.finished()
1083
830
 
1084
 
    @needs_write_lock
1085
 
    def reconcile_canonicalize_chks(self):
1086
 
        """Reconcile this repository to make sure all CHKs are in canonical
1087
 
        form.
1088
 
        """
1089
 
        from bzrlib.reconcile import PackReconciler
1090
 
        reconciler = PackReconciler(self, thorough=True, canonicalize_chks=True)
1091
 
        reconciler.reconcile()
1092
 
        return reconciler
1093
 
 
1094
831
    def _reconcile_pack(self, collection, packs, extension, revs, pb):
1095
832
        packer = GCCHKReconcilePacker(collection, packs, extension)
1096
833
        return packer.pack(pb)
1097
834
 
1098
 
    def _canonicalize_chks_pack(self, collection, packs, extension, revs, pb):
1099
 
        packer = GCCHKCanonicalizingPacker(collection, packs, extension, revs)
1100
 
        return packer.pack(pb)
1101
 
 
1102
835
    def _get_source(self, to_format):
1103
836
        """Return a source for streaming from this repository."""
1104
 
        if self._format._serializer == to_format._serializer:
 
837
        if isinstance(to_format, remote.RemoteRepositoryFormat):
 
838
            # Can't just check attributes on to_format with the current code,
 
839
            # work around this:
 
840
            to_format._ensure_real()
 
841
            to_format = to_format._custom_format
 
842
        if to_format.__class__ is self._format.__class__:
1105
843
            # We must be exactly the same format, otherwise stuff like the chk
1106
 
            # page layout might be different.
1107
 
            # Actually, this test is just slightly looser than exact so that
1108
 
            # CHK2 <-> 2a transfers will work.
 
844
            # page layout might be different
1109
845
            return GroupCHKStreamSource(self, to_format)
1110
846
        return super(CHKInventoryRepository, self)._get_source(to_format)
1111
847
 
1112
 
    def _find_inconsistent_revision_parents(self, revisions_iterator=None):
1113
 
        """Find revisions with different parent lists in the revision object
1114
 
        and in the index graph.
1115
 
 
1116
 
        :param revisions_iterator: None, or an iterator of (revid,
1117
 
            Revision-or-None). This iterator controls the revisions checked.
1118
 
        :returns: an iterator yielding tuples of (revison-id, parents-in-index,
1119
 
            parents-in-revision).
1120
 
        """
1121
 
        if not self.is_locked():
1122
 
            raise AssertionError()
1123
 
        vf = self.revisions
1124
 
        if revisions_iterator is None:
1125
 
            revisions_iterator = self._iter_revisions(None)
1126
 
        for revid, revision in revisions_iterator:
1127
 
            if revision is None:
1128
 
                pass
1129
 
            parent_map = vf.get_parent_map([(revid,)])
1130
 
            parents_according_to_index = tuple(parent[-1] for parent in
1131
 
                parent_map[(revid,)])
1132
 
            parents_according_to_revision = tuple(revision.parent_ids)
1133
 
            if parents_according_to_index != parents_according_to_revision:
1134
 
                yield (revid, parents_according_to_index,
1135
 
                    parents_according_to_revision)
1136
 
 
1137
 
    def _check_for_inconsistent_revision_parents(self):
1138
 
        inconsistencies = list(self._find_inconsistent_revision_parents())
1139
 
        if inconsistencies:
1140
 
            raise errors.BzrCheckError(
1141
 
                "Revision index has inconsistent parents.")
1142
 
 
1143
 
 
1144
 
class GroupCHKStreamSource(StreamSource):
 
848
 
 
849
class GroupCHKStreamSource(repository.StreamSource):
1145
850
    """Used when both the source and target repo are GroupCHK repos."""
1146
851
 
1147
852
    def __init__(self, from_repository, to_format):
1149
854
        super(GroupCHKStreamSource, self).__init__(from_repository, to_format)
1150
855
        self._revision_keys = None
1151
856
        self._text_keys = None
1152
 
        self._text_fetch_order = 'groupcompress'
1153
857
        self._chk_id_roots = None
1154
858
        self._chk_p_id_roots = None
1155
859
 
1194
898
            p_id_roots_set.clear()
1195
899
        return ('inventories', _filtered_inv_stream())
1196
900
 
1197
 
    def _get_filtered_chk_streams(self, excluded_revision_keys):
 
901
    def _find_present_inventories(self, revision_ids):
 
902
        revision_keys = [(r,) for r in revision_ids]
 
903
        inventories = self.from_repository.inventories
 
904
        present_inventories = inventories.get_parent_map(revision_keys)
 
905
        return [p[-1] for p in present_inventories]
 
906
 
 
907
    def _get_filtered_chk_streams(self, excluded_revision_ids):
1198
908
        self._text_keys = set()
1199
 
        excluded_revision_keys.discard(_mod_revision.NULL_REVISION)
1200
 
        if not excluded_revision_keys:
 
909
        excluded_revision_ids.discard(_mod_revision.NULL_REVISION)
 
910
        if not excluded_revision_ids:
1201
911
            uninteresting_root_keys = set()
1202
912
            uninteresting_pid_root_keys = set()
1203
913
        else:
1205
915
            # actually present
1206
916
            # TODO: Update Repository.iter_inventories() to add
1207
917
            #       ignore_missing=True
1208
 
            present_keys = self.from_repository._find_present_inventory_keys(
1209
 
                            excluded_revision_keys)
1210
 
            present_ids = [k[-1] for k in present_keys]
 
918
            present_ids = self.from_repository._find_present_inventory_ids(
 
919
                            excluded_revision_ids)
 
920
            present_ids = self._find_present_inventories(excluded_revision_ids)
1211
921
            uninteresting_root_keys = set()
1212
922
            uninteresting_pid_root_keys = set()
1213
923
            for inv in self.from_repository.iter_inventories(present_ids):
1214
924
                uninteresting_root_keys.add(inv.id_to_entry.key())
1215
925
                uninteresting_pid_root_keys.add(
1216
926
                    inv.parent_id_basename_to_file_id.key())
 
927
        bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
1217
928
        chk_bytes = self.from_repository.chk_bytes
1218
929
        def _filter_id_to_entry():
1219
 
            interesting_nodes = chk_map.iter_interesting_nodes(chk_bytes,
1220
 
                        self._chk_id_roots, uninteresting_root_keys)
1221
 
            for record in _filter_text_keys(interesting_nodes, self._text_keys,
1222
 
                    chk_map._bytes_to_text_key):
 
930
            for record, items in chk_map.iter_interesting_nodes(chk_bytes,
 
931
                        self._chk_id_roots, uninteresting_root_keys):
 
932
                for name, bytes in items:
 
933
                    # Note: we don't care about name_utf8, because we are always
 
934
                    # rich-root = True
 
935
                    _, file_id, revision_id = bytes_to_info(bytes)
 
936
                    self._text_keys.add((file_id, revision_id))
1223
937
                if record is not None:
1224
938
                    yield record
1225
939
            # Consumed
1236
950
 
1237
951
    def _get_text_stream(self):
1238
952
        # Note: We know we don't have to handle adding root keys, because both
1239
 
        # the source and target are the identical network name.
1240
 
        text_stream = self.from_repository.texts.get_record_stream(
1241
 
                        self._text_keys, self._text_fetch_order, False)
1242
 
        return ('texts', text_stream)
 
953
        # the source and target are GCCHK, and those always support rich-roots
 
954
        # We may want to request as 'unordered', in case the source has done a
 
955
        # 'split' packing
 
956
        return ('texts', self.from_repository.texts.get_record_stream(
 
957
                            self._text_keys, 'groupcompress', False))
1243
958
 
1244
959
    def get_stream(self, search):
1245
 
        def wrap_and_count(pb, rc, stream):
1246
 
            """Yield records from stream while showing progress."""
1247
 
            count = 0
1248
 
            for record in stream:
1249
 
                if count == rc.STEP:
1250
 
                    rc.increment(count)
1251
 
                    pb.update('Estimate', rc.current, rc.max)
1252
 
                    count = 0
1253
 
                count += 1
1254
 
                yield record
1255
 
 
1256
960
        revision_ids = search.get_keys()
1257
 
        pb = ui.ui_factory.nested_progress_bar()
1258
 
        rc = self._record_counter
1259
 
        self._record_counter.setup(len(revision_ids))
1260
961
        for stream_info in self._fetch_revision_texts(revision_ids):
1261
 
            yield (stream_info[0],
1262
 
                wrap_and_count(pb, rc, stream_info[1]))
 
962
            yield stream_info
1263
963
        self._revision_keys = [(rev_id,) for rev_id in revision_ids]
 
964
        yield self._get_inventory_stream(self._revision_keys)
1264
965
        # TODO: The keys to exclude might be part of the search recipe
1265
966
        # For now, exclude all parents that are at the edge of ancestry, for
1266
967
        # which we have inventories
1267
968
        from_repo = self.from_repository
1268
 
        parent_keys = from_repo._find_parent_keys_of_revisions(
1269
 
                        self._revision_keys)
1270
 
        self.from_repository.revisions.clear_cache()
1271
 
        self.from_repository.signatures.clear_cache()
1272
 
        # Clear the repo's get_parent_map cache too.
1273
 
        self.from_repository._unstacked_provider.disable_cache()
1274
 
        self.from_repository._unstacked_provider.enable_cache()
1275
 
        s = self._get_inventory_stream(self._revision_keys)
1276
 
        yield (s[0], wrap_and_count(pb, rc, s[1]))
1277
 
        self.from_repository.inventories.clear_cache()
1278
 
        for stream_info in self._get_filtered_chk_streams(parent_keys):
1279
 
            yield (stream_info[0], wrap_and_count(pb, rc, stream_info[1]))
1280
 
        self.from_repository.chk_bytes.clear_cache()
1281
 
        s = self._get_text_stream()
1282
 
        yield (s[0], wrap_and_count(pb, rc, s[1]))
1283
 
        self.from_repository.texts.clear_cache()
1284
 
        pb.update('Done', rc.max, rc.max)
1285
 
        pb.finished()
 
969
        parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
 
970
        for stream_info in self._get_filtered_chk_streams(parent_ids):
 
971
            yield stream_info
 
972
        yield self._get_text_stream()
1286
973
 
1287
974
    def get_stream_for_missing_keys(self, missing_keys):
1288
975
        # missing keys can only occur when we are byte copying and not
1297
984
            missing_inventory_keys.add(key[1:])
1298
985
        if self._chk_id_roots or self._chk_p_id_roots:
1299
986
            raise AssertionError('Cannot call get_stream_for_missing_keys'
1300
 
                ' until all of get_stream() has been consumed.')
 
987
                ' untill all of get_stream() has been consumed.')
1301
988
        # Yield the inventory stream, so we can find the chk stream
1302
989
        # Some of the missing_keys will be missing because they are ghosts.
1303
990
        # As such, we can ignore them. The Sink is required to verify there are
1304
991
        # no unavailable texts when the ghost inventories are not filled in.
1305
992
        yield self._get_inventory_stream(missing_inventory_keys,
1306
993
                                         allow_absent=True)
1307
 
        # We use the empty set for excluded_revision_keys, to make it clear
1308
 
        # that we want to transmit all referenced chk pages.
 
994
        # We use the empty set for excluded_revision_ids, to make it clear that
 
995
        # we want to transmit all referenced chk pages.
1309
996
        for stream_info in self._get_filtered_chk_streams(set()):
1310
997
            yield stream_info
1311
998
 
1312
999
 
1313
 
class _InterestingKeyInfo(object):
1314
 
    def __init__(self):
1315
 
        self.interesting_root_keys = set()
1316
 
        self.interesting_pid_root_keys = set()
1317
 
        self.uninteresting_root_keys = set()
1318
 
        self.uninteresting_pid_root_keys = set()
1319
 
 
1320
 
    def all_interesting(self):
1321
 
        return self.interesting_root_keys.union(self.interesting_pid_root_keys)
1322
 
 
1323
 
    def all_uninteresting(self):
1324
 
        return self.uninteresting_root_keys.union(
1325
 
            self.uninteresting_pid_root_keys)
1326
 
 
1327
 
    def all_keys(self):
1328
 
        return self.all_interesting().union(self.all_uninteresting())
1329
 
 
1330
 
 
1331
 
def _build_interesting_key_sets(repo, inventory_ids, parent_only_inv_ids):
1332
 
    result = _InterestingKeyInfo()
1333
 
    for inv in repo.iter_inventories(inventory_ids, 'unordered'):
1334
 
        root_key = inv.id_to_entry.key()
1335
 
        pid_root_key = inv.parent_id_basename_to_file_id.key()
1336
 
        if inv.revision_id in parent_only_inv_ids:
1337
 
            result.uninteresting_root_keys.add(root_key)
1338
 
            result.uninteresting_pid_root_keys.add(pid_root_key)
1339
 
        else:
1340
 
            result.interesting_root_keys.add(root_key)
1341
 
            result.interesting_pid_root_keys.add(pid_root_key)
1342
 
    return result
1343
 
 
1344
 
 
1345
 
def _filter_text_keys(interesting_nodes_iterable, text_keys, bytes_to_text_key):
1346
 
    """Iterate the result of iter_interesting_nodes, yielding the records
1347
 
    and adding to text_keys.
1348
 
    """
1349
 
    text_keys_update = text_keys.update
1350
 
    for record, items in interesting_nodes_iterable:
1351
 
        text_keys_update([bytes_to_text_key(b) for n,b in items])
1352
 
        yield record
1353
 
 
1354
 
 
1355
 
class RepositoryFormat2a(RepositoryFormatPack):
1356
 
    """A CHK repository that uses the bencode revision serializer."""
 
1000
class RepositoryFormatCHK1(RepositoryFormatPack):
 
1001
    """A hashed CHK+group compress pack repository."""
1357
1002
 
1358
1003
    repository_class = CHKInventoryRepository
1359
1004
    supports_external_lookups = True
1360
1005
    supports_chks = True
 
1006
    # For right now, setting this to True gives us InterModel1And2 rather
 
1007
    # than InterDifferingSerializer
1361
1008
    _commit_builder_class = PackRootCommitBuilder
1362
1009
    rich_root_data = True
1363
 
    _serializer = chk_serializer.chk_bencode_serializer
 
1010
    _serializer = chk_serializer.chk_serializer_255_bigpage
1364
1011
    _commit_inv_deltas = True
1365
1012
    # What index classes to use
1366
1013
    index_builder_class = BTreeBuilder
1374
1021
    _fetch_order = 'unordered'
1375
1022
    _fetch_uses_deltas = False # essentially ignored by the groupcompress code.
1376
1023
    fast_deltas = True
1377
 
    pack_compresses = True
1378
1024
 
1379
1025
    def _get_matching_bzrdir(self):
1380
 
        return bzrdir.format_registry.make_bzrdir('2a')
 
1026
        return bzrdir.format_registry.make_bzrdir('development6-rich-root')
1381
1027
 
1382
1028
    def _ignore_setting_bzrdir(self, format):
1383
1029
        pass
1385
1031
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1386
1032
 
1387
1033
    def get_format_string(self):
1388
 
        return ('Bazaar repository format 2a (needs bzr 1.16 or later)\n')
 
1034
        """See RepositoryFormat.get_format_string()."""
 
1035
        return ('Bazaar development format - group compression and chk inventory'
 
1036
                ' (needs bzr.dev from 1.14)\n')
1389
1037
 
1390
1038
    def get_format_description(self):
1391
1039
        """See RepositoryFormat.get_format_description()."""
1392
 
        return ("Repository format 2a - rich roots, group compression"
 
1040
        return ("Development repository format - rich roots, group compression"
1393
1041
            " and chk inventories")
1394
1042
 
1395
 
 
1396
 
class RepositoryFormat2aSubtree(RepositoryFormat2a):
1397
 
    """A 2a repository format that supports nested trees.
1398
 
 
1399
 
    """
 
1043
    def check_conversion_target(self, target_format):
 
1044
        if not target_format.rich_root_data:
 
1045
            raise errors.BadConversionTarget(
 
1046
                'Does not support rich root data.', target_format)
 
1047
        if not getattr(target_format, 'supports_tree_reference', False):
 
1048
            raise errors.BadConversionTarget(
 
1049
                'Does not support nested trees', target_format)
 
1050
 
 
1051
 
 
1052
 
 
1053
class RepositoryFormatCHK2(RepositoryFormatCHK1):
 
1054
    """A CHK repository that uses the bencode revision serializer."""
 
1055
 
 
1056
    _serializer = chk_serializer.chk_bencode_serializer
1400
1057
 
1401
1058
    def _get_matching_bzrdir(self):
1402
 
        return bzrdir.format_registry.make_bzrdir('development-subtree')
 
1059
        return bzrdir.format_registry.make_bzrdir('development7-rich-root')
1403
1060
 
1404
1061
    def _ignore_setting_bzrdir(self, format):
1405
1062
        pass
1407
1064
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1408
1065
 
1409
1066
    def get_format_string(self):
1410
 
        return ('Bazaar development format 8\n')
1411
 
 
1412
 
    def get_format_description(self):
1413
 
        """See RepositoryFormat.get_format_description()."""
1414
 
        return ("Development repository format 8 - nested trees, "
1415
 
                "group compression and chk inventories")
1416
 
 
1417
 
    experimental = True
1418
 
    supports_tree_reference = True
 
1067
        """See RepositoryFormat.get_format_string()."""
 
1068
        return ('Bazaar development format - chk repository with bencode '
 
1069
                'revision serialization (needs bzr.dev from 1.15)\n')
 
1070
 
 
1071