/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/repofmt/pack_repo.py

  • Committer: Martin Pool
  • Date: 2009-06-11 04:04:14 UTC
  • mto: This revision was merged to the branch mainline in revision 4437.
  • Revision ID: mbp@sourcefrog.net-20090611040414-ippyu6blszi7fhux
Add 2a format

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007, 2008 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
24
24
 
25
25
from bzrlib import (
26
26
    chk_map,
27
 
    cleanup,
28
27
    debug,
29
28
    graph,
30
29
    osutils,
37
36
    )
38
37
from bzrlib.index import (
39
38
    CombinedGraphIndex,
 
39
    GraphIndex,
 
40
    GraphIndexBuilder,
40
41
    GraphIndexPrefixAdapter,
 
42
    InMemoryGraphIndex,
41
43
    )
42
44
from bzrlib.knit import (
43
45
    KnitPlainFactory,
53
55
    lockable_files,
54
56
    lockdir,
55
57
    revision as _mod_revision,
 
58
    symbol_versioning,
56
59
    )
57
60
 
58
 
from bzrlib.decorators import needs_write_lock, only_raises
 
61
from bzrlib.decorators import needs_write_lock
59
62
from bzrlib.btree_index import (
60
63
    BTreeGraphIndex,
61
64
    BTreeBuilder,
64
67
    GraphIndex,
65
68
    InMemoryGraphIndex,
66
69
    )
67
 
from bzrlib.lock import LogicalLockResult
68
70
from bzrlib.repofmt.knitrepo import KnitRepository
69
71
from bzrlib.repository import (
70
72
    CommitBuilder,
71
73
    MetaDirRepositoryFormat,
72
74
    RepositoryFormat,
73
 
    RepositoryWriteLockResult,
74
75
    RootCommitBuilder,
75
 
    StreamSource,
76
76
    )
 
77
import bzrlib.revision as _mod_revision
77
78
from bzrlib.trace import (
78
79
    mutter,
79
 
    note,
80
80
    warning,
81
81
    )
82
82
 
228
228
        return self.index_name('text', name)
229
229
 
230
230
    def _replace_index_with_readonly(self, index_type):
231
 
        unlimited_cache = False
232
 
        if index_type == 'chk':
233
 
            unlimited_cache = True
234
231
        setattr(self, index_type + '_index',
235
232
            self.index_class(self.index_transport,
236
233
                self.index_name(index_type, self.name),
237
 
                self.index_sizes[self.index_offset(index_type)],
238
 
                unlimited_cache=unlimited_cache))
 
234
                self.index_sizes[self.index_offset(index_type)]))
239
235
 
240
236
 
241
237
class ExistingPack(Pack):
316
312
 
317
313
    def finish(self):
318
314
        self._check_references()
 
315
        new_name = '../packs/' + self.file_name()
 
316
        self.upload_transport.rename(self.file_name(), new_name)
319
317
        index_types = ['revision', 'inventory', 'text', 'signature']
320
318
        if self.chk_index is not None:
321
319
            index_types.append('chk')
324
322
            new_name = '../indices/' + old_name
325
323
            self.upload_transport.rename(old_name, new_name)
326
324
            self._replace_index_with_readonly(index_type)
327
 
        new_name = '../packs/' + self.file_name()
328
 
        self.upload_transport.rename(self.file_name(), new_name)
329
325
        self._state = 'finished'
330
326
 
331
327
    def _get_external_refs(self, index):
430
426
        self._writer.begin()
431
427
        # what state is the pack in? (open, finished, aborted)
432
428
        self._state = 'open'
433
 
        # no name until we finish writing the content
434
 
        self.name = None
435
429
 
436
430
    def abort(self):
437
431
        """Cancel creating this pack."""
458
452
            self.signature_index.key_count() or
459
453
            (self.chk_index is not None and self.chk_index.key_count()))
460
454
 
461
 
    def finish_content(self):
462
 
        if self.name is not None:
463
 
            return
464
 
        self._writer.end()
465
 
        if self._buffer[1]:
466
 
            self._write_data('', flush=True)
467
 
        self.name = self._hash.hexdigest()
468
 
 
469
455
    def finish(self, suspend=False):
470
456
        """Finish the new pack.
471
457
 
477
463
         - stores the index size tuple for the pack in the index_sizes
478
464
           attribute.
479
465
        """
480
 
        self.finish_content()
 
466
        self._writer.end()
 
467
        if self._buffer[1]:
 
468
            self._write_data('', flush=True)
 
469
        self.name = self._hash.hexdigest()
481
470
        if not suspend:
482
471
            self._check_references()
483
472
        # write indices
589
578
                                             flush_func=flush_func)
590
579
        self.add_callback = None
591
580
 
 
581
    def replace_indices(self, index_to_pack, indices):
 
582
        """Replace the current mappings with fresh ones.
 
583
 
 
584
        This should probably not be used eventually, rather incremental add and
 
585
        removal of indices. It has been added during refactoring of existing
 
586
        code.
 
587
 
 
588
        :param index_to_pack: A mapping from index objects to
 
589
            (transport, name) tuples for the pack file data.
 
590
        :param indices: A list of indices.
 
591
        """
 
592
        # refresh the revision pack map dict without replacing the instance.
 
593
        self.index_to_pack.clear()
 
594
        self.index_to_pack.update(index_to_pack)
 
595
        # XXX: API break - clearly a 'replace' method would be good?
 
596
        self.combined_index._indices[:] = indices
 
597
        # the current add nodes callback for the current writable index if
 
598
        # there is one.
 
599
        self.add_callback = None
 
600
 
592
601
    def add_index(self, index, pack):
593
602
        """Add index to the aggregate, which is an index for Pack pack.
594
603
 
601
610
        # expose it to the index map
602
611
        self.index_to_pack[index] = pack.access_tuple()
603
612
        # put it at the front of the linear index list
604
 
        self.combined_index.insert_index(0, index, pack.name)
 
613
        self.combined_index.insert_index(0, index)
605
614
 
606
615
    def add_writable_index(self, index, pack):
607
616
        """Add an index which is able to have data added to it.
627
636
        self.data_access.set_writer(None, None, (None, None))
628
637
        self.index_to_pack.clear()
629
638
        del self.combined_index._indices[:]
630
 
        del self.combined_index._index_names[:]
631
639
        self.add_callback = None
632
640
 
633
 
    def remove_index(self, index):
 
641
    def remove_index(self, index, pack):
634
642
        """Remove index from the indices used to answer queries.
635
643
 
636
644
        :param index: An index from the pack parameter.
 
645
        :param pack: A Pack instance.
637
646
        """
638
647
        del self.index_to_pack[index]
639
 
        pos = self.combined_index._indices.index(index)
640
 
        del self.combined_index._indices[pos]
641
 
        del self.combined_index._index_names[pos]
 
648
        self.combined_index._indices.remove(index)
642
649
        if (self.add_callback is not None and
643
650
            getattr(index, 'add_nodes', None) == self.add_callback):
644
651
            self.add_callback = None
1102
1109
            iterator is a tuple with:
1103
1110
            index, readv_vector, node_vector. readv_vector is a list ready to
1104
1111
            hand to the transport readv method, and node_vector is a list of
1105
 
            (key, eol_flag, references) for the node retrieved by the
 
1112
            (key, eol_flag, references) for the the node retrieved by the
1106
1113
            matching readv_vector.
1107
1114
        """
1108
1115
        # group by pack so we do one readv per pack
1400
1407
        self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1401
1408
        self.text_index = AggregateIndex(self.reload_pack_names, flush)
1402
1409
        self.signature_index = AggregateIndex(self.reload_pack_names, flush)
1403
 
        all_indices = [self.revision_index, self.inventory_index,
1404
 
                self.text_index, self.signature_index]
1405
1410
        if use_chk_index:
1406
1411
            self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1407
 
            all_indices.append(self.chk_index)
1408
1412
        else:
1409
1413
            # used to determine if we're using a chk_index elsewhere.
1410
1414
            self.chk_index = None
1411
 
        # Tell all the CombinedGraphIndex objects about each other, so they can
1412
 
        # share hints about which pack names to search first.
1413
 
        all_combined = [agg_idx.combined_index for agg_idx in all_indices]
1414
 
        for combined_idx in all_combined:
1415
 
            combined_idx.set_sibling_indices(
1416
 
                set(all_combined).difference([combined_idx]))
1417
1415
        # resumed packs
1418
1416
        self._resumed_packs = []
1419
1417
 
1420
 
    def __repr__(self):
1421
 
        return '%s(%r)' % (self.__class__.__name__, self.repo)
1422
 
 
1423
1418
    def add_pack_to_memory(self, pack):
1424
1419
        """Make a Pack object available to the repository to satisfy queries.
1425
1420
 
1463
1458
        in synchronisation with certain steps. Otherwise the names collection
1464
1459
        is not flushed.
1465
1460
 
1466
 
        :return: Something evaluating true if packing took place.
 
1461
        :return: True if packing took place.
1467
1462
        """
1468
1463
        while True:
1469
1464
            try:
1470
1465
                return self._do_autopack()
1471
 
            except errors.RetryAutopack:
 
1466
            except errors.RetryAutopack, e:
1472
1467
                # If we get a RetryAutopack exception, we should abort the
1473
1468
                # current action, and retry.
1474
1469
                pass
1478
1473
        total_revisions = self.revision_index.combined_index.key_count()
1479
1474
        total_packs = len(self._names)
1480
1475
        if self._max_pack_count(total_revisions) >= total_packs:
1481
 
            return None
 
1476
            return False
1482
1477
        # determine which packs need changing
1483
1478
        pack_distribution = self.pack_distribution(total_revisions)
1484
1479
        existing_packs = []
1506
1501
            'containing %d revisions. Packing %d files into %d affecting %d'
1507
1502
            ' revisions', self, total_packs, total_revisions, num_old_packs,
1508
1503
            num_new_packs, num_revs_affected)
1509
 
        result = self._execute_pack_operations(pack_operations,
 
1504
        self._execute_pack_operations(pack_operations,
1510
1505
                                      reload_func=self._restart_autopack)
1511
1506
        mutter('Auto-packing repository %s completed', self)
1512
 
        return result
 
1507
        return True
1513
1508
 
1514
1509
    def _execute_pack_operations(self, pack_operations, _packer_class=Packer,
1515
1510
                                 reload_func=None):
1517
1512
 
1518
1513
        :param pack_operations: A list of [revision_count, packs_to_combine].
1519
1514
        :param _packer_class: The class of packer to use (default: Packer).
1520
 
        :return: The new pack names.
 
1515
        :return: None.
1521
1516
        """
1522
1517
        for revision_count, packs in pack_operations:
1523
1518
            # we may have no-ops from the setup logic
1539
1534
                self._remove_pack_from_memory(pack)
1540
1535
        # record the newly available packs and stop advertising the old
1541
1536
        # packs
1542
 
        to_be_obsoleted = []
1543
 
        for _, packs in pack_operations:
1544
 
            to_be_obsoleted.extend(packs)
1545
 
        result = self._save_pack_names(clear_obsolete_packs=True,
1546
 
                                       obsolete_packs=to_be_obsoleted)
1547
 
        return result
 
1537
        self._save_pack_names(clear_obsolete_packs=True)
 
1538
        # Move the old packs out of the way now they are no longer referenced.
 
1539
        for revision_count, packs in pack_operations:
 
1540
            self._obsolete_packs(packs)
1548
1541
 
1549
1542
    def _flush_new_pack(self):
1550
1543
        if self._new_pack is not None:
1560
1553
 
1561
1554
    def _already_packed(self):
1562
1555
        """Is the collection already packed?"""
1563
 
        return not (self.repo._format.pack_compresses or (len(self._names) > 1))
 
1556
        return len(self._names) < 2
1564
1557
 
1565
 
    def pack(self, hint=None, clean_obsolete_packs=False):
 
1558
    def pack(self):
1566
1559
        """Pack the pack collection totally."""
1567
1560
        self.ensure_loaded()
1568
1561
        total_packs = len(self._names)
1569
1562
        if self._already_packed():
 
1563
            # This is arguably wrong because we might not be optimal, but for
 
1564
            # now lets leave it in. (e.g. reconcile -> one pack. But not
 
1565
            # optimal.
1570
1566
            return
1571
1567
        total_revisions = self.revision_index.combined_index.key_count()
1572
1568
        # XXX: the following may want to be a class, to pack with a given
1573
1569
        # policy.
1574
1570
        mutter('Packing repository %s, which has %d pack files, '
1575
 
            'containing %d revisions with hint %r.', self, total_packs,
1576
 
            total_revisions, hint)
 
1571
            'containing %d revisions into 1 packs.', self, total_packs,
 
1572
            total_revisions)
1577
1573
        # determine which packs need changing
 
1574
        pack_distribution = [1]
1578
1575
        pack_operations = [[0, []]]
1579
1576
        for pack in self.all_packs():
1580
 
            if hint is None or pack.name in hint:
1581
 
                # Either no hint was provided (so we are packing everything),
1582
 
                # or this pack was included in the hint.
1583
 
                pack_operations[-1][0] += pack.get_revision_count()
1584
 
                pack_operations[-1][1].append(pack)
 
1577
            pack_operations[-1][0] += pack.get_revision_count()
 
1578
            pack_operations[-1][1].append(pack)
1585
1579
        self._execute_pack_operations(pack_operations, OptimisingPacker)
1586
1580
 
1587
 
        if clean_obsolete_packs:
1588
 
            self._clear_obsolete_packs()
1589
 
 
1590
1581
    def plan_autopack_combinations(self, existing_packs, pack_distribution):
1591
1582
        """Plan a pack operation.
1592
1583
 
1680
1671
            txt_index = self._make_index(name, '.tix')
1681
1672
            sig_index = self._make_index(name, '.six')
1682
1673
            if self.chk_index is not None:
1683
 
                chk_index = self._make_index(name, '.cix', unlimited_cache=True)
 
1674
                chk_index = self._make_index(name, '.cix')
1684
1675
            else:
1685
1676
                chk_index = None
1686
1677
            result = ExistingPack(self._pack_transport, name, rev_index,
1705
1696
            txt_index = self._make_index(name, '.tix', resume=True)
1706
1697
            sig_index = self._make_index(name, '.six', resume=True)
1707
1698
            if self.chk_index is not None:
1708
 
                chk_index = self._make_index(name, '.cix', resume=True,
1709
 
                                             unlimited_cache=True)
 
1699
                chk_index = self._make_index(name, '.cix', resume=True)
1710
1700
            else:
1711
1701
                chk_index = None
1712
1702
            result = self.resumed_pack_factory(name, rev_index, inv_index,
1742
1732
        return self._index_class(self.transport, 'pack-names', None
1743
1733
                ).iter_all_entries()
1744
1734
 
1745
 
    def _make_index(self, name, suffix, resume=False, unlimited_cache=False):
 
1735
    def _make_index(self, name, suffix, resume=False):
1746
1736
        size_offset = self._suffix_offsets[suffix]
1747
1737
        index_name = name + suffix
1748
1738
        if resume:
1751
1741
        else:
1752
1742
            transport = self._index_transport
1753
1743
            index_size = self._names[name][size_offset]
1754
 
        return self._index_class(transport, index_name, index_size,
1755
 
                                 unlimited_cache=unlimited_cache)
 
1744
        return self._index_class(transport, index_name, index_size)
1756
1745
 
1757
1746
    def _max_pack_count(self, total_revisions):
1758
1747
        """Return the maximum number of packs to use for total revisions.
1786
1775
        :param return: None.
1787
1776
        """
1788
1777
        for pack in packs:
1789
 
            try:
1790
 
                pack.pack_transport.rename(pack.file_name(),
1791
 
                    '../obsolete_packs/' + pack.file_name())
1792
 
            except (errors.PathError, errors.TransportError), e:
1793
 
                # TODO: Should these be warnings or mutters?
1794
 
                mutter("couldn't rename obsolete pack, skipping it:\n%s"
1795
 
                       % (e,))
 
1778
            pack.pack_transport.rename(pack.file_name(),
 
1779
                '../obsolete_packs/' + pack.file_name())
1796
1780
            # TODO: Probably needs to know all possible indices for this pack
1797
1781
            # - or maybe list the directory and move all indices matching this
1798
1782
            # name whether we recognize it or not?
1800
1784
            if self.chk_index is not None:
1801
1785
                suffixes.append('.cix')
1802
1786
            for suffix in suffixes:
1803
 
                try:
1804
 
                    self._index_transport.rename(pack.name + suffix,
1805
 
                        '../obsolete_packs/' + pack.name + suffix)
1806
 
                except (errors.PathError, errors.TransportError), e:
1807
 
                    mutter("couldn't rename obsolete index, skipping it:\n%s"
1808
 
                           % (e,))
 
1787
                self._index_transport.rename(pack.name + suffix,
 
1788
                    '../obsolete_packs/' + pack.name + suffix)
1809
1789
 
1810
1790
    def pack_distribution(self, total_revisions):
1811
1791
        """Generate a list of the number of revisions to put in each pack.
1837
1817
        self._remove_pack_indices(pack)
1838
1818
        self.packs.remove(pack)
1839
1819
 
1840
 
    def _remove_pack_indices(self, pack, ignore_missing=False):
1841
 
        """Remove the indices for pack from the aggregated indices.
1842
 
        
1843
 
        :param ignore_missing: Suppress KeyErrors from calling remove_index.
1844
 
        """
1845
 
        for index_type in Pack.index_definitions.keys():
1846
 
            attr_name = index_type + '_index'
1847
 
            aggregate_index = getattr(self, attr_name)
1848
 
            if aggregate_index is not None:
1849
 
                pack_index = getattr(pack, attr_name)
1850
 
                try:
1851
 
                    aggregate_index.remove_index(pack_index)
1852
 
                except KeyError:
1853
 
                    if ignore_missing:
1854
 
                        continue
1855
 
                    raise
 
1820
    def _remove_pack_indices(self, pack):
 
1821
        """Remove the indices for pack from the aggregated indices."""
 
1822
        self.revision_index.remove_index(pack.revision_index, pack)
 
1823
        self.inventory_index.remove_index(pack.inventory_index, pack)
 
1824
        self.text_index.remove_index(pack.text_index, pack)
 
1825
        self.signature_index.remove_index(pack.signature_index, pack)
 
1826
        if self.chk_index is not None:
 
1827
            self.chk_index.remove_index(pack.chk_index, pack)
1856
1828
 
1857
1829
    def reset(self):
1858
1830
        """Clear all cached data."""
1891
1863
        disk_nodes = set()
1892
1864
        for index, key, value in self._iter_disk_pack_index():
1893
1865
            disk_nodes.add((key, value))
1894
 
        orig_disk_nodes = set(disk_nodes)
1895
1866
 
1896
1867
        # do a two-way diff against our original content
1897
1868
        current_nodes = set()
1910
1881
        disk_nodes.difference_update(deleted_nodes)
1911
1882
        disk_nodes.update(new_nodes)
1912
1883
 
1913
 
        return disk_nodes, deleted_nodes, new_nodes, orig_disk_nodes
 
1884
        return disk_nodes, deleted_nodes, new_nodes
1914
1885
 
1915
1886
    def _syncronize_pack_names_from_disk_nodes(self, disk_nodes):
1916
1887
        """Given the correct set of pack files, update our saved info.
1956
1927
                added.append(name)
1957
1928
        return removed, added, modified
1958
1929
 
1959
 
    def _save_pack_names(self, clear_obsolete_packs=False, obsolete_packs=None):
 
1930
    def _save_pack_names(self, clear_obsolete_packs=False):
1960
1931
        """Save the list of packs.
1961
1932
 
1962
1933
        This will take out the mutex around the pack names list for the
1966
1937
 
1967
1938
        :param clear_obsolete_packs: If True, clear out the contents of the
1968
1939
            obsolete_packs directory.
1969
 
        :param obsolete_packs: Packs that are obsolete once the new pack-names
1970
 
            file has been written.
1971
 
        :return: A list of the names saved that were not previously on disk.
1972
1940
        """
1973
 
        already_obsolete = []
1974
1941
        self.lock_names()
1975
1942
        try:
1976
1943
            builder = self._index_builder_class()
1977
 
            (disk_nodes, deleted_nodes, new_nodes,
1978
 
             orig_disk_nodes) = self._diff_pack_names()
 
1944
            disk_nodes, deleted_nodes, new_nodes = self._diff_pack_names()
1979
1945
            # TODO: handle same-name, index-size-changes here -
1980
1946
            # e.g. use the value from disk, not ours, *unless* we're the one
1981
1947
            # changing it.
1983
1949
                builder.add_node(key, value)
1984
1950
            self.transport.put_file('pack-names', builder.finish(),
1985
1951
                mode=self.repo.bzrdir._get_file_mode())
 
1952
            # move the baseline forward
1986
1953
            self._packs_at_load = disk_nodes
1987
1954
            if clear_obsolete_packs:
1988
 
                to_preserve = None
1989
 
                if obsolete_packs:
1990
 
                    to_preserve = set([o.name for o in obsolete_packs])
1991
 
                already_obsolete = self._clear_obsolete_packs(to_preserve)
 
1955
                self._clear_obsolete_packs()
1992
1956
        finally:
1993
1957
            self._unlock_names()
1994
1958
        # synchronise the memory packs list with what we just wrote:
1995
1959
        self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1996
 
        if obsolete_packs:
1997
 
            # TODO: We could add one more condition here. "if o.name not in
1998
 
            #       orig_disk_nodes and o != the new_pack we haven't written to
1999
 
            #       disk yet. However, the new pack object is not easily
2000
 
            #       accessible here (it would have to be passed through the
2001
 
            #       autopacking code, etc.)
2002
 
            obsolete_packs = [o for o in obsolete_packs
2003
 
                              if o.name not in already_obsolete]
2004
 
            self._obsolete_packs(obsolete_packs)
2005
 
        return [new_node[0][0] for new_node in new_nodes]
2006
1960
 
2007
1961
    def reload_pack_names(self):
2008
1962
        """Sync our pack listing with what is present in the repository.
2022
1976
        if first_read:
2023
1977
            return True
2024
1978
        # out the new value.
2025
 
        (disk_nodes, deleted_nodes, new_nodes,
2026
 
         orig_disk_nodes) = self._diff_pack_names()
2027
 
        # _packs_at_load is meant to be the explicit list of names in
2028
 
        # 'pack-names' at then start. As such, it should not contain any
2029
 
        # pending names that haven't been written out yet.
2030
 
        self._packs_at_load = orig_disk_nodes
 
1979
        disk_nodes, _, _ = self._diff_pack_names()
 
1980
        self._packs_at_load = disk_nodes
2031
1981
        (removed, added,
2032
1982
         modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2033
1983
        if removed or added or modified:
2042
1992
            raise
2043
1993
        raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2044
1994
 
2045
 
    def _clear_obsolete_packs(self, preserve=None):
 
1995
    def _clear_obsolete_packs(self):
2046
1996
        """Delete everything from the obsolete-packs directory.
2047
 
 
2048
 
        :return: A list of pack identifiers (the filename without '.pack') that
2049
 
            were found in obsolete_packs.
2050
1997
        """
2051
 
        found = []
2052
1998
        obsolete_pack_transport = self.transport.clone('obsolete_packs')
2053
 
        if preserve is None:
2054
 
            preserve = set()
2055
1999
        for filename in obsolete_pack_transport.list_dir('.'):
2056
 
            name, ext = osutils.splitext(filename)
2057
 
            if ext == '.pack':
2058
 
                found.append(name)
2059
 
            if name in preserve:
2060
 
                continue
2061
2000
            try:
2062
2001
                obsolete_pack_transport.delete(filename)
2063
2002
            except (errors.PathError, errors.TransportError), e:
2064
 
                warning("couldn't delete obsolete pack, skipping it:\n%s"
2065
 
                        % (e,))
2066
 
        return found
 
2003
                warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2067
2004
 
2068
2005
    def _start_write_group(self):
2069
2006
        # Do not permit preparation for writing if we're not in a 'write lock'.
2096
2033
        # FIXME: just drop the transient index.
2097
2034
        # forget what names there are
2098
2035
        if self._new_pack is not None:
2099
 
            operation = cleanup.OperationWithCleanups(self._new_pack.abort)
2100
 
            operation.add_cleanup(setattr, self, '_new_pack', None)
2101
 
            # If we aborted while in the middle of finishing the write
2102
 
            # group, _remove_pack_indices could fail because the indexes are
2103
 
            # already gone.  But they're not there we shouldn't fail in this
2104
 
            # case, so we pass ignore_missing=True.
2105
 
            operation.add_cleanup(self._remove_pack_indices, self._new_pack,
2106
 
                ignore_missing=True)
2107
 
            operation.run_simple()
 
2036
            try:
 
2037
                self._new_pack.abort()
 
2038
            finally:
 
2039
                # XXX: If we aborted while in the middle of finishing the write
 
2040
                # group, _remove_pack_indices can fail because the indexes are
 
2041
                # already gone.  If they're not there we shouldn't fail in this
 
2042
                # case.  -- mbp 20081113
 
2043
                self._remove_pack_indices(self._new_pack)
 
2044
                self._new_pack = None
2108
2045
        for resumed_pack in self._resumed_packs:
2109
 
            operation = cleanup.OperationWithCleanups(resumed_pack.abort)
2110
 
            # See comment in previous finally block.
2111
 
            operation.add_cleanup(self._remove_pack_indices, resumed_pack,
2112
 
                ignore_missing=True)
2113
 
            operation.run_simple()
 
2046
            try:
 
2047
                resumed_pack.abort()
 
2048
            finally:
 
2049
                # See comment in previous finally block.
 
2050
                try:
 
2051
                    self._remove_pack_indices(resumed_pack)
 
2052
                except KeyError:
 
2053
                    pass
2114
2054
        del self._resumed_packs[:]
2115
2055
 
2116
2056
    def _remove_resumed_pack_indices(self):
2118
2058
            self._remove_pack_indices(resumed_pack)
2119
2059
        del self._resumed_packs[:]
2120
2060
 
2121
 
    def _check_new_inventories(self):
2122
 
        """Detect missing inventories in this write group.
2123
 
 
2124
 
        :returns: list of strs, summarising any problems found.  If the list is
2125
 
            empty no problems were found.
2126
 
        """
2127
 
        # The base implementation does no checks.  GCRepositoryPackCollection
2128
 
        # overrides this.
2129
 
        return []
2130
 
        
2131
2061
    def _commit_write_group(self):
2132
2062
        all_missing = set()
2133
2063
        for prefix, versioned_file in (
2142
2072
            raise errors.BzrCheckError(
2143
2073
                "Repository %s has missing compression parent(s) %r "
2144
2074
                 % (self.repo, sorted(all_missing)))
2145
 
        problems = self._check_new_inventories()
2146
 
        if problems:
2147
 
            problems_summary = '\n'.join(problems)
2148
 
            raise errors.BzrCheckError(
2149
 
                "Cannot add revision(s) to repository: " + problems_summary)
2150
2075
        self._remove_pack_indices(self._new_pack)
2151
 
        any_new_content = False
 
2076
        should_autopack = False
2152
2077
        if self._new_pack.data_inserted():
2153
2078
            # get all the data to disk and read to use
2154
2079
            self._new_pack.finish()
2155
2080
            self.allocate(self._new_pack)
2156
2081
            self._new_pack = None
2157
 
            any_new_content = True
 
2082
            should_autopack = True
2158
2083
        else:
2159
2084
            self._new_pack.abort()
2160
2085
            self._new_pack = None
2165
2090
            self._remove_pack_from_memory(resumed_pack)
2166
2091
            resumed_pack.finish()
2167
2092
            self.allocate(resumed_pack)
2168
 
            any_new_content = True
 
2093
            should_autopack = True
2169
2094
        del self._resumed_packs[:]
2170
 
        if any_new_content:
2171
 
            result = self.autopack()
2172
 
            if not result:
 
2095
        if should_autopack:
 
2096
            if not self.autopack():
2173
2097
                # when autopack takes no steps, the names list is still
2174
2098
                # unsaved.
2175
 
                return self._save_pack_names()
2176
 
            return result
2177
 
        return []
 
2099
                self._save_pack_names()
2178
2100
 
2179
2101
    def _suspend_write_group(self):
2180
2102
        tokens = [pack.name for pack in self._resumed_packs]
2282
2204
        self._reconcile_fixes_text_parents = True
2283
2205
        self._reconcile_backsup_inventory = False
2284
2206
 
2285
 
    def _warn_if_deprecated(self, branch=None):
 
2207
    def _warn_if_deprecated(self):
2286
2208
        # This class isn't deprecated, but one sub-format is
2287
2209
        if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2288
 
            super(KnitPackRepository, self)._warn_if_deprecated(branch)
 
2210
            from bzrlib import repository
 
2211
            if repository._deprecation_warning_done:
 
2212
                return
 
2213
            repository._deprecation_warning_done = True
 
2214
            warning("Format %s for %s is deprecated - please use"
 
2215
                    " 'bzr upgrade --1.6.1-rich-root'"
 
2216
                    % (self._format, self.bzrdir.transport.base))
2289
2217
 
2290
2218
    def _abort_write_group(self):
2291
 
        self.revisions._index._key_dependencies.clear()
 
2219
        self.revisions._index._key_dependencies.refs.clear()
2292
2220
        self._pack_collection._abort_write_group()
2293
2221
 
2294
 
    def _get_source(self, to_format):
2295
 
        if to_format.network_name() == self._format.network_name():
2296
 
            return KnitPackStreamSource(self, to_format)
2297
 
        return super(KnitPackRepository, self)._get_source(to_format)
 
2222
    def _find_inconsistent_revision_parents(self):
 
2223
        """Find revisions with incorrectly cached parents.
 
2224
 
 
2225
        :returns: an iterator yielding tuples of (revison-id, parents-in-index,
 
2226
            parents-in-revision).
 
2227
        """
 
2228
        if not self.is_locked():
 
2229
            raise errors.ObjectNotLocked(self)
 
2230
        pb = ui.ui_factory.nested_progress_bar()
 
2231
        result = []
 
2232
        try:
 
2233
            revision_nodes = self._pack_collection.revision_index \
 
2234
                .combined_index.iter_all_entries()
 
2235
            index_positions = []
 
2236
            # Get the cached index values for all revisions, and also the
 
2237
            # location in each index of the revision text so we can perform
 
2238
            # linear IO.
 
2239
            for index, key, value, refs in revision_nodes:
 
2240
                node = (index, key, value, refs)
 
2241
                index_memo = self.revisions._index._node_to_position(node)
 
2242
                if index_memo[0] != index:
 
2243
                    raise AssertionError('%r != %r' % (index_memo[0], index))
 
2244
                index_positions.append((index_memo, key[0],
 
2245
                                       tuple(parent[0] for parent in refs[0])))
 
2246
                pb.update("Reading revision index", 0, 0)
 
2247
            index_positions.sort()
 
2248
            batch_size = 1000
 
2249
            pb.update("Checking cached revision graph", 0,
 
2250
                      len(index_positions))
 
2251
            for offset in xrange(0, len(index_positions), 1000):
 
2252
                pb.update("Checking cached revision graph", offset)
 
2253
                to_query = index_positions[offset:offset + batch_size]
 
2254
                if not to_query:
 
2255
                    break
 
2256
                rev_ids = [item[1] for item in to_query]
 
2257
                revs = self.get_revisions(rev_ids)
 
2258
                for revision, item in zip(revs, to_query):
 
2259
                    index_parents = item[2]
 
2260
                    rev_parents = tuple(revision.parent_ids)
 
2261
                    if index_parents != rev_parents:
 
2262
                        result.append((revision.revision_id, index_parents,
 
2263
                                       rev_parents))
 
2264
        finally:
 
2265
            pb.finished()
 
2266
        return result
2298
2267
 
2299
2268
    def _make_parents_provider(self):
2300
2269
        return graph.CachingParentsProvider(self)
2308
2277
        self._pack_collection._start_write_group()
2309
2278
 
2310
2279
    def _commit_write_group(self):
2311
 
        hint = self._pack_collection._commit_write_group()
2312
 
        self.revisions._index._key_dependencies.clear()
2313
 
        return hint
 
2280
        self.revisions._index._key_dependencies.refs.clear()
 
2281
        return self._pack_collection._commit_write_group()
2314
2282
 
2315
2283
    def suspend_write_group(self):
2316
2284
        # XXX check self._write_group is self.get_transaction()?
2317
2285
        tokens = self._pack_collection._suspend_write_group()
2318
 
        self.revisions._index._key_dependencies.clear()
 
2286
        self.revisions._index._key_dependencies.refs.clear()
2319
2287
        self._write_group = None
2320
2288
        return tokens
2321
2289
 
2342
2310
        return self._write_lock_count
2343
2311
 
2344
2312
    def lock_write(self, token=None):
2345
 
        """Lock the repository for writes.
2346
 
 
2347
 
        :return: A bzrlib.repository.RepositoryWriteLockResult.
2348
 
        """
2349
2313
        locked = self.is_locked()
2350
2314
        if not self._write_lock_count and locked:
2351
2315
            raise errors.ReadOnlyError(self)
2353
2317
        if self._write_lock_count == 1:
2354
2318
            self._transaction = transactions.WriteTransaction()
2355
2319
        if not locked:
2356
 
            if 'relock' in debug.debug_flags and self._prev_lock == 'w':
2357
 
                note('%r was write locked again', self)
2358
 
            self._prev_lock = 'w'
2359
2320
            for repo in self._fallback_repositories:
2360
2321
                # Writes don't affect fallback repos
2361
2322
                repo.lock_read()
2362
2323
            self._refresh_data()
2363
 
        return RepositoryWriteLockResult(self.unlock, None)
2364
2324
 
2365
2325
    def lock_read(self):
2366
 
        """Lock the repository for reads.
2367
 
 
2368
 
        :return: A bzrlib.lock.LogicalLockResult.
2369
 
        """
2370
2326
        locked = self.is_locked()
2371
2327
        if self._write_lock_count:
2372
2328
            self._write_lock_count += 1
2373
2329
        else:
2374
2330
            self.control_files.lock_read()
2375
2331
        if not locked:
2376
 
            if 'relock' in debug.debug_flags and self._prev_lock == 'r':
2377
 
                note('%r was read locked again', self)
2378
 
            self._prev_lock = 'r'
2379
2332
            for repo in self._fallback_repositories:
2380
2333
                repo.lock_read()
2381
2334
            self._refresh_data()
2382
 
        return LogicalLockResult(self.unlock)
2383
2335
 
2384
2336
    def leave_lock_in_place(self):
2385
2337
        # not supported - raise an error
2390
2342
        raise NotImplementedError(self.dont_leave_lock_in_place)
2391
2343
 
2392
2344
    @needs_write_lock
2393
 
    def pack(self, hint=None, clean_obsolete_packs=False):
 
2345
    def pack(self):
2394
2346
        """Compress the data within the repository.
2395
2347
 
2396
2348
        This will pack all the data to a single pack. In future it may
2397
2349
        recompress deltas or do other such expensive operations.
2398
2350
        """
2399
 
        self._pack_collection.pack(hint=hint, clean_obsolete_packs=clean_obsolete_packs)
 
2351
        self._pack_collection.pack()
2400
2352
 
2401
2353
    @needs_write_lock
2402
2354
    def reconcile(self, other=None, thorough=False):
2410
2362
        packer = ReconcilePacker(collection, packs, extension, revs)
2411
2363
        return packer.pack(pb)
2412
2364
 
2413
 
    @only_raises(errors.LockNotHeld, errors.LockBroken)
2414
2365
    def unlock(self):
2415
2366
        if self._write_lock_count == 1 and self._write_group is not None:
2416
2367
            self.abort_write_group()
2433
2384
                repo.unlock()
2434
2385
 
2435
2386
 
2436
 
class KnitPackStreamSource(StreamSource):
2437
 
    """A StreamSource used to transfer data between same-format KnitPack repos.
2438
 
 
2439
 
    This source assumes:
2440
 
        1) Same serialization format for all objects
2441
 
        2) Same root information
2442
 
        3) XML format inventories
2443
 
        4) Atomic inserts (so we can stream inventory texts before text
2444
 
           content)
2445
 
        5) No chk_bytes
2446
 
    """
2447
 
 
2448
 
    def __init__(self, from_repository, to_format):
2449
 
        super(KnitPackStreamSource, self).__init__(from_repository, to_format)
2450
 
        self._text_keys = None
2451
 
        self._text_fetch_order = 'unordered'
2452
 
 
2453
 
    def _get_filtered_inv_stream(self, revision_ids):
2454
 
        from_repo = self.from_repository
2455
 
        parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
2456
 
        parent_keys = [(p,) for p in parent_ids]
2457
 
        find_text_keys = from_repo._find_text_key_references_from_xml_inventory_lines
2458
 
        parent_text_keys = set(find_text_keys(
2459
 
            from_repo._inventory_xml_lines_for_keys(parent_keys)))
2460
 
        content_text_keys = set()
2461
 
        knit = KnitVersionedFiles(None, None)
2462
 
        factory = KnitPlainFactory()
2463
 
        def find_text_keys_from_content(record):
2464
 
            if record.storage_kind not in ('knit-delta-gz', 'knit-ft-gz'):
2465
 
                raise ValueError("Unknown content storage kind for"
2466
 
                    " inventory text: %s" % (record.storage_kind,))
2467
 
            # It's a knit record, it has a _raw_record field (even if it was
2468
 
            # reconstituted from a network stream).
2469
 
            raw_data = record._raw_record
2470
 
            # read the entire thing
2471
 
            revision_id = record.key[-1]
2472
 
            content, _ = knit._parse_record(revision_id, raw_data)
2473
 
            if record.storage_kind == 'knit-delta-gz':
2474
 
                line_iterator = factory.get_linedelta_content(content)
2475
 
            elif record.storage_kind == 'knit-ft-gz':
2476
 
                line_iterator = factory.get_fulltext_content(content)
2477
 
            content_text_keys.update(find_text_keys(
2478
 
                [(line, revision_id) for line in line_iterator]))
2479
 
        revision_keys = [(r,) for r in revision_ids]
2480
 
        def _filtered_inv_stream():
2481
 
            source_vf = from_repo.inventories
2482
 
            stream = source_vf.get_record_stream(revision_keys,
2483
 
                                                 'unordered', False)
2484
 
            for record in stream:
2485
 
                if record.storage_kind == 'absent':
2486
 
                    raise errors.NoSuchRevision(from_repo, record.key)
2487
 
                find_text_keys_from_content(record)
2488
 
                yield record
2489
 
            self._text_keys = content_text_keys - parent_text_keys
2490
 
        return ('inventories', _filtered_inv_stream())
2491
 
 
2492
 
    def _get_text_stream(self):
2493
 
        # Note: We know we don't have to handle adding root keys, because both
2494
 
        # the source and target are the identical network name.
2495
 
        text_stream = self.from_repository.texts.get_record_stream(
2496
 
                        self._text_keys, self._text_fetch_order, False)
2497
 
        return ('texts', text_stream)
2498
 
 
2499
 
    def get_stream(self, search):
2500
 
        revision_ids = search.get_keys()
2501
 
        for stream_info in self._fetch_revision_texts(revision_ids):
2502
 
            yield stream_info
2503
 
        self._revision_keys = [(rev_id,) for rev_id in revision_ids]
2504
 
        yield self._get_filtered_inv_stream(revision_ids)
2505
 
        yield self._get_text_stream()
2506
 
 
2507
 
 
2508
 
 
2509
2387
class RepositoryFormatPack(MetaDirRepositoryFormat):
2510
2388
    """Format logic for pack structured repositories.
2511
2389
 
2558
2436
        utf8_files = [('format', self.get_format_string())]
2559
2437
 
2560
2438
        self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
2561
 
        repository = self.open(a_bzrdir=a_bzrdir, _found=True)
2562
 
        self._run_post_repo_init_hooks(repository, a_bzrdir, shared)
2563
 
        return repository
 
2439
        return self.open(a_bzrdir=a_bzrdir, _found=True)
2564
2440
 
2565
2441
    def open(self, a_bzrdir, _found=False, _override_transport=None):
2566
2442
        """See RepositoryFormat.open().
2615
2491
        """See RepositoryFormat.get_format_description()."""
2616
2492
        return "Packs containing knits without subtree support"
2617
2493
 
 
2494
    def check_conversion_target(self, target_format):
 
2495
        pass
 
2496
 
2618
2497
 
2619
2498
class RepositoryFormatKnitPack3(RepositoryFormatPack):
2620
2499
    """A subtrees parameterized Pack repository.
2629
2508
    repository_class = KnitPackRepository
2630
2509
    _commit_builder_class = PackRootCommitBuilder
2631
2510
    rich_root_data = True
2632
 
    experimental = True
2633
2511
    supports_tree_reference = True
2634
2512
    @property
2635
2513
    def _serializer(self):
2647
2525
 
2648
2526
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2649
2527
 
 
2528
    def check_conversion_target(self, target_format):
 
2529
        if not target_format.rich_root_data:
 
2530
            raise errors.BadConversionTarget(
 
2531
                'Does not support rich root data.', target_format)
 
2532
        if not getattr(target_format, 'supports_tree_reference', False):
 
2533
            raise errors.BadConversionTarget(
 
2534
                'Does not support nested trees', target_format)
 
2535
 
2650
2536
    def get_format_string(self):
2651
2537
        """See RepositoryFormat.get_format_string()."""
2652
2538
        return "Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n"
2685
2571
 
2686
2572
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2687
2573
 
 
2574
    def check_conversion_target(self, target_format):
 
2575
        if not target_format.rich_root_data:
 
2576
            raise errors.BadConversionTarget(
 
2577
                'Does not support rich root data.', target_format)
 
2578
 
2688
2579
    def get_format_string(self):
2689
2580
        """See RepositoryFormat.get_format_string()."""
2690
2581
        return ("Bazaar pack repository format 1 with rich root"
2731
2622
        """See RepositoryFormat.get_format_description()."""
2732
2623
        return "Packs 5 (adds stacking support, requires bzr 1.6)"
2733
2624
 
 
2625
    def check_conversion_target(self, target_format):
 
2626
        pass
 
2627
 
2734
2628
 
2735
2629
class RepositoryFormatKnitPack5RichRoot(RepositoryFormatPack):
2736
2630
    """A repository with rich roots and stacking.
2763
2657
 
2764
2658
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2765
2659
 
 
2660
    def check_conversion_target(self, target_format):
 
2661
        if not target_format.rich_root_data:
 
2662
            raise errors.BadConversionTarget(
 
2663
                'Does not support rich root data.', target_format)
 
2664
 
2766
2665
    def get_format_string(self):
2767
2666
        """See RepositoryFormat.get_format_string()."""
2768
2667
        return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n"
2809
2708
 
2810
2709
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2811
2710
 
 
2711
    def check_conversion_target(self, target_format):
 
2712
        if not target_format.rich_root_data:
 
2713
            raise errors.BadConversionTarget(
 
2714
                'Does not support rich root data.', target_format)
 
2715
 
2812
2716
    def get_format_string(self):
2813
2717
        """See RepositoryFormat.get_format_string()."""
2814
2718
        return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n"
2852
2756
        """See RepositoryFormat.get_format_description()."""
2853
2757
        return "Packs 6 (uses btree indexes, requires bzr 1.9)"
2854
2758
 
 
2759
    def check_conversion_target(self, target_format):
 
2760
        pass
 
2761
 
2855
2762
 
2856
2763
class RepositoryFormatKnitPack6RichRoot(RepositoryFormatPack):
2857
2764
    """A repository with rich roots, no subtrees, stacking and btree indexes.
2881
2788
 
2882
2789
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2883
2790
 
 
2791
    def check_conversion_target(self, target_format):
 
2792
        if not target_format.rich_root_data:
 
2793
            raise errors.BadConversionTarget(
 
2794
                'Does not support rich root data.', target_format)
 
2795
 
2884
2796
    def get_format_string(self):
2885
2797
        """See RepositoryFormat.get_format_string()."""
2886
2798
        return "Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n"
2903
2815
    repository_class = KnitPackRepository
2904
2816
    _commit_builder_class = PackRootCommitBuilder
2905
2817
    rich_root_data = True
2906
 
    experimental = True
2907
2818
    supports_tree_reference = True
2908
2819
    supports_external_lookups = True
2909
2820
    # What index classes to use
2923
2834
 
2924
2835
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2925
2836
 
 
2837
    def check_conversion_target(self, target_format):
 
2838
        if not target_format.rich_root_data:
 
2839
            raise errors.BadConversionTarget(
 
2840
                'Does not support rich root data.', target_format)
 
2841
        if not getattr(target_format, 'supports_tree_reference', False):
 
2842
            raise errors.BadConversionTarget(
 
2843
                'Does not support nested trees', target_format)
 
2844
 
2926
2845
    def get_format_string(self):
2927
2846
        """See RepositoryFormat.get_format_string()."""
2928
2847
        return ("Bazaar development format 2 with subtree support "