/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: Sidnei da Silva
  • Date: 2009-06-05 01:37:21 UTC
  • mto: (4531.1.1 integration)
  • mto: This revision was merged to the branch mainline in revision 4532.
  • Revision ID: sidnei.da.silva@canonical.com-20090605013721-w0urfus3il6tbxoi
- Start of a batch file

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,
69
72
    CommitBuilder,
70
73
    MetaDirRepositoryFormat,
71
74
    RepositoryFormat,
72
 
    RepositoryWriteLockResult,
73
75
    RootCommitBuilder,
74
 
    StreamSource,
75
76
    )
 
77
import bzrlib.revision as _mod_revision
76
78
from bzrlib.trace import (
77
79
    mutter,
78
 
    note,
79
80
    warning,
80
81
    )
81
82
 
227
228
        return self.index_name('text', name)
228
229
 
229
230
    def _replace_index_with_readonly(self, index_type):
230
 
        unlimited_cache = False
231
 
        if index_type == 'chk':
232
 
            unlimited_cache = True
233
231
        setattr(self, index_type + '_index',
234
232
            self.index_class(self.index_transport,
235
233
                self.index_name(index_type, self.name),
236
 
                self.index_sizes[self.index_offset(index_type)],
237
 
                unlimited_cache=unlimited_cache))
 
234
                self.index_sizes[self.index_offset(index_type)]))
238
235
 
239
236
 
240
237
class ExistingPack(Pack):
315
312
 
316
313
    def finish(self):
317
314
        self._check_references()
 
315
        new_name = '../packs/' + self.file_name()
 
316
        self.upload_transport.rename(self.file_name(), new_name)
318
317
        index_types = ['revision', 'inventory', 'text', 'signature']
319
318
        if self.chk_index is not None:
320
319
            index_types.append('chk')
323
322
            new_name = '../indices/' + old_name
324
323
            self.upload_transport.rename(old_name, new_name)
325
324
            self._replace_index_with_readonly(index_type)
326
 
        new_name = '../packs/' + self.file_name()
327
 
        self.upload_transport.rename(self.file_name(), new_name)
328
325
        self._state = 'finished'
329
326
 
330
327
    def _get_external_refs(self, index):
429
426
        self._writer.begin()
430
427
        # what state is the pack in? (open, finished, aborted)
431
428
        self._state = 'open'
432
 
        # no name until we finish writing the content
433
 
        self.name = None
434
429
 
435
430
    def abort(self):
436
431
        """Cancel creating this pack."""
457
452
            self.signature_index.key_count() or
458
453
            (self.chk_index is not None and self.chk_index.key_count()))
459
454
 
460
 
    def finish_content(self):
461
 
        if self.name is not None:
462
 
            return
463
 
        self._writer.end()
464
 
        if self._buffer[1]:
465
 
            self._write_data('', flush=True)
466
 
        self.name = self._hash.hexdigest()
467
 
 
468
455
    def finish(self, suspend=False):
469
456
        """Finish the new pack.
470
457
 
476
463
         - stores the index size tuple for the pack in the index_sizes
477
464
           attribute.
478
465
        """
479
 
        self.finish_content()
 
466
        self._writer.end()
 
467
        if self._buffer[1]:
 
468
            self._write_data('', flush=True)
 
469
        self.name = self._hash.hexdigest()
480
470
        if not suspend:
481
471
            self._check_references()
482
472
        # write indices
588
578
                                             flush_func=flush_func)
589
579
        self.add_callback = None
590
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
 
591
601
    def add_index(self, index, pack):
592
602
        """Add index to the aggregate, which is an index for Pack pack.
593
603
 
600
610
        # expose it to the index map
601
611
        self.index_to_pack[index] = pack.access_tuple()
602
612
        # put it at the front of the linear index list
603
 
        self.combined_index.insert_index(0, index, pack.name)
 
613
        self.combined_index.insert_index(0, index)
604
614
 
605
615
    def add_writable_index(self, index, pack):
606
616
        """Add an index which is able to have data added to it.
626
636
        self.data_access.set_writer(None, None, (None, None))
627
637
        self.index_to_pack.clear()
628
638
        del self.combined_index._indices[:]
629
 
        del self.combined_index._index_names[:]
630
639
        self.add_callback = None
631
640
 
632
 
    def remove_index(self, index):
 
641
    def remove_index(self, index, pack):
633
642
        """Remove index from the indices used to answer queries.
634
643
 
635
644
        :param index: An index from the pack parameter.
 
645
        :param pack: A Pack instance.
636
646
        """
637
647
        del self.index_to_pack[index]
638
 
        pos = self.combined_index._indices.index(index)
639
 
        del self.combined_index._indices[pos]
640
 
        del self.combined_index._index_names[pos]
 
648
        self.combined_index._indices.remove(index)
641
649
        if (self.add_callback is not None and
642
650
            getattr(index, 'add_nodes', None) == self.add_callback):
643
651
            self.add_callback = None
1101
1109
            iterator is a tuple with:
1102
1110
            index, readv_vector, node_vector. readv_vector is a list ready to
1103
1111
            hand to the transport readv method, and node_vector is a list of
1104
 
            (key, eol_flag, references) for the node retrieved by the
 
1112
            (key, eol_flag, references) for the the node retrieved by the
1105
1113
            matching readv_vector.
1106
1114
        """
1107
1115
        # group by pack so we do one readv per pack
1298
1306
        # space (we only topo sort the revisions, which is smaller).
1299
1307
        topo_order = tsort.topo_sort(ancestors)
1300
1308
        rev_order = dict(zip(topo_order, range(len(topo_order))))
1301
 
        bad_texts.sort(key=lambda key:rev_order.get(key[0][1], 0))
 
1309
        bad_texts.sort(key=lambda key:rev_order[key[0][1]])
1302
1310
        transaction = repo.get_transaction()
1303
1311
        file_id_index = GraphIndexPrefixAdapter(
1304
1312
            self.new_pack.text_index,
1399
1407
        self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1400
1408
        self.text_index = AggregateIndex(self.reload_pack_names, flush)
1401
1409
        self.signature_index = AggregateIndex(self.reload_pack_names, flush)
1402
 
        all_indices = [self.revision_index, self.inventory_index,
1403
 
                self.text_index, self.signature_index]
1404
1410
        if use_chk_index:
1405
1411
            self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1406
 
            all_indices.append(self.chk_index)
1407
1412
        else:
1408
1413
            # used to determine if we're using a chk_index elsewhere.
1409
1414
            self.chk_index = None
1410
 
        # Tell all the CombinedGraphIndex objects about each other, so they can
1411
 
        # share hints about which pack names to search first.
1412
 
        all_combined = [agg_idx.combined_index for agg_idx in all_indices]
1413
 
        for combined_idx in all_combined:
1414
 
            combined_idx.set_sibling_indices(
1415
 
                set(all_combined).difference([combined_idx]))
1416
1415
        # resumed packs
1417
1416
        self._resumed_packs = []
1418
1417
 
1419
 
    def __repr__(self):
1420
 
        return '%s(%r)' % (self.__class__.__name__, self.repo)
1421
 
 
1422
1418
    def add_pack_to_memory(self, pack):
1423
1419
        """Make a Pack object available to the repository to satisfy queries.
1424
1420
 
1462
1458
        in synchronisation with certain steps. Otherwise the names collection
1463
1459
        is not flushed.
1464
1460
 
1465
 
        :return: Something evaluating true if packing took place.
 
1461
        :return: True if packing took place.
1466
1462
        """
1467
1463
        while True:
1468
1464
            try:
1469
1465
                return self._do_autopack()
1470
 
            except errors.RetryAutopack:
 
1466
            except errors.RetryAutopack, e:
1471
1467
                # If we get a RetryAutopack exception, we should abort the
1472
1468
                # current action, and retry.
1473
1469
                pass
1477
1473
        total_revisions = self.revision_index.combined_index.key_count()
1478
1474
        total_packs = len(self._names)
1479
1475
        if self._max_pack_count(total_revisions) >= total_packs:
1480
 
            return None
 
1476
            return False
1481
1477
        # determine which packs need changing
1482
1478
        pack_distribution = self.pack_distribution(total_revisions)
1483
1479
        existing_packs = []
1505
1501
            'containing %d revisions. Packing %d files into %d affecting %d'
1506
1502
            ' revisions', self, total_packs, total_revisions, num_old_packs,
1507
1503
            num_new_packs, num_revs_affected)
1508
 
        result = self._execute_pack_operations(pack_operations,
 
1504
        self._execute_pack_operations(pack_operations,
1509
1505
                                      reload_func=self._restart_autopack)
1510
1506
        mutter('Auto-packing repository %s completed', self)
1511
 
        return result
 
1507
        return True
1512
1508
 
1513
1509
    def _execute_pack_operations(self, pack_operations, _packer_class=Packer,
1514
1510
                                 reload_func=None):
1516
1512
 
1517
1513
        :param pack_operations: A list of [revision_count, packs_to_combine].
1518
1514
        :param _packer_class: The class of packer to use (default: Packer).
1519
 
        :return: The new pack names.
 
1515
        :return: None.
1520
1516
        """
1521
1517
        for revision_count, packs in pack_operations:
1522
1518
            # we may have no-ops from the setup logic
1538
1534
                self._remove_pack_from_memory(pack)
1539
1535
        # record the newly available packs and stop advertising the old
1540
1536
        # packs
1541
 
        to_be_obsoleted = []
1542
 
        for _, packs in pack_operations:
1543
 
            to_be_obsoleted.extend(packs)
1544
 
        result = self._save_pack_names(clear_obsolete_packs=True,
1545
 
                                       obsolete_packs=to_be_obsoleted)
1546
 
        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)
1547
1541
 
1548
1542
    def _flush_new_pack(self):
1549
1543
        if self._new_pack is not None:
1559
1553
 
1560
1554
    def _already_packed(self):
1561
1555
        """Is the collection already packed?"""
1562
 
        return not (self.repo._format.pack_compresses or (len(self._names) > 1))
 
1556
        return len(self._names) < 2
1563
1557
 
1564
 
    def pack(self, hint=None, clean_obsolete_packs=False):
 
1558
    def pack(self):
1565
1559
        """Pack the pack collection totally."""
1566
1560
        self.ensure_loaded()
1567
1561
        total_packs = len(self._names)
1568
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.
1569
1566
            return
1570
1567
        total_revisions = self.revision_index.combined_index.key_count()
1571
1568
        # XXX: the following may want to be a class, to pack with a given
1572
1569
        # policy.
1573
1570
        mutter('Packing repository %s, which has %d pack files, '
1574
 
            'containing %d revisions with hint %r.', self, total_packs,
1575
 
            total_revisions, hint)
 
1571
            'containing %d revisions into 1 packs.', self, total_packs,
 
1572
            total_revisions)
1576
1573
        # determine which packs need changing
 
1574
        pack_distribution = [1]
1577
1575
        pack_operations = [[0, []]]
1578
1576
        for pack in self.all_packs():
1579
 
            if hint is None or pack.name in hint:
1580
 
                # Either no hint was provided (so we are packing everything),
1581
 
                # or this pack was included in the hint.
1582
 
                pack_operations[-1][0] += pack.get_revision_count()
1583
 
                pack_operations[-1][1].append(pack)
 
1577
            pack_operations[-1][0] += pack.get_revision_count()
 
1578
            pack_operations[-1][1].append(pack)
1584
1579
        self._execute_pack_operations(pack_operations, OptimisingPacker)
1585
1580
 
1586
 
        if clean_obsolete_packs:
1587
 
            self._clear_obsolete_packs()
1588
 
 
1589
1581
    def plan_autopack_combinations(self, existing_packs, pack_distribution):
1590
1582
        """Plan a pack operation.
1591
1583
 
1679
1671
            txt_index = self._make_index(name, '.tix')
1680
1672
            sig_index = self._make_index(name, '.six')
1681
1673
            if self.chk_index is not None:
1682
 
                chk_index = self._make_index(name, '.cix', unlimited_cache=True)
 
1674
                chk_index = self._make_index(name, '.cix')
1683
1675
            else:
1684
1676
                chk_index = None
1685
1677
            result = ExistingPack(self._pack_transport, name, rev_index,
1704
1696
            txt_index = self._make_index(name, '.tix', resume=True)
1705
1697
            sig_index = self._make_index(name, '.six', resume=True)
1706
1698
            if self.chk_index is not None:
1707
 
                chk_index = self._make_index(name, '.cix', resume=True,
1708
 
                                             unlimited_cache=True)
 
1699
                chk_index = self._make_index(name, '.cix', resume=True)
1709
1700
            else:
1710
1701
                chk_index = None
1711
1702
            result = self.resumed_pack_factory(name, rev_index, inv_index,
1741
1732
        return self._index_class(self.transport, 'pack-names', None
1742
1733
                ).iter_all_entries()
1743
1734
 
1744
 
    def _make_index(self, name, suffix, resume=False, unlimited_cache=False):
 
1735
    def _make_index(self, name, suffix, resume=False):
1745
1736
        size_offset = self._suffix_offsets[suffix]
1746
1737
        index_name = name + suffix
1747
1738
        if resume:
1750
1741
        else:
1751
1742
            transport = self._index_transport
1752
1743
            index_size = self._names[name][size_offset]
1753
 
        return self._index_class(transport, index_name, index_size,
1754
 
                                 unlimited_cache=unlimited_cache)
 
1744
        return self._index_class(transport, index_name, index_size)
1755
1745
 
1756
1746
    def _max_pack_count(self, total_revisions):
1757
1747
        """Return the maximum number of packs to use for total revisions.
1785
1775
        :param return: None.
1786
1776
        """
1787
1777
        for pack in packs:
1788
 
            try:
1789
 
                pack.pack_transport.rename(pack.file_name(),
1790
 
                    '../obsolete_packs/' + pack.file_name())
1791
 
            except (errors.PathError, errors.TransportError), e:
1792
 
                # TODO: Should these be warnings or mutters?
1793
 
                mutter("couldn't rename obsolete pack, skipping it:\n%s"
1794
 
                       % (e,))
 
1778
            pack.pack_transport.rename(pack.file_name(),
 
1779
                '../obsolete_packs/' + pack.file_name())
1795
1780
            # TODO: Probably needs to know all possible indices for this pack
1796
1781
            # - or maybe list the directory and move all indices matching this
1797
1782
            # name whether we recognize it or not?
1799
1784
            if self.chk_index is not None:
1800
1785
                suffixes.append('.cix')
1801
1786
            for suffix in suffixes:
1802
 
                try:
1803
 
                    self._index_transport.rename(pack.name + suffix,
1804
 
                        '../obsolete_packs/' + pack.name + suffix)
1805
 
                except (errors.PathError, errors.TransportError), e:
1806
 
                    mutter("couldn't rename obsolete index, skipping it:\n%s"
1807
 
                           % (e,))
 
1787
                self._index_transport.rename(pack.name + suffix,
 
1788
                    '../obsolete_packs/' + pack.name + suffix)
1808
1789
 
1809
1790
    def pack_distribution(self, total_revisions):
1810
1791
        """Generate a list of the number of revisions to put in each pack.
1836
1817
        self._remove_pack_indices(pack)
1837
1818
        self.packs.remove(pack)
1838
1819
 
1839
 
    def _remove_pack_indices(self, pack, ignore_missing=False):
1840
 
        """Remove the indices for pack from the aggregated indices.
1841
 
        
1842
 
        :param ignore_missing: Suppress KeyErrors from calling remove_index.
1843
 
        """
1844
 
        for index_type in Pack.index_definitions.keys():
1845
 
            attr_name = index_type + '_index'
1846
 
            aggregate_index = getattr(self, attr_name)
1847
 
            if aggregate_index is not None:
1848
 
                pack_index = getattr(pack, attr_name)
1849
 
                try:
1850
 
                    aggregate_index.remove_index(pack_index)
1851
 
                except KeyError:
1852
 
                    if ignore_missing:
1853
 
                        continue
1854
 
                    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)
1855
1828
 
1856
1829
    def reset(self):
1857
1830
        """Clear all cached data."""
1890
1863
        disk_nodes = set()
1891
1864
        for index, key, value in self._iter_disk_pack_index():
1892
1865
            disk_nodes.add((key, value))
1893
 
        orig_disk_nodes = set(disk_nodes)
1894
1866
 
1895
1867
        # do a two-way diff against our original content
1896
1868
        current_nodes = set()
1909
1881
        disk_nodes.difference_update(deleted_nodes)
1910
1882
        disk_nodes.update(new_nodes)
1911
1883
 
1912
 
        return disk_nodes, deleted_nodes, new_nodes, orig_disk_nodes
 
1884
        return disk_nodes, deleted_nodes, new_nodes
1913
1885
 
1914
1886
    def _syncronize_pack_names_from_disk_nodes(self, disk_nodes):
1915
1887
        """Given the correct set of pack files, update our saved info.
1955
1927
                added.append(name)
1956
1928
        return removed, added, modified
1957
1929
 
1958
 
    def _save_pack_names(self, clear_obsolete_packs=False, obsolete_packs=None):
 
1930
    def _save_pack_names(self, clear_obsolete_packs=False):
1959
1931
        """Save the list of packs.
1960
1932
 
1961
1933
        This will take out the mutex around the pack names list for the
1965
1937
 
1966
1938
        :param clear_obsolete_packs: If True, clear out the contents of the
1967
1939
            obsolete_packs directory.
1968
 
        :param obsolete_packs: Packs that are obsolete once the new pack-names
1969
 
            file has been written.
1970
 
        :return: A list of the names saved that were not previously on disk.
1971
1940
        """
1972
 
        already_obsolete = []
1973
1941
        self.lock_names()
1974
1942
        try:
1975
1943
            builder = self._index_builder_class()
1976
 
            (disk_nodes, deleted_nodes, new_nodes,
1977
 
             orig_disk_nodes) = self._diff_pack_names()
 
1944
            disk_nodes, deleted_nodes, new_nodes = self._diff_pack_names()
1978
1945
            # TODO: handle same-name, index-size-changes here -
1979
1946
            # e.g. use the value from disk, not ours, *unless* we're the one
1980
1947
            # changing it.
1982
1949
                builder.add_node(key, value)
1983
1950
            self.transport.put_file('pack-names', builder.finish(),
1984
1951
                mode=self.repo.bzrdir._get_file_mode())
 
1952
            # move the baseline forward
1985
1953
            self._packs_at_load = disk_nodes
1986
1954
            if clear_obsolete_packs:
1987
 
                to_preserve = None
1988
 
                if obsolete_packs:
1989
 
                    to_preserve = set([o.name for o in obsolete_packs])
1990
 
                already_obsolete = self._clear_obsolete_packs(to_preserve)
 
1955
                self._clear_obsolete_packs()
1991
1956
        finally:
1992
1957
            self._unlock_names()
1993
1958
        # synchronise the memory packs list with what we just wrote:
1994
1959
        self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1995
 
        if obsolete_packs:
1996
 
            # TODO: We could add one more condition here. "if o.name not in
1997
 
            #       orig_disk_nodes and o != the new_pack we haven't written to
1998
 
            #       disk yet. However, the new pack object is not easily
1999
 
            #       accessible here (it would have to be passed through the
2000
 
            #       autopacking code, etc.)
2001
 
            obsolete_packs = [o for o in obsolete_packs
2002
 
                              if o.name not in already_obsolete]
2003
 
            self._obsolete_packs(obsolete_packs)
2004
 
        return [new_node[0][0] for new_node in new_nodes]
2005
1960
 
2006
1961
    def reload_pack_names(self):
2007
1962
        """Sync our pack listing with what is present in the repository.
2021
1976
        if first_read:
2022
1977
            return True
2023
1978
        # out the new value.
2024
 
        (disk_nodes, deleted_nodes, new_nodes,
2025
 
         orig_disk_nodes) = self._diff_pack_names()
2026
 
        # _packs_at_load is meant to be the explicit list of names in
2027
 
        # 'pack-names' at then start. As such, it should not contain any
2028
 
        # pending names that haven't been written out yet.
2029
 
        self._packs_at_load = orig_disk_nodes
 
1979
        disk_nodes, _, _ = self._diff_pack_names()
 
1980
        self._packs_at_load = disk_nodes
2030
1981
        (removed, added,
2031
1982
         modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2032
1983
        if removed or added or modified:
2041
1992
            raise
2042
1993
        raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2043
1994
 
2044
 
    def _clear_obsolete_packs(self, preserve=None):
 
1995
    def _clear_obsolete_packs(self):
2045
1996
        """Delete everything from the obsolete-packs directory.
2046
 
 
2047
 
        :return: A list of pack identifiers (the filename without '.pack') that
2048
 
            were found in obsolete_packs.
2049
1997
        """
2050
 
        found = []
2051
1998
        obsolete_pack_transport = self.transport.clone('obsolete_packs')
2052
 
        if preserve is None:
2053
 
            preserve = set()
2054
1999
        for filename in obsolete_pack_transport.list_dir('.'):
2055
 
            name, ext = osutils.splitext(filename)
2056
 
            if ext == '.pack':
2057
 
                found.append(name)
2058
 
            if name in preserve:
2059
 
                continue
2060
2000
            try:
2061
2001
                obsolete_pack_transport.delete(filename)
2062
2002
            except (errors.PathError, errors.TransportError), e:
2063
 
                warning("couldn't delete obsolete pack, skipping it:\n%s"
2064
 
                        % (e,))
2065
 
        return found
 
2003
                warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2066
2004
 
2067
2005
    def _start_write_group(self):
2068
2006
        # Do not permit preparation for writing if we're not in a 'write lock'.
2095
2033
        # FIXME: just drop the transient index.
2096
2034
        # forget what names there are
2097
2035
        if self._new_pack is not None:
2098
 
            operation = cleanup.OperationWithCleanups(self._new_pack.abort)
2099
 
            operation.add_cleanup(setattr, self, '_new_pack', None)
2100
 
            # If we aborted while in the middle of finishing the write
2101
 
            # group, _remove_pack_indices could fail because the indexes are
2102
 
            # already gone.  But they're not there we shouldn't fail in this
2103
 
            # case, so we pass ignore_missing=True.
2104
 
            operation.add_cleanup(self._remove_pack_indices, self._new_pack,
2105
 
                ignore_missing=True)
2106
 
            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
2107
2045
        for resumed_pack in self._resumed_packs:
2108
 
            operation = cleanup.OperationWithCleanups(resumed_pack.abort)
2109
 
            # See comment in previous finally block.
2110
 
            operation.add_cleanup(self._remove_pack_indices, resumed_pack,
2111
 
                ignore_missing=True)
2112
 
            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
2113
2054
        del self._resumed_packs[:]
2114
2055
 
2115
2056
    def _remove_resumed_pack_indices(self):
2117
2058
            self._remove_pack_indices(resumed_pack)
2118
2059
        del self._resumed_packs[:]
2119
2060
 
2120
 
    def _check_new_inventories(self):
2121
 
        """Detect missing inventories in this write group.
2122
 
 
2123
 
        :returns: list of strs, summarising any problems found.  If the list is
2124
 
            empty no problems were found.
2125
 
        """
2126
 
        # The base implementation does no checks.  GCRepositoryPackCollection
2127
 
        # overrides this.
2128
 
        return []
2129
 
        
2130
2061
    def _commit_write_group(self):
2131
2062
        all_missing = set()
2132
2063
        for prefix, versioned_file in (
2141
2072
            raise errors.BzrCheckError(
2142
2073
                "Repository %s has missing compression parent(s) %r "
2143
2074
                 % (self.repo, sorted(all_missing)))
2144
 
        problems = self._check_new_inventories()
2145
 
        if problems:
2146
 
            problems_summary = '\n'.join(problems)
2147
 
            raise errors.BzrCheckError(
2148
 
                "Cannot add revision(s) to repository: " + problems_summary)
2149
2075
        self._remove_pack_indices(self._new_pack)
2150
 
        any_new_content = False
 
2076
        should_autopack = False
2151
2077
        if self._new_pack.data_inserted():
2152
2078
            # get all the data to disk and read to use
2153
2079
            self._new_pack.finish()
2154
2080
            self.allocate(self._new_pack)
2155
2081
            self._new_pack = None
2156
 
            any_new_content = True
 
2082
            should_autopack = True
2157
2083
        else:
2158
2084
            self._new_pack.abort()
2159
2085
            self._new_pack = None
2164
2090
            self._remove_pack_from_memory(resumed_pack)
2165
2091
            resumed_pack.finish()
2166
2092
            self.allocate(resumed_pack)
2167
 
            any_new_content = True
 
2093
            should_autopack = True
2168
2094
        del self._resumed_packs[:]
2169
 
        if any_new_content:
2170
 
            result = self.autopack()
2171
 
            if not result:
 
2095
        if should_autopack:
 
2096
            if not self.autopack():
2172
2097
                # when autopack takes no steps, the names list is still
2173
2098
                # unsaved.
2174
 
                return self._save_pack_names()
2175
 
            return result
2176
 
        return []
 
2099
                self._save_pack_names()
2177
2100
 
2178
2101
    def _suspend_write_group(self):
2179
2102
        tokens = [pack.name for pack in self._resumed_packs]
2281
2204
        self._reconcile_fixes_text_parents = True
2282
2205
        self._reconcile_backsup_inventory = False
2283
2206
 
2284
 
    def _warn_if_deprecated(self, branch=None):
 
2207
    def _warn_if_deprecated(self):
2285
2208
        # This class isn't deprecated, but one sub-format is
2286
2209
        if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2287
 
            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))
2288
2217
 
2289
2218
    def _abort_write_group(self):
2290
 
        self.revisions._index._key_dependencies.clear()
 
2219
        self.revisions._index._key_dependencies.refs.clear()
2291
2220
        self._pack_collection._abort_write_group()
2292
2221
 
2293
 
    def _get_source(self, to_format):
2294
 
        if to_format.network_name() == self._format.network_name():
2295
 
            return KnitPackStreamSource(self, to_format)
2296
 
        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
2297
2267
 
2298
2268
    def _make_parents_provider(self):
2299
2269
        return graph.CachingParentsProvider(self)
2307
2277
        self._pack_collection._start_write_group()
2308
2278
 
2309
2279
    def _commit_write_group(self):
2310
 
        hint = self._pack_collection._commit_write_group()
2311
 
        self.revisions._index._key_dependencies.clear()
2312
 
        return hint
 
2280
        self.revisions._index._key_dependencies.refs.clear()
 
2281
        return self._pack_collection._commit_write_group()
2313
2282
 
2314
2283
    def suspend_write_group(self):
2315
2284
        # XXX check self._write_group is self.get_transaction()?
2316
2285
        tokens = self._pack_collection._suspend_write_group()
2317
 
        self.revisions._index._key_dependencies.clear()
 
2286
        self.revisions._index._key_dependencies.refs.clear()
2318
2287
        self._write_group = None
2319
2288
        return tokens
2320
2289
 
2321
2290
    def _resume_write_group(self, tokens):
2322
2291
        self._start_write_group()
2323
 
        try:
2324
 
            self._pack_collection._resume_write_group(tokens)
2325
 
        except errors.UnresumableWriteGroup:
2326
 
            self._abort_write_group()
2327
 
            raise
 
2292
        self._pack_collection._resume_write_group(tokens)
2328
2293
        for pack in self._pack_collection._resumed_packs:
2329
2294
            self.revisions._index.scan_unvalidated_index(pack.revision_index)
2330
2295
 
2348
2313
        if self._write_lock_count == 1:
2349
2314
            self._transaction = transactions.WriteTransaction()
2350
2315
        if not locked:
2351
 
            if 'relock' in debug.debug_flags and self._prev_lock == 'w':
2352
 
                note('%r was write locked again', self)
2353
 
            self._prev_lock = 'w'
2354
2316
            for repo in self._fallback_repositories:
2355
2317
                # Writes don't affect fallback repos
2356
2318
                repo.lock_read()
2357
2319
            self._refresh_data()
2358
 
        return RepositoryWriteLockResult(self.unlock, None)
2359
2320
 
2360
2321
    def lock_read(self):
2361
2322
        locked = self.is_locked()
2364
2325
        else:
2365
2326
            self.control_files.lock_read()
2366
2327
        if not locked:
2367
 
            if 'relock' in debug.debug_flags and self._prev_lock == 'r':
2368
 
                note('%r was read locked again', self)
2369
 
            self._prev_lock = 'r'
2370
2328
            for repo in self._fallback_repositories:
2371
2329
                repo.lock_read()
2372
2330
            self._refresh_data()
2373
 
        return self
2374
2331
 
2375
2332
    def leave_lock_in_place(self):
2376
2333
        # not supported - raise an error
2381
2338
        raise NotImplementedError(self.dont_leave_lock_in_place)
2382
2339
 
2383
2340
    @needs_write_lock
2384
 
    def pack(self, hint=None, clean_obsolete_packs=False):
 
2341
    def pack(self):
2385
2342
        """Compress the data within the repository.
2386
2343
 
2387
2344
        This will pack all the data to a single pack. In future it may
2388
2345
        recompress deltas or do other such expensive operations.
2389
2346
        """
2390
 
        self._pack_collection.pack(hint=hint, clean_obsolete_packs=clean_obsolete_packs)
 
2347
        self._pack_collection.pack()
2391
2348
 
2392
2349
    @needs_write_lock
2393
2350
    def reconcile(self, other=None, thorough=False):
2401
2358
        packer = ReconcilePacker(collection, packs, extension, revs)
2402
2359
        return packer.pack(pb)
2403
2360
 
2404
 
    @only_raises(errors.LockNotHeld, errors.LockBroken)
2405
2361
    def unlock(self):
2406
2362
        if self._write_lock_count == 1 and self._write_group is not None:
2407
2363
            self.abort_write_group()
2424
2380
                repo.unlock()
2425
2381
 
2426
2382
 
2427
 
class KnitPackStreamSource(StreamSource):
2428
 
    """A StreamSource used to transfer data between same-format KnitPack repos.
2429
 
 
2430
 
    This source assumes:
2431
 
        1) Same serialization format for all objects
2432
 
        2) Same root information
2433
 
        3) XML format inventories
2434
 
        4) Atomic inserts (so we can stream inventory texts before text
2435
 
           content)
2436
 
        5) No chk_bytes
2437
 
    """
2438
 
 
2439
 
    def __init__(self, from_repository, to_format):
2440
 
        super(KnitPackStreamSource, self).__init__(from_repository, to_format)
2441
 
        self._text_keys = None
2442
 
        self._text_fetch_order = 'unordered'
2443
 
 
2444
 
    def _get_filtered_inv_stream(self, revision_ids):
2445
 
        from_repo = self.from_repository
2446
 
        parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
2447
 
        parent_keys = [(p,) for p in parent_ids]
2448
 
        find_text_keys = from_repo._find_text_key_references_from_xml_inventory_lines
2449
 
        parent_text_keys = set(find_text_keys(
2450
 
            from_repo._inventory_xml_lines_for_keys(parent_keys)))
2451
 
        content_text_keys = set()
2452
 
        knit = KnitVersionedFiles(None, None)
2453
 
        factory = KnitPlainFactory()
2454
 
        def find_text_keys_from_content(record):
2455
 
            if record.storage_kind not in ('knit-delta-gz', 'knit-ft-gz'):
2456
 
                raise ValueError("Unknown content storage kind for"
2457
 
                    " inventory text: %s" % (record.storage_kind,))
2458
 
            # It's a knit record, it has a _raw_record field (even if it was
2459
 
            # reconstituted from a network stream).
2460
 
            raw_data = record._raw_record
2461
 
            # read the entire thing
2462
 
            revision_id = record.key[-1]
2463
 
            content, _ = knit._parse_record(revision_id, raw_data)
2464
 
            if record.storage_kind == 'knit-delta-gz':
2465
 
                line_iterator = factory.get_linedelta_content(content)
2466
 
            elif record.storage_kind == 'knit-ft-gz':
2467
 
                line_iterator = factory.get_fulltext_content(content)
2468
 
            content_text_keys.update(find_text_keys(
2469
 
                [(line, revision_id) for line in line_iterator]))
2470
 
        revision_keys = [(r,) for r in revision_ids]
2471
 
        def _filtered_inv_stream():
2472
 
            source_vf = from_repo.inventories
2473
 
            stream = source_vf.get_record_stream(revision_keys,
2474
 
                                                 'unordered', False)
2475
 
            for record in stream:
2476
 
                if record.storage_kind == 'absent':
2477
 
                    raise errors.NoSuchRevision(from_repo, record.key)
2478
 
                find_text_keys_from_content(record)
2479
 
                yield record
2480
 
            self._text_keys = content_text_keys - parent_text_keys
2481
 
        return ('inventories', _filtered_inv_stream())
2482
 
 
2483
 
    def _get_text_stream(self):
2484
 
        # Note: We know we don't have to handle adding root keys, because both
2485
 
        # the source and target are the identical network name.
2486
 
        text_stream = self.from_repository.texts.get_record_stream(
2487
 
                        self._text_keys, self._text_fetch_order, False)
2488
 
        return ('texts', text_stream)
2489
 
 
2490
 
    def get_stream(self, search):
2491
 
        revision_ids = search.get_keys()
2492
 
        for stream_info in self._fetch_revision_texts(revision_ids):
2493
 
            yield stream_info
2494
 
        self._revision_keys = [(rev_id,) for rev_id in revision_ids]
2495
 
        yield self._get_filtered_inv_stream(revision_ids)
2496
 
        yield self._get_text_stream()
2497
 
 
2498
 
 
2499
 
 
2500
2383
class RepositoryFormatPack(MetaDirRepositoryFormat):
2501
2384
    """Format logic for pack structured repositories.
2502
2385
 
2549
2432
        utf8_files = [('format', self.get_format_string())]
2550
2433
 
2551
2434
        self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
2552
 
        repository = self.open(a_bzrdir=a_bzrdir, _found=True)
2553
 
        self._run_post_repo_init_hooks(repository, a_bzrdir, shared)
2554
 
        return repository
 
2435
        return self.open(a_bzrdir=a_bzrdir, _found=True)
2555
2436
 
2556
2437
    def open(self, a_bzrdir, _found=False, _override_transport=None):
2557
2438
        """See RepositoryFormat.open().
2606
2487
        """See RepositoryFormat.get_format_description()."""
2607
2488
        return "Packs containing knits without subtree support"
2608
2489
 
 
2490
    def check_conversion_target(self, target_format):
 
2491
        pass
 
2492
 
2609
2493
 
2610
2494
class RepositoryFormatKnitPack3(RepositoryFormatPack):
2611
2495
    """A subtrees parameterized Pack repository.
2620
2504
    repository_class = KnitPackRepository
2621
2505
    _commit_builder_class = PackRootCommitBuilder
2622
2506
    rich_root_data = True
2623
 
    experimental = True
2624
2507
    supports_tree_reference = True
2625
2508
    @property
2626
2509
    def _serializer(self):
2638
2521
 
2639
2522
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2640
2523
 
 
2524
    def check_conversion_target(self, target_format):
 
2525
        if not target_format.rich_root_data:
 
2526
            raise errors.BadConversionTarget(
 
2527
                'Does not support rich root data.', target_format)
 
2528
        if not getattr(target_format, 'supports_tree_reference', False):
 
2529
            raise errors.BadConversionTarget(
 
2530
                'Does not support nested trees', target_format)
 
2531
 
2641
2532
    def get_format_string(self):
2642
2533
        """See RepositoryFormat.get_format_string()."""
2643
2534
        return "Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n"
2676
2567
 
2677
2568
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2678
2569
 
 
2570
    def check_conversion_target(self, target_format):
 
2571
        if not target_format.rich_root_data:
 
2572
            raise errors.BadConversionTarget(
 
2573
                'Does not support rich root data.', target_format)
 
2574
 
2679
2575
    def get_format_string(self):
2680
2576
        """See RepositoryFormat.get_format_string()."""
2681
2577
        return ("Bazaar pack repository format 1 with rich root"
2722
2618
        """See RepositoryFormat.get_format_description()."""
2723
2619
        return "Packs 5 (adds stacking support, requires bzr 1.6)"
2724
2620
 
 
2621
    def check_conversion_target(self, target_format):
 
2622
        pass
 
2623
 
2725
2624
 
2726
2625
class RepositoryFormatKnitPack5RichRoot(RepositoryFormatPack):
2727
2626
    """A repository with rich roots and stacking.
2754
2653
 
2755
2654
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2756
2655
 
 
2656
    def check_conversion_target(self, target_format):
 
2657
        if not target_format.rich_root_data:
 
2658
            raise errors.BadConversionTarget(
 
2659
                'Does not support rich root data.', target_format)
 
2660
 
2757
2661
    def get_format_string(self):
2758
2662
        """See RepositoryFormat.get_format_string()."""
2759
2663
        return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n"
2800
2704
 
2801
2705
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2802
2706
 
 
2707
    def check_conversion_target(self, target_format):
 
2708
        if not target_format.rich_root_data:
 
2709
            raise errors.BadConversionTarget(
 
2710
                'Does not support rich root data.', target_format)
 
2711
 
2803
2712
    def get_format_string(self):
2804
2713
        """See RepositoryFormat.get_format_string()."""
2805
2714
        return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n"
2843
2752
        """See RepositoryFormat.get_format_description()."""
2844
2753
        return "Packs 6 (uses btree indexes, requires bzr 1.9)"
2845
2754
 
 
2755
    def check_conversion_target(self, target_format):
 
2756
        pass
 
2757
 
2846
2758
 
2847
2759
class RepositoryFormatKnitPack6RichRoot(RepositoryFormatPack):
2848
2760
    """A repository with rich roots, no subtrees, stacking and btree indexes.
2872
2784
 
2873
2785
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2874
2786
 
 
2787
    def check_conversion_target(self, target_format):
 
2788
        if not target_format.rich_root_data:
 
2789
            raise errors.BadConversionTarget(
 
2790
                'Does not support rich root data.', target_format)
 
2791
 
2875
2792
    def get_format_string(self):
2876
2793
        """See RepositoryFormat.get_format_string()."""
2877
2794
        return "Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n"
2894
2811
    repository_class = KnitPackRepository
2895
2812
    _commit_builder_class = PackRootCommitBuilder
2896
2813
    rich_root_data = True
2897
 
    experimental = True
2898
2814
    supports_tree_reference = True
2899
2815
    supports_external_lookups = True
2900
2816
    # What index classes to use
2914
2830
 
2915
2831
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2916
2832
 
 
2833
    def check_conversion_target(self, target_format):
 
2834
        if not target_format.rich_root_data:
 
2835
            raise errors.BadConversionTarget(
 
2836
                'Does not support rich root data.', target_format)
 
2837
        if not getattr(target_format, 'supports_tree_reference', False):
 
2838
            raise errors.BadConversionTarget(
 
2839
                'Does not support nested trees', target_format)
 
2840
 
2917
2841
    def get_format_string(self):
2918
2842
        """See RepositoryFormat.get_format_string()."""
2919
2843
        return ("Bazaar development format 2 with subtree support "