1
# Copyright (C) 2007-2010 Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
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
55
54
revision as _mod_revision,
58
from bzrlib.decorators import needs_write_lock, only_raises
57
from bzrlib.decorators import needs_write_lock
59
58
from bzrlib.btree_index import (
226
224
return self.index_name('text', name)
228
226
def _replace_index_with_readonly(self, index_type):
229
unlimited_cache = False
230
if index_type == 'chk':
231
unlimited_cache = True
232
227
setattr(self, index_type + '_index',
233
228
self.index_class(self.index_transport,
234
229
self.index_name(index_type, self.name),
235
self.index_sizes[self.index_offset(index_type)],
236
unlimited_cache=unlimited_cache))
230
self.index_sizes[self.index_offset(index_type)]))
239
233
class ExistingPack(Pack):
587
581
flush_func=flush_func)
588
582
self.add_callback = None
584
def replace_indices(self, index_to_pack, indices):
585
"""Replace the current mappings with fresh ones.
587
This should probably not be used eventually, rather incremental add and
588
removal of indices. It has been added during refactoring of existing
591
:param index_to_pack: A mapping from index objects to
592
(transport, name) tuples for the pack file data.
593
:param indices: A list of indices.
595
# refresh the revision pack map dict without replacing the instance.
596
self.index_to_pack.clear()
597
self.index_to_pack.update(index_to_pack)
598
# XXX: API break - clearly a 'replace' method would be good?
599
self.combined_index._indices[:] = indices
600
# the current add nodes callback for the current writable index if
602
self.add_callback = None
590
604
def add_index(self, index, pack):
591
605
"""Add index to the aggregate, which is an index for Pack pack.
599
613
# expose it to the index map
600
614
self.index_to_pack[index] = pack.access_tuple()
601
615
# put it at the front of the linear index list
602
self.combined_index.insert_index(0, index, pack.name)
616
self.combined_index.insert_index(0, index)
604
618
def add_writable_index(self, index, pack):
605
619
"""Add an index which is able to have data added to it.
625
639
self.data_access.set_writer(None, None, (None, None))
626
640
self.index_to_pack.clear()
627
641
del self.combined_index._indices[:]
628
del self.combined_index._index_names[:]
629
642
self.add_callback = None
631
def remove_index(self, index):
644
def remove_index(self, index, pack):
632
645
"""Remove index from the indices used to answer queries.
634
647
:param index: An index from the pack parameter.
648
:param pack: A Pack instance.
636
650
del self.index_to_pack[index]
637
pos = self.combined_index._indices.index(index)
638
del self.combined_index._indices[pos]
639
del self.combined_index._index_names[pos]
651
self.combined_index._indices.remove(index)
640
652
if (self.add_callback is not None and
641
653
getattr(index, 'add_nodes', None) == self.add_callback):
642
654
self.add_callback = None
1100
1112
iterator is a tuple with:
1101
1113
index, readv_vector, node_vector. readv_vector is a list ready to
1102
1114
hand to the transport readv method, and node_vector is a list of
1103
(key, eol_flag, references) for the node retrieved by the
1115
(key, eol_flag, references) for the the node retrieved by the
1104
1116
matching readv_vector.
1106
1118
# group by pack so we do one readv per pack
1398
1410
self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1399
1411
self.text_index = AggregateIndex(self.reload_pack_names, flush)
1400
1412
self.signature_index = AggregateIndex(self.reload_pack_names, flush)
1401
all_indices = [self.revision_index, self.inventory_index,
1402
self.text_index, self.signature_index]
1403
1413
if use_chk_index:
1404
1414
self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1405
all_indices.append(self.chk_index)
1407
1416
# used to determine if we're using a chk_index elsewhere.
1408
1417
self.chk_index = None
1409
# Tell all the CombinedGraphIndex objects about each other, so they can
1410
# share hints about which pack names to search first.
1411
all_combined = [agg_idx.combined_index for agg_idx in all_indices]
1412
for combined_idx in all_combined:
1413
combined_idx.set_sibling_indices(
1414
set(all_combined).difference([combined_idx]))
1415
1418
# resumed packs
1416
1419
self._resumed_packs = []
1419
return '%s(%r)' % (self.__class__.__name__, self.repo)
1421
1421
def add_pack_to_memory(self, pack):
1422
1422
"""Make a Pack object available to the repository to satisfy queries.
1537
1537
self._remove_pack_from_memory(pack)
1538
1538
# record the newly available packs and stop advertising the old
1540
to_be_obsoleted = []
1541
for _, packs in pack_operations:
1542
to_be_obsoleted.extend(packs)
1543
result = self._save_pack_names(clear_obsolete_packs=True,
1544
obsolete_packs=to_be_obsoleted)
1540
result = self._save_pack_names(clear_obsolete_packs=True)
1541
# Move the old packs out of the way now they are no longer referenced.
1542
for revision_count, packs in pack_operations:
1543
self._obsolete_packs(packs)
1547
1546
def _flush_new_pack(self):
1560
1559
"""Is the collection already packed?"""
1561
1560
return not (self.repo._format.pack_compresses or (len(self._names) > 1))
1563
def pack(self, hint=None, clean_obsolete_packs=False):
1562
def pack(self, hint=None):
1564
1563
"""Pack the pack collection totally."""
1565
1564
self.ensure_loaded()
1566
1565
total_packs = len(self._names)
1582
1581
pack_operations[-1][1].append(pack)
1583
1582
self._execute_pack_operations(pack_operations, OptimisingPacker)
1585
if clean_obsolete_packs:
1586
self._clear_obsolete_packs()
1588
1584
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1589
1585
"""Plan a pack operation.
1678
1674
txt_index = self._make_index(name, '.tix')
1679
1675
sig_index = self._make_index(name, '.six')
1680
1676
if self.chk_index is not None:
1681
chk_index = self._make_index(name, '.cix', unlimited_cache=True)
1677
chk_index = self._make_index(name, '.cix')
1683
1679
chk_index = None
1684
1680
result = ExistingPack(self._pack_transport, name, rev_index,
1703
1699
txt_index = self._make_index(name, '.tix', resume=True)
1704
1700
sig_index = self._make_index(name, '.six', resume=True)
1705
1701
if self.chk_index is not None:
1706
chk_index = self._make_index(name, '.cix', resume=True,
1707
unlimited_cache=True)
1702
chk_index = self._make_index(name, '.cix', resume=True)
1709
1704
chk_index = None
1710
1705
result = self.resumed_pack_factory(name, rev_index, inv_index,
1740
1735
return self._index_class(self.transport, 'pack-names', None
1741
1736
).iter_all_entries()
1743
def _make_index(self, name, suffix, resume=False, unlimited_cache=False):
1738
def _make_index(self, name, suffix, resume=False):
1744
1739
size_offset = self._suffix_offsets[suffix]
1745
1740
index_name = name + suffix
1750
1745
transport = self._index_transport
1751
1746
index_size = self._names[name][size_offset]
1752
return self._index_class(transport, index_name, index_size,
1753
unlimited_cache=unlimited_cache)
1747
return self._index_class(transport, index_name, index_size)
1755
1749
def _max_pack_count(self, total_revisions):
1756
1750
"""Return the maximum number of packs to use for total revisions.
1784
1778
:param return: None.
1786
1780
for pack in packs:
1788
pack.pack_transport.rename(pack.file_name(),
1789
'../obsolete_packs/' + pack.file_name())
1790
except (errors.PathError, errors.TransportError), e:
1791
# TODO: Should these be warnings or mutters?
1792
mutter("couldn't rename obsolete pack, skipping it:\n%s"
1781
pack.pack_transport.rename(pack.file_name(),
1782
'../obsolete_packs/' + pack.file_name())
1794
1783
# TODO: Probably needs to know all possible indices for this pack
1795
1784
# - or maybe list the directory and move all indices matching this
1796
1785
# name whether we recognize it or not?
1798
1787
if self.chk_index is not None:
1799
1788
suffixes.append('.cix')
1800
1789
for suffix in suffixes:
1802
self._index_transport.rename(pack.name + suffix,
1803
'../obsolete_packs/' + pack.name + suffix)
1804
except (errors.PathError, errors.TransportError), e:
1805
mutter("couldn't rename obsolete index, skipping it:\n%s"
1790
self._index_transport.rename(pack.name + suffix,
1791
'../obsolete_packs/' + pack.name + suffix)
1808
1793
def pack_distribution(self, total_revisions):
1809
1794
"""Generate a list of the number of revisions to put in each pack.
1835
1820
self._remove_pack_indices(pack)
1836
1821
self.packs.remove(pack)
1838
def _remove_pack_indices(self, pack, ignore_missing=False):
1839
"""Remove the indices for pack from the aggregated indices.
1841
:param ignore_missing: Suppress KeyErrors from calling remove_index.
1843
for index_type in Pack.index_definitions.keys():
1844
attr_name = index_type + '_index'
1845
aggregate_index = getattr(self, attr_name)
1846
if aggregate_index is not None:
1847
pack_index = getattr(pack, attr_name)
1849
aggregate_index.remove_index(pack_index)
1823
def _remove_pack_indices(self, pack):
1824
"""Remove the indices for pack from the aggregated indices."""
1825
self.revision_index.remove_index(pack.revision_index, pack)
1826
self.inventory_index.remove_index(pack.inventory_index, pack)
1827
self.text_index.remove_index(pack.text_index, pack)
1828
self.signature_index.remove_index(pack.signature_index, pack)
1829
if self.chk_index is not None:
1830
self.chk_index.remove_index(pack.chk_index, pack)
1855
1832
def reset(self):
1856
1833
"""Clear all cached data."""
1889
1866
disk_nodes = set()
1890
1867
for index, key, value in self._iter_disk_pack_index():
1891
1868
disk_nodes.add((key, value))
1892
orig_disk_nodes = set(disk_nodes)
1894
1870
# do a two-way diff against our original content
1895
1871
current_nodes = set()
1908
1884
disk_nodes.difference_update(deleted_nodes)
1909
1885
disk_nodes.update(new_nodes)
1911
return disk_nodes, deleted_nodes, new_nodes, orig_disk_nodes
1887
return disk_nodes, deleted_nodes, new_nodes
1913
1889
def _syncronize_pack_names_from_disk_nodes(self, disk_nodes):
1914
1890
"""Given the correct set of pack files, update our saved info.
1954
1930
added.append(name)
1955
1931
return removed, added, modified
1957
def _save_pack_names(self, clear_obsolete_packs=False, obsolete_packs=None):
1933
def _save_pack_names(self, clear_obsolete_packs=False):
1958
1934
"""Save the list of packs.
1960
1936
This will take out the mutex around the pack names list for the
1965
1941
:param clear_obsolete_packs: If True, clear out the contents of the
1966
1942
obsolete_packs directory.
1967
:param obsolete_packs: Packs that are obsolete once the new pack-names
1968
file has been written.
1969
1943
:return: A list of the names saved that were not previously on disk.
1971
already_obsolete = []
1972
1945
self.lock_names()
1974
1947
builder = self._index_builder_class()
1975
(disk_nodes, deleted_nodes, new_nodes,
1976
orig_disk_nodes) = self._diff_pack_names()
1948
disk_nodes, deleted_nodes, new_nodes = self._diff_pack_names()
1977
1949
# TODO: handle same-name, index-size-changes here -
1978
1950
# e.g. use the value from disk, not ours, *unless* we're the one
1981
1953
builder.add_node(key, value)
1982
1954
self.transport.put_file('pack-names', builder.finish(),
1983
1955
mode=self.repo.bzrdir._get_file_mode())
1956
# move the baseline forward
1984
1957
self._packs_at_load = disk_nodes
1985
1958
if clear_obsolete_packs:
1988
to_preserve = set([o.name for o in obsolete_packs])
1989
already_obsolete = self._clear_obsolete_packs(to_preserve)
1959
self._clear_obsolete_packs()
1991
1961
self._unlock_names()
1992
1962
# synchronise the memory packs list with what we just wrote:
1993
1963
self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1995
# TODO: We could add one more condition here. "if o.name not in
1996
# orig_disk_nodes and o != the new_pack we haven't written to
1997
# disk yet. However, the new pack object is not easily
1998
# accessible here (it would have to be passed through the
1999
# autopacking code, etc.)
2000
obsolete_packs = [o for o in obsolete_packs
2001
if o.name not in already_obsolete]
2002
self._obsolete_packs(obsolete_packs)
2003
1964
return [new_node[0][0] for new_node in new_nodes]
2005
1966
def reload_pack_names(self):
2022
1983
# out the new value.
2023
(disk_nodes, deleted_nodes, new_nodes,
2024
orig_disk_nodes) = self._diff_pack_names()
2025
# _packs_at_load is meant to be the explicit list of names in
2026
# 'pack-names' at then start. As such, it should not contain any
2027
# pending names that haven't been written out yet.
2028
self._packs_at_load = orig_disk_nodes
1984
disk_nodes, _, _ = self._diff_pack_names()
1985
self._packs_at_load = disk_nodes
2029
1986
(removed, added,
2030
1987
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2031
1988
if removed or added or modified:
2041
1998
raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2043
def _clear_obsolete_packs(self, preserve=None):
2000
def _clear_obsolete_packs(self):
2044
2001
"""Delete everything from the obsolete-packs directory.
2046
:return: A list of pack identifiers (the filename without '.pack') that
2047
were found in obsolete_packs.
2050
2003
obsolete_pack_transport = self.transport.clone('obsolete_packs')
2051
if preserve is None:
2053
2004
for filename in obsolete_pack_transport.list_dir('.'):
2054
name, ext = osutils.splitext(filename)
2057
if name in preserve:
2060
2006
obsolete_pack_transport.delete(filename)
2061
2007
except (errors.PathError, errors.TransportError), e:
2062
warning("couldn't delete obsolete pack, skipping it:\n%s"
2008
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2066
2010
def _start_write_group(self):
2067
2011
# Do not permit preparation for writing if we're not in a 'write lock'.
2094
2038
# FIXME: just drop the transient index.
2095
2039
# forget what names there are
2096
2040
if self._new_pack is not None:
2097
operation = cleanup.OperationWithCleanups(self._new_pack.abort)
2098
operation.add_cleanup(setattr, self, '_new_pack', None)
2099
# If we aborted while in the middle of finishing the write
2100
# group, _remove_pack_indices could fail because the indexes are
2101
# already gone. But they're not there we shouldn't fail in this
2102
# case, so we pass ignore_missing=True.
2103
operation.add_cleanup(self._remove_pack_indices, self._new_pack,
2104
ignore_missing=True)
2105
operation.run_simple()
2042
self._new_pack.abort()
2044
# XXX: If we aborted while in the middle of finishing the write
2045
# group, _remove_pack_indices can fail because the indexes are
2046
# already gone. If they're not there we shouldn't fail in this
2047
# case. -- mbp 20081113
2048
self._remove_pack_indices(self._new_pack)
2049
self._new_pack = None
2106
2050
for resumed_pack in self._resumed_packs:
2107
operation = cleanup.OperationWithCleanups(resumed_pack.abort)
2108
# See comment in previous finally block.
2109
operation.add_cleanup(self._remove_pack_indices, resumed_pack,
2110
ignore_missing=True)
2111
operation.run_simple()
2052
resumed_pack.abort()
2054
# See comment in previous finally block.
2056
self._remove_pack_indices(resumed_pack)
2112
2059
del self._resumed_packs[:]
2114
2061
def _remove_resumed_pack_indices(self):
2116
2063
self._remove_pack_indices(resumed_pack)
2117
2064
del self._resumed_packs[:]
2119
def _check_new_inventories(self):
2120
"""Detect missing inventories in this write group.
2122
:returns: list of strs, summarising any problems found. If the list is
2123
empty no problems were found.
2125
# The base implementation does no checks. GCRepositoryPackCollection
2129
2066
def _commit_write_group(self):
2130
2067
all_missing = set()
2131
2068
for prefix, versioned_file in (
2140
2077
raise errors.BzrCheckError(
2141
2078
"Repository %s has missing compression parent(s) %r "
2142
2079
% (self.repo, sorted(all_missing)))
2143
problems = self._check_new_inventories()
2145
problems_summary = '\n'.join(problems)
2146
raise errors.BzrCheckError(
2147
"Cannot add revision(s) to repository: " + problems_summary)
2148
2080
self._remove_pack_indices(self._new_pack)
2149
any_new_content = False
2081
should_autopack = False
2150
2082
if self._new_pack.data_inserted():
2151
2083
# get all the data to disk and read to use
2152
2084
self._new_pack.finish()
2153
2085
self.allocate(self._new_pack)
2154
2086
self._new_pack = None
2155
any_new_content = True
2087
should_autopack = True
2157
2089
self._new_pack.abort()
2158
2090
self._new_pack = None
2163
2095
self._remove_pack_from_memory(resumed_pack)
2164
2096
resumed_pack.finish()
2165
2097
self.allocate(resumed_pack)
2166
any_new_content = True
2098
should_autopack = True
2167
2099
del self._resumed_packs[:]
2169
result = self.autopack()
2101
if not self.autopack():
2171
2102
# when autopack takes no steps, the names list is still
2173
2104
return self._save_pack_names()
2177
2107
def _suspend_write_group(self):
2280
2210
self._reconcile_fixes_text_parents = True
2281
2211
self._reconcile_backsup_inventory = False
2283
def _warn_if_deprecated(self, branch=None):
2213
def _warn_if_deprecated(self):
2284
2214
# This class isn't deprecated, but one sub-format is
2285
2215
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2286
super(KnitPackRepository, self)._warn_if_deprecated(branch)
2216
from bzrlib import repository
2217
if repository._deprecation_warning_done:
2219
repository._deprecation_warning_done = True
2220
warning("Format %s for %s is deprecated - please use"
2221
" 'bzr upgrade --1.6.1-rich-root'"
2222
% (self._format, self.bzrdir.transport.base))
2288
2224
def _abort_write_group(self):
2289
self.revisions._index._key_dependencies.clear()
2225
self.revisions._index._key_dependencies.refs.clear()
2290
2226
self._pack_collection._abort_write_group()
2292
2228
def _get_source(self, to_format):
2306
2242
self._pack_collection._start_write_group()
2308
2244
def _commit_write_group(self):
2309
hint = self._pack_collection._commit_write_group()
2310
self.revisions._index._key_dependencies.clear()
2245
self.revisions._index._key_dependencies.refs.clear()
2246
return self._pack_collection._commit_write_group()
2313
2248
def suspend_write_group(self):
2314
2249
# XXX check self._write_group is self.get_transaction()?
2315
2250
tokens = self._pack_collection._suspend_write_group()
2316
self.revisions._index._key_dependencies.clear()
2251
self.revisions._index._key_dependencies.refs.clear()
2317
2252
self._write_group = None
2347
2282
if self._write_lock_count == 1:
2348
2283
self._transaction = transactions.WriteTransaction()
2350
if 'relock' in debug.debug_flags and self._prev_lock == 'w':
2351
note('%r was write locked again', self)
2352
self._prev_lock = 'w'
2353
2285
for repo in self._fallback_repositories:
2354
2286
# Writes don't affect fallback repos
2355
2287
repo.lock_read()
2363
2295
self.control_files.lock_read()
2365
if 'relock' in debug.debug_flags and self._prev_lock == 'r':
2366
note('%r was read locked again', self)
2367
self._prev_lock = 'r'
2368
2297
for repo in self._fallback_repositories:
2369
2298
repo.lock_read()
2370
2299
self._refresh_data()
2378
2307
raise NotImplementedError(self.dont_leave_lock_in_place)
2380
2309
@needs_write_lock
2381
def pack(self, hint=None, clean_obsolete_packs=False):
2310
def pack(self, hint=None):
2382
2311
"""Compress the data within the repository.
2384
2313
This will pack all the data to a single pack. In future it may
2385
2314
recompress deltas or do other such expensive operations.
2387
self._pack_collection.pack(hint=hint, clean_obsolete_packs=clean_obsolete_packs)
2316
self._pack_collection.pack(hint=hint)
2389
2318
@needs_write_lock
2390
2319
def reconcile(self, other=None, thorough=False):
2398
2327
packer = ReconcilePacker(collection, packs, extension, revs)
2399
2328
return packer.pack(pb)
2401
@only_raises(errors.LockNotHeld, errors.LockBroken)
2402
2330
def unlock(self):
2403
2331
if self._write_lock_count == 1 and self._write_group is not None:
2404
2332
self.abort_write_group()
2546
2474
utf8_files = [('format', self.get_format_string())]
2548
2476
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
2549
repository = self.open(a_bzrdir=a_bzrdir, _found=True)
2550
self._run_post_repo_init_hooks(repository, a_bzrdir, shared)
2477
return self.open(a_bzrdir=a_bzrdir, _found=True)
2553
2479
def open(self, a_bzrdir, _found=False, _override_transport=None):
2554
2480
"""See RepositoryFormat.open().
2891
2816
repository_class = KnitPackRepository
2892
2817
_commit_builder_class = PackRootCommitBuilder
2893
2818
rich_root_data = True
2895
2819
supports_tree_reference = True
2896
2820
supports_external_lookups = True
2897
2821
# What index classes to use
added
removed
bzrlib/repofmt/pack_repo.py
